Page 45 - Wire Bonding in Microelectronics

24 Cha pte r T w o

Figs. 2-10 and 2-11 except that as the bonds mature the center also
became welded [2-12].
The Au-Au crescent-wedge (after ball bond) weld formation pat-
terns were studied by Zhou [2-13] and found to be similar to US Al
wedge growth (starting around the perimeter and generally growing
inward with power/time). Thus most bonds start around the perim-
eter and mature inwardly, but there are many reported variations.
Although the initial welds on TC-ball bonds form around the
perimeter, as with US wedge bonds, TS bonds appear to follow a more
random pattern, as shown in Fig. 2-12 [2-22]. The microwelds may be
elongated in the direction of the ultrasonic vibration motion. The weld-
ing time in this figure ranged from 2 to 16 ms at 100 kHz. No equivalent
comparison has been made at 60 kHz bonding frequency. However a
fully welded bond from a recent mfg’s bonder is shown for compari-
son in Fig 2-12(E), where the intermetallic appeared in swirl patterns,
and these have been observed by others as well. Thus each bonder can
develop welding in different patterns (as revealed by etching ∗ ).
2.3.1. Brief Phenomenological Explanation of the Ultrasonic
and Thermosonic Bonding Process
Ultrasonic welding is a type of deformation weld in which the metal
is first softened by the ultrasonic energy. The clamping force deforms
the softened wire or ball against the equivalently softened bonding
pad, sweeping aside brittle surface oxides and contaminants, leaving
clean surfaces in contact. A pictorial cartoon demonstrating this pro-
cess for a ball bond is given in Fig. 2-13, but it is also appropriate for
a wedge bond as in Fig. 2-14. The modeling of the ultrasonic bonding
process is detailed in Chap. 11.
Little deformation takes place in the center of the weld, so the
oxides and contaminants mostly remain there, and this area is often
observed to remain unwelded, as shown for wedge bonding in
Fig. 2-14. Presumably, the same energy transfer mechanism that soft-
ens the metals without significant heat generation also supplies the
required activation energy for metal-metal interdiffusion and, for
Au-Al bonds, the formation of intermetallic compounds. This forms
these metal-to-metal (atomic) bonds within a few milliseconds. Note
∗ The amount of Au-Al intermetallic formation is often used to demonstrate
the extent of welding [2-25, 2-36, 2-52,] to ensure reproducibility. The amount
of intermetallic in the bond interface is implied by observing the disruption
(roughness) of the Au-ball bonded surface after etching the Al pad (from under
the ball) with a 20% solution of KOH or other etchant that does not attack gold.
In some cases, however, more sophisticated failure-analysis techniques are used
to examine, identify and analyze the actual intermetallics that forms. Although
these techniques give excellent visual evidence of the welded area, they are too
slow and expensive to be used to execute a designed experiment or even to set up
a bonder. The ball shear test is a quicker, more quantitative method to use for these
purposes, and its results have been correlated with the amount of welded area in
as-made bond interfaces [2-36, 2-52] to assure reproducibility.