Journal of ELECTRONIC MATERIALS, Vol. 33, No.

10, 2004 Special Issue Paper

Bromine- and Chlorine-Induced Degradation of

Gold-Aluminum Bonds
MIN-HSIEN LUE,1 CHEN-TOWN HUANG,1 SHENG-TZUNG HUANG,1 and
KER-CHANG HSIEH1,2

1.—National Sun Yat-sen University, Institute of Materials Science and Engineering, Kaohsiung,
804 Taiwan, Republic of China. 2.—E-mail: khsieh@mail.nsysu.edu.tw

The presence of bromine (Br) in flame retardant epoxies accelerates the degra-
dation of gold-aluminum wire bonds. This experiment tests degradation by
adding 5 wt.% 2,6-dibromophenol to the regular molding compound and hold-
ing thermal aging at 175°C and 205°C in ovens for up to 1,008 h. The inter-
metallic degrading microstructure was examined at different aging times. In
order to better understand the mechanism behind the degradation, bulk Au4Al
and Au5Al2 single-intermetallic phases were prepared and reacted with mold-
ing compound at 205°C. The reactions were analyzed by a JEOL (Japan Elec-
tron Optics Ltd., Tokyo) Superprobe JXA-8900R under wavelength-dispersive
spectrometry of x-ray mode. A similar chloride degrading study was performed
by adding Tetrachlorobisphenol A to the regular molding compound. The re-
sults show that Br attacked the Au4Al phase first and then the Au5Al2 phase.
The chloride reacted with the Au4Al phase only.

Key words: Halide degradation, wire bonding, reliability

accelerates bond failure. High moisture, too, has been

INTRODUCTION
Thermal reliability of integrated circuit (IC) de-
vices in plastic packaging is a major problem because
the molding resin used has low thermal stability, and
the temperatures are high under car hoods. With the
power consumption of these IC devices, their surface
temperature could rise to 150°C, making their relia-
bility even more important.
“Purple-plague,” “Kirkendall void,” and corrosion
have been suspected of being causes of bond failure.
Purple-plague, found by Philofsky, was not a cause,
but a result of bond failure.1 Kirkendall voids were
found by Uno to be caused by nonuniform inter-
metallic formation,2 and such void formation is ac-
cepted as one of the actual causes of bond failures.3
Corrosion is known to be a reaction between Au-Al
intermetallic phases and molding compounds.
Molding compounds have halogen elements, such
as bromine (Br) and chlorine (Cl), and Br is thought
by some to cause the reactions.4–8 They believe that,
under heat brominated flame retardant, Br is re-
leased primarily in the form of methyl bromide4 and
(Received February 27, 2004; accepted April 21, 2004)
reported by Gustafsson and Lindborg to bring about
ball bond degradation.9 They ascribed the bond fail-
ures to the high Cl content of the molding compounds
they were examining. However, the mechanisms of
the corrosion-induced bond failure have not been
clarified in detail, as the failure is thought to be
caused by a combination of intermetallic formation,
diffusion, mass transportation in resin, and reaction
around the wire bonds in molding resin.
This study aims to bring to light the role of such
factors in bond failure in an effort to improve bond
reliability. We studied the corrosion behaviors that
took place over time in gold wire bonds by adding
flame retardants with higher Br and Cl content as
the process of bond failure was activated under high
temperatures.
EXPERIMENTAL
The 4N type (Au
99.99%) 25-µm-diameter gold
wire was our sample wire. Test dies were attached to
copper-type lead frames. The bond-pad metallization
structure consisted of an
1-µm-thick aluminum
layer (Al-1% Si-0.5% Cu) and a thin titanium layer.
Samples were bonded on a K&S Model 1488 bonder

