Asml Masktools B.v.

SangBong Park

Jang Fung Chen

Armin Liebchen

Tamer Coskun

Gabriel Berger

Ting Chen

A method of optimizing a process for use with a plurality of lithography systems. The method includes the steps of: (a) determining a calibrated resist model for a given process and a target pattern utilizing a first lithography system; (b) selecting a second lithography system to be utilized to image the target pattern utilizing the given process, the second lithography system capable of being configured with one of a plurality of diffractive optical elements, each of the plurality of diffractive optical elements having corresponding variable parameters for optimizing performance of the given diffractive optical element; (c) selecting one of the plurality of diffractive optical elements and simulating the imaging performance of the second lithography system utilizing the selected one of the plurality of diffractive optical elements, the calibrated resist model and the target pattern; and (d) optimizing the imaging performance of the selected one of the plurality of diffractive optical elements by executing a genetic algorithm which identifies the values of the parameters of the selected one of the plurality of diffractive optical elements that optimizes the imaging of the target pattern.

Method of performing resist process calibration/optimization and DOE optimization for providing OPE matching between different lithography systems

Improved calibration of a resist model used in critical dimension (CD) calculation is disclosed. A dose function is obtained based on optical tool to be used form the resist on a wafer. The dose function indicates the amount of energy in a resist. The dose function is convolved with a convolution kernel to obtain a modified dose function. The convolution kernel has variable diffusion lengths in different directions. The convolution kernel may include multiple Gaussian kernels each having variable diffusion lengths in different directions. The modified dose function is converted into a CD value which is compared with a target value. If necessary, the diffusion lengths of the Gaussian kernels are adjusted based on the comparison result.

Growing community of inventors

Fremont, CA, United States of America

SangBong Park