Kla Tencor Corporation

Kla Corporation

Saravanan Paramasivam

Martin Plihal

Prasanti Uppaluri

Niveditha Lakshmi Narasimhan

Sairam Ravu

Ankit Jain

Jacob George

Erfan Soltanmohammadi

Vidyasagar Anantha

Raghavan Konuru

Santosh Bhattacharyya

Patrick Huet

Arpit Yati

Ardis Liang

Chris W Lee

Sandeep Bhagwat

Vijay Ramachandran

Sarath Shekkizhar

Luc Debarge

Somesh Challapalli

Aditya Gulati

Jincheng Lin

Methods and systems for determining information for a specimen are provided. One system includes a computer subsystem and one or more components executed by the computer subsystem. The one or more components include a deep learning model configured for denoising an image of a specimen generated by an imaging subsystem. The computer subsystem is configured for determining information for the specimen from the denoised image.

Deep learning image denoising for semiconductor-based applications

A system for characterizing a specimen is disclosed. In one embodiment, the system includes a characterization sub-system configured to acquire one or more images a specimen, and a controller communicatively coupled to the characterization sub-system. The controller may be configured to: receive from the characterization sub-system one or more training images of one or more defects of a training specimen; generate one or more augmented images of the one or more defects of the training specimen; generate a machine learning classifier based on the one or more augmented images of the one or more defects of the training specimen; receive from the characterization sub-system one or more target images of one or more target features of a target specimen; and determine one or more defects of the one or more target features with the machine learning classifier.

System and method for determining defects using physics-based image perturbations

Machine learning approaches provide additional information about semiconductor wafer inspection stability issues that makes it possible to distinguish consequential process variations like process excursions from minor process variations that are within specification. The effect of variable defect of interest (DOI) capture rates in the inspection result and the effect of variable defect count on the wafer can be monitored independently.

Method for process monitoring with optical inspections

Methods and systems for improved detection and classification of defects of interest (DOI) is realized based on values of one or more automatically generated attributes derived from images of a candidate defect. Automatically generated attributes are determined by iteratively training, reducing, and retraining a deep learning model. The deep learning model relates optical images of candidate defects to a known classification of those defects. After model reduction, attributes of the reduced model are identified which strongly relate the optical images of candidate defects to the known classification of the defects. The reduced model is subsequently employed to generate values of the identified attributes associated with images of candidate defects having unknown classification. In another aspect, a statistical classifier is employed to classify defects based on automatically generated attributes and attributes identified manually.

Methods and systems for inspection of semiconductor structures with automatically generated defect features

Wafer inspection with stable nuisance rates and defect of interest capture rates are disclosed. This technique can be used for discovery of newly appearing defects that occur during the manufacturing process. Based on a first wafer, defects of interest are identified based on the classified filtered inspection results. For each remaining wafer, the defect classifier is updated and defects of interest in the next wafer are identified based on the classified filtered inspection results.

Diagnostic methods for the classifiers and the defects captured by optical tools

A system for defect review and classification is disclosed. The system may include a controller, wherein the controller may be configured to receive one or more training images of a specimen. The one or more training images including a plurality of training defects. The controller may be further configured to apply a plurality of difference filters to the one or more training images, and receive a signal indicative of a classification of a difference filter effectiveness metric for at least a portion of the plurality of difference filters. The controller may be further configured to generate a deep learning network classifier based on the received classification and the attributes of the plurality of training defects. The controller may be further configured to extract convolution layer filters of the deep learning network classifier, and generate one or more difference filter recipes based on the extracted convolution layer filters.

System and method for difference filter and aperture selection using shallow deep learning

To evaluate a semiconductor-fabrication process, a semiconductor wafer is obtained that includes die grouped into modulation sets. Each modulation set is fabricated using distinct process parameters. The wafer is optically inspected to identify defects. A nuisance filter is trained to classify the defects as DOI or nuisance defects. Based on results of the training, a first, preliminary process window for the wafer is determined and die structures having DOI are identified in a first group of modulation sets bordering the first process window. The trained nuisance filter is applied to the identified defects to determine a second, revised process window for the wafer. A third, further revised process window for the wafer is determined based on SEM images of specified care areas in one or more modulation sets within the second, revised process window. A report is generated that specifies the third process window.

