# US 2019/0377849 A1 · Generative adversarial networks for generating physical design layout patterns

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2019/0377849 A1
- Document ID: US20190377849A1
- Status: Published application
- Kind: A1
- Application number: 16/005,999
- Filed: 2018-06-12
- Published: 2019-12-12
- Earliest claimed: 2018-06-12
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Jing Sha; Michael A. Guillorn; Martin Burkhardt; Derren N. Dunn
- Web page: https://patentgrep.com/patent/US20190377849A1

## CPC

- G06F 30/398: Design verification or optimisation, e.g. using design rule check [DRC], layout versus schematics [LVS] or finite element methods [FEM] (optical proximity correction [OPC] design processes G03F1/36)
- G06N 3/08: Learning methods
- G06F 30/27: using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
- G06N 3/045: Combinations of networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/047: Probabilistic or stochastic networks
- G06N 3/0475: Generative networks
- G06N 3/088: Non-supervised learning, e.g. competitive learning
- G06N 3/094: Adversarial learning
- G06F 2111/20: Configuration CAD, e.g. designing by assembling or positioning modules selected from libraries of predesigned modules
- G06F 2119/18: Manufacturability analysis or optimisation for manufacturability
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/048: Activation functions
- Y02P 90/02: Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- Y02P: CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS

## Abstract

A method for generating physical design layout patterns includes selecting as training data a set of physical design layout patterns of features in a given layer of a given patterned structure and converting the physical design layout patterns into two-dimensional (2D) arrays comprising entries for different locations in the given layer of the given patterned structure with values representing presence of the features at the different locations. The method also includes training, utilizing the 2D arrays, a generative adversarial network (GAN) comprising a discriminator neural network and a generator neural network. The method further includes generating one or more synthetic 2D arrays utilizing the trained generator neural network of the GAN, a given synthetic 2D array comprising entries for different locations in the given layer of a new physical design layout pattern with values representing presence of the features at the different locations of the new physical design layout pattern.

## Claims

### Claim 1

A method for generating physical design layout patterns, comprising steps of:
selecting as training data a set of physical design layout patterns of features in a given layer of a given patterned structure;
converting the physical design layout patterns into two-dimensional arrays, a given two-dimensional array for a given physical design layout pattern comprising entries for different locations in the given layer of the given patterned structure with values representing presence of the features at the different locations;
training, utilizing the two-dimensional arrays, a generative adversarial network (GAN) comprising a discriminator neural network and a generator neural network; and
generating one or more synthetic two-dimensional arrays utilizing the trained generator neural network of the GAN, a given one of the synthetic two-dimensional arrays comprising entries for different locations in the given layer of a new physical design layout pattern with values representing presence of the features at the different locations of the new physical design layout pattern;
wherein the method is performed by at least one processing device comprising a processor coupled to a memory.

### Claim 2 (depends on 1)

The method of claim 1, wherein the set of physical design layout patterns selected as the training data comprises a first set of physical design layout patterns having the features aligned primarily in a first direction and a second set of physical design layout patterns having the features aligned primarily in a second direction.

### Claim 3 (depends on 2)

The method of claim 2, wherein the first direction comprises a horizontal direction and the second direction comprises a vertical direction.

### Claim 4 (depends on 2)

The method of claim 2, wherein the new physical design layout pattern has a first set of the features aligned in the first direction and a second set of the features aligned in the second direction.

### Claim 5 (depends on 1)

The method of claim 1, wherein the features of the given patterned structure comprise metal lines.

### Claim 6 (depends on 1)

The method of claim 1, wherein the feature of the given patterned structure comprise vias.

### Claim 7 (depends on 1)

The method of claim 1, wherein the given two-dimensional array has a size of W×H, where W represents a width of a field of view of the given layer of the given physical design layout pattern and H represents a height of the field of view.

### Claim 8 (depends on 1)

The method of claim 1, wherein converting the physical design layout patterns into the two-dimensional arrays comprises, for the given physical design layout pattern, capturing an image of the given physical design layout pattern from a physical design layout and converting the captured image to the given two-dimensional array.

### Claim 9 (depends on 8)

The method of claim 8, wherein the captured image comprises a black-and-white image, and wherein the entries of the given two-dimensional array comprise binary values representing presence or absence of the features at the different locations in the given layer.

### Claim 10 (depends on 1)

The method of claim 1, wherein the discriminator neural network comprises a convolutional neural network.

