Sie sind auf Seite 1von 14

Mechanics of Materials 38 (2006) 1124

www.elsevier.com/locate/mechmat

Numerical analysis of ultrasonic wire bonding: Eects


of bonding parameters on contact pressure
and frictional energy
Yong Ding, Jang-Kyo Kim *, Pin Tong
Department of Mechanical Engineering, Hong Kong University of Science and Technology, Clear Water Bay,
Kowloon, Hong Kong
Received in revised form 27 April 2004

Abstract
The elasto-plastic large deformation taking place in ultrasonic wire bonding is analysed by means of 2-D and 3-D
nite element method. A special focus has been placed on how the important wire bonding parameters, such as bond
force and power, aect the contact pressure along the wire-bond pad interface. It is shown that the contact interface had
a long elliptical shape, and the maximum contact pressure occurred always at the periphery of the contact interface,
which is consistent in the current 2-D and 3-D nite element analyses. The normalised real contact area as well as
the maximum frictional energy intensity varied in a similar manner to the contact pressure, with the maximum values
occurring at the periphery of contact interface, where weld is preferentially formed in practical wire bonding. A higher
bond force does not result in a higher contact pressure, or higher frictional energy intensity, suggesting that a high bond
force is not directly correlated to better wire bondability.
 2005 Elsevier Ltd. All rights reserved.
Keywords: Ultrasonic wire bonding; Wire bondability; Finite element method; Contact pressure; Frictional energy intensity

1. Introduction
Wire bonding is by far the most popular rstlevel interconnection technology used between
*

Corresponding author. Tel.: +852 23587207; fax: +852


23581543l.
E-mail address: mejkkim@ust.hk (J.-K. Kim).

the die and package terminals. Amongst various


wire bonding processes, the ultrasonic wire bonding (UWB) has the advantages of fast bonding
process, high productivity, excellent electrical
performance, good heat conductivity and good
corrosion resistance (Tummala, 2001). It has been
widely used in various electronic packages, such as
chip on board (COB), chip scale package (CSP)

0167-6636/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mechmat.2005.05.007

12

Y. Ding et al. / Mechanics of Materials 38 (2006) 1124

and ball grid array (BGA) packages. The UWB is


carried out by pressing the metal wire on a bond
pad of semiconductor and vibrating it using high
frequency ultrasound, generating frictional energy.
The energy applied onto the bond pad breaks
down the surface oxide lm, allowing the formation of intermetallics between the wire and the
bond pad (Tummala, 2001; Harman, 1997). Advances in wire bonding technology require higher
reliability, bonding speed and product yield (Harman, 1997; Winchell and Berg, 1978; Hamidi et al.,
1999). However, the details of bonding mechanisms taking place in UWB remain ill-understood
and most our understanding is qualitative and/or
experimental.
The wire bond strength is one of the main criteria used to ensure the reliability of UWB. The
bond strength is dependent on many process and
material variables, such as ultrasonic power, applied force, welding time, bond pad surface hardness and roughness, interface temperature (Kim
and Au, 2001; Chan et al., in press). The bond
strength can be improved by increasing the ultrasonic power, but too high a power can lead to premature failure at the wire neck (Sheaer and
Levine, 1991). Ultrasonic power is a dominant factor for bond formation (Sheaer and Levine, 1991;
Hu et al., 1991) and a proper range of bond force
is required for ecient transmission of ultrasonic
power. The inuences of hardness and surface
roughness of bond pad have been extensively studied (Jeng and Hong, 2001; Chan et al., 2004; Jeng
et al., 2001). Although a high surface roughness
can reduce the bonding time for successful bonds
at a given bond force and power (Jeng and Hong,
2001), too high a surface roughness did not further
improve the wire bondability (Chan et al., 2004;
Jeng et al., 2001). The bond pad temperature also
played an important role in determining the wire
pull strength of metallization on an organic substrate (Hu et al., 1991; Jeng et al., 2001). An optimal preheat temperature produced the maximum
bond strength, and too high a temperature was
detrimental to wire bonding (Chan et al., 2004).
For such a complex problem with many parameters aecting the wire bond quality, to identify
key mechanisms is essential to understanding the
bonding mechanisms. Fortunately, the lift-o fail-

