Methods of determining process models by machine learning
Methods of determining, and using, a patterning process model that is a machine learning model. The process model is trained partially based on simulation or based on a non-machine learning model. The training data may include inputs obtained from a design layout, patterning process measurements, and image measurements.
1. A method comprising:
obtaining values of inputs of a process model of a patterning process;
obtaining values of a first set of one or more physical characteristics of an image formed on a substrate under a processing condition represented by the values of the inputs;
obtaining simulated values of a second set of one or more characteristics of the image; and
determining, by a hardware computer, the process model based on training data comprising samples whose feature vector comprises the values of the inputs and whose supervisory signal comprises the values of the first set, and based on the simulated values of the second set, the process model configured to output one or more physical characteristics of an image formed on a substrate using the patterning process or output one or more characteristics of the patterning process to achieve a particular result of the patterning process.
2. The method of claim 1 , wherein the values of the inputs are obtained from a design layout or by measurements on the patterning process.
3. The method of claim 1 , further comprising computing a cost function of the inputs.
4. The method of claim 3 , wherein the cost function represents a deviation between an output of the process model and a target value.
5. The method of claim 3 , further comprising, upon determination that a termination condition is not satisfied, adjusting one or more values of the values of the inputs.
6. The method claim 5 , wherein the adjusting the values of the inputs is based on a gradient of the cost function with respect to one or more values of the values of the inputs.
7. The method of claim 5 , wherein the adjusting the values of the inputs comprises retraining the process model.
8. The method of claim 1 , wherein the first set of one or more characteristics of the image is a CD in the image.
9. The method of claim 1 , wherein the first set of one or more characteristics of the image is a location of an edge in the image.
10. The method of claim 1 , wherein the values of the inputs comprise values of one or more characteristics of an illumination, of one or more characteristics of projection optics, of one or more characteristics of a patterning device, and/or of one or more characteristics of one or more post-exposure procedures.
11. The method of claim 1 , wherein the values of the first set are obtained by measuring the image or simulating the image.
12. The method of claim 1 , wherein the first set and the second set are different.
13. The method of claim 1 , wherein the simulated values are obtained based on the values of the inputs.
14. The method of claim 1 , wherein determining the process model comprises using the simulated values as a constraint.
15. A computer program product comprising a non-transitory computer readable medium having instructions therein, the instructions, upon execution by a computer system, configured to cause the computer system to at least:
obtain values of inputs of a process model of a patterning process;
obtain values of a first set of one or more physical characteristics of an image formed on a substrate under a processing condition represented by the values of the inputs;
obtain simulated values of a second set of one or more characteristics of the image; and
determine the process model based on training data comprising samples whose feature vector comprises the values of the inputs and whose supervisory signal comprises the values of the first set, and based on the simulated values of the second set, the process model configured to output one or more physical characteristics of an image formed on a substrate using the patterning process or output one or more characteristics of the patterning process to achieve a particular result of the patterning process.
16. The computer program product of claim 15 , wherein the values of the inputs are obtained from a design layout or by measurements on the patterning process.
17. The computer program product of claim 15 , wherein the instructions are further configured to cause the computer system to compute a cost function of the inputs.
18. The computer program product of claim 15 , wherein the values of the first set are obtained by measuring the image or simulating the image.
19. The computer program product of claim 15 , wherein the first set and the second set are different.
20. The computer program product of claim 15 , wherein the simulated values are obtained based on the values of the inputs.
21. The computer program product of claim 15 , wherein the simulated values of the second set are used as a constraint in training of the process model using the training data, wherein a retraining of the process model is triggered based on the simulated values of the second set, and/or wherein a combining of the process model with another model is triggered based on the simulated values of the second set.
22. The method of claim 1 , wherein the simulated values of the second set are used as a constraint in training of the process model using the training data, wherein a retraining of the process model is triggered based on the simulated values of the second set, and/or wherein a combining of the process model with another model is triggered based on the simulated values of the second set.