Resist process characterization and optimization for ArF lithography

This study presents the results of the characterization and optimization of the coat process of a positive-tone, chemically amplified (CA) photoresist at a 193 nm wavelength, Which is employed for printing sub-100 nm and beyond CDs. A comprehensive approach utilized a statistical design of experiments method, comprised of thirty-four processing variables, to screen the main factors in a typical coat program. A fractional factorial experiment subsequently examined the interaction effects with respect to the resist film thickness and uniformity responses, in addition to residual analysis diagnostics, Which revealed curvature in the data. Using the response surface methodology (RSM) optimization technique, the responses were modeled in design-space by means of a nonlinear stepwise-regression approach. The most significant parameters with corresponding high-order interactions affecting resist thickness and uniformity variations were identified and optimized. With a 95-percent confidence level, the nonlinear statistical-regression model predicted the resist thickness more accurately than it predicted the uniformity at best-recommended process factor settings, matching film-thickness target values while minimizing variations. Additionally, particulate defects were analyzed for confirmation runs with advanced, sophisticated particle-detection metrology equipment. The inability of the model to predict uniformity accurately may be the result of intrinsic experimental error, which has a greater effect on the variation of the resist thickness than it does on the mean value. Such variations contribute to CD variations that impact yield and costs in advanced IC fabrication. An improvement in the noise level will undoubtedly improve the predictability of the resist-coat process uniformity.