Page 107 - Wire Bonding in Microelectronics

86 Cha pte r F o u r

assuming a normal distribution, the prediction was that 2.27% of the
bonds should have pull forces in the range of less than or equal to
34.3 mN (≤3.5 gf). However, experimental data revealed only 0.4% in
that low range. The Chi Square statistic confirmed that normality was
absent [4-8].

4.2.4 Effect of Metallurgy and Bonding Processes
on the Bond Pull Force
In a production-line environment where speed is essential, pull-test
operators seldom ascertain that the hook is at the exact center of the
bond loop. Often, the hook will slip toward the highest point of the
loop. This point is determined by the type of bonding machine or by
the device package. If the package has one very high or low bond pad,
then hook slippage∗ can lead to peel-mode failures as previously
described. If both bonds are well made, however, the method of bond-
ing will generally dominate the results, as discussed below.
Gold ball bonds (thermosonic) are normally bonded with a capil-
lary-type tool. Assuming a normal loop, the wire rises straight up
from the center of the ball to a peak near the ball, bends, and pro-
gresses linearly downward towards the second bond, which is the
wedge or crescent bond (see Fig. 4-4). If the pulling hook rises to the
peak, most of the force is applied directly to the ball, which, because
of its large bonded area, is stronger than the wire. [Above the ball
bond peeling or tearing does not occur with off-center hook placement
as it does for wedge bonds (but it could if the bond pitch spacing is
below ~50 µm—see Cu/Lo-k, Chap. 10).] Typically, the wire breaks in
the recrystallized (heat-affected) zone immediately above the ball.
The wedge or crescent bond is usually weaker than the ball bond.
However, when the hook is located near the peak of the loop (nearer the
ball), relatively little force is applied to the wedge bond, and it seldom
breaks. Thus, only the heat-affected zone (neck) of the stronger bond
(the ball) is tested (see Sec. 4.3 on shear testing of ball bonds). For
single-level ultrasonic wedge bonds, Fig. 4-4, the case is reversed. The
wire rises from the edge of the first bond (which is the weaker bond),
peaks somewhat before the center of the relatively low loop, and then
goes down continuously to the second, stronger bond. Thus, if the hook
rises to the peak of the loop, more of the force is applied to the weaker
bond, which breaks. In this case, the stronger bond remains untested. It
is apparent that the combination of a high-bond loop as well as a force
distribution that tests the stronger bond is one reason why Au ball
bonds are specified to have, and do give, a higher pull force than Al

∗ Most modern pull testers have stiff hooks that eliminate slippage and pull
vertically regardless of the shape of the loop. These are preferable but might lead to
some other problems in dealing with complex loops generated by autobonders.