Page 122 - Wire Bonding in Microelectronics

W ir e Bond Testing 101

Cleanliness and wear of shear tools: Since a shear tool should be posi-
tioned within ~2.5 to 5 µm (~ 0.1 to 0.2 mil) from the chip or sub-

strate (and even closer for fine-pitch situations), its bottom surface
must be kept clean. Metallic smears, bits of Al, Au, glassivation,
Si, etc., can stick to the bottom and prevent proper vertical posi-
tioning. This problem is most significant when shearing very fl at

(low-profile) balls with tools that are several times wider than the
diameter of those balls. A narrow tool (about 1-ball-diameter wide)
tends to clean itself off as it shears through metal, but not if the
entire ball pushes off.
Shear tools wear during long use and/or mishandling. The front
and bottom surfaces must be smooth, polished, and rectangular in
shape, with no chipouts permitted on the shearing edge. To obtain
reproducible test results, shear tools must be monitored and replaced
with the same urgency as bonding tools and capillaries. Tools used
for finer-pitch balls require more frequent monitoring than tools for
coarser pitch, but the test becomes increasingly impractical for pitches
less than 50 µm.

4.3.5 Ball-Shear Force versus Bonded Area
Experimental shear force data versus actual bonded area for ball bonds
have been published by two independent sources [4-16, 4-17] and

verified [4-11]. The shear strength of both Cu- and Ag-doped (<10 ppm)
Au wires was measured to be 90 MPa (13,100 psi). Beryllium-doped
and recent proprietary Au alloy wires could be 10 to 20% higher. The
ultimate shear strength for the hard and annealed 1% Si, Al wires,
respectively, were 139 and 84 MPa (20,200 and 12,200 psi). Some shear
strength values of the hard and annealed Al wires bracketed those of
Au, indicating that a ball-bond weld can fail either on the Al or Au side
depending on the particular characteristics of the Al pad. These data are
plotted in Fig. 4-17 as shear force versus diameter of the bonded area.
The wires used in this experiment contained Cu and Ag impuri-
ties at the parts per million level. Gold wire used in automatic bond-
ing machines is usually stabilized with Be, Ca, and proprietary ele-
ments, which results in a 10 to 30% stronger wire in a stress-strain
test. One might assume that such Au wire would produce higher
gold-ball shear values, but this has not been experimentally observed. ∗
Values obtained from Fig. 4-17 should be considered maximum theo-
retically obtainable values, rather than expected values, since all of
the area under a ball is not welded. Thermosonic ball bonds on an Al

∗ Comparisons using 25 µm wire with breaking loads ranging from 8 to 13 gf
showed no statistical difference in ball-shear values for the same size balls [4-26].
Apparently, the “burnout annealing” that occurs during EFO ball formation leaves
the balls in a similar metallurgical condition.