Content uploaded by Bernhard Czerny
Author content
All content in this area was uploaded by Bernhard Czerny on Nov 27, 2020
Content may be subject to copyright.
Contents lists available at ScienceDirect
Microelectronics Reliability
journal homepage: www.elsevier.com/locate/microrel
Interface characterization of CueCu ball bonds by a fast shear fatigue
method
B. Czerny
⁎
, G. Khatibi
Christian Doppler Laboratory for Lifetime and Reliability of Interfaces in Complex Multi-Material Electronics, Chemical Technologies and Analytics, TU Wien,
Getreidemarkt 9/CT-164, 1060 Vienna, Austria
ABSTRACT
A highly accelerated shear fatigue testing method is presented to test the long-term reliability and reveal the bonded interface of thermosonic CueCu ball bonds. The
method is an adaptation to a new industrial fatigue tester (BAMFIT) and can be conducted without an intricate specimen preparation. This method induces
mechanical cyclic shear stresses to the Cu nailhead in order to initiate fatigue fracture until lift-off, revealing the actual bonded interface. This study compares the
fatigue resistance of Cu wire bonded to coarse and fine grained Cu and Al metallization. The fatigue experiments are accompanied by nano indentation tests, shear
tests and finite element analysis. The fatigue results showed the best performance for Cu bonds on coarse grained Cu pads (metallization), followed by those bonded
on fine grained Cu while the CueAl nailheads failed at least a decade earlier than CueCu bonds. Annealing the specimens prior to testing resulted in slight increases
in the number of loading cycles to failure (N
f
) for Cu bonds as well as for CueAl bonds, while the scattering in N
f
for Cu bonds increased. Nevertheless the calculated
endurance limit of the fatigue data decreases with increasing annealing stages, due to a change in the fracture probability curve. With the ability to compare the
fatigue behaviour of the bonded interface within minutes, this method is most suitable for rapid qualification at an early stage of development.
1. Introduction
The semiconductor industry is constantly developing new wire
material combinations and innovative design concepts in order to im-
prove the performance of devices at higher operation power, elevated
temperatures and less electrical losses to increase their reliability. That
is also the case for ball bond connections in semiconductor modules
switching to Cu wires bonded onto Cu metallization [1]. But determi-
nation of the quality of these bond connections by means of standard
static ball shear/pull tests, does not draw any conclusions about the
cyclic loading capacity. At present their long-term reliability can only
be estimated by time-consuming tests such as temperature cycling or
high temperature storage tests [2]. Thus fast and reliable testing
methods for rapid evaluation of the bonding quality subjected to cyclic
loads are essential.
The only known fatigue investigation regarding the interface was
done by Lassnig et al. [3]. They used as well a highly accelerated me-
chanical testing setup that worked at 20 kHz and obtained lifetime
curves of CueAl nailheads with the failure mode being wire bond lift-
off. The method is based on the mechanical stresses which are induced
at the interface of two vibrating coupled parts. As described in detail in
[3], two different sample preparation methods consisting of single bond
and multiple bond testing procedures were developed to increase the
mass of the nail heads and induce sufficient cyclic stresses at the
bonding interface during the vibration. Applying these two methods
resulted in a slight difference between the loading modes of the nail-
heads which was analyzed by using FEM. This was an interesting study
where the fatigue resistance of thermosonic ball bonds was investigated
for the first time. The only drawback in this method is the intricate and
time consuming sample preparation.
In this study the interface reliability and bonding quality of Cu wire
bonds to Cu and Al pads was evaluated by using a highly accelerated
fatigue test (BAMFIT) [4–6,18] adapted for thermosonic ball bonds.
Shear tests and nano indentation tests were performed to compare the
hardness and static material behaviour to the obtained fatigue values.
The BAMFIT method applies a vibrational load via the resonating
tweezers to the nailhead in a certain direction and with a constant
displacement amplitude to facilitate a fatigue crack growth in the in-
terface. This mechanically induced load is not exactly the same as a
multiaxial thermo-mechanical load due to the CTE mismatch. This
difference and the influence of the testing frequency for the BAMFIT
tests were investigated in [5,18] for heavy Al wedge bonds and com-
pared to thermally induced fatigue failure by power cycling tests. In
order to put the results of this method into perspective and to compare
different nailhead sizes and materials, finite element analysis (FEA)
were conducted to calculate the stress and strain conditions in the in-
terface.
https://doi.org/10.1016/j.microrel.2020.113831
Received 31 May 2020; Received in revised form 10 July 2020; Accepted 17 July 2020
⁎
Corresponding author.
