Kla Tencor Corporation

Kla-tencor Technologies Corporation

Other

Kla Corporation

Kla-tenor Corp.

Mark J Roulo

Krishnamurthy Bhaskar

Jason Z Lin

Lawrence Robert Miller

Richard Wallingford

Eliezer Rosengaus

John S Taylor

Paul T Russell

Kishore Bubna

Kris Bhaskar

Jing Zhang

Zvi Dubiner

Bjorn Brauer

Brian Duffy

Scott Allen Young

Mohan Mahadevan

Laurent Karsenti

Michael John Van Riet

Stewart Hill

Ashok Mathew

Arrigo Benedetti

Ravi Chandra Donapati

Methods and systems for performing diagnostic functions for a deep learning model are provided. One system includes one or more components executed by one or more computer subsystems. The one or more components include a deep learning model configured for determining information from an image generated for a specimen by an imaging tool. The one or more components also include a diagnostic component configured for determining one or more causal portions of the image that resulted in the information being determined and for performing one or more functions based on the determined one or more causal portions of the image.

Diagnostic systems and methods for deep learning models configured for semiconductor applications

An inspection system is disclosed. The inspection system includes a shared memory configured to receive image data from a defect inspection tool and a controller communicatively coupled to the shared memory. The controller includes a host image module configured to apply one or more general-purpose defect-inspection algorithms to the image data using central-processing unit (CPU) architectures, a results module configured to generate inspection data for defects identified by the host image module, and secondary image module(s) configured to apply one or more targeted defect-inspection algorithms to the image data. The secondary image module(s) employ flexible sampling of the image data to match a data processing rate of the host image module within a selected tolerance. The flexible sampling of the image data is adjusted responsive to the inspection data generated by the results module and the host image module.

Multi-controller inspection system

Methods and systems for performing one or more functions for a specimen using output simulated for the specimen are provided. One system includes one or more computer subsystems configured for acquiring output generated for a specimen by one or more detectors included in a tool configured to perform a process on the specimen. The system also includes one or more components executed by the one or more computer subsystems. The one or more components include a learning based model configured for performing one or more first functions using the acquired output as input to thereby generate simulated output for the specimen. The one or more computer subsystems are also configured for performing one or more second functions for the specimen using the simulated output.

Accelerating semiconductor-related computations using learning based models

An inspection system that receives image data corresponding to an image and processes the image data to produce a report corresponding to characteristics of the image. Interface cards receive the image data in a flow, where each interface card receives image data corresponding to a different portion of the image. Process nodes connect to the interface cards, and receive the image data from the interface cards. A host computer is connected to the process nodes, and job managers implemented in the host computer manage the flow of image data to and from the process nodes. The job managers remain operable during a crash of one of the process nodes. Process node programs are implemented in the process nodes, and analyze a portion of the image data and produce the report corresponding to the characteristics of the analyzed portion of the image data. At least one process node program is implemented in each process node. The process node programs rapidly analyze the image. The process node programs are isolated from the job managers so that a crash of a process node program does not crash the job manager.

Image processing architecture

An inspection system that receives image data corresponding to an image and processes the image data to produce a report corresponding to characteristics of the image. Interface cards receive the image data in a flow, where each interface card receives image data corresponding to a different portion of the image. Process nodes connect to the interface cards, and receive the image data from the interface cards. A host computer is connected to the process nodes, and job managers implemented in the host computer manage the flow of image data to and from the process nodes. The job managers are written in a relatively stable and slow programming environment, to remain operable during a crash of one of the process nodes. Process node programs are implemented in the process nodes, and analyze a portion of the image data and produce the report corresponding to the characteristics of the analyzed portion of the image data. At least one process node program is implemented in each process node. The process node programs are written in a relatively instable and fast programming environment, and rapidly analyze the image. The process node programs are isolated from the job managers so that a crash of a process node program does not crash the job manager.

