JOURNAL OF MATERIALS SCIENCE LETTERS 22, 2003, 1373 – 1376
Crystallographic texture of drawn gold bonding wires
using electron backscattered diffraction (EBSD)
F. WULFF, C. D. BREACH, K. DITTMER
Kulicke and Soffa (S.E.A.) Pte. Ltd., #04-05 TECHplace II, Block 5002, Ang Mo Kio, Avenue 5, Singapore 569871
E-mail: fwulff@kns.com; cbreach@kns.com
Gold ball bonding wire is used as an interconnection be-
tween die pads and substrates in over 90% of worldwide
semiconductor assemblies. Typical wire thicknesses in
consumer applications are 20–25 µm with an increas-
ing trend towards the use of 12–15 µm wires forecast
for the near future. The end of the wire is melted into
a ball using an electric arc on a wirebonding machine
and the ball is then thermosonically welded to the die
pad metallization (which is usually 0.7–1.0 µm thick Al
with Cu and/or Si minor additions). The thermosonic
welding is performed at frequencies from 60–140 kHz
and at temperatures of 150–250
◦
C. The melting of the
end of the wire results in a temperature gradient along
the wire length that causes a variation in properties over
a region called the heat-affected zone (HAZ). The re-
gion after the HAZ has the normal properties of the
wire. After the ball has been bonded, the wire is then
formed into a controlled-profile loop that is thermoson-
ically welded to the substrate, thus forming an electrical
connection between die and substrate. A significant fac-
tor in loop control is the HAZ length that determines
where the wire bends and characterization of the HAZ
is an essential part of bonding wire development.
Gold bonding wire is produced by a wire drawing
process that develops a preferred 111 texture in the
drawing direction. In wire form, gold, being anisotropic
therefore tends to be stronger in the drawing direction
[1]. In addition to the mechanically induced texture of
gold bonding wire, the wire has additions of various
elements to a maximum of 100 ppm by weight (4N or
99.99% purity gold) to control strength, modulus and
recrystallization behavior.
Understanding how the processing of gold bonding
wires affects texture can enable better microstructural
engineering of the wire properties for performance im-
provements. Specific concerns are the stiffness of the
wire that affects wire swaying during molding, me-
chanical property effects on long loops and low loops,
chemistry effects on the HAZ length, to name only a
few of the wire properties that can affect the assembly
yield of semiconductor devices. While many reports
have used macrotexture measurements such as X-ray
techniques to measure the average texture of bundled
wire [2], semiconductor assemblies use single wires
and therefore bonding wire development demands the
characterisation of single wire strands. Electron back-
scattered diffraction (EBSD) is a technique that allows
rapid characterization of overall and localized texture
of single wire strands. In this letter, the use of EBSD
is demonstrated in the characterization of gold bond-
ing wires at various stages of processing, from casting
through wire drawing. EBSD experiments were carried
out using an Oxford Instruments Link OPAL EBSD sys-
tem on a LEO Stereoscan S440 SEM. Cross-sectioned
samples were mounted on a special fixture to ensure
the 71
◦
tilt to the electron beam. A set of experiments
was designed to explore the crystallographic texture of
the gold at several stages of the manufacturing process.
Each material was investigated at the following stages:
(a) As cast
(b) At intermediate drawn diameter before annealing
(c) At intermediate drawn diameter after annealing
(d) Finished drawn wire (25 µm diameter) after an-
nealing
(e) Free air ball on 25 µm wire
The gold was cast into 12 mm diameter rods by vertical
continuous casting. Two 4N alloys were manufactured
by this route, A1 and A2. The cast rods were drawn
to a finished wire size of 25 µm. Samples were metal-
lographically sectioned in longitudinal and transverse
directions followed by chemical etching to remove the
final deformation layer and reveal the microstructure.
Fig. 1 shows cross-sections of the two different rods to-
gether with the inverse pole figures in the axial direction
determined by EBSD.
The microstructure of A1 in Fig. 1a shows large elon-
gated grains at the center and smaller grains at the edges
of the cast rod with an average shape orientation along
the major axis of the grain. The microstructure of A2 in
Fig. 1b shows a different structure consisting of longer
grains oriented more along the rod axis, with very lit-
tle difference in size between the edge and center. The
appearance of the cross-sections suggests a significant
difference in the crystallographic texture of the alloys.
Although the grains of the A1 rod appear randomly
distributed, the inverse pole figure, Fig. 1a shows that
in the center where the long grains are observed, ap-
proximately 50% of the rod consists of 100 texture.
The texture of A2 in Fig. 1b is much stronger and shows
an exclusive 100 texture in agreement with the highly
oriented appearance of the grains. Both alloys were cast
under the same conditions, and the differences in texture
show the strong effect of chemistry on the microstruc-
ture of the cast rod.
The general effects of drawing and annealing are il-
lustrated by Fig. 2, where the microstructure of A1 is
0261–8028
C
2003 Kluwer Academic Publishers
1373