Optical Proximity Correction Based On Semi-implicit Addition Operator Splitting

Photography is a key process in integrated circuit manufacturing,which transfers the target graphics on the mask to the photoresist through the projection of the lithography system and the photochemical reaction characteristics of the photoresist.Due to the imperfection of the optical system and the diffraction effect of the light,the figures projected on the photoresist are not exactly consistent.Optical proximity correction(OPC)is compensation on the mask using computational methods.Reverse lithography is an important method to compensate for image distortion,which relies on the iterative optimization of computational lithography,causing poor computational efficiency due to the small time steps of the explicit iteration equation.Firstly,this paper adopts a semi-implicit time discrete scheme,which increases the time step while ensuring the stability of the mask synthesis process.The proposed method achieves additive operator splitting(AOS)for the coordinate axis,reduces the mask synthesis to continuous one-dimensional updates,expressed in three-diagonal linear equations,and is solved quickly by the Thomas algorithm.Combining simplicity and stability greatly reduces the Diego algebra and the overall computational load.Secondly,based on a partially coherent light projection imaging system is built,and the imaging effect and difference of the two are compared and analyzed in the same lighting system,and then the two groups of experiments are inversely optimized and compared the output experiment results before and after the optimization.Finally,the respective optimization results are obtained by using the mask optimization(MO)method and the light source mask optimization(SMO)methods under the same vector model and the same lighting system.By comparing the difference analysis of the experimental results and generating the corresponding optimization cost function,the SMO method has great advantages in both image quality and degrees of freedom.