ure of bond area provides a valuable insight: it is


shown (Harman, 1997; Winchell and Berg, 1978)
that a strong bond was formed preferentially at
the periphery of the contact interface, whereas
the central region remained largely unbonded.
This phenomenon is not unique for the ultrasonic
bonding and is also observed in the thermocompression wire bonding (Harman and Albers,
1977), suggesting a non-uniform application of
energies across the contact interface. An improved
understanding is needed as to what mechanical
conditions in terms of stress, contact pressure
and frictional energy at the wire-bond pad interface produce such a non-uniform bond.
The analytical solutions for the displacement
and stress distribution in the wire and bond pad
during wire bonding are very complicated due to
the complexity in bonding geometry and material
properties. Recognizing the signicance of stress
distribution along the wire-bond pad interface,
increasing eorts are directed towards employing
numerical methods. The use of the nite element
method (FEM) in particular allows a more accurate description of wire deformation. In addition,
the specic loading geometries as well as varying
mechanical properties of bond pads can be properly taken into account. The stress distribution at
the wire-bond pad interface has been studied (Ikeda et al., 1999) for the ball bonding based on the
axisymmetric 2-D FEM (Ikeda et al., 1999; Takahashi and Inoue, 2002). The maximum equivalent
stresses were determined for the periphery of the
interface assuming a rigid bond pad in the contact
analysis (Ikeda et al., 1999). Few studies have hitherto been reported based on the 3-D FEM analysis
that can reect accurately the global deformation
and stress distribution of wire bonding.
Although the above FEM analyses were able to
calculate the stress distributions at the wire-bond
pad interface, the changes in bond pad surface
characteristics and ultrasonic energy cannot be
incorporated in the analysis. Experiments (Harman and Albers, 1977; Joshi, 1971; Harman and
Leedy, 1972) suggest that the UWB is a frictional
bonding process. A micro-contact approach was
adopted previously (Jeng and Hong, 2001; Jeng
et al., 2001) to take into account the surface roughness of bond pad. Real contact area and frictional

Y. Ding et al. / Mechanics of Materials 38 (2006) 1124

20,352 brick elements and 24,551 nodes. This ensures sucient resolution and thus accuracy of
the results while maintaining a reasonable time
needed for computation. The model consisted of
a wedge tool, wire, bond pad and substrate. The
wedge tool was assumed to be a rigid body because
it is usually made of titanium carbide, which is
much harder than the Au wire. The Au wire had
a cylindrical shape of radius 25.4 lm. The bond
pad had a layered structure consisting of Au, Ni
and Cu layers and was electrolytically plated on
the FR-4 substrate that was made from glass woven fabric reinforced epoxy laminates. The dimensions and material properties of these components
are presented in Fig. 2 and Table 1 (Callister, 2003;
ASM International Handbook Committee, 1990).
The wire bonding process is carried out at 25 C
in this study. When the Au wire was being pressed
by the wedge tool toward the bond pad, there exists two contact interfaces, one between the wire
and bond pad, the other between the wire and
wedge. A typical bond forcetime history of ultrasonic wire bonding is shown in Fig. 3 (Harman,
1997; Ikeda et al., 1999), where the applied force
increases linearly in the rst few ms (Step 1), followed by a constant force for the rest of wire
bonding (Step 2). The rst step takes place only
for a very short period of time, say about 1 ms
(Ikeda et al., 1999), whereas the second step takes
up the most time of wire bonding. During the rst
step, the wire undergoes large elasto-plastic deformation under the applied force and the interface is

energy were calculated to interpret the experimental results of wire bondability. However, the nonuniform stress distribution which in turn forms
the bond along the periphery, have not been specically studied. The non-uniform stress distributions at the interface may cause non-uniform
contact pressure, uneven real contact area and frictional energy intensity, all of which aect the
bondability.
With the above ndings in mind, the objectives
of this paper are (i) to develop a 3-D and a precise
2-D models to simulate the ultrasonic wire bonding, and (ii) to investigate the bonding mechanisms
in terms of contact pressure, real contact area and
frictional energy intensity generating at the wirebond pad interface. Elasto-plastic large deformation contact analyses are performed based on the
FEM.