1111
1112
(Kulicke and Soffa Corp., PA), and the process
checked using the ball shear and ball size tests as well
as others using typical and relevant production-moni-
toring procedures. Bonded samples were encapsu-
lated in commercially available molding compounds
to which was added either 5 wt.% 2,6-dibromophenol
or Tetrachlorobisphenol A to accelerate the corrosion
effect. The samples were aged by heating at 175°C
and 205°C in an oven free of protective gas and exam-
ined from 1 h to 1,008 h.
Sample cross sections were mounted and ground
to reveal the ball bond. For metallographic inspec-
tion, samples were first ground with various grits of
SiC paper and then polished with 3-µm and 1-µm
grit diamond pastes. An electron probe microana-
lyzer (EPMA) was used to study alloy concentra-
tions in diffusion layers and intermetallic phases.
Tables in Figs. 4 and 6 show the weight and atomic
percentages found by EPMA. The weight percentage
measured with pure Au and Al as the standard
reference tested the accuracy of the measured data.
The atomic percentage normalized from the mea-
sured weight percentage was used to identify the
intermetallic phase.
The Au4Al and Au5Al2 single intermetallic phases,
prepared through vacuum melting, encapsulated
a b
d e
Fig. 1. The whole wire-bond microstructure of samples encapsulated with brominated compound aging at 175°C for various time periods: (a)
192 h, (b) 240 h, (3) 336 h, (4) 480 h, (5) 528 h, and (6) 1,008 h.
Lue, Huang, Huang, and Hsieh
with halide-added molding compound and heated
at 205°C, were used to confirm the corrosion effect
of the two phases and observe the corrosion mi-
crostructure.
RESULTS AND DISCUSSION
A series of elementary steps leading to the degra-
dation of the Au-Al junction in the presence of ionic
and organic halides was proposed in a previous
study.10 In that study, the degradation was started at
the interface between gold and the gold-aluminum
intermetallic phases.
Bromine
Figure 1, the whole microstructure, shows the
wire-bond microstructure attacked by Br at 175°C.
Bromine corrosion started on the interface of the Au
ball and intermetallic phase (Fig. 1a), and it gradu-
ally extended through all the intermetallic phase
layers (Fig. 1a–f). Figure 2, the edge region of the
microstructure, shows Br ion diffusion into the
wire bond from the peripheral gap around the
wire bond and the detailed microstructures in the
ball-bond edge regions. Bromine began reacting
with the intermetallic phases, forming a crack in
the upper part, which separated the Au and Au-Al
c
f
Bromine- and Chlorine-Induced Degradation of
Gold-Aluminum Bonds
a b
d e
Fig. 2. The edge region, wire-bond microstructure of samples encapsulated with brominated compound aging at 175°C for various time periods:
(a) 192 h, (b) 240 h, (3) 336 h, (4) 480 h, (5) 528 h, and (6) 1,008 h.
inter-metallic phases only after a long period of
aging time (Fig. 2f).
The Br-corrosion mechanism is described in Fig. 3.
The first stage involves the thermal dissolution of
Br ion; the second, the transportation of Br ion
through the peripheral gap around the ball bond; the
third, the corrosion reaction with Au4Al and Au5Al2
phases and decomposition of both intermetallic
phases to AlBr3 volatile species leaving Au rich
phase; and the fourth, the volatile AlBr3 species re-
action with oxygen and formation of a free Br ion and
continued reaction with the intermetallic phases, as
had happened during the third stage. The alternat-
ing reactions between the third and fourth stages
could build up the layer-type corrosion microstruc-
ture. This layer-type corrosion microstructure was
confirmed by using bulk Au4Al and Au5Al2 single-
phase samples.
Figure 4 shows the corrosion microstructure and
the results of EPMA. The Au-rich and porosity lay-
ers formed alternately as depicted in the corrosion
schematic diagram (Fig. 4). It was difficult to deter-
mine whether Au5Al2 was directly corroded by Br in
the wire bond samples or whether it had at first
phase transferred became Au4Al, which was cor-
roded by Br. Usually, because of the massive supply
1113
c
f
of Al from the outside bond-pad area, the Au5Al2
phase stabilizes in the edge region of the ball bond.
However, our study did not find any Au5Al2 phase
left in the edge region (Fig. 2f). On the other hand,
we found the Au5Al2 phase in the edge region in our
Cl corrosion tests (Fig. 5). Additional evidence that
the Au5Al2 phase can be directly attacked by bro-
mide is that, in our test with bulk Au5Al2 phase, Br
corrosion was found (Fig. 4b).
In a previous study, using transmission electron
microscopy, the species found at the intermetallic
region were oxides of aluminum, traces of bromi-
nated species, and precipitated gold crystallites.5 It
appears from these facts that gold-aluminum inter-
metallic phases, generically represented as Au4Al,
are particularly reactive with halogen ions and halo-
genated organic species, leading to the formation of
aluminum oxide and elemental gold as corrosion
products.
Based upon the foregoing considerations, the pro-
posed sequence of reactions in this study could be
the mechanism behind bond degradation. According
to one of our previous studies,10 wire bond degrad-
ing started with the formation of the Au4Al phase,
with a crack developing at the interface of Au and
the Au4Al phase. In the present similar study, Br or
1114
Fig. 3. The schematic Br-corrosion sequences diagram and possible reactions.
Cl reacted with Au4Al phase. Therefore, the Au4Al
phase formation may determine the rate that wire
bond degrades.
Chlorine
Figure 5 shows the Cl-corroded microstructure
aging at 205°C for 300 h and 500 h. The layer-type
corrosion of Cl, unlike what had happened with
Br, formed in the middle region only. We found
that the Au5Al2 phase did not attack by the Cl ion,
and the Au5Al2 phase can stand in the edge region
for a longer period mainly because there was an
Lue, Huang, Huang, and Hsieh
excess supply of aluminum from the immediate pe-
riph-eral region, as was proved by the bulk Au5Al2-
alloy corrosion test. That test, which was holding
the alloy at 205°C for 345 h, found that chloride was
not attacking or degrading the Au5Al2 alloy, and it
found that, much in the same way that happened
with the Br ion, chloride decomposed the Au4Al alloy
(Fig. 6).
In all reactions, Cl and Br ions have been assumed.
Based on what we have found, it seems very likely
that by eliminating low molecular weight organic
halides, the aluminum metallization bond pad and
Bromine- and Chlorine-Induced Degradation of
Gold-Aluminum Bonds 1115

