Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Priority
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d). The certified copy has been filed in parent Application No. DE10 2023 200 851.3, filed on 02/02/2023.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 02/29/2024 is being considered by the examiner.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-10 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
The term “distinct” in claim 1 is a relative term which renders the claim indefinite. The term “distinct” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. Examiner suggests to applicant’s attorney to briefly disclose in the claims how some clusters are distinct from each other.
The term “partially” in claim 2 is a relative term which renders the claim indefinite. The term “partially” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. Examiner suggests to applicant’s attorney to briefly disclose in the claims what is considered partial.
Claim 4 recites the limitation "the same data" in line 1. There is insufficient antecedent basis for this limitation in the claim.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claim 9 rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claim(s) does/do not fall within at least one of the four categories of patent eligible subject matter because it claims a computer program, which does not fall within at least one of the four categories. Examiner suggests to applicant’s attorney to amend claim 9 to include a non-transitory computer readable medium. Support can be found in [0015] of applicant’s specification.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claim(s) 1-5, 7-10 are rejected under 35 U.S.C. 102 (a)(1) as being anticipated by Bidault (US 20210150688 A1).
Regarding claims 1, 9 and 10, Bidault discloses determining multiple wafer maps comprising indications of anomalies of the wafers ([0033] During a wafer manufacturing process, a statistically significant sampling of wafers may be performed at various steps of the process. For example, after each deposition, etching, stripping, cleaning, etc., process. Individual inspection stations may be added to the processing line to sample the wafers. The wafers may be manually examined using visual inspection. Wafer defect maps may be generated and operators may examine the wafer defect maps and consider additional production parameters.);
performing an evaluation based on the determined wafer maps ([0036] An automated inspection system may generate a wafer defect map containing coordinates of each defect of a wafer die of the sampling. The defect map may take the form of a data file, such as a Klarf™ file. A WDM may include defect patterns, such as a specific spatial arrangement of defects within the wafer. The patterns may be analyzed to determine a root cause of a defect or defects within the wafer.); and
performing a failure analysis based on the evaluation performed ("[0005]The wafers may include physical defects which may cause one or more chips of the wafer to fail. Defects may include voids, craters, protrusions, bridges, particles, etc.
[0039] Deep-learning techniques also may be employed. However, conventionally such techniques are bound to the dimensions of the WDMs, do not distinguish between multiple classes, and may employ thousands of failure patterns and causes."),
wherein performing the evaluation comprises:
multiple execution of a cluster analysis based on the determined wafer maps using different parameters ("[0037] For example, clustering algorithms may be employed. Clusters of wafers may be created with maximize intraclass similarity and maximize interclass diversity. Clustering may be viewed as related to template matching.
[0043] Conventionally, a training set of actual or real WDMs is provided which has labels and may have tags. FIG. 2 illustrates an embodiment of a set of defined classes into which WDMs may be classified. As illustrated, the set of classes comprises twelve defined classes including a normal class, a big cluster class, a half-moon class, a donut class, a grid class, a ring class, a fingerprints class, an incomplete class, a geometric scratch class, a small cluster class, a zig-zig class, and a basketball class. The labels identify a class of a defined class of WDM classes (e.g., one of the twelve classes of FIG. 2). The tags identify a root cause associated with WDM of the training set (e.g., a particular problem with a particular machine), and may include other information as well, such as identifiers of the machines used to process the wafer."), and
identifying distinct clusters determined by the differently parameterized cluster analyses ("[0014] In an embodiment, the contents comprise parameters of the data-driven model. In an embodiment, the data driven model associates AWDIs with root causes of wafer defects and the generating the defect classification associated with the WDM comprises generating a label identifying a class of the defined set of classes associated with the WDM and a tag identifying a root cause associated with the WDM.
[0046] The training process of a CNN, such as a deep CNN, to predict classes may produce a model which suffers from overfitting (e.g., the CNN learns from the training set so well that the CNN cannot generalize to new data), or from over-prediction of the majority class (e.g., the CNN is more likely to predict the majority class and yet maintain a high accuracy rate). In addition, representing the WDMs as images can result in large data files (e.g., 20,000 by 20,000 pixel images), which may be difficult to analyze using a CNN. Embodiments may employ various techniques to address or reduce the impact of such issues.").
Regarding claim 2, Bidault discloses the different parameterization of the cluster analyses is performed such that clusters determined from the different executions of the cluster analysis are partially different from one another in order to identify the distinct clusters ("[0014] In an embodiment, the contents comprise parameters of the data-driven model. In an embodiment, the data driven model associates AWDIs with root causes of wafer defects and the generating the defect classification associated with the WDM comprises generating a label identifying a class of the defined set of classes associated with the WDM and a tag identifying a root cause associated with the WDM.