2. Finite element analysis (FEA)


Both the 3-D and 2-D models of ultrasonic
wedge bonding are considered in the present FE
analysis. The geometry, the loading method and
the boundary conditions were selected to represent
those of the actual experimental technique, as
shown in Fig. 1. A mesh was created for the symmetric loading geometry using the FE code ABAQUS, as shown in Fig. 2. Due to the symmetry,
only one quarter of the structure was adopted
for the 3-D model which was composed of

Load

bond wedge

l = 51um
R

13

R = 15um

Wire
Pad
Substrate

Fig. 1. Schematic of wedge wire bonding.

14

Y. Ding et al. / Mechanics of Materials 38 (2006) 1124

Fig. 2. 3-D FEM model for wire bonding.

Table 1
Material properties used in the present analyses (Ikeda et al., 1999; Callister, 2003; ASM International Handbook Committee, 1990)
Material

Youngs modulus (GPa)

Poissons ratio

Rate of strain hardening (MPa)

Initial yield stress (MPa)

Au
Ni
Cu

68.6
207
115

0.44
0.31
0.308

Eq. (1)

32.7

Bond Force
(Newton)
Ultrasound application
Fb
Bond Time
(msec)
0
Step 1

20
Step 2

Fig. 3. Bond force-time history during ultrasonic wire bonding.

formed between the wire and bond pad. During


the second step, a bond is formed between the wire
and bond pad due to the applied ultrasonic vibration which in turn generates frictional energy. The

large elasto-plastic deformation contact analysis


was performed for the second step using the
FEM to simulate wire deformation and stress distribution at the contact interface. The real contact
area was also calculated based on the micro-contact approach. The frictional energy consumed at
the interface during the second step was adopted
as the criterion for bondability.
One of the most important material properties
required for proper simulations of wire bonding
is the plastic property of Au wire. Judging from
the maximum bond force of 0.2 N and the loading
time of 1 ms for Step 1, the loading speed approximates to 0.2 N/ms. The extremely fast wire bonding process may result in the deformation of Au
wire at a very high strain rate, with a corresponding increase in yield strength of Au wire due to the
strain rate eect. According the Hopkinson pres-

Y. Ding et al. / Mechanics of Materials 38 (2006) 1124

15

Table 2
Material properties of FR-4 substrate (Yao and Qu, 1999)
E1 (GPa)

E2 (GPa)

E3 (GPa)

G12 (GPa)

G13 (GPa)

G23 (GPa)

m12

m32

m13

22.4

1.6

22.4

0.2

0.63

0.2

0.14

0.14

0.002

sure bar test (Ikeda et al., 1999), the yield stress,


ry, of Au wire is expressed as:

32:7 0:057_e MPa for Step 1;
ry
1
32:7 1459e MPa for Step 2.
The rst equation takes into account the strain
rate eect upon linear increase in bond force (Step
1), whereas the second equation considers the plastic strain hardening eect during constant loading
(Step 2). Ni and Cu were assumed linearly elastic
in this study because they are much harder than
Au. The FR-4 substrate was also assumed linearly
elastic because it has a large supporting area and is
far away from the contact interface. Its elastic
properties in the in-plane and out-of-plane directions are shown in Table 2 (Yao and Qu, 1999).
To compare the accuracy of stress distribution
obtained from the 3-D model as well as to accurately calculate the real contact area, a 2-D analysis was also performed using a rened mesh
especially for the wire and bond pad. Due to the
symmetry, only one half the structure was modelled with a total of 23,559 quadrilateral elements
and 23,965 nodes, as shown in Fig. 4. The layered

Fig. 4. 2-D FEM model of wire bonding.

