Training method for machine learning assisted optical proximity error correction

1. A method of determining representative patterns for training a machine learning model to predict optical proximity corrections, the method comprising:

obtaining a design layout comprising a set of groups of patterns, each group of patterns includes one or more sub-groups, each sub-group of the one or sub-groups of that group comprising one or more patterns of that group;

determining a set of representative patterns of the set of groups of patterns, a representative pattern being a sub-group whose instances appear in the set of groups of patterns;

obtaining, via simulating an optical proximity correction process using the set of representative patterns, optical proximity correction data associated with the set of representative patterns; and

training a machine learning model to predict optical proximity corrections for the design layout, based on the set of representative patterns and the optical proximity correction data.

2. The method of claim 1 , wherein the groups of patterns are arranged in a hierarchy in which each group of patterns includes the one or more sub-groups.

3. The method of claim 2 , wherein the determining the set of representative patterns is an iterative process, an iteration comprising:

searching for the instances of a given sub-group within the hierarchy of the set of groups of patterns;

categorizing the instances of the given sub-group as the representative pattern; and

extracting, from the design layout, pattern information associated with the representative pattern.

4. The method of claim 1 , wherein each group of the set of groups of patterns is associated with a first identifier, and the one or more sub-groups is associated with a second identifier.

5. The method of claim 4 , wherein the determining the set of representative patterns comprises:

comparing the second identifier associated with a given sub-group with identifiers within a hierarchy of the set of groups of patterns; and

identifying, based on the comparison, instances of the given sub-group within the set of groups of patterns having the same second identifier; and

categorizing the instances of the given sub-group as a representative pattern of the set of representative patterns.

6. The method of claim 1 , wherein the obtaining of the optical proximity correction data comprises:

simulating the optical proximity correction process using pattern information associated with a representative pattern of the set of representative patterns; and

providing the optical proximity correction data associated with the representative pattern for a patterning process.

7. The method of claim 1 , wherein the determining the set of representative patterns comprising searching for a given sub-group of the one or more sub-groups within the set of groups of patterns and the searching for the given sub-group does not directly compare the given sub-group's one or more patterns' shapes and sizes with pattern shapes and sizes within the set of groups of patterns.

8. The method of claim 1 , wherein the optical proximity corrections comprise placement of one or more assist features associated with a desired pattern.

9. The method of claim 1 , wherein the optical proximity correction data is in the form of images and the training is based on the images or pixel data of the images.

10. The method of claim 9 , wherein the images are continuous transmission mask (CTM) images, and/or assist feature guidance maps, wherein the CTM images and/or assist feature guidance maps provide assist feature locations associated with the set of representative patterns, and/or wherein the images are assist feature guidance maps generated by model-based optical proximity correction (OPC) simulation, or rule-based OPC simulation.

11. The method of claim 1 , further comprising:

determining, via the trained machine learning model, mask pattern data; and

outputting the mask pattern data to be used in a patterning process to image a substrate.

12. The method of claim 11 , further comprising fabricating, via a mask-making apparatus using the mask pattern data, a mask to be used in the patterning process to image the substrate, and/or

wherein the mask pattern data comprises characteristics upon which the patterning process adjusts one or more process parameters including dose, focus, illumination intensity, and/or an illumination pupil.

13. The method of claim 1 , further comprising:

determining, via simulating using a process model in cooperation with the trained machine learning model, a process condition associated with a desired pattern; and

exposing, via a lithographic apparatus configured according to the process condition and employing a mask corresponding to the desired pattern, a substrate.

14. The method of claim 13 , wherein the process condition comprises a value of one or more process parameters including dose, focus, illumination intensity, and/or an illumination pupil.

15. A computer program product comprising a non-transitory computer readable medium having instructions therein, the instructions, when executed by a computer system, configured to cause the computer system to at least:

obtain a design layout comprising a set of groups of patterns, each group of patterns includes one or more sub-groups, each sub-group of the one or sub-groups of that group comprising one or more patterns of that group;

determine a set of representative patterns of the set of groups of patterns, a representative pattern being a sub-group whose instances appear in the set of groups of patterns;

obtain, via simulating an optical proximity correction process using the set of representative patterns, reference optical proximity correction data associated with the set of representative patterns; and

train a machine learning model to predict optical proximity corrections for the design layout, based on the set of representative patterns and the set of reference optical proximity correction data.

16. The computer program product of claim 15 , wherein the groups of patterns are arranged in a hierarchy in which each group of patterns includes the one or more sub-groups.

17. The computer program product of claim 16 , wherein the instructions configured to cause the computer system to determine the set of representative patterns are further configured to cause the computer system to do as an iterative process, an iteration comprising:

search for the instances of a given sub-group within the hierarchy of the set of groups of patterns;

categorization of the instances of the given sub-group as the representative pattern; and

extraction, from the design layout, of pattern information associated with the representative pattern.

18. The computer program product of claim 15 , wherein the instructions configured to cause the computer system to determine the set of representative patterns are further configured to cause the computer system to:

compare an identifier associated with a given sub-group with identifiers within a hierarchy of the set of groups of patterns; and

identify, based on the comparison, instances of the given sub-group within the set of groups of patterns having the same identifier; and

categorize the instances of the given sub-group as a representative pattern of the set of representative patterns.

19. The computer program product of claim 15 , wherein the instructions configured to cause the computer system to determine the set of representative patterns are further configured to cause the computer system to search for a given sub-group of the one or more sub-groups within the set of groups of patterns, wherein the search for the given sub-group does not directly compare the given sub-group's one or more patterns' shapes and sizes with pattern shapes and sizes within the set of groups of patterns.

20. The computer program product of claim 15 , wherein the instructions are further configured to cause the computer system to:

determine, via the trained machine learning model, mask pattern data; and

output the mask pattern data to be used in a patterning process to image a substrate.