Generative adversarial networks for generating physical design layout patterns of integrated multi-layers

Patent metadata

Field	Value
Publication number	US-10706200-B2
Application number	US-201816000243-A
Country	US
Kind code	B2
Filing date	Jun 5, 2018
Priority date	Jun 5, 2018
Publication date	Jul 7, 2020
Grant date	Jul 7, 2020

Abstract

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

First claim

What is claimed is:

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

2. The method of claim 1 , wherein the features in at least a first layer of the given patterned structure comprise vias and the features in at least a second layer of the given patterned structure comprise metal lines.

3. The method of claim 1 , wherein the given three-dimensional array has a size of W×H×N, where W represents a width of a field of view of the at least two layers of the given physical design layout pattern, H represents a height of the field of view, and N represents a number of layers in the given physical design layout pattern.

4. The method of claim 3 , wherein N is 3.

5. The method of claim 3 , wherein N is greater than 3.

6. The method of claim 3 , wherein entries of the given three-dimensional array comprise values indicating presence of a given feature at respective positions in the field of view for a corresponding layer of the given physical design layout pattern.

7. The method of claim 3 , wherein the converting the physical design layout patterns into the three-dimensional arrays comprises, for the given physical design layout pattern:
capturing an image of each individual layer of the given physical design layout pattern;
converting the image of said each individual layer into an array of size W×H×1; and
generating the given three-dimensional array of size of W×H×N as a combination of N arrays of size W×H×1 for said each individual layer of the given physical design layout pattern.

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

9. The method of claim 8 , wherein the convolutional neural network comprises: an input layer; a series of one or more convolutional layers, a batch normalization and a leaky rectified linear unit (L-ReLU) activation functions; a fully connected layer; and a softmax output layer.

10. The method of claim 9 , wherein the three-dimensional arrays are provided to the input layer.

11. 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.

12. The method of claim 11 , wherein the input layer receives a noise vector and the output layer provides the synthetic three-dimensional arrays.

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

14. The method of claim 13 , wherein the converting the physical design layout patterns into the three-dimensional arrays comprises, for the given physical design layout pattern:
converting the layout format of each layer of layers of the given physical design layout pattern to a two-dimensional array; and
structuring the two-dimensional arrays into a three-dimensional array of values, the values representing presence of designated features at locations in said each layer of the layers of the given physical design layout pattern.

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

16. The method of claim 15 , wherein the converting the synthetic three-dimensional arrays to the layout format supported by the EDA software comprises, for the given synthetic three-dimensional array comprising a set of two-dimensional arrays each representing a layer of the new physical design layout pattern, translating each of the two-dimensional arrays to coordinates of vertices of polygons of features in a corresponding layer of the new physical design layout pattern.

17. The method of claim 15 , further comprising applying a post-processing using one or more design rule checks.

18. The method of claim 15 , further comprising applying a post-processing to remove one or more rule-violating shapes.

19. The method of claim 15 , further comprising utilizing the synthetic three-dimensional arrays converted to the layout format to evaluate manufacturability of the new physical design layout pattern.

20. A computer program product, the computer program product comprising a non-transitory 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 one or more physical design layout patterns of integrated multi-layers for features in at least two layers of a given patterned structure;
converting the physical design layout patterns into three-dimensional arrays, a given three-dimensional array comprising a set of two-dimensional arrays each representing features of one layer of the at least two layers in a given one of the physical design layout patterns;
training, utilizing the three-dimensional arrays, a generative adversarial network (GAN) comprising a discriminator neural network and a generator neural network; and
generating one or more synthetic three-dimensional arrays utilizing the generator neural network of the trained GAN, a given one of the synthetic three-dimensional arrays comprising a set of two-dimensional arrays each representing features for one new layer of at least two new layers of a new physical design layout pattern.

21. The computer program product of claim 20 , wherein the given three-dimensional array has a size of W×H×N, where W represents a width of a field of view of the at least two layers of the given physical design layout pattern, H represents a height of the field of view, and N represents a number of layers in the given physical design layout pattern.

22. The computer program product of claim 21 , wherein the converting the physical design layout patterns into the three-dimensional arrays comprises, for the given physical design layout pattern:
capturing an image of each individual layer of the given physical design layout pattern;
converting the image of said each individual layer into an array of size W×H×1; and
generating the given three-dimensional array of size of W×H×N as a combination of N arrays of size W×H×1 for said each individual layer of the given physical design layout pattern.

23. An apparatus comp

Assignees

• IBM

Inventors

• Sha Jing
• Guillorn Michael A
• Burkhardt Martin
• Dunn Derren N

Classifications

• G06F30/398

  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) · CPC title

• G06F30/27

  using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model · CPC title

• G06F30/39 Primary

  Circuit design at the physical level (physical level design for reconfigurable circuits G06F30/347) · CPC title

• G06F30/392 Primary

  Floor-planning or layout, e.g. partitioning or placement · CPC title