Page 90 - Wire Bonding in Microelectronics

Bonding W ir e Metallur gy and Characteristics 69

out are the wire resistivity, thermal conductivity, temperature coeffi-
cient of resistance, and melting point. Other possible factors such as
the deformation of the wedge bond and the quality of the bond (the
percent of welded interface) can also affect the burn out of short
wires. These latter have never been evaluated or even considered in
the published literature.
Heat conductivity from the wire to the chip or package, as well as
2
the I R heat generated by the given length of wire are important fac-
tors in changing the burnout current. Thus, in an open cavity package,
assuming a minimally deformed perfectly welded bond interface, the
2
longer the wire, the lower the burnout current (more I R heating and
less of it conducted out the ends), up to some length in which thermal
conduction out the bonded ends is insignificant. As the wire is length-
ened, convection and radiation losses into the ambient, control the
heat loss process more than thermal conduction out the bonded ends.
However, this is very different for plastic encapsulated devices.
Aluminum wire responds differently in oxygen (or air) than in
inert gasses or in vacuum. A rapid burn out (in a millisecond or less)
in oxygen may result in distorted ball formation on each side of the
wire. However, when heated slowly by a current ramp-up (longer
than a few seconds), a thick aluminum oxide sheath is produced
which changes the heat transfer into the ambient, protects the liquid
metal from further oxidation, and holds it in place. Thus, the Al wire
temperature can rise hundreds of degrees above its melting point,
which results in an apparent artificially high burnout current. When
the current is removed, the liquid metal cools and contracts. Some-
times an open circuit will result (if no continuous metal remains
inside the Al O sheath). At other times, the wire survives and, for
2 3
practical purposes, has a higher burnout current than its temperature
would predict. The first observation of this phenomenon was reported
by Kessler [3-25] and later verified [3-29].
Gold wire, which does not oxidize, burns out neatly at its melting
point, leaving gold balls on each open wire end. Assuming that the
heat conducted out through each weld is the same, such wire will
burn out approximately in the center of the span. Gold wire has both
a higher melting point and a lower resistivity, and thus has a higher
burnout current than Al.∗ Unfortunately, there is little experimental

∗Al, 1% Si, 25 µm diameter, bonding wire has a resistivity of approximately
3.1 mΩ-cm @ 20°C, which results in a resistance of about 60 Ω/m or 0.06 Ω/mm
(40 mil length). Its melting point is in the range of 600 to 655°C. The same diameter
gold wire, 99.99% pure, has a resistivity of approximately 2.4 mΩ-cm @ 20°C, and a
resistance of about 45 Ω/m or 0.045 Ω/mm. Its melting point is 1063°C. In practice,
the exact resistivity varies somewhat with added impurity, especially when at the
99.9% level. Also, the measured resistance is very dependant on the accuracy
of the actual wire diameter. (Most specifications allow ∼5% variation.) Since the
2
resistance varies as 1/r , these specifications can result in about a 10% variation of
burnout current for a 25-µm diameter wire, everything else being equal.