An inspection system for detecting anomalies on a substrate. The inspection system has a sensor array for generating image data. A first high speed network is coupled to the sensor array and receives and communicates the image data. An array of process nodes is coupled to the first high speed network, and receives and processes the image data to produce anomaly reports. Each process node has an interface card coupled to the first high speed network, that receives the image data from the first high speed network and formats the image data according to a high speed interface bus protocol. The interface card sets a register indicating whether a predetermined amount of image data has been stored in a memory, and the process node reads the register to determine whether the predetermined amount of image data has been stored in the memory, and initiates image processing when the register indicates that the predetermined amount of image data has been stored in the memory.

Status polling

An inspection system for detecting anomalies on a substrate. A first network is coupled to a sensor array and communicates data. Process nodes are coupled to the first network, and process the data to produce reports. Each process node includes memory sufficient to buffer the data until it can process the data. Each process node has an interface card that formats the data for a high speed interface bus that is coupled to the interface card. A computer receives and processes the data to produce the report. A second network receives the reports. A job manager is coupled to the second network, receives the reports, and sends information to the process nodes to coordinate processing of the data.

Memory load balancing

An inspection image analysis system. At least one image processing computer is configured to receive and analyze at least one portion of an image. At least one test computer is configured to receive at least one common portion of the image also received by the at least one image processing computer, and to analyze the at least one common portion, using equivalent image processes as the corresponding at least one image processing computer. A job manager is configured to assign the common portion and to configure the corresponding image processing computer and the test computer to run equivalent image processes.

Mirror node process verification

An inspection system. The inspection system has a sensor for generating data. A first network is coupled to the sensor and communicates the data. An array of nodes is coupled to the first network, and processes the data to produce reports. Each node has an interface coupled to the first network, and formats the data according to protocol. A bus is coupled to the interface A computer is coupled to the bus, and process the data according to an algorithm, to produce the report. The array of nodes is coupled to the first network in a daisy chain topology, and each node within a column of nodes receives common data with other nodes within the column. A second network is coupled to the nodes the second network, and receives the anomaly reports from the nodes and sends information to the nodes to coordinate processing of the data.

Daisy chained topology

A system for inspecting a substrate. An inspector includes a sensor that inspects the substrate and produces a video stream. A control interface sends and receives a control stream, and a network receives and transports the video stream and the control stream as two separate data streams. A desktop receives the video stream and the control stream as two separate data streams. The desktop has a display that presents the video stream, and a user interface controls that control the operation of the inspector, using the control stream across the network.

Remote live video inspection

A system having a sensor array adapted to provide image data. A process node is connected to the sensor array, and analyzes portions of the image data. A job manager is connected to the process node, and instructs the process node to send the portions of the image data to a storage node. The storage node receives and stores the sent portions of the image data. In this manner, image data can be stored and run multiple times in order to tune a recipe for the analysis system. Because the same image data can be used over and over again, the recipe creating process can be more finely tuned to detect both the types of defects and at a level that is desired.

Image data storage

An inspection system having a sensor array that provides image data. A process node includes a memory to receive the image data, a commercially available central processing unit to receive and coprocess at least a first portion of the image data within the memory, and a field programmable gate array to receive and coprocess at least a second portion of the image data within the memory. In this manner, there are two elements in the process node that are used to simultaneously process the image data, and the image data analysis thereby proceeds at a much faster rate than it would with just a single processor in a commercially available computer. However, the system as described has very little custom hardware, and thus is relatively inexpensive, and highly versatile.

Fast bus image coprocessing

An inspection system for detecting anomalies on a substrate. A first network is coupled to a sensor array and communicates data. Process nodes are coupled to the first network, and process the data to produce reports. Each process node has an interface card that formats the data for a high speed interface bus that is coupled to the interface card. A computer receives and processes the data to produce the report. A second network receives the reports. A job manager is coupled to the second network, receives the reports, and sends information to the process nodes to coordinate processing of the data.

Programmable image computer

An inspection system for detecting anomalies on a substrate. A first network is coupled to the sensor array and communicates image data. Process nodes are couple to the first network, and process the data to produce reports. Each process node has an interface card that formats the data for a high speed interface bus that is coupled to the interface card. A computer receives and processes the data to produce the report. A second network receives the reports from the process nodes. A job manager is coupled to the second network, and receives the reports from the process nodes and sends information to the process nodes to coordinate the processing of the data in the process nodes.

Growing community of inventors

Mountain View, CA, United States of America

Mark J Roulo