Page 88 - Wire Bonding in Microelectronics

Bonding W ir e Metallur gy and Characteristics 67

compared to the above Cornell study, which used more precise equip-
ment. Considering all of these uncertainties discussed previously,
wires do fail during thermal cycles, whether they are in open cavity
or plastic packages. Therefore, this problem must be considered when
designing interconnection systems subject to significant temperature
variations, and the designer may be forced to make fatigue life esti-
mates based on limited data. Clearly more work is needed to better
characterize the relative mechanical properties of bonding wires by
using similar equipment [3-20]. The practical solution to wire fatigue
in open cavity packages is to increase the loop-height to bond-length
ratio. This minimizes the amount of flexing in a given ∆T situation
(see Sec. 8.2), and as such affects both Au and Al wires. Other envi-
ronmental factors such as humidity have been shown to reduce the
fatigue life of Al wires [3-17]. However, since most chips/wires are
plastic encapsulated, where the strain must be estimated and can
vary around a chip/package, the only practical method is to use tem-
perature cycling to establish a realistic life. This has become standard
in all packaging development.

3.9 Copper Wire for Ball Bonding
Copper wire has been studied for over 20 years, but only recently has
been incorporated in actual volume production (see Chap. 5). It is
very different from Au wire. It oxidizes and thus requires an inert
atmosphere during “EFO” ball formation, it is harder and more crater
prone etc. However, the renewed interest has been driven primarily
by the high cost of Au, Cu’s lower resistivity (for carrying more cur-
rent), as well as its resistance to wire sweep in plastic encapsulation.
Its intermetallic reliability and increased hardness on Al pads have
been well established and are discussed in Chap. 5 with appropriate
references. In addition, thermosonic bonded Cu balls reveal a much
higher hardness of 111 HVN than that of initial Cu balls (84 HVN).
This can and does lead to cratering (see Chap. 8) [3-21]. Most Cu ball
bond production is at 50 µm (2 mil) diameter and larger wire sizes
used for small power devices. This author is not aware of high-volume
production at 25 µm sizes or less. However, with Au approaching
$1000 per troy ounce (2008), it surely will be when metallurgical prob-
lems, such as neck and stitch-fatigue in plastic encapsulation, have
been solved. All bonding wire manufacturers make Cu wire, and
autobonders designed for it are readily available. There are no ASTM
or international standards for it, as yet, but many of those listed below
for Au wire should be helpful.
Several comments from the packaging industry indicate that
problems have occurred in plastic encapsulated Cu wire less than
50 µm (generally <40 µm) diameter. Cu is sensitive to cold working
or recrystallization. It has been observed that a higher incidence of
heel cracks as well as ball neck cracks result from temperature cycling