E-mail address: bernhard.czerny@tuwien.ac.at (B. Czerny).
Microelectronics Reliability 114 (2020) 113831
Available online 31 October 2020
0026-2714/ © 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
T
2. Specimen and testing methods
Cu ballbond connections from decapsulated devices, depicted in
Fig. 1, with two different types of Cu metallization and as a reference an
Al pad metallization were subject of this study. 4 N Cu wires with a
diameter of 65 μm were thermo-sonically bonded to a Cu metallization
of 20 μm thickness and 50 μm Cu wires to a 5 μm Al pad metallization.
Two kind of CueCu bonds with different electrochemically deposited
Cu metallization layers, one showing a fine grain (~3 μm) and the other
a coarse grain structure (~20 μm) were investigated. The former is
further mentioned as Cu-Cu
fine
, the latter as Cu-Cu
coarse
and the ball-
bonds on the Al pad metallization as CueAl. The etched cross section of
the initial state of the two CueCu specimens is shown in Fig. 2. Each
thermo-sonic bonding process was optimized to their type of metalli-
zation and wire. Those three types of bond connections were tested in
the initial state at room temperature and after annealing for 100 h and
200 h, exposed to 200 °C. CueAl bonds performance are well covered in
literature regarding the intermetallic phase formation [2,7], shear
performance [8,9], and recently also fatigue resistance [3,10]. The
same CueAl specimen as Lassnig et al. [3] used, only with an additional
shelf life of 5 years, were tested in this study as a direct reference.
Conventional shear tests and nano indentation tests were conducted
to evaluate the properties of the ball bond connections at conditions
stated above and compare them with the proposed BAMFIT test.
2.1. Shear test
The shear test is the most common test to validate a bond quality
and to optimize the bonding process parameters. It is a static evaluation
of the shear resistance of the interconnect, which can result in different
failure modes such as ball lift, pad lift, ball shear, and cratering. The test
procedure is described in detail in the ASTM, JEDEC standard [11,12]
for Au and Cu ball bonds.
In this study shear tests were performed in order to compare the
static performance of the ballbond interconnects with an established
qualification tests to the presented accelerated fatigue test. The shear
tests were performed on a BONDTEC 5600 at a shear height of 10 μm
and a velocity of 200 μm/s. The results are shown in Fig. 3 for each type
of connection and annealing stage. Cu-Cu
fine
and Cu-Cu
coarse
show si-
milar levels, which increases after annealing up to 200 h at 200 °C. The
measured shear force for CueAl with a peak value at 100 h at 200 °C
are comparable to the results given in [7], but showing an overall re-
duction of ~40 cN, which is still in the 3 sigma limit. This can be ex-
plained due to a five years shelf storage of the samples at ambient
condition. Shear test results of CueAl bond and the influence of inter-
facial intermetallic phase growth was not perused in this study, since
they are well covered literature [7–9].
Since the bond area varies for each type, due to the wire diameter,
material properties and bonding parameters, it is more appropriate to
consider the shear stress for a direct comparison. The shear stress is
calculated by the max. shear force divided by the total area of the bond
interface, which itself was measured by outlining the perimeter visible
in the fracture surface after BAMFIT testing for each type. The CueAl
bonds seem to have an equal shear resistance compared to Cu-Cu
fine
.
with respect to the size difference. Cu-Cu
coarse
shows a slightly smaller
interface area and hence enhanced shear resistance comparing with Cu-
Cu
fine
.
Shear fracture occurred in the Al pad for CueAl and for the CueCu
in the interface and through the Cu wire nailhead. With increasing
aging time, the area of the remaining Cu nailhead on the pad increases
and the shear tool cuts more through the wire material, as illustrated in
Fig. 4 for Cu-Cu
coarse
bonds. This improvement of the bond quality may
be caused by a stimulated interdiffusion due to the heat treatment and
possible local recrystallization, as reported for CueCu bonding in
[13,14].
2.2. Nano indentation
The material properties in the region of the nailhead differ greatly
from the wire, due to free air ball forming and thermosonic bonding.
Hence nano indentation tests were performed around the bond con-
nection in rows of ~10 measurement starting from the metallization
layer into the nailhead in the cross section, by an ASMEC Unat with a
Berkovich indenter. In Table 1 the average hardness values for the in-
vestigated Cu metallization and the nailheads are listed. These mea-
surements and the grain size structure were used to select the material
model for the FEM simulations of the fatigue test.