structure of the bond pad and their material properties were the same as those used in the 3-D model. A plane strain condition was assumed for the
cross-section of the structure.
3. Results and discussion
3.1. 3-D model
In the 3-D analysis, the bond force was increased linearly until it reached a plateau value,
Fb, which remained constant until the end of wire
bonding (Fig. 3). Fig. 5 shows the deformation
and von Mises stress distributions for bond forces
Fb = 0.1 and 0.2 N. At the low bond force of
0.1 N, the wire has just started to deform with high
stress concentrations taking place at the periphery
of the contact interface between the wedge and
wire, as well as between the wire and bond pad.
The minimum stress occurred in the central region
of contact interface, and there were hardly stresses
built up in the substrate underneath the bond pad.
When the bond force was increased to 0.2 N, the
stress concentrations occurred at the same contact
regions although the magnitudes of stress distribution generally increased.
The stress distributions are further studied in
more details with reference to Fig. 6, which plots
the contact pressure distribution over the whole
contact interface between the wire and bond pad
in dierent view directions. Fig. 6(a) presents a
3-D perspective view, whereas Fig. 6(b)(d) are
the 2-D views seen in the z, x and y axes, respectively. The wire-bond pad contact interface had a
long elliptical shape, with the contact width in
the lateral direction being larger in the central area
than at the wire ends (Fig. 6(b)). It is clearly seen
that the maximum contact pressure occurred at
the periphery of the contact interface, which is
consistent with the previous analysis as far as the
bond pad is suciently thick (Takahashi and

16

Y. Ding et al. / Mechanics of Materials 38 (2006) 1124

Fig. 5. Deformation and Mises stress (in 106 MPa) distributions at bond forces of (a) 0.1 N and (b) 0.2 N.

Inoue, 2002). There is striking analogy between the


region corresponding to maximum contact pressure and the real bonded area identied by the
microscopic analysis (Harman, 1997; Winchell
and Berg, 1978). Fig. 7 shows the SEM photographs taken of a typical wedge bond fracture
surface. Much of the central area was unbonded,
which is consistent with the generally low contact
pressure seen in Fig. 6(a) and (b). It is also noted
that the contact pressure in the central region of
contact interface (near y = 0) was even higher than

that at the wire edges (y = 4050 m), see Fig. 6(c).


The large contact pressure concentrated in the
mid-region of 25 lm < y < 25 lm along the wire
direction, which is a reection of the initial contact
between the wire and wedge over the wedge length
of l = 51 lm (see Fig. 1).
3.2. 2-D model
The eects of bond force on bonded interface
area and stress distributions were further studied

Y. Ding et al. / Mechanics of Materials 38 (2006) 1124

17

Fig. 6. Contact pressure distributions at a bond force of 0.2 N: (a) 3-D view; (b) top view; (c) view from the lateral direction and (d)
view from the wire longitudinal direction.

based on the 2-D model. Fig. 8 shows the wire


deformation and the contact pressure distributions
along the contact length between the wire and
bond pad at varying bond force, Fb. It is obvious
that the maximum contact pressure occurred always at the periphery of the contact interface, consistent with the results from the 3-D analysis. The
absolute magnitudes of the maximum contact
pressure were also almost identical between the
2-D and 3-D models, justifying the applicability
of 2-D analysis in liu of the more time-consuming
3-D analysis.
Fig. 9 shows the changes in contact length, bn,
when the bond force, Fb, varied from 0 to 0.2 N,
indicating that the nominal contact length is line-

arly proportional to bond force. Fig. 10 shows


the maximum and average contact pressures at
the contact interface taken from Fig. 8. It is found
that the maximum contact pressure seen at the
periphery increased initially with increasing the
bond force, followed by a constant plateau value
for bond forces above 0.07 N, whereas the mean
contact pressure remained almost a constant over
the whole bond force increment. There were slight
uctuations in the maximum and average contact
pressure values due to marginal changes in contact
area for these bond forces. Judging from the fact
that the real bonded area is conned to the periphery of the contact interface (Harman, 1997; Winchell and Berg, 1978) and the majority of central

18

Y. Ding et al. / Mechanics of Materials 38 (2006) 1124

Fig. 7. SEM photographs of wedge wire bonds at dierent bond time: after (a) 0 ms, (b) 4 ms, (c) 7 ms and (d) 10 ms. After Harman
(1997).

