Observation of silicon

1
Observation of silicon debris in chips
cut-before-thinning
A set of 9 chips (from wafers VB58MPX and
VT5MW6X) cut with the new method
(cut-before-thinning) has been sent to Genoa in a
new gel-pak. These chips are electrically
bad. The chips have been moved out of the gel-pak
after arrival in Genoa and we looked for Si
debris on the gel-pak, especially along the chip
periphery. For this measurement we have used a
digital microscope (Keyence VHX 100,
magnification175) which also allows to measure
the dimensions of the debris.
2
Debris are along a (horizontal) line. Typical
dimension is very small (few mm), some are 25-30
mm in radius. The center of the two large debris
is 989 mm apart.
3
Here another dangerous debris (same picture as in
previous page), 20 mm in diameter.
4
53x24mm
62x22mm
There are debris for 1mm, then few mm without
appreciable number of debris.
5
40x19mm
Another example, still following an horizontal
line. We could not see as clearly lines of
debris along the vertical edges of the chips,
only some debris are visible here and there.
6
In the next two pages, the situation with the old
cutting procedure is shown for comparison. These
pictures, taken with a standard high
magnification microscope, have been already
circulated in July04
7
X 1000
8
Pictures taken at x1000 (see bump for comparison)
2 pictures showing fragments along the edge of a
chip (now taken out).
X 200
X 100
9
We find a couple of gt10 mm dia (i.e. dangerous)
debris per mm along the horizontal line close to
the EoC logic. The distribution is not uniform
(i.e. there are regions of debris). Some
debris are present around the other sides, but
there is no evidence of lines of debris as
those shown in this report. The number of
dangerous debris per mm is one order of magnitude
less. This may be related to the way the chip is
taken in/out from the gel-pak. The comparison
with the previous cutting procedure indicate
smaller debris and possibly less per mm. The
improvement does not appear to be very
significant because the smaller debris are in
several cases still large enough to create VDDA
problems. The edge of an old and a new chip
has been compared then.
10
New cutting
Old cutting
40x
Bump side
Bump side
11
Again there is no evidence of difference between
the edges. The new cut procedure seems, in this
case, slightly worse than the old one
contraddicting other observations. It would be
good to observe the same chip edge before leaving
LBNL and after arrival in Genoa. A larger detail
(next page) indicate that silicon debris of
several tens of microns can fly off the edge on
the side opposite to the bumps (so no surprise if
we find them on the gel-pak).
12
A 1000x picture of the edge of a new-cut chip
(0208A of wafer VT5MW6X), The size of the
chipping can be compared with a bump (below).