[0046] The training process of a CNN, such as a deep CNN, to predict classes may produce a model which suffers from overfitting (e.g., the CNN learns from the training set so well that the CNN cannot generalize to new data), or from over-prediction of the majority class (e.g., the CNN is more likely to predict the majority class and yet maintain a high accuracy rate). In addition, representing the WDMs as images can result in large data files (e.g., 20,000 by 20,000 pixel images), which may be difficult to analyze using a CNN. Embodiments may employ various techniques to address or reduce the impact of such issues.");
further clusters are determined multiple times from the determined clusters ("[0043] As illustrated, the set of classes comprises twelve defined classes including a normal class, a big cluster class, a half-moon class, a donut class, a grid class, a ring class, a fingerprints class, an incomplete class, a geometric scratch class, a small cluster class, a zig-zig class, and a basketball class. The labels identify a class of a defined class of WDM classes (e.g., one of the twelve classes of FIG. 2). The tags identify a root cause associated with WDM of the training set (e.g., a particular problem with a particular machine), and may include other information as well, such as identifiers of the machines used to process the wafer.
[0044] Conventionally, the training set of WDMs comprises actual WDMs which may be augmented or oversampled to address imbalances in the number of samples of each class and overtraining issues. The WDMs are converted to images which are used to train the CNN. The CNN generates a data-driven model which matches an input WDM to a label corresponding to a class. In an embodiment, the training set also or instead trains the CNN to generate a data-driven model which matches an input WDM to a tag. A testing phase may be employed in which the CNN is tested with a new set of WDMs.
[0061] FIG. 5 illustrates example classes of AWDIs that may be employed to train a WDM classification system, such as the classification system 300 of FIG. 3. As illustrated, fifty-two classes of artificial images have been coded. For classifying wafer defect maps, it may be desirable to have 100 artificial image classes or more. For each coded class, a plurality of AWDI images, for example, 2000 images, may be generated. "); and
the identified clusters and the identified further clusters are used for fault analysis ([0045] After the training (and testing), WDMs generated during a fabrication process are represented as image data and provided to the CNN. The data-driven model learned by the CNN is used to predict/identify defect root causes of defects associated with the WDMs generated during the fabrication process. For example, the CNN may predict a class (or a plurality of class) and a tag to associate with a WDM produced during a fabrication process using the trained model. In another example, the CNN may predict/identify a class (or a plurality of classes) to associate with a WDM generated during the fabrication process based on the trained model, and use a similarity test to associate a tag of a training WDM having the predicted class(s) which is most similar to the WDM generated during the fabrication process with the WDM generated during the fabrication process.).
Regarding claim 3, Bidault discloses clusters determined from the different executions of the cluster analysis partially differ from one another ("[0036] An automated inspection system may generate a wafer defect map containing coordinates of each defect of a wafer die of the sampling. The defect map may take the form of a data file, such as a Klarf™ file. A WDM may include defect patterns, such as a specific spatial arrangement of defects within the wafer. The patterns may be analyzed to determine a root cause of a defect or defects within the wafer.
[0037] For example, clustering algorithms may be employed. Clusters of wafers may be created with maximize intraclass similarity and maximize interclass diversity. Clustering may be viewed as related to template matching.
[0038] In another example, feature extraction with classification based on defined features extracted from the WDMs may be employed. A feature is a discriminative characteristic that a classifier can learn to distinguish WDMs."); and
clusters determined multiple times are combined ([0050] The system 300 may include more components than illustrated, may include fewer components that illustrated, may combine or split components in various manners, and may have configurations other than the illustrated configuration. For example, in some embodiments, AWDIs output by the artificial image generation circuitry 302 may be provided to the WDM to image generation circuitry 304 for processing before being provided to the classification circuitry 308.).
Regarding claim 4, Bidault discloses the same data are analyzed by the different cluster analyses ("[0055] The original data values of a WDM (or of an AWDI in a training phase) are replaced by a count of defects that fall into a small region, or bin, of the WDM. Fixed binning, where the size of each bin is fixed, or adaptive binning may be employed. In fixed binning, the wafer map may be divided into a uniformly spaced grid.
[0056] In adaptive binning, the wafer is split into intervals of different dimensions with the expected value of defects constant in each bin. The size of each bin is inversely proportional to the density of defects and smaller bins are used to describe high-density defect regions of the WDM, which results in higher resolution images.
[0079] In contrast, when layer of inspection information is provided to a multi-input CNN 1110, the accurate of the prediction 1112 is much higher, 96%.
