Effect of slurry on Vapor Deposition Polymerization (VDP)Chemical Mechanical Planarization (CMP) in Through-Silicon Via (TSV) Applications

The effect of various types of slurry on vapor deposition polymerization (VDP) polishing in through-silicon via (TSV) applications was studied. The effort involved three learning cycles: one with blanket VDP wafers and two with patterned VDP wafers. During the early stages of the experimentation, the removal rate and non-uniformity of 300mm blanket VDP wafers on a bare silicon substrate using a slurry A were found to be high (94%) within wafer (WIW) and 115% wafer-to-wafer (WTW). In contrast, measurements of pre-chemical mechanical planarization (CMP) blanket VDP thickness showed excellent WIW (±1.3%) and WTW (±1.5%) uniformity (Figure 1). The removal rate of blanket VDP was 6000 A/min, whereas the rate for TEOS was ~150 A/min. Fig 1: Pre-CMP and post-CMP 49-point diameter blanket VDP thickness measurements The difference in order of magnitude raised concerns about controlled CMP of TSVs with a VDP liner. For patterned VDP wafers, CMP was challenging using the baseline process: the Cu overburden was removed with slurry A after up to 150 sec at 3 psi with endpoint detection; Ru/Ta/TaN layers were removed with slurry B barrier slurry after 180 sec at 3 psi. Results with the barrier slurry polish on the Ebara F-REX300S demonstrated that the VDP layer separated from the underlying hardmask and was not uniformly removed. This was in line with results of the CMP evaluation of blanket VDP wafers. The process generated severe defects, primarily scratches on the wafer surface. Figure 2 shows an example of post-TSV CMP defects captured by a Leica INS 3300 inspection microscope. Fig 2: Post-CMP Defects Figure 2 clearly demonstrates that the alumina-based slurry designed for the Ru seed layer is too aggressive for VDP films. The second phase of the patterned VDP CMP evaluation focused on a three-step CMP process using a silica-based barrier slurry. The copper overburden was removed with slurry A after ~150 sec at 3 psi with endpoint detection; the Ru seed layer was removed with barrier slurry B after 60 sec at 3 psi, and the Ta and TaN barrier layers were removed with slurry C after 60 sec at 3 psi. Figure 3 shows a healthy post-CMP TSV top-down inspection. Following silica-based Ru polish experiments with slurry D showed that the silica-based D removed Ru/Ta/TaN with only negligible defects on the post-CMP surface. The copper overburden was removed with slurry A after 150 sec polish at 3 psi. Figure 4 is a healthy post-CMP TSV top-down inspection showing a defect-free surface. Fig 3: Defect-Free Post-TSV CMP Inspection with ThreeStep CMP Process. Fig 4: Defect-Free Post-TSV CMP Inspection with TwoStep CMP Process. Abstract #2498, Honolulu PRiME 2012, © 2012 The Electrochemical Society