Page 171 - Wire Bonding in Microelectronics

148 Cha pte r F i v e

5.2 Impurity-Accelerated Au-Al Bond Failures
The previous section (Sec. 5.1) described metallurgical diffusion,
intermetallic compound formation, and Kirkendall voiding in pure
bulk welds, as well as wire bonds made to uncontaminated pads.
These can result in Au-Al weld failures, but usually only when sub-
jected to very high temperatures for long times. Horsting [5-28] was
the first to discover that voiding-type wire-bond failures can be accel-
erated by impurities. He found that a number of impurities (e.g., Ni,
Fe, Co, B) in Au-plated films may result in rapid Kirkendall-like-voiding
and Al wire-bond failures. His model proposed that during a high-
temperature bake (for an impurity-free Au-Al bond interface), the
intermetallic diffusion front moves through the Au plating down to
the Ni underplating, and the bond remains strong. For impure Au,
the impurities became concentrated ahead of the intermetallic growth.
At some concentration, precipitation of these impurities occurs. These
particles then act as sinks for vacancies produced by the diffusion
reaction, resulting in Kirkendall-like voids and leading to weak or
zero-strength bonds. He introduced a thermal-stress test (390°C for 1 h
followed by a pull test) as a pragmatic means of detecting Au films
containing impurities. Horsting’s failure model was derived from
plated films and is treated more completely in Chap. 6A, see Fig. 6A-1.
Comparisons of his and other thermal-stress tests for bond reliability
are given in Table 4-4 in Chap. 4.
After Horsting’s work showed that contaminants can accelerate
bond failure, a number of other contaminants in Au-plated films, as
well as from plastics, ambient atmospheres, etc., have been shown to
degrade bond reliability.

5.2.1 The Effect of Halogens on the Au-Al Bond System
Halogens are pervasive and are well known to corrode Al metalliza-
tion in integrated circuits [5-29] (see App. 5B). However, the first
observation that halogen compounds could degrade the strength of
previously made Au bonds on Al metallization was by Thomas [5-30].
He cured various epoxies in the caps of TO-18 headers and sealed
them to the package base that contained wire-bonded devices. Groups
of these sealed packages were then stored at 150, 180, and 200°C for up
to 1000 h. Massive wire-bond failures occurred within 24 h at 200°C in
devices with epoxies containing brominated flame retardants (tetra-
bromobispheonol-A). The Au-Al bonds failed after developing a weak
lamellar microstructure as shown in Fig. 5-11. This structure is not
characteristic of normal, intermetallic growth, but is more characteris-
tic of a single-phase alloy that has grown unstable and separated.
Apparently, the outgassing products from the epoxy attacked the
intermetallic compound, diffusing in from the sides or other areas
where the compound was exposed. No corrosion of the Al-bond pad