Applied Materials, Inc.

Pdf Solutions, Incorporated

Stream Mosaic, Inc.

Jeffrey Drue David

Dominic J Benvegnu

Boguslaw A Swedek

Harry Q Lee

Jun Qian

Bogdan Swedek

Lakshmanan Karuppiah

Tomonori Honda

Thomas H Osterheld

Benjamin Cherian

Sivakumar Dhandapani

Lin Lee Cheong

Doyle E Bennett

Ingemar Carlsson

Sidney P Huey

Jimin Zhang

Richard Gene Burch

Abraham Ravid

Zhize Zhu

Qing Zhu

Manoocher Birang

Wen-Chiang Tu

David J Lischka

Charles C Garretson

Tomohiko Kitajima

Garlen C Leung

Taketo Sekine

Xiaoyuan Hu

Michael Keleher

Nils Johansson

Rohan Deodatta Kekatpure

Steven M Zuniga

Fred Conrad Redeker

Dennis Ciplickas

John Kibarian

Gregory E Menk

Allen L D'Ambra

Brian E Stine

Zhihong Wang

Jagan Rangarajan

Kiran Lall Shrestha

Benjamin A Bonner

Thomas Li

Garrett Ho Yee Sin

Lakshmikar Kuravi

Gary Ka Ho Lam

Raymond R Jin

Thian Choi Lim

Dirk De Roover

Sandeep R Koppikar

Kenneth Harris

Said Akar

Vaishnavi Reddipalli

A template for assigning the most probable root causes for wafer defects. The bin map data for a subject wafer can be compared with bin map data for prior wafers to find wafers with similar issues. A probability can be determined as to whether the same root cause should be applied to the subject wafer, and if so, the wafer can be labeled with that root cause accordingly.

Wafer bin map based root cause analysis

Classifying wafers using Collaborative Learning. An initial wafer classification is determined by a rule-based model. A predicted wafer classification is determined by a machine learning model. Multiple users can manually review the classifications to confirm or modify, or to add user classifications. All of the classifications are input to the machine learning model to continuously update its scheme for detection and classification.

Collaborative learning model for semiconductor applications

A semiconductor image classifier. Convolution functions are applied to modify the wafer images in order to extract key information about the image. The modified images are condensed then processed through a series of pairwise classifiers, each classifier configured to determine that the image is more like one of the pair than the other. Probabilities from each classifier are collected to form a prediction for each image.

Abnormal wafer image classification

Among other things, a computer-based method is described. The method comprises receiving, by one or more computers, a plurality of measured spectra reflected from a substrate at a plurality of different positions on the substrate. The substrate comprises at least two regions having different structural features. The method also comprises performing, by the one or more computers, a clustering algorithm on the plurality of measured spectra to separate the plurality of measured spectra into a number of groups based on the spectral characteristics of the plurality of measured spectra; selecting one of the number of groups to provide a selected group having a subset of spectra from the plurality of measured spectra; and determining, in the one or more computers, at least one characterizing value for the substrate based on the subset of spectra of the selected group.

Grouping spectral data from polishing substrates

A method of detecting a polishing endpoint includes storing a plurality of library spectra, measuring a sequence of spectra from the substrate in-situ during polishing, and for each measured spectrum of the sequence of spectra, finding a best matching library spectrum from the plurality of library spectra to generate a sequence of best matching library spectra. Each library spectrum has a stored associated value representing a degree of progress through a polishing process, and the stored associated value for the best matching library spectrum is determined for each best matching library spectrum to generate a sequence of values representing a progression of polishing of the substrate. The sequence of values is compared to a target value, and a polishing endpoint is triggered when the sequence of values reaches the target value.

Endpoint detection for chemical mechanical polishing based on spectrometry

Scheme for detection and classification of semiconductor equipment faults. Sensor traces are monitored and processed to separate known abnormal operating conditions from unknown abnormal operating conditions. Feature engineering permits focus on relevant traces for a targeted feature. A machine learning model is built to detect and classify based on an initial classification set of anomalies. The machine learning model is continuously updated as more traces are processed and learned.

Anomalous equipment trace detection and classification

A maintenance tool for semiconductor process equipment and components. Sensor data is evaluated by machine learning tools to determine when to schedule maintenance action.

Maintenance scheduling for semiconductor manufacturing equipment

Endpointing detection for chemical mechanical polishing based on spectrometry

Robust machine learning predictions. Temporal dependencies of process targets for different machine learning models can be captured and evaluated for the impact on process performance for target. The most robust of these different models is selected for deployment based on minimizing variance for the desired performance characteristic.

Generating robust machine learning predictions for semiconductor manufacturing processes

A method for predicting yield for a semiconductor process. A particular type of wafer is fabricated to have a first set of features disposed on the wafer, with a wafer map identifying a location for each of the first set of features on the wafer. Data from wafer acceptance tests and circuit probe tests is collected over time for wafers of that particular type as made in a semiconductor fabrication process, and at least one training dataset and a least one validation dataset are created from the collected data. A second set of 'engineered' features are created and also incorporated onto the wafer and wafer map. Important features from the first and second sets of features are identified and selected, and using those important features as inputs, a number of different process models are run, with yield as the target. The results of the different models can be combined, for example, statistically.

Semiconductor yield prediction

A method to assist in identifying a spectral feature and a characteristic of the selected spectral feature to monitor during polishing includes polishing a test substrate and measuring a sequence of spectra of light reflected from a substrate while the substrate is being polished, where at least some of the spectra of the sequence differ due to material being removed during the polishing. The sequence of spectra are visually displayed as a contour plot.

Display of spectra contour plots versus time for semiconductor processing system control

Testing data is evaluated by machine learning tools to determine whether to include or exclude chips from further testing.

Selective inclusion/exclusion of semiconductor chips in accelerated failure tests

A computer implemented method of monitoring a polishing process includes, for each sweep of a plurality of sweeps of an optical sensor across a substrate undergoing polishing, obtaining a plurality of current spectra, each current spectrum of the plurality of current spectra being a spectrum resulting from reflection of white light from the substrate, for each sweep of the plurality of sweeps, determining a difference between each current spectrum and each reference spectrum of a plurality of reference spectra to generate a plurality of differences, for each sweep of the plurality of sweeps, determining a smallest difference of the plurality of differences, thus generating a sequence of smallest difference, and determining a polishing endpoint based on the sequence of smallest differences.

Spectra based endpointing for chemical mechanical polishing

Techniques for measuring and/or compensating for process variations in a semiconductor manufacturing processes. Machine learning algorithms are used on extensive sets of input data, including upstream data, to organize and pre-process the input data, and to correlate the input data to specific features of interest. The correlations can then be used to make process adjustments. The techniques may be applied to any feature or step of the semiconductor manufacturing process, such as overlay, critical dimension, and yield prediction.

Process control techniques for semiconductor manufacturing processes

Growing community of inventors

San Jose, CA, United States of America

Jeffrey Drue David