Aberration Analysis and Control Based on Fully Connected Neural Network

Thermal aberration of a projection lens, which is caused by lens heating, leads to degradation of imaging quality in the Deep Ultraviolet Lithography (DUVL). Two tasks have been done in this article. One is to analyze the impact of single aberration on imaging quality, and the other is to use machine learning to calculate the combination of Zernike aberration coefficients to make the process window larger. The impact of aberration on DUVL imaging has been analyzed individually for 16 Zernike terms. It is found that Zernike terms Z
₅
and Z
₁₂
are determined to be the prominent terms affecting the Depth of Focus (DOF). Moreover, Z
₁₀
aggravates the Image Log Slope (ILS) and the Zernike Z
₁₅
term increases the value of Mask Error Enhancement Factor (MEEF). However, there is more than one Zernike terms in the aberration of a real DUV optics. To make reasonable budget for aberration, using a fully connected neural network (FCNN) model to find some combinations of Zernike coefficients, so that the value of DOF, MEEF and ILS is close to the value of zero aberration. This implies that the FCNN models possess the ability to screen out better Zernike combination regardless of each individual Zernike term value in the range of -25 mλ to 25 mλ, which will play a guiding role in DUV optics design.