contact region is not bonded, the maximum contact pressure at the perimeter is more important
than the average value. According to the observation that there were negligible changes in maximum contact pressure at bond forces above
0.07 N, a moderately high bond force was already
sucient to make wire bonds, and too high a bond
force is not necessarily benecial.
To verify the accuracy of 2-D model, the pressure distribution obtained from the 2-D FEM
analysis was compared with that obtained from
the 3-D FEM model. Two cross-sections at y = 0
and y = 17 lm were chosen in the 3-D model for
this purpose, representing the axis of symmetry
and the region with moderately high contact pressure, both within the main contact interface,
respectively. The contact pressures shown in
Fig. 11 indicates that the results obtained from
the 2-D and 3-D FEM analyses were almost identical in terms of both pressure distribution and

maximum value. This conrms the validity of the


2-D analysis used in this study, which can provide
the advantage of a drastically reduced computation time.
3.3. Real contact area
Although it was assumed in the present analysis
that the bond pad surface is perfectly at, it, in
reality, is rough on a microscopic scale. The surface roughness causes the real contact area during
wire bonding to become much less than the nominal contact area. The micro-contact phenomenon
cannot be simulated properly in the FEM model
because of the microscale and random distribution
of the roughness. Thus, a micro-contact model
needs to be introduced here. Since the early development of a basic elastic model (Greenwood and
Willamson, 1966), there have been many studies
on contact of rough surfaces (Abbott and Fire-

Y. Ding et al. / Mechanics of Materials 38 (2006) 1124

19

Fig. 8. Deformations in the wire and the corresponding contact pressure between the wire and bond pad at dierent bond forces.

14
12

bn (um)

10
8
6
4
2
0
0.00

0.05

0.10

0.15

0.20

Bond Force (N)


Fig. 9. Variation of contact length, bn, as a function of bond
force.

stone, 1933; Pullen and Williamson, 1972; Chang


et al., 1987; Zhao et al., 2000; Tabor, 1959; Johnson, 1968; Lim and Ashby, 1987; Ashby et al.,
1991). The rough-surface contact was characterized by full plastic deformation of asperities (Abbott and Firestone, 1933; Pullen and Williamson,
1972), whereas some (Chang et al., 1987; Zhao
et al., 2000) developed elasto-plastic contact models. The elasto-plastic models were adopted later in
the study of wire bonding (Jeng and Hong, 2001;
Jeng et al., 2001). Although the above contact
models considered surface proles on a microscopic scale, the eect of shear stress was not

20

Y. Ding et al. / Mechanics of Materials 38 (2006) 1124

determined from experiment (Tabor, 1959; Johnson, 1968). Eq. (2) was further modied (Lim
and Ashby, 1987; Ashby et al., 1991) to explain
the wear mechanisms assuming that the real contact area for plastic contact is given by

Contact Pressure (MPa)

250
200
Max Contact Pressure
Average Contact Pressure

150

Ar
F
;
An F s

100
50
0
0.00

0.05

0.10

0.15

0.20

Bond Force (N)


Fig. 10. Variations of maximum and average contact pressures
as a function of bond force.

where F is the normal contact force. When F > Fs,


Ar becomes identical to An. In 2-D model, An and
Ar are represented by the lengths of bn and br,
respectively. Considering the non-uniform pressure distribution between the wire and bond pad,
Eq. (2) can be reduced:
ps

Contact Pressure (MPa)

250
2-D model
3-D model at the cross-section of y=0
3-D model at the cross-section of y=17um

200

1 at l2 1=2

mx; y

100

50

0
4

x (um)

10

12

14

Fig. 11. Comparisons of contact pressure between 2-D and 3-D


FEM analyses.

specically taken into account. In ultrasonic or


thermosonic wire bonding, the contact surfaces
slide each other, generating signicant shear stresses at the asperities of the contact surfaces. The seizure load, Fs), developed at the surface asperities
at the real contact area is expressed as (Tabor,
1959):
Fs

H0

H 0 An
1 at l2

1=2

where ps is the seizure pressure. Here, m(x, y) is


dened as the ratio of the real contact area to
the nominal contact area:

150

An is the nominal contact area, which was assumed


to be identical to the real contact area, Ar in their
analysis (Tabor, 1959); H0 is the hardness of the
softer material of the two contact surfaces; and l
is the coecient of friction. at = 12 is a constant

