Page 49 - Wire Bonding in Microelectronics

28 Cha pte r T w o

for copper wire [2-26]. However, Krazanowski [2-19] observed that a
wide range of structures can remain after bonding, depending on the
metals involved, so this hardening may not be universal in all materials
and/or produced by all bonding equipment.
There is little experimental evidence that US bonding process
occurs because of heat generated by the US energy. Various in-process
bonding measurements have shown less than ~10 to 100°C maximum
temperature rise [2-1]. More recently, Mayer [2-27] as well as Suman
[2-28] measured only small changes in interface temperatures (~10°C)
using surrounding sensors. (M), or Al, topped under-the-pad sensors
that allowed experiments to take place at bonding temperatures of
~150°C for Au ball bonds.
However, one study inferred an interface temperature of 250°C
from observations of crystallographic defect microstructures in wedge-
bonded larger diameter Al wires [2-19]. Another study, bonding Au
balls directly to sputtered type-K thin film thermocouples, obtained a
rise from 240 to 320°C. Other bonding conditions were omitted [2-29].
The temperature rise was attributed to friction; however, it is unlikely
that any welding (bonding) ever took place, considering the bond
was to the thermocouple as the pad. These results cannot be com-
pared to actual bonding measurements where microwelds would
presumably begin to minimize slippage/friction-generated heat after
a few milliseconds.
In no case where bonding actually took place, did the interface
temperature approach the melting point of either weldment or more
than or equal to 300°C normally required for TC bonding. Also, US
Al-Al wedge bonds have been made immersed in liquid N at a tem-
2
perature of 77 K [2-1, 2-9]. Since high heat is not necessary, the activa-
tion energy to form the metal-metal bond may possibly be transmit-
ted in the form of phonon-lattice interactions. Lower clamping forces
have been shown to increase the interface temperature by ~25°C,
implying that there is increased interfacial scrubbing during some
portion of the bonding cycle (note that similar low bonding forces
have also been shown to result in cratering and poor bonding). Some
papers have attributed the entire bonding process to the frictional
rubbing and propose melting temperatures in the interface, [2-30]
but there is little analytical evidence for such where actual welding
occurred.
When Langenecker’s softening mechanism is incorporated into
the above observations, Al wedge bonding to a typical Al IC bond
pad can be summarized as follows: Some wire-to-pad interfacial
motion (scrubbing) certainly occurs upon first application of ultra-
sonic energy, resulting in some interfacial cleaning action and modest
frictional temperature rise. A few milliseconds later, small microwelds
form just inside the perimeter of the mated surfaces as in the lift-off
patterns of Figs. 2-10 and 2-12, and, at this point, the wire-to-pad