Modeling post-lithography stochastic critical dimension variation with multi-task neural networks

Patent metadata

Field	Value
Publication number	US-10657420-B2
Application number	US-201816037454-A
Country	US
Kind code	B2
Filing date	Jul 17, 2018
Priority date	Jul 17, 2018
Publication date	May 19, 2020
Grant date	May 19, 2020

Abstract

A method of modeling distributions of post-lithography critical dimensions includes the following steps. A plurality of aerial images of respective portions of a physical design layout of a semiconductor wafer are generated, and the plurality of aerial images are employed as training data. In the method, first and second portions of a neural network architecture are generated. The first portion includes a neural network which is shared by a plurality of output channels, and the second portion includes a plurality of neural networks, wherein each of the plurality of neural networks respectively correspond to one of the plurality of output channels. The method further includes training the first and second portions of the neural network architecture with the training data, and outputting the distributions of the post-lithography critical dimensions based on the plurality of output channels.

First claim

What is claimed is:

1. A method of modeling distributions of post-lithography critical dimensions, comprising:
generating a plurality of aerial images of respective portions of a physical design layout of a semiconductor wafer;
employing the plurality of aerial images as training data;
generating a first portion of a neural network architecture, wherein the first portion comprises a neural network which is shared by a plurality of output channels;
generating a second portion of the neural network architecture, wherein the second portion comprises a plurality of neural networks, each of the plurality of neural networks respectively corresponding to one of the plurality of output channels;
training the first and second portions of the neural network architecture with the training data; and
outputting the distributions of the post-lithography critical dimensions based on the plurality of output channels;
wherein the method is performed by at least one computer system comprising at least one memory and at least one processor coupled to the memory.

2. The method according to claim 1 , wherein the first portion of the neural network architecture comprises a convolutional neural network.

3. The method according to claim 1 , wherein each of the plurality of neural networks of the second portion of the neural network architecture comprises a fully connected neural network.

4. The method according to claim 3 , wherein each of the plurality of neural networks of the second portion comprises a plurality of hidden layers.

5. The method according to claim 1 , wherein the plurality of output channels comprise at least three output channels.

6. The method according to claim 1 , wherein the plurality of output channels respectively comprise different parameters of the distributions of the post-lithography critical dimensions, wherein the distributions are asymmetric.

7. The method according to claim 6 , wherein the different parameters comprise critical dimension mean, critical dimension variation, and critical dimension skewness.

8. The method according to claim 6 , further comprising simultaneously modeling the different parameters using the first and second portions of the neural network architecture.

9. The method according to claim 1 , wherein the plurality of output channels respectively comprise at least critical dimension mean and critical dimension stochastic variation, and the method further comprises simultaneously modeling the critical dimension mean and the critical dimension stochastic variation using the first and second portions of the neural network architecture.

10. The method according to claim 1 , wherein the distributions of the post-lithography critical dimensions are asymmetric.

11. The method according to claim 1 , further comprising fine-tuning hyperparameters of the first and second portions of the neural network architecture.

12. The method according to claim 1 , wherein the plurality of aerial images comprise a distribution of light intensity as a function of position.

13. The method according to claim 1 , wherein the respective portions of the physical design layout of the semiconductor wafer comprise respective layout patterns from different locations of the physical design layout.

14. The method of claim 1 , further comprising analyzing stochastic effect of EUV lithography with the outputted distributions of the post-lithography critical dimensions.

15. The method of claim 1 , further comprising applying the trained first and second portions of the neural network architecture to full-chip modeling of the distributions of the post-lithography critical dimensions.

16. A system for modeling distributions of post-lithography critical dimensions, comprising:
a memory and at least one processor coupled to the memory, wherein the at least one processor is configured to:
generate a plurality of aerial images of respective portions of a physical design layout of a semiconductor wafer;
employ the plurality of aerial images as training data;
generate a first portion of a neural network architecture, wherein the first portion comprises a neural network which is shared by a plurality of output channels;
generate a second portion of the neural network architecture, wherein the second portion comprises a plurality of neural networks, each of the plurality of neural networks respectively corresponding to one of the plurality of output channels;
train the first and second portions of the neural network architecture with the training data; and
output the distributions of the post-lithography critical dimensions based on the plurality of output channels.

17. The system according to claim 16 , wherein the first portion of the neural network architecture comprises a convolutional neural network.

18. The system according to claim 16 , each of the plurality of neural networks of the second portion of the neural network architecture comprises a fully connected neural network.

19. The system according to claim 16 , wherein the plurality of output channels respectively comprise different parameters of the distributions of the post-lithography critical dimensions, wherein the distributions are asymmetric.

20. A computer program product for modeling distributions of post-lithography critical dimensions, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to perform a method comprising:
generating a plurality of aerial images of respective portions of a physical design layout of a semiconductor wafer;
employing the plurality of aerial images as training data;
generating a first portion of a neural network architecture, wherein the first portion comprises a neural network which is shared by a plurality of output channels;
generating a second portion of the neural network architecture, wherein the second portion comprises a plurality of neural networks, each of the plurality of neural networks respectively corresponding to one of the plurality of output channels;
training the first and second portions of the neural network architecture with the training data; and
outputting the distributions of the post-lithography critical dimensions based on the plurality of output channels.

21. A method of modeling distributions of critical dimensions, comprising:
receiving as inputs a plurality of aerial images of respective portions of a physical design layout of a semiconductor wafer;
employing the plurality of aerial images as training data;
generating a first neural network which is shared by at least three output channels;
generating a plurality of second neural networks respectively corresponding to each one of the at least three output channels;
training the first and second neural networks with the training data; and
outputting the distributions of the critical dimensions based on the at least three output channels;
wherein the method is performed by at least one computer system comprising at least one memory and at least one processor coupled to the memory.

22. The method according to claim 21 , wherein the at least three output channels respectively comprise different parameters of the distributions of the critical dimensions, and wherein the different parameters comprise critical dimension mean, critical dimension variation, and critical dimension skewness.

23. The method according to claim 21 , wherein

Assignees

• IBM

Inventors

• Sha Jing
• De Silva Ekmini A
• Dunn Derren N

Classifications

• G06N3/08

  Learning methods · CPC title

• G06K9/6262 Primary

  Physics · mapped topic

• G06N3/0454

  Physics · mapped topic

• G03F7/70625 Primary

  Dimensions, e.g. line width, critical dimension [CD], profile, sidewall angle or edge roughness · CPC title

• G06V10/454

  Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN] · CPC title