Semiconductor hot-spot and process-window discovery combining optical and electron-beam inspection

Use of care areas in scanning electron microscopes or other review tools can provide improved sensitivity and throughput. A care area is received at a controller of a scanning electron microscope from, for example, an inspector tool. The inspector tool may be a broad band plasma tool. The care area is applied to a field of view of a scanning electron microscope image to identify at least one area of interest. Defects are detected only within the area of interest using the scanning electron microscope. The care areas can be design-based or some other type of care area. Use of care areas in SEM tools can provide improved sensitivity and throughput.

Care areas for improved electron beam defect detection

Methods and systems for selecting a mode for inspection of a specimen are provided. One method includes determining how separable defects of interest (DOIs) and nuisances detected on a specimen are in one or more modes of an inspection subsystem. The separability of the modes for the Dais and nuisances is used to select a subset of the modes for inspection of other specimens of the same type. Other characteristics of the performance of the modes may be used in combination with the separability to select the modes. The subset of modes selected based on the separability may also be an initial subset of modes for which additional analysis is performed to determine the final subset of the modes.

Mode selection for inspection

Methods and systems for training an inspection-related algorithm are provided. One system includes one or more computer subsystems configured for performing an initial training of an inspection-related algorithm with a labeled set of defects thereby generating an initial version of the inspection-related algorithm and applying the initial version of the inspection-related algorithm to an unlabeled set of defects. The computer subsystem(s) are also configured for altering the labeled set of defects based on results of the applying. The computer subsystem(s) may then iteratively re-train the inspection-related algorithm and alter the labeled set of defects until one or more differences between results produced by a most recent version and a previous version of the algorithm meet one or more criteria. When the one or more differences meet the one or more criteria, the most recent version of the inspection-related algorithm is outputted as the trained algorithm.

Optimizing training sets used for setting up inspection-related algorithms

A method for production line monitoring during semiconductor device fabrication includes acquiring a plurality of inspection results from a plurality of reference samples with an inspection sub-system. The method includes storing the acquired inspection results and geometric pattern codes for each of the reference samples in a database. The method includes acquiring an additional inspection result from an additional sample, where the additional inspection result includes an additional set of geometric pattern codes for identifying each defect identified within the additional inspection result from the additional sample. The method also includes correlating the set of geometric pattern codes of the additional sample with the geometric pattern codes from the reference set of samples to identify at least one of one or more new patterns or one or more patterns displaying a frequency of occurrence above a selected threshold.

System and method for production line monitoring

Methods and systems for generating unbiased wafer defect samples are provided. One method includes selecting the defects detected by each of multiple scans performed on a wafer that have the most diversity in one or more defect attributes such that a diverse set of defects are selected across each scan. In addition, the method may include selecting the defects such that any defect that is selected and is common to two or more of the scans is not selected twice and any defects that are selected are diverse with respect to the common, selected defect. Furthermore, no sampling, binning, or classifying of the defects may be performed prior to selection of the defects such that the sampled defects are unbiased by any sampling, binning, or classifying method.

Unbiased wafer defect samples

A computerized method for categorizing defects on a substrate. A list of defects on the substrate is received as input to a processor, where each defect is represented by a defect location and an associated micro-defect code. The input is analyzed with the processor to detect spatial clusters of defects on the substrate. The spatial clusters are analyzed with the processor to determine which of the spatial clusters represent known macro-defects and which of the spatial clusters represent unknown macro-defects. The micro-defect code associated with each defect that is included in one of the spatial clusters that is determined to be a known macro-defect is changed with the processor with a macro-defect code that is associated solely with the known macro-defect. The processor analyzes the defects that are included in one of the spatial clusters that is determined to be an unknown macro-defect to determine a predominantly occurring micro-defect code. The processor changes the micro-defect code associated with each defect that is included in the one spatial cluster that is determined to be an unknown macro-defect with the predominantly-occurring micro-defect code. The processor sends the changed list of defects on the substrate as output.

Growing community of inventors

Chennai, India

Saravanan Paramasivam