a b

d
Fig. 4. The bulk alloy, Br-corrosion microstructures: (a) Au4Al alloy at 205°C for 120 h, (b) Au5Al2 alloy at 205°C for 120 h, (c) EPMA results for
the Au4Al alloy at 205°C for 480 h, and (d) the schematic description of the layer-type corrosion morphology.

gold-aluminum intermetallic phase can be protected
from halide-induced degradation. By reducing the
amount of such species, the level of ions generated
will be reduced, resulting in better thermal stability
of the resin. This improved stability would mean a
slower release of halide ion and, thus, less active
halide ions in the encapsulant.
CONCLUSIONS
• Degradation is caused by halogen-aluminum
corrosion reactions formed between gold and
Au-Al intermetallic phases.
• The liberation of Br from epoxy degrades Au4Al
and Au5Al2 phases and forms brittle layer-type
corrosion products.
• The Cl reacts with the Au4Al phase without
affecting the Au5Al2 phase.
ACKNOWLEDGEMENTS
The author Ker-Chang Hsieh thanks NSC for its
financial support (NSC 90-2216-E-110-009). The
authors also thank the Technical Development
Division of Philips at Taiwan for their preparation
of the samples.
1116 Lue, Huang, Huang, and Hsieh

a b

Fig. 5. The chloride-corrosion microstructure aging at 205°C for 300 h and 500 h, respectively.
Bromine- and Chlorine-Induced Degradation of
Gold-Aluminum Bonds 1117

b
Fig. 6. The bulk alloy, chloride-corrosion microstructures: (a) Au4Al alloy at 205°C for 345 h with EPMA results
and (b) Au5Al2 alloy at 205°C for 345 h without attacking by chloride.

REFERENCES 6. M. Khan, H. Fatemi, J. Romero, and E. Delenia, IEEE Int.

1. E. Philofsky, Proc. IEEE Reliability Physics Symp.
(New York: IEEE, 1970), pp. 177–185.
2. T. Uno, K. Tatsumi, and Y. Ohno, Proc. Joint ASME/JSME
Advanced Electronic Packaging (Englewood, CO: Rogers
Pub., 1992), pp. 771–777.
3. H.-S. Chang, K.-C. Hsieh, T. Martens, and A. Yang, IEEE
Trans. Components Packaging Technol. 27, 155 (2004).
4. T. Uno and K. Tatsumi, Microelectron. Reliab. 40, 145 (2000).
5. E. Imasato, M. Araki, I. Shimizu, and Y. Ohno,
IEEE/CPMT Electronic Packaging Technology Conf.
(Piscataway, NJ: IEEE, 1998), pp. 338–344.
Reliability Physics Symp. (New York: IEEE, 1988),
pp. 40–49.
7. S. Ahmad, R. Blish, T. Corbett, L. King, and G. Shirley,
Comp. Hybrid Manufacturing Technol. 9, 379 (1986).
8. A.A. Gallo, IEEE Int. Reliability Physics Symp. (New York:
IEEE, 1990), pp. 244-251.
9. K. Gustafsson and U. Lindborg (Paper presented at 37th
Proc. ECC, Boston, MA, May 11–13, 1987).
10. H.-S. Chang, J. P. Pon, K.-C. Hsieh, and C. C. Chen,
J. Electron. Mater. 30, 1171 (2001).