Prediction of process-sensitive geometries with machine learning

1. A method comprising:

predicting, via an integrated circuit (IC) modeling program of a computer system, process-sensitive geometries (PSGs) in a proposed IC layout based on violations of a set of processing constraints for the proposed IC layout, the set of processing constraints including a minimum pattern width and a minimum pattern separation, a pattern density, a topography-caused effect, and an inter-layer size in the proposed IC layout, the processing constraints being calculated with a predictive model based on a training data repository including an artificial neural network (ANN) with training data from a plurality of optical rule check (ORC) simulations for different IC layouts;

manufacturing a circuit using the proposed IC layout;

identifying, via the IC modeling program, actual PSGs in a scanning electron microscope (SEM) image of the circuit manufactured using the proposed IC layout;

determining, via the IC modeling program, whether the predicted PSGs correspond to the actual PSGs in the SEM image of the manufactured circuit as being correct, wherein at least one of the actual PSG not corresponding with one of the predicted PSGs indicates the predicting as being incorrect;

in response to the predicting being incorrect:

adjusting, via the IC modeling program, the predictive model based on the actual PSGs in the SEM image of the manufactured circuit, wherein the adjusting includes:

submitting additional ORC data to the training data repository, wherein the additional ORC data represents at least one actual PSG, not included in the predicted PSGs, to the training data repository, and

recalculating the set of spatial-domain processing constraints based on the training data repository with the additional ORC data; and

flagging the proposed IC layout as incorrectly predicted via the IC modeling program;

in response to the predicting being correct, flagging the proposed IC layout as correctly predicted via the IC modeling program;

adjusting a mask shape for the proposed IC layout based on the actual PSGs; and

manufacturing an additional circuit using the proposed IC layout with the adjusted mask shape.

2. The method of claim 1 , further comprising:

selecting, via the IC modeling program, a set of related ORC data for the proposed IC layout based on the manufactured circuit; and

in response to the proposed IC layout or the related ORC data not being represented in the training data repository, storing the proposed IC layout or the related ORC data in the training data repository.

3. The method of claim 1 , further comprising repeating the predicting, manufacturing, and determining for the proposed IC layout, via the IC modeling program, with the adjusted mask shape.

4. The method of claim 1 , further comprising adding ORC simulation data of the proposed IC layout with the adjusted mask shape to the training data repository.

5. The method of claim 1 , wherein the training data repository includes at least one correctly predicted IC layout and at least one incorrectly predicted IC layout.

6. A computer program product stored on a computer readable storage medium, the computer program product comprising program code, which, when being executed by at least one computing device, causes the at least one computing device to:

predict, via an integrated circuit (IC) modeling program, process-sensitive geometries (PSGs) in a proposed IC layout based on violations of a set of processing constraints for the proposed IC layout, the set of processing constraints including a minimum pattern width and a minimum pattern separation, a pattern density, a topography-caused effect, and an inter-layer size in the proposed IC layout, the processing constraints being calculated with a predictive model based on a training data repository including an artificial neural network (ANN) with training data from a plurality of optical rule check (ORC) simulations for different IC layouts;

cause a manufacturing device to manufacture a circuit using the proposed IC layout;

identify, via the IC modeling program, actual PSGs in a scanning electron microscope (SEM) image of the circuit manufactured using the proposed IC layout;

determine, via the IC modeling program, whether the predicted PSGs correspond to the actual PSGs in SEM image of the manufactured circuit as being correct, wherein at least one actual PSG not being one of the predicted PSGs indicates the predicting as being incorrect;

in response to the predicted PSGs being incorrect:

adjust, via the IC modeling program, the predictive model based on the actual PSGs in the SEM image of the manufactured circuit, wherein the adjusting includes:

submitting additional ORC data to the training data repository, wherein the additional ORC data represents at least one actual PSG, not included in the predicted PSGs, to the training data repository, and

recalculating the set of processing constraints based on the training data repository with the additional ORC data; and

flag the proposed IC layout as incorrectly predicted via the IC modeling program;

in response to the predicted PSGs being correct, flagging the proposed IC layout as correctly predicted via the IC modeling program;

adjust a mask shape for the proposed IC layout based on the actual PSGs; and

instruct the manufacturing device to manufacture an additional circuit using the proposed IC layout with the adjusted mask shape.

7. The computer program product of claim 6 , the computer program product further comprising program code for causing the at least one computing device to:

select, via the IC modeling program, a set of related ORC data for the proposed IC layout based on the manufactured circuit; and

in response to the proposed IC layout or the related ORC data not being represented in the training data repository, store the proposed IC layout or the related ORC data in the training data repository.

8. The computer program product of claim 6 , the computer program product further comprising program code for causing the at least one computing device to repeat the predicting, manufacturing, and determining for the proposed IC layout, via the IC modeling program, with the adjusted mask shape.

9. The computer program product of claim 6 , the computer program product further comprising program code for causing the at least one computing device to add ORC simulation data of the proposed IC layout with the adjusted mask shape to the training data repository.

10. The computer program product of claim 6 , wherein the training data repository includes at least one correctly predicted IC layout and at least one incorrectly predicted IC layout.

11. A system comprising at least one computing device configured to perform a method by performing actions including:

predict, via an integrated circuit (IC) modeling program of the at least one computing device, process-sensitive geometries (PSGs) in a proposed IC layout based on violations of a set of processing constraints for the proposed IC layout, the set of processing constraints including a minimum pattern width and a minimum pattern separation, a pattern density, a topography-caused effect, and an inter-layer size in the proposed IC layout, the processing constraints and being calculated with a predictive model based on a training data repository including an artificial neural network (ANN) with training data from a plurality of optical rule check (ORC) simulations for different IC layouts;

cause a manufacturing device to manufacture a circuit using the proposed IC layout;

identify, via the IC modeling program, actual PSGs in a scanning electron microscope (SEM) image of the circuit manufactured using the proposed IC layout;

determine, via the IC modeling program, whether the predicted PSGs correspond to the actual PSGs in the SEM image of the manufactured circuit as being correct, wherein at least one actual PSG not being one of the predicted PSGs indicates the predicting as being incorrect;

in response to the predicted PSGs being incorrect:

adjust, via the IC modeling program, the predictive model based on the actual PSGs in the SEM image of the manufactured circuit, wherein the adjusting includes:

submitting additional ORC data to the training data repository, wherein the additional ORC data represents at least one actual PSG, not included in the predicted PSGs, to the training data repository, and

recalculating the set of spatial-domain processing constraints based on the training data repository with the additional ORC data; and

flag the proposed IC layout as incorrectly predicted via the IC modeling program; and

in response to the predicted PSGs being correct, flagging the proposed IC layout as correctly predicted via the IC modeling program;

adjust a mask shape for the proposed IC layout based on the actual PSGs; and

instruct the manufacturing device to manufacture an additional circuit using the proposed IC layout with the adjusted mask shape.

12. The system of claim 11 , wherein the at least one computing device is further configured to:

select, via the IC modeling program, a set of related ORC data for the proposed IC layout based on the manufactured circuit; and

in response to the proposed IC layout or the related ORC data not being represented in the training data repository, store the proposed IC layout or the related ORC data in the training data repository.

13. The system of claim 11 , wherein the at least one computing device is further configured to repeat the predicting, manufacturing, and determining for the proposed IC layout with the adjusted mask shape.

14. The system of claim 11 , wherein the at least one computing device is further configured to add ORC simulation data of the proposed IC layout with the adjusted mask shape to the training data repository.