[0080] Type of inspection information indicates to a layer of a CNN, such as layer Fc2 of FIG. 10, that a particular type of inspection is to be performed, e.g., a full map inspection, one row over two rows, etc. FIG. 12 illustrates a full map inspection and a one row over two rows inspection. Indicating a one row over two inspection indicates to the CNN that the inspection provides only a partial view of a defect pattern.");
the different cluster analyses differ in terms of their parameterization ("[0055] The original data values of a WDM (or of an AWDI in a training phase) are replaced by a count of defects that fall into a small region, or bin, of the WDM. Fixed binning, where the size of each bin is fixed, or adaptive binning may be employed. In fixed binning, the wafer map may be divided into a uniformly spaced grid.
[0056] In adaptive binning, the wafer is split into intervals of different dimensions with the expected value of defects constant in each bin. The size of each bin is inversely proportional to the density of defects and smaller bins are used to describe high-density defect regions of the WDM, which results in higher resolution images.
[0058] In an embodiment, the saturation parameter s and the transform parameter γ may be manually selected. In an embodiment, the classifier may learn to select its own thresholds. For example, the saturation parameter s may be set to 255, and additional convolutional layers (e.g., two layers) may be added to the CNN so that the model learns a contrast enhancement function. (clusters are put in different categories or “bins” according to the measured parameters.)"); and
the data results from the determined wafer maps ("[0085] At 1310, the method 1300 optionally performs further processing based on the classification of the image. For example, when the image is classified into particular classes, a warning signal may be generated. In another example, a count of images in the class may be incremented, and further action taken when the count exceeds a threshold value. Statistical analysis may be performed in determining whether to perform further processing based on the classification of the image. The method proceeds from 1310 to 1312.
[0086] At 1312, the method 1300 determines whether the classification of WDMs is finished (e.g., whether there are additional WDMs to process). When it is determined at 1312 that classification of WDMs is not finished, the method 1300 returns to 1304 to receive or retrieve the next WDM. When it is determined at 1312 that the processing of WDMs is finished, the method 1300 proceeds to 1314, where further processing, such as a return of the results of classifying a set of WDMs, may be performed.").
Regarding claim 5, Bidault discloses multiple different wafers from a wafer production process are provided as the wafers (fig. 2 and [0041] In an embodiment, WDMs may be generated from representative wafers (e.g., a statistically significant sampling) at various points during a wafer fabrication process. The WDMs are represented as image data and a deep neural network (DNN), such as a convolutional neural network (CNN), employs image classification techniques to identify root causes of defects associated with the wafers.); and
the clusters are each identified as signatures specific for consistent anomalies in multiple different wafers (fig.2 and fig. 7).
Regarding claim 7, Bidault discloses the respective wafer map results from a metrological detection of the anomalies on a wafer during wafer production ([0033] During a wafer manufacturing process, a statistically significant sampling of wafers may be performed at various steps of the process. For example, after each deposition, etching, stripping, cleaning, etc., process. Individual inspection stations may be added to the processing line to sample the wafers. The wafers may be manually examined using visual inspection. Wafer defect maps may be generated and operators may examine the wafer defect maps and consider additional production parameters.).
Regarding claim 8, Bidault discloses wherein the respective wafer is designed as a silicon wafer and/or a metal wafer ([0005] DCNNs can be applied to manufacturing processes. For example, semiconductor wafers are generally manufactured using production lines. Each semiconductor wafer may include a plurality of chips, which are separated or cut from one another as part of the production process. (a semiconductor is made of silicon)).
Allowable Subject Matter
Claim 6 would be allowable if rewritten to overcome the rejection(s) under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), 2nd paragraph, set forth in this Office action and to include all of the limitations of the base claim and any intervening claims.
The following is a statement of reasons for the indication of allowable subject matter: Prior art does not disclose or teach the unique combinations of at least one of the wafer maps are combined to form a matrix, the wafer maps are combined using a dimensional extension, the wafer maps are combined using a subsequent dimensional reduction, and the wafer maps are combined using hyperparameters, in order to analyze the combined wafer maps of the different wafers using the different cluster analyses.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US 11741596 B2 of claim 4: col 12, lines 20-28: The first to n-th augmenting modules 143_1 to 143_n may be configured to expand reference maps of different fault types, respectively. In other words, the first to n-th augmenting modules 143_1 to 143_n may be configured to expand reference maps based on different parameters corresponding to the different fault types. The pre-processing device 140-3 may select an augmenting module based on a fault type of a reference map.
US 20230289949 A1 of claim 1: Abstract: The processor controls to learn the acquired image data to output a defect attribute corresponding to the detected edge defect, and verifies an accuracy of the output defect attribute of the edge defect based on the determined defect attribute of the edge defect.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to AHMED A NASHER whose telephone number is (571)272-1885. The examiner can normally be reached Mon - Fri 0800 - 1700.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Andrew Moyer can be reached at (571) 272-9523. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/AHMED A NASHER/Examiner, Art Unit 2675
/ANDREW M MOYER/Supervisory Patent Examiner, Art Unit 2675