DAr px; y

;
ps
DAn

where DAr and DAn are the micro real and micro
nominal contact areas at the point (x, y); and
p(x, y) is the corresponding contact pressure.
When p(x, y) > ps, m(x, y) becomes unity. For a
2-D analysis, m(x, y) can be simplied to m(x).
When H0 = 245 MPa (ASM International Handbook Committee, 1990) for the Au wire, l = 0.38
(Mayer et al., 2002), and the corresponding seizure
pressure, ps = 170 MPa.
The normalised real contact area, m(x, y) presented in Fig. 12 varied along the wire lateral
direction in a similar manner to the contact pressure shown in Fig. 8, except the very edge where
the normalised contact area consistently showed
a plateau constant for a length of 0.50.8 lm for
all bond forces studied. The plateau constant value
represents the most intimate real contact where
weld is most likely to be made. The maximum value of the normalised real contact area was almost
identical for all bond forces studied, which is consistent with the summary drawn from Fig. 10 in
that a high bond force does not necessarily mean
a large area of intimate contact. Fig. 13 summarises the real contact length, br, which increased
linearly with increasing the bond force.

Y. Ding et al. / Mechanics of Materials 38 (2006) 1124


1.0

1.0

0.8

0.8

Bond Force = 0.05 N

Bond Force = 0.10 N

m(x)

m(x)

0.6
0.4
0.2

0.6
0.4
0.2

0.0
0

10

12

0.0

14

x (um)
(a)

10

12

14

12

14

x (um)
(b)
1.0

1.0

Bond Force = 0.15 N

0.8

Bond Force = 0.20 N

0.8

m(x)

0.6

m(x)

21

0.4

0.6
0.4
0.2

0.2

0.0

0.0
0

10

12

14

10

x (um)
(d)

x (um)
(c)

Fig. 12. Normalised real contact area, m(x), along the wire lateral direction for dierent bond forces.

14

periphery of contact interface and the vast majority of central region remains unwelded, see Fig. 7
(Harman, 1997; Winchell and Berg, 1978). This
observation indicates that the local frictional energy intensity plays an important role in wire
bondabilty. Considering the non-uniform pressure
distribution at the contact interface, the frictional
energy intensity, Eif, in 3-D model is dened:

12

br (um)

10
8
6
4
2
0
0.00

Eif ulpx; y;
0.05

0.10

0.15

0.20

Bond Force (N)


Fig. 13. Variation of real contact length, br, as a function of
bond force.

3.4. Frictional energy intensity


A key factor essential to successful bonds in
ultrasonic wire bonding is the sucient frictional
energy between the wire and bond pad to form a
solid state weld through local melting. It was conrmed that bond is made preferentially along the

where u is the average speed due to the reciprocating motion in ultrasonic wire bonding. The average speed can be dened as u = 4bf, where b and
f refer the amplitude and frequency of vibration,
respectively. p(x, y) is the contact pressure that
can be obtained from the FEM analysis. For a
2-D model, p(x, y) can be simplied to p(x).
The friction energy intensities were calculated
for an ultrasonic frequency of 60 kHz and a vibration amplitude of 1 lm (Harman, 1997), which are
plotted for dierent bond forces as shown in
Fig. 14. The maximum frictional energy intensity

Y. Ding et al. / Mechanics of Materials 38 (2006) 1124


4 x105
3 x105

Friction energy intensity (J/m2)

Friction energy intensity (J/m2)

22

Bond Force = 0.05 N

2 x105
1 x105
0

6
8
x (um)

10

12

4.0 x10

Bond Force = 0.10 N

3.0 x10

2.0 x10

1.0 x10

0.0

14

4 x10

Bond Force = 0.15 N

3 x10

2 x10

1 x10

6
8
x (um)

10

12

14

(b)
Friction energy intensity (J/m2)

Friction energy intensity (J/m2)

(a)

6
8
x (um)

10

12

4 x10

Bond Force = 0.20 N

3 x10

2 x10

1 x10

14

6
8
x (um)

10

12

14

(d)

(c)

Fig. 14. Variations of frictional energy intensity along the wire lateral direction for dierent bond forces.