### Claim 11 (depends on 10)

The method of claim 10, wherein the convolutional neural network comprises: an input layer; a series of one or more convolutional layers, batch normalization and leaky rectified linear unit (L-ReLU) activation functions; a fully connected layer; and a softmax output layer.

### Claim 12 (depends on 11)

The method of claim 11, wherein the two-dimensional arrays are provided to the input layer.

### Claim 13 (depends on 1)

The method of claim 1, wherein the generator neural network comprises an input layer, a projection layer, a series of deconvolutional layers and an output layer.

### Claim 14 (depends on 13)

The method of claim 13, wherein the input layer receives a noise vector and the output layer provides the synthetic two-dimensional arrays.

### Claim 15 (depends on 1)

The method of claim 1, wherein the given physical design layout pattern is in a layout format supported by an electronic design automation (EDA) software, the layout format comprising coordinates of polygons representing the features in the given layer of the given physical design layout pattern.

### Claim 16 (depends on 15)

The method of claim 15, wherein converting the physical design layout patterns into the two-dimensional arrays comprises, for the given physical design layout pattern, converting the layout format of the given physical design layout pattern to the given two-dimensional array.

### Claim 17 (depends on 1)

The method of claim 1, further comprising converting the synthetic two-dimensional arrays to a layout format supported by an electronic design automation (EDA) software.

### Claim 18 (depends on 17)

The method of claim 17, wherein converting the synthetic two-dimensional arrays to the layout format supported by the EDA software comprises, for the given synthetic two-dimensional array, translating values of the given synthetic two-dimensional array to coordinates of vertices of polygons of the features in the given layer of the new physical design layout pattern.

### Claim 19 (depends on 17)

The method of claim 17, further comprising applying post-processing using one or more design rule checks.

### Claim 20 (depends on 17)

The method of claim 17, further comprising applying post-processing to remove one or more rule-violating shapes.

### Claim 21 (depends on 17)

The method of claim 17, further comprising utilizing the given synthetic two-dimensional array converted to the layout format to evaluate manufacturability of the new physical design layout pattern.

### Claim 22

A computer program product, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by at least one computing device to cause the at least one computing device to perform steps of:
selecting as training data a set of physical design layout patterns of features in a given layer of a given patterned structure;
converting the physical design layout patterns into two-dimensional arrays, a given two-dimensional array for a given physical design layout pattern comprising entries for different locations in the given layer of the given patterned structure with values representing presence of the features at the different locations;
training, utilizing the two-dimensional arrays, a generative adversarial network (GAN) comprising a discriminator neural network and a generator neural network; and
generating one or more synthetic two-dimensional arrays utilizing the trained generator neural network of the GAN, a given one of the synthetic two-dimensional arrays comprising entries for different locations in the given layer of a new physical design layout pattern with values representing presence of the features at the different locations of the new physical design layout pattern.

### Claim 23 (depends on 22)

The computer program product of claim 22, wherein the set of physical design layout patterns selected as the training data comprises a first set of physical design layout patterns having the features aligned primarily in a first direction and a second set of physical design layout patterns having the features aligned primarily in a second direction, and wherein the new physical design layout pattern has a first set of the features aligned in the first direction and a second set of the features aligned in the second direction.

### Claim 24

An apparatus comprising:
a memory; and
at least one processor coupled to the memory and configured for:
selecting as training data a set of physical design layout patterns of features in a given layer of a given patterned structure;
converting the physical design layout patterns into two-dimensional arrays, a given two-dimensional array for a given physical design layout pattern comprising entries for different locations in the given layer of the given patterned structure with values representing presence of the features at the different locations;
training, utilizing the two-dimensional arrays, a generative adversarial network (GAN) comprising a discriminator neural network and a generator neural network; and
generating one or more synthetic two-dimensional arrays utilizing the trained generator neural network of the GAN, a given one of the synthetic two-dimensional arrays comprising entries for different locations in the given layer of a new physical design layout pattern with values representing presence of the features at the different locations of the new physical design layout pattern.

### Claim 25 (depends on 24)

The apparatus of claim 24, wherein the set of physical design layout patterns selected as the training data comprises a first set of physical design layout patterns having the features aligned primarily in a first direction and a second set of physical design layout patterns having the features aligned primarily in a second direction, and wherein the new physical design layout pattern has a first set of the features aligned in the first direction and a second set of the features aligned in the second direction.
