Method for optimizing a number of kernels used in a sum of coherent sources for optical proximity correction in an optical microlithography process

A method for optical proximity correction of a mask used in an optical microlithography process, comprising:

• automatically optimizing a number of kernels N used in a sum of coherent sources (SOCS) based on accuracy and asymmetry;
• simulating shapes formed by the mask using the optimized value for N;
• comparing the simulated shapes with desired shapes;
• and
• correcting the mask based on the comparison between the simulated shapes and the desired shapes.
• wherein the step of automatically optimizing the number of kernels N comprises:
• setting the number of kernels N to a value at least as large as a predetermined minimum value Nmin;
• determining an added X/Y asymmetry estimate of calculated intensity by switching between X and Y using the number of kernels N, comparing the X-Y results with Y-X results in a certain norm, and computing the X/Y asymmetry;
• determining whether the added X/Y asymmetry estimate of calculated intensity is within a first tolerable value;
• and
• determining whether an accuracy estimate of the calculated intensity is within a second tolerable value.

A method for optimizing a number of kernels N used in a sum of coherent systems (SOCS) for optical proximity correction in an optical microlithography process, comprising:

• setting the number of kernels N to a value at least as large as a predetermined minimum value Nmin;
• determining an added X/Y asymmetry estimate of calculated intensity by switching between X and Y using the number of kernels N, comparing the X-Y results with Y-X results in a certain norm, and computing the X/Y asymmetry;
• determining whether the added X/Y asymmetry estimate of calculated intensity is within a first tolerable value;
• increasing the value of N if the added X/Y asymmetry estimate of calculated intensity is not within the first tolerable value, where a 1st through (N+1)th SOCS eigenvalue correspond to the 1st through (N+1)th kernel, respectively, and determining whether the Nth SOCS eigenvalue is larger than the (N+1)th eigenvalue;
• and
• determining whether an accuracy estimate of the calculated intensity is within a second tolerable value.