An

Fig. 15 clearly shows that the total frictional energy increases linearly with bond force. It was
shown that while the total frictional energy increased consistently with bond force, the high frictional energy intensity obtained at the periphery of
contact interface did not show a similar increase.
Thus, it can be said that a high bond force is not
necessarily benecial for wire bondabilty. Instead,
too high a bond force may impair the eciency of
ultrasonic energy transfer and may cause cratering
failures (Harman, 1997). Besides the bond force,

2.5

Total friction energy (J/m)

always occurred at the perimeter of the contact


interface regardless of bond force, consistent with
the pressure distribution in Fig. 8. The frictional
energy intensity at the perimeter was much higher
than that at the central area.
The total frictional energy per unit length was
calculated by integrating the friction energy intensity of Eq. (6) over the contact interface:
Z
Ef
4fblpx; ydA.
7

2.0

1.5

1.0

0.5

0.0
0.00

0.05

0.10

0.15

0.20

Bond Force (N)


Fig. 15. Total frictional energy generated at the contact
interface as a function of bond force.

the eects of other wire bonding parameters can


also be envisaged from Eq. (7). A long welding
time, a high frictional coecient and a high vibrational speed may also increase the frictional energy
intensity, which in turn increase the bondability.

Y. Ding et al. / Mechanics of Materials 38 (2006) 1124

The vibrational speed is the product of ultrasonic


frequency and amplitude, u = 4bf, and a high
vibrational speed normally requires a high ultrasonic power (Ramsey et al., 1997). However, to increase the ultrasonic frequency while maintaining
the constant ultrasonic power does not necessarily
enhance the vibration speed.

23

entially made as shown by many experimental evidence. The total frictional energy
increased linearly with bond force, but the
high frictional energy intensity obtained at
the periphery of the contact interface did
not show a similar increase.

4. Conclusions

Acknowledgements

The deformation and stress distributions in the


wire and bond pad during the ultrasonic wire
bonding are analysed using the 2-D and 3-D
FEM analyses. The following can be highlighted
from the numerical study.

This paper is presented at the International


Symposium on Macro-, Meso-, Micro- and
Nano-Mechanics of Materials (MM2003), which
is dedicated to Professor Pin Tong on the occasion
of his 65th birthday. This work has been supported by the Research Grant Council (RGC) of
Hong Kong and the postdoctoral matching fund
of HKUST. The rst author (YD) was a visiting
scholar when this work was performed.

(1) The wire-bond pad contact interface had a


long elliptical shape, with the contact width
in the lateral direction being larger in the
central area than at the wire ends. The maximum contact pressure between the wire and
bond pad occurred at the perimeter of the
contact interface, which is consistent with
the previous analysis. The absolute magnitudes of the maximum contact pressure
obtained from the 2-D and 3-D analyses
agreed well, conrming the validity of the
present 2-D model. A higher bond force does
not mean a higher contact pressure, suggesting that a high bond force is not necessarily
benecial to high wire bondability.
(2) The normalised real contact area calculated
based on the 2-D analysis varied along the
wire lateral direction in a similar manner to
the contact pressure. The plateau maximum
value occurred always at the periphery over
a length of 0.50.8 lm for all bond forces
studied, which represents the most intimate
real contact area.
(3) The frictional energy is thought to be one of
the most important factors determining successful bonding. The non-uniform pressure
distribution at the wire-bond pad contact
interface resulted in a non-uniform frictional
energy intensity. The maximum frictional
energy intensity occurred at the periphery
of the contact interface, where weld is prefer-

References
Abbott, E.J., Firestone, F.A., 1933. Specifying surface
qualitya method based on accurate measurement and
comparison. Mech. Eng. 55, 569.
Ashby, M.F., Abulawi, J., Kong, H.S., 1991. Temperature
maps for frictional heating in dry sliding. Tribol. Trans. 34,
577587.
ASM International Handbook Committee, 1990. ASM Handbook. ASM International, Materials Park, OH.
Callister Jr., W.D., 2003. Material Science and Engineering: An
Introduction. John Wiley & Sons, NY.
Chan, Y.H., Kim, J.K.D.M., Liu, P.C.K., Liu, Y.M., Cheung,
M.W.Ng, 2004. Process windows for low temperature Au
wire bonding. J. Electron. Mater. 33, 146155.
Chan, Y.H., Kim, J.K., Liu D.M., Liu, P.C.K., Cheung, Y.M.,
Ng, M.W., in press. Improvement of wire bond process
window via plasma treatment. IEEE Trans. CPMT Part A:
Advanced Packaging.
Chang, W.R., Etsion, I., Bogy, D.B., 1987. An elasticplastic
model for the contact of rough surfaces. J. Tribol. 109, 263.
Greenwood, J.A., Willamson, J.B.P., 1966. Contact of nominally at surface. Proc. R. Soc. London, Ser. A 295, 300
319.
Hamidi, A., Beck, N., Thomas, K., Herr, E., 1999. Reliability
and lifetime evaluation of dierent wire bonding technologies for high power IGBT modules. Microelectron. Reliab.
39, 11531158.
Harman, G.G., 1997. Wire Bonding in Microelectronics.
McGraw-Hill, NY.

