Detecting hotspots using machine learning on diffraction patterns

Patent metadata

Field	Value
Publication number	US-9626459-B2
Application number	US-201414162889-A
Country	US
Kind code	B2
Filing date	Jan 24, 2014
Priority date	Jan 24, 2014
Publication date	Apr 18, 2017
Grant date	Apr 18, 2017

Abstract

A mechanism is provided in a data processing system for detecting lithographic hotspots. The mechanism receives a design layout. The mechanism generates spatial pattern clips from the design layout. The mechanism performs a transform on the spatial pattern clips to form frequency domain pattern clips. The mechanism performs feature extraction on the frequency domain pattern clips to form frequency domain features. The mechanism utilizes the extracted features on a set of training samples to train a machine learning classifier model. The mechanism classifies a set of previously unseen patterns, based on frequency domain features of the previously unseen patterns using the trained machine learning classifier model, into hotspots and non-hotspots.

First claim

What is claimed is:

1. A method, in a data processing system, for detecting lithographic hotspots, the method comprising:
receiving a design layout;
generating spatial pattern clips from the design layout using a sliding window to split the design layout into overlapping windows;
performing a transform on the spatial pattern clips to form frequency domain pattern clips;
performing feature extraction on the frequency domain pattern clips to form frequency domain features;
utilizing the extracted features on a set of training samples to train a machine learning classifier model; and
classifying a set of previously unseen patterns, based on frequency domain features of the previously unseen patterns using the trained machine learning classifier model, into hotspots and non-hotspots.

2. The method of claim 1 , wherein performing feature extraction comprises extracting a square window of N by N samples from a given frequency domain pattern clip, wherein N is based on a numerical aperture of an imaging lens used in a photolithography process.

3. The method of claim 1 , wherein performing feature extraction comprises extracting a frequency domain feature based on a subset of a given frequency domain pattern clip using layer, illumination, and optics information.

4. The method of claim 1 , further comprising:
training the machine learning classifier model using a support vector machine.

5. The method of claim 4 , wherein training the machine learning, classifier model comprises:
finding a hyperplane w such that features partitioned using maximum separation.

6. The method of claim 5 , wherein finding the hyperplane comprises solving the following quadric equation:
min 1/2 w T w,
s.t.:w.v+b≧ 1,∀ v εnonhotspot,

w.v+b≦− 1,∀ v εhotspot,

where T is the matrix transpose and b is a constant representing the intercept of the hyperplane with the axes.

7. The method of claim 1 , wherein the trained machine learning classifier model comprises a support vector machine or a neural networks.

8. The method of claim 1 , further comprising:
performing resolution enhancement technique optimization or ground rules to fix identified hotspots.

9. A computer program product comprising a computer readable storage medium having a computer readable program stored therein, wherein the computer readable program, when executed on a computing device, causes the computing device to:
receive a design layout;
generate spatial pattern clips from the design layout using a sliding window to split the design layout into overlapping windows;
perform a transform on the spatial pattern clips to form frequency domain pattern clips;
perform feature extraction on the frequency domain pattern clips to form frequency domain features;
utilize the extracted features on a set of training samples to train a machine learning classifier model; and
classify a set of previously unseen patterns, based on frequency domain features of the previously unseen patterns using the trained machine learning classifier model, into hotspots and non-hotspots.

10. The computer program product of claim 9 , wherein performing feature extraction comprises extracting a square window of N by N samples from a given frequency domain pattern clip, wherein N is based on a numerical aperture of an imaging lens used in a photolithography process.

11. The computer program product of claim 9 , wherein the computer readable program further causes the computing device to:
training the machine learning classifier model using a support vector machine.

12. The computer program product of claim 11 , wherein training the machine learning classifier model comprises:
finding a hyperplane w such that features are partitioned using maximum separation.

13. The computer program product of claim 12 , wherein finding the hyperplane comprises solving the following quadric equation:
min 1/2 w T w,
s.t.:w.v+b≧ 1,∀ v εnonhotspot,

w.v+b≦− 1,∀ v εhotspot,

where T is the matrix transpose and b is a constant representing the intercept of the hyperplane with the axes.

14. An apparatus, comprising:
a processor; and
a memory coupled to the processor, wherein the memory stores instructions which, when executed by the processor, cause the processor to:
receive a design layout;
generate spatial pattern clips from the design layout using a sliding window to split the design layout into overlapping windows;
perform a transform on the spatial pattern clips to form frequency domain pattern clips;
perform feature extraction on the frequency domain pattern clips to form frequency domain features;
utilize the extracted features on a set of training samples to train a machine learning classifier model; and
classify a set of previously unseen patterns, based on frequency domain features of the previously unseen patterns using the trained machine learning classifier model, into hotspots and non-hotspots.

15. The apparatus of claim 14 , wherein performing feature extraction comprises extracting a square window of N by N samples from a given frequency domain pattern clip, wherein N is based on a numerical aperture of an imaging lens used in a photolithography process.

16. The apparatus of claim 14 , wherein the computer readable program further causes the computing device to:
training the machine learning classifier model using a support vector machine.

17. The apparatus of claim 16 , wherein training the machine learning classifier model comprises:
finding a hyperplane w such that features are partitioned using maximum separation by solving the following quadric equation:
min1/2 w T w,
s.t.:w.v+b≧ 1,∀ v εnonhotspot,

w.v+b≦− 1,∀ v εhotspot,

where T is the matrix transpose and b is a constant representing the intercept of the hyperplane with the axes.

18. The computer program product of claim 9 , wherein performing feature extraction comprises extracting a frequency domain feature based on a subset of a given frequency domain pattern(clip using layer, illumination, and optics information.

19. The computer program product of claim 9 , wherein the computer readable program further causes the computing device to:
performing resolution enhancement technique optimization or ground rules to fix identified hotspots.

20. The apparatus of claim 14 , wherein performing feature extraction comprises extracting a frequency domain feature based on a subset of a given frequency domain pattern clip using layer, illumination, and optics information.

Assignees

• IBM

Inventors

• Agarwal Kanak B
• Banerjee Shayak
• Pathak Piyush

Classifications

• G06F30/00

  Computer-aided design [CAD] · CPC title

• G06V10/245 Primary

  by locating a pattern; Special marks for positioning · CPC title

• G06F18/2411

  based on the proximity to a decision surface, e.g. support vector machines · CPC title

• G06F30/398 Primary

  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

• G06K2009/3225

  Physics · mapped topic