Archives
Contribute
|
Technology: SiC Wafer Polishing With Gas Cluster Ion Beams - Part II
|
|
Anil Saigal 05/01/2006
A novel technology utilizing energetic ionized
gas cluster ion beams (GCIB) has been successfully used to reduce the
surface roughness of SiC for electronic applications. Part I of the
article dealt with the basic concepts and GCIB equipment.
(http://www.lokvani.com/lokvani/article.php?article_id=3212)
EXPERIMENTAL CONDITIONS Two,
electronic grade, SiC wafers were used. One wafer (A) was processed
with only Ar clusters while the second wafer (B) was polished with both
Ar and O2 clusters. A series of ion doses at different cluster
energy levels was applied to separate process areas on each
wafer. On the second wafer in particular, a number of process
areas were exposed to two consecutive ion doses in which the cluster
energy and gas were varied. Cluster dose and cluster energy were the
two GCIB process parameters that were varied.
Surfaces
were measured with a Digital Instruments atomic force microscope (AFM)
equipped with analysis software, that was used to calculate surface
roughness values such as arithmetic mean roughness (Ra) and maximum
peak to valley height (Rmax). Rutherford backscattering
(RBS) channeling measurements were also performed on the wafers to
measure the extent of near surface crystal lattice damage. Ion
channeling measurements were performed with He+ ions at 1.8 MeV after
etching with hydrofluoric acid to remove any oxide layer.
RESULTS Figure
3 shows a 5 um x 5 um AFM scan, which is typical of the surfaces of the
SiC wafers. This surface distinctly shows the effects of
mechanical polishing. The large number of surface scratches
should be noted. The more pronounced scratches are on the order
of 50 Ã… deep. The average virgin surface roughness Ra was 8.4 Ã…
for wafer A and 9.6 Ã… for wafer B.
Figure 3: 5mm x 5mm AFM image of untreated SiC wafer B surface. (Ra = 9.6 Ã…, Rmax = 248 Ã…) As
described above and shown in Figure 3, many polishing scratches are
visible on the as-received surfaces of the SiC wafers. Figure 4
shows processing with Ar clusters greatly improve surface topography by
removing CMP scratches, but failed to improve the measured surface
roughness. Ar cluster processing added high frequency roughness
components to the surface, but a dual energy Ar process was able to
slightly improve Ra. Figure 4: 5mm x 5mm AFM images of SiC surfaces exposed to different GCIB Ar cluster energies: a) high energy low dose, Ra = 9.5 Ã…, Rmax = 178 Ã…, scratches still visible; b) high energy greater dose, Ra = 8.2 Ã…, Rmax =
105 Ã…, scratches gone but some high frequency roughness; c) dual dose,
high energy dose followed by low energy dose (condition A5) Ra = 7.0 Ã…, Rmax = 86 Ã…, smoothest argon result. Figure
5 shows the effect of GCIB processing on the SiC wafer surface with O2
clusters. At low cluster energy, the polishing scratches are less
distinct. At a higher energy the polishing scratches are
completely gone, but a higher frequency roughness is now present.
Figure 5 shows that the low energy treatment slightly lowered the
roughness while the high-energy process increased it.
A plot of
Ra normalized to the untreated roughness value versus ion dose is shown
in Figure 6. It shows the effect on surface roughness as a function of
ion dose for 1) Ar single step processes and 2) Ar and O2 dual energy
processes. In the case of Ar alone, there is an initial is an
increase in roughness from 8.4 Ã… (as received) to 11 Ã… and then a
gradual smoothing with increasing dose.
Figure 5: 5mm x 5mm AFM images of SiC surfaces exposed to different GCIB O2 cluster energies: a) low energy, Ra = 9.1 Ã…, Rmax = 157 Ã…, scratches unaffected; b) high energy, Ra = 11.6 Ã…, Rmax = 355 Ã… Figure 6: Plot of relative Ra
(relative to as received roughness) vs. dose at high cluster energy (25
kV) with Ar clusters. Also shown is the effect of a dual energy
smoothing process, consisting of a dose at high energy followed by a
dose at lower energy for O2 and Ar. (The lines are only a guide to the eye) The
greatest reduction in Ra and Rmax while utilizing only Ar was obtained
using a dual energy smoothing process consisting of an initial dose at
high energy followed by a dose at a moderate energy. As shown in
Figure 7, this dual energy process produced a greater reduction in
surface roughness than an equivalent dose using only high energy
clusters. The result was a final Ra of 7.1 Ã… and Rmax of 86 Ã….
Figure 7: 5mm x 5mm AFM images of SiC surfaces exposed to GCIB polishing with O2 clusters at two energy levels. (Ra = 3.8 Ã…, Rmax = 69 Ã…), smoothest result. Processing
with O2 clusters produced similar results. Low energy oxygen
cluster doses alone did not significantly reduce surface roughness or
remove polishing scratches. High energy oxygen doses were needed
to alter the surface morphology and remove scratches. As with Ar
clusters, these high-energy oxygen doses introduce a high frequency
roughness component, which needs to be reduced by a subsequent lower
energy oxygen dose. As with the argon dual energy process both
low frequency polishing scratches and high frequency roughness were
reduced. However, the oxygen process produced an overall smoother
surface than the argon process with a final Ra of 3.8 Ã… and Rmax of 69
Ã… (Figure 7).
Defect density (Nsp) and Chi Min values (Xmin)
from RBS measurements of crystal lattice damage. Areas exposed to lower
energy clusters had lower levels of damage than those exposed to higher
energy clusters. The areas exposed to a dual energy process (high
energy dose followed by lower energy dose) had still lower damage
levels, similar to the unprocessed areas. Regions processed with
only oxygen clusters had significantly lower damage levels than those
processed with argon.
CONCLUSIONS Gas
cluster ion beam smoothing appears to be a viable process for improving
the surface quality of electronic grade SiC. CMP polishing
scratches were removed and maximum peak to valley height was decreased
overall by over 60% by using a dual energy oxygen cluster ion beam
process.
(Anil Saigal is Chair and Professor of Mechanical Engineering at Tufts University. He can be reached at anil@lokvani.com. )
|
You may also access this article through our web-site http://www.lokvani.com/
|
|