Online information source for semiconductor professionals

KATRINA MIKHAYLICHENKO & MIKE RAVKIN, Lam Research Corp., Fremont, CA, USA
DAVE STEIN & DALE HETHERINGTON, Sandia National Laboratory, Albuquerque, NM, USA

ABSTRACT

Lam Research and Sandia National Laboratory have conducted studies that compare the cleaning capabilities of non-contact megasonic cleaning with contact cleaning that uses brush scrubbing. We found that partially planarised features are far more difficult to clean than polished blanket dielectric or metal surfaces. Features with larger step heights and/or small horizontal dimensions may efficiently trap slurry and are, of course, difficult to clean. To investigate the effect of topography size and aspect ratio on the relative cleaning efficiency of spin-rinse-dry (SRD), megasonic, and brush scrubbing techniques, special masks with topography of variable size and aspect ratio were developed. Trenches, holes, and corners from 0.6 to 4.0 microns in size with aspect ratio changing from 1:1 to 4:1 were etched in PETEOS film. Two groups of wafers were then used for the tests. Group I wafers were dipped in fumed silica, colloidal silica or Al2O3 slurry and cleaned using either a megasonic cleaning or a brush scrubber. Group II wafers were polished with the same slurries to remove various amounts of material and then cleaned using the same tool sets. In this paper we present the results obtained with different cleaning parameters and chemicals. Our results indicate that it is extremely difficult to remove slurry from deep features after CMP regardless of the cleaning technique. In conclusion, we show that both the step height and the horizontal dimension of the feature are crucial to the cleaning capability of a specific cleaning technique.