24

Y. Ding et al. / Mechanics of Materials 38 (2006) 1124

Harman, G.G., Albers, J., 1977. The ultrasonic welding


mechanism as applied to aluminium- and gold-wire bonding
in microelectronics. IEEE Trans. PHP 13, 406412.
Harman, G.G., Leedy, K.O., 1972. An experimental model of
the microelectronics ultrasonic bonding mechanisms. In:
10th Annu. Proc. Reliab. Phys Symp., pp. 4956.
Hu, S.J., Lim, G.E., Foong, K.P., 1991. Study of temperature
parameter on the thermosonic gold wire bonding of highspeed CMOS. IEEE Trans. Compon. Hybrids, Manuf.
Technol. 14, 855858.
Ikeda, T., Miyazaki, N., Kudo, K., Arita, K., Yakiyama, H.,
1999. Failure estimation of semiconductor chip during wire
bonding process. J. Electron. Packag. 121, 8591.
Jeng, Y.R., Hong, J.H., 2001. A microcontact approach for
ultrasonic bonding in microelectronics. J. Tribol. 123, 3515
3521.
Jeng, Y.R., Aoh, J.H., Wang, C.M., 2001. Thermosonic wire
bonding of gold wire onto copper pad using the saturated
interfacial phenomena. J. Phys. D: Appl. Phys. 34, 35153521.
Johnson, K.L., 1968. Deformation of a plastic wedge by a rigid
at die under the action of a tangential force. J. Mech. Phys.
Solids 16, 395402.
Joshi, K.C., 1971. The formation of ultrasonic wire bonding
between metals. Weld. J. 50, 840858.
Kim, J.K., Au, B.P.L., 2001. Eects of metallisation surface
characteristics on gold wire bondability of organic printed
circuit boards. J. Electron. Mater. 30, 10011011.
Lim, S.C., Ashby, M.F., 1987. Wear mechanism maps. Acta
Metal. 35, 124.

Mayer, M., Schwizer, J. 2002. Ultrasonic bonding: understanding how process parameters determine the strength of Au
Al bonds. In: Proc. Intl. Symp. Microelectron., IMAPS,
Denver, CO, USA, pp. 626631.
Pullen, J., Williamson, J.B.P., 1972. On the plastic contact of
rough surfaces. Proc. Roy. Soc. Lond. 327, 159173.
Ramsey, Thomas, H., Alfaro, Cesar, 1997. High-frequency
enhancement for ambient temperature ball bonding. Semicond. Intern. 20, 9396.
Sheaer, M., Levine, L., 1991. How to optimize and control the
wire bonding process: Part II. Solid State Technol. 34, 54
62.
Tabor, D., 1959. Junction growth in metallic friction: the role of
combined stresses and surface contamination. Proc. Roy.
Soc. 251, 378393.
Takahashi, Y., Inoue, M., 2002. Numerical study of wire
bondinganalysis of interfacial deformation between wire
and pad. J. Electron. Packag. 124, 2736.
Tummala, R.R., 2001. Fundamental of Microsystems Packaging. McGraw-Hill, NY.
Winchell, V.H., Berg, H.M., 1978. Enhancing ultrasonic bond
development. IEEE Trans. CHMT 1, 211219.
Yao, Q., Qu, J., 1999. Three-dimensional versus two-dimensional nite element modelling of ip-chip packages. J.
Electron. Packag. 121, 196201.
Zhao, Y.W., Maietta, D.M., Chang, L., 2000. An asperity
microcontact model incorporation the transition from
elastic deformation to fully plastic ow. J. Tribol. 122, 86
93.

Das könnte Ihnen auch gefallen