Synopsys, Inc.

Luminescent Technologies, Inc.

Thomas Christopher Cecil

Kevin Hooker

Michael Lawrence Rieger

Benjamin David Painter

Linyong Pang

Robert E Gleason

Amyn A Poonawala

Changqing Hu

Guangming Xiao

Tatung Chow

Jason Jiale Shu

John F McCarty

Donghwan Son

David H Kim

David W Thomas

Marco Guajardo

A system uses machine learning models, such as neural networks for generating mask design from a circuit design. The machine learning models have inputs and outputs which are localized to a small region of the circuit design. The machine learning model takes as input features describing the circuit design in the neighborhood of a location and generates an offset distance as output. The system uses the offset distance to generate features of the mask design, for example, main features or assist features corresponding to a circuit design polygon. The system may use the offset distance for target optimization by modifying the circuit design polygon to obtain a circuit design polygon that has improved manufacturability.

Neural network based mask synthesis for integrated circuits

A system generates a mask for a circuit design while enforcing symmetry and consistency across random areas of the mask. The system builds a mask solutions database mapping circuit patterns to mask patterns. The system uses the mask solutions database to replace circuit patterns of the circuit design with mask patterns. The system identifies properties in circuit patterns of the circuit design and enforces the same property in the corresponding mask patterns. Examples of properties enforced include symmetry within circuit patterns and similarity across circuit patterns. The system combines mask patterns in different regions of the circuit and resolves conflicts that occur when there are multiple masks within a region.

Enforcing mask synthesis consistency across random areas of integrated circuit chips

In certain embodiments, a method includes the following steps. A layout used in a lithographic mask development process is accessed. For example, the layout may be the layout of the mask itself, or it may be the layout of the resulting printed pattern on the wafer. The layout includes a number of disjoint shapes. Skeleton representations for at least some of the disjoint shapes in the layout are determined. The skeleton representation of an individual shape has elements of two or more nodes connected by edges. It also includes size parameters for at least some of the elements. The skeleton representations of the shapes are used in the mask development process.

Skeleton representation of layouts for the development of lithographic masks

A method comprises receiving an integrated circuit (IC) design file and determining, by one or more processors, dose information from the IC design file. The method further comprises determining, by the one or more processors, a mask vector file from the IC design file, and converting, by the one or more processors, the dose information to a vector file format. Further, the method comprises outputting the dose information in the vector file format and the mask vector file to a mask writer device.

Dose information generation and communication for lithography manufacturing systems

Techniques relating to synthesizing masks for use in manufacturing a semiconductor device are disclosed. A plurality of training masks, for a machine learning (ML) model, are generated by synthesizing one or more polygons, relating to a design pattern for the semiconductor device, using Inverse Lithography Technology (ILT) (). The ML model is trained using both the plurality of training masks generated using ILT, and the design pattern for the semiconductor device, as inputs (). The trained ML model is configured to synthesize one or more masks, for use in manufacturing the semiconductor device, based on the design pattern ().

Inverse lithography and machine learning for mask synthesis

A system performs mask rule checks (MRC) for curvilinear shapes. The width of a curvilinear shape is different along different parts of the shape. A medial axis for a curvilinear shape is determined. The medial axis is trimmed to exclude portions that are within a threshold distance from corners or too far from edges. The trimmed medial axis is used to perform width checks for mask rules. The system generates medial axis between geometric shapes and uses it to determine whether two geometric shapes are at least a threshold distance apart. The system performs acute angle checks for sharp corners. The system determines angles using lines drawn from vertices to end points on the boundary of the shape that are at a threshold distance. These angles are used for checking acute angle mask rule violations.

Mask rule checking for curvilinear masks for electronic circuits

Aspects described herein relate to mask synthesis using design guided offsets. A target shape on an image surface to be fabricated using a mask based on a design of an integrated circuit is obtained. Rays are generated emanating from respective anchor points. The anchor points are on a boundary of the target shape or a boundary of a mask shape of the mask. For each ray of the rays, a distance is defined between a first intersection of the respective ray and the boundary of the target shape and a second intersection of the respective ray and the boundary of the mask shape. An analysis is performed by one or more processors, where the analysis is configured to modify the distances based on an error between the target shape and a resulting shape simulated to be on the image surface resulting from the mask shape.

Mask synthesis using design guided offsets

A system performs distributed or parallel pattern extraction and clustering for pattern classification of large layouts of electronic circuits. The system identifies circuit patterns with a layout representation. The system encodes the circuit patterns using a neural network based autoencoder to generate encoded circuit patterns that can be stored efficiently. The system clusters the encoded circuit patterns into an arbitrary number of clusters based upon a high degree of similarity. The clusters of circuit patterns may be used for training and evaluation of machine learning based models.

Classification of patterns in an electronic circuit layout using machine learning based encoding

A method comprises receiving an integrated circuit (IC) chip design, and generating, by one or more processors, a wafer image and a wafer target from the IC chip design. The method further comprises generating, by the one or more processors, sensitivity information based on a determination that the wafer image and the wafer target converge, and outputting the sensitivity information. The sensitivity information is associated with writing a mask written for the IC chip design.

Wafer sensitivity determination and communication

A method comprises receiving an integrated circuit (IC) chip design, and generating, by one or more processors and based on the IC chip design, dose information, a wafer image, and a wafer target. Further, the method comprises modifying, by the one or more processors, the dose information based on a comparison of the wafer image and the wafer target. Further, the method comprises outputting the modified dose information to a mask writing device.

Dose optimization techniques for mask synthesis tools

Aspects described herein relate to obtaining a mask pattern using a cost function gradient (CFG) generated from a Jacobian matrix generated from a perturbation look-up table (PLT). In an example method, a PLT is populated (). Each table entry of the PLT is based on a respective perturbed intensity signal. The respective perturbed intensity signal is based on a simulated signal received at an image surface using a mask pattern having a perturbed element of the mask pattern. The mask pattern is for a design of an integrated circuit. A matrix is populated () using the PLT and a target intensity signal. The target intensity signal is based on a signal received at the image surface to form target features at the image surface. A CFG is defined () based on the matrix. An analysis is performed () on the mask pattern based on the CFG.

Obtaining a mask using a cost function gradient from a jacobian matrix generated from a perturbation look-up table

An example is a method. An electronic representation of a design of an integrated circuit to be manufactured on a semiconductor die is obtained. The design of the integrated circuit includes layers. The electronic representation includes initial polygons. Polygon topological skeletons of the initial polygons of the target layer are generated. A space topological skeleton in a space between the polygon topological skeletons is generated. A connected network comprising network edges is generated. Each network edge is connected between a respective polygon topological skeleton and the space topological skeleton. A transformation of the polygon topological skeletons is performed, by one or more processors, based on the network edges, a spacing specification for a spacing between polygons, and respective specified widths associated with the initial polygons by perturbing the polygon topological skeletons.

Growing community of inventors

Menlo Park, CA, United States of America

Thomas Christopher Cecil