Characterization of Via Etching in CHF3/CF4 Magnetically Enhanced Reactive Ion Etching Using Neural Networks

This study characterizes an oxide etching process in a magnetically enhanced reactive ion etching (MERIE) reactor with a CHF3/CF4 gas chemistry. We use a statistical 2^4‐1 experimental design plus one center point to characterize the relationships between the process factors and etch responses. The factors that we varied in the design include RF power, pressure, and gas composition, and the modeled etch responses were the etch rate, etch selectivity to TiN, and uniformity. The developed models produced 3D response plots. Etching of SiO₂ mainly depends on F density and ion bombardment. SiO₂ etch selectivity to TiN sensitively depends on the F density in the plasma and the effects of ion bombardment. The process conditions for a high etch selectivity are a 0.3 to 0.5 CF₄ flow ratio and a –600 V to –650 V DC bias voltage according to the process pressure in our experiment. Etching uniformity was improved with an increase in the CF4 flow ratio in the gas mixture, an increase in the source power, and a higher pressure. Our characterization of via etching in a CHF₃/CF₄ MERIE using neural networks was successful, economical, and effective. The results provide highly valuable information about etching mechanisms and optimum etching conditions.