A comparative study of ceria-based and silica-based slurries for 32 nm shallow trench isolation chemical mechanical planarization

Abstract

We evaluate and compare the performances of two slurries for STI CMP. One, a ceria-based high selectivity slurry, exhibits a polish rate that declines rapidly with polish time. The second is a silica-based non-selective slurry with a sustained intermediate rate. These opposing behaviors are tested on 32 nm device structures in the first of a two-step STI CMP process. The planarization efficiency of the silica-based slurry is observed to be superior to that of the ceria-based slurry, as suggested by the difference in active oxide and field oxide removal rates and manifested by the final measure of step-height between active and field regions after slurry polish. The silica-based slurry also yields better within wafer and within die uniformity. Evaluation of the repeatability of slurry performance reveals that the ceria-based slurry exhibits a rate deviation more than four times larger than that of the silica-based slurry. Post-CMP defectivity measurements evidence the tendency of the silica-based slurry to scratch less than the ceria-based slurry. This study indicates that the silica-based slurry process for step one of 32 nm STI CMP will produce consistent results from lot to lot, with maximum STI oxide remaining (i.e., minimal dishing), low defectivity, and tight uniformity.