Method and system for simulating and verifying layout based on distribution
A method for simulating a layout of an integrated circuit manufactured by a semiconductor process includes extracting a plurality of pattern layouts from layout data that defines the layout, generating training data by amplifying the plurality of pattern layouts and at least one parameter provided from the semiconductor process, generating sample data by sampling the training data, generating feature data including a three-dimensional array from the sample data, providing the sample data and the feature data to a simulator and a machine learning model, respectively, and training the machine learning model based on an output of the machine learning model and an output of the simulator.
1 . A method for simulating a layout of an integrated circuit manufactured by a semiconductor process, the method comprising:
extracting a plurality of pattern layouts from layout data that defines the layout;
generating training data by amplifying the plurality of pattern layouts and at least one parameter provided from the semiconductor process;
generating sample data by sampling the training data;
generating feature data including a three-dimensional array from the sample data;
providing the sample data and the feature data to a simulator and a machine learning model, respectively; and
training the machine learning model based on an output of the machine learning model and an output of the simulator.
2 . The method of claim 1 , wherein the extracting of the plurality of pattern layouts comprises:
pre-processing the layout data based on information about a plurality of reference patterns;
grouping patterns respectively corresponding to the plurality of reference patterns from the pre-processed layout data into a plurality of groups; and
extracting coordinates of the plurality of pattern layouts respectively corresponding to the plurality of groups.
3 . The method of claim 2 , wherein the pre-processing of the layout data comprises adjusting a resolution of the layout data to correspond to information about the plurality of reference patterns or a resolution of the feature data.
4 . The method of claim 2 , wherein the pre-processing of the layout data comprises flattening a hierarchy included in the layout data.
5 . The method of claim 1 , wherein the generating of the training data comprises performing a design of experiments (DOE) by sampling the at least one parameter.
6 . The method of claim 1 , wherein the generating of the sample data comprises:
providing the feature data corresponding to the training data to the machine learning model and collecting a plurality of outputs of a hidden layer of the machine learning model; and
grouping the plurality of outputs of the hidden layer into a plurality of groups; and
sampling the sample data from the training data based on the plurality of groups.
7 . The method of claim 6 , wherein the generating of the sample data further comprises:
training the machine learning model so that a Lipschitz constant in a latent space of the plurality of outputs decreases.
8 . The method of claim 1 , wherein the generating of the feature data comprises:
transforming a pattern layout included in the sample data based on at least one parameter included in the sample data;
generating a plurality of two-dimensional arrays respectively corresponding to a plurality of layers of the transformed pattern layout; and
generating the three-dimensional array including the plurality of two-dimensional arrays.
9 . The method of claim 8 , wherein the generating of the feature data further comprises:
generating a new layer from at least one of the plurality of layers based on the at least one parameter included in the sample data; and
generating a two-dimensional array corresponding to the new layer.
10 . The method of claim 1 , wherein the machine learning model comprises:
a first sub-model configured to receive the three-dimensional array;
a second sub-model configured to receive at least one parameter included in the feature data; and
a third sub-model configured to generate the output of the machine learning model from an output of the first sub-model and an output of the second sub-model.
11 . The method of claim 10 , wherein
the third sub-model comprises a deconvolution layer, and
the output of the machine learning model is a two-dimensional array.
12 . The method of claim 11 , wherein
the feature data comprises:
a first three-dimensional array comprising two-dimensional arrays of same size as the output of the machine learning model; and
a second three-dimensional array comprising two-dimensional arrays of a greater size than the output of the machine learning model, and
the first sub-model comprises a model receiving the first three-dimensional array and a model receiving the second three-dimensional array.
13 . A method for simulating a layout of an integrated circuit manufactured by a semiconductor process, the method comprising:
extracting a plurality of pattern layouts from layout data that defines the layout;
obtaining at least one distribution of parameters of the semiconductor process;
generating at least one input parameter by sampling the at least one distribution;
generating feature data including a three-dimensional array from the plurality of pattern layouts and the at least one input parameter;
providing the feature data to a machine learning model trained based on an output of a simulator; and
verifying the layout based on an output of the machine learning model, wherein verifying the layout comprises generating verification data, and the verification data includes a value indicating reliability of a pattern layout of the plurality of pattern layouts.
14 . The method of claim 13 , wherein the obtaining of the at least one distribution comprises obtaining at least one distribution from a process model that models the semiconductor process.
15 . The method of claim 13 , wherein the generating of the at least one input parameter comprises generating the at least one input parameter by performing Monte Carlo sampling on the at least one distribution.
16 . The method of claim 13 , wherein the verifying of the layout comprises calculating a standard score from the output of the machine learning model based on a threshold value.
17 . The method of claim 16 , wherein the calculating of the standard score comprises:
counting an output of the machine learning model that is less than or equal to the threshold value;
calculating a probability based on a result of the counting; and
calculating the standard score based on the probability.
18 . The method of claim 16 , wherein the calculating of the standard score comprises:
calculating a probability based on the threshold value and importance sampling; and
calculating the standard score based on the probability.
19 . The method of claim 16 , wherein the verifying of the layout comprises:
collecting standard scores corresponding to a plurality of pattern layouts; and
calculating a reliability index of the integrated circuit, based on the collected standard scores.
20 . A system comprising:
at least one processor; and
a non-transitory storage medium storing instructions which, when executed by the at least one processor, allow the at least one processor to perform the method of claim 13 .