Fig. 1. Overview of the decapsulated samples (a) CueAl, (b) CueCu wire bonds.
50µm
50µm
(a)
(b)
Fig. 2. SEM images of the etched cross sections of (a) Cu-Cu
fine
and (b) Cu-
Cu
coarse
nailhead bonds in the initial state.
B. Czerny and G. Khatibi Microelectronics Reliability 114 (2020) 113831
2
2.3. BAMFIT
The BAMFIT method is designed for a commercial bondtester
(BONDTEC 5600) to reproduce the thermo-mechanical shear stresses by
mechanical means and invoke lift-off failure at the bond interface. This
method, which was originally developed for heavy Al wire wedge bonds
as described in [4], is further developed for evaluation of fine ther-
mosonic wire ballbonds in the current study, by adapting the resonance
tweezers gripping tool.
The mechanical stresses are induced by mechanically gripping and
exciting the bond wire near the bonded interface in one direction with
very small cyclic shear loads and simultaneously holding the substrate
static. Gripping the ball-bond is realized by a special tweezers fabri-
cated out of one piece of spring steel, which operates in resonance at
60 kHz. Due to the tweezers design the gripping force of < 1 N is al-
ways constant. A camera pattern recognition unit provides precise po-
sitioning with less than 5 μm offset error and a touch-down sensor keeps
a defined gripping height of 5 μm above the surface. The tip is ma-
chined in a way to maximize the pinching contact to the nailhead, as
illustrated in Fig. 5. The oval, micro scaled shape was cut into the tip
with a plasma focus ion beam. The grippers are excited at a constant
sinusoidal amplitude at 60 kHz during fatigue testing until a complete
wire bond lift-off occurs, while a small tensile preload is applied in Z-
direction in order to prevent re-bonding or grinding in the interface.
Each step of the described BAMFIT test process, the positioning and
opening of the tweezers (I), the touchdown (II), the retraction to the
clamping height and closing of the tweezers (III), the static preload and
ultrasonic excitation (IV) and the final lift off (V), is illustrated in Fig. 6.
The excitation amplitude was measured using a differential laser Dop-
pler vibrometer (LDV) at the nailhead and the metallization to
determine the displacement amplitude (dx).
The pinching of the ball bond and the ultrasonic softening effect
during testing causes a slight deformation of Cu nailhead within the
first few ms of the fatigue test. This deformation can be seen in Fig. 7a
and remains in most cases above the unbonded region. The few cases
(< 10%), where the nailhead deformed significantly during BAMFIT
testing, were excluded for the evaluation, since very high unrealistic N
f
and occasional slipping of the tweezers tool were observed.
The tests were conducted for nailhead bonds with cut wires at the
neck for a better clearance and visibility of the tweezers tool during
testing to the neighbouring bonds. Nevertheless with the designed
tweezers tip shape it is well possible to reach and test the nailheads
with intact wires (Fig. 5), in which case the testing direction has to be
aligned with the wire direction. Hence investigating the fatigue re-
sistance of the bonded interface of such ~50 μm Cu ballbonds is pos-
sible without special specimen design and can be performed directly
after bonding. This BAMFIT method is an extremely fast reliability test,
designed to reveal the interface structure and determine the fatigue
behaviour of ballbond connections.
Fig. 3. Shear test results plotted as (a) the maximum reached shear forces and (b) calculated shear stresses for all wire bonds and conditions.
(a) (b) (c)
100µm 100µm
100µm
Fig. 4. Shear fracture surface for Cu-Cu
coarse
in (a) the initial state, after annealing for (b) 100 h and (c) 200 h at 200 °C.
Table 1
Hardness investigation of Cu-Cu ballbonds and metallization.
H (GPa) Initial state 100 h @ 200 °C 200 h @ 200 °C
Fine metallization 1.45 ± 0.18 0.95 ± 0.06 0.94 ± 0.08
Coarse metallization 0.77 ± 0.03 0.93 ± 0.05 1.08 ± 0.07
Nailhead fine met. 0.95 ± 0.07 0.79 ± 0.22 1.03 ± 0.12
Nailhead coarse met. 0.96 ± 0.08 0.96 ± 0.09 0.96 ± 0.05
B. Czerny and G. Khatibi Microelectronics Reliability 114 (2020) 113831
3
3. BAMFIT results
The cyclic excitation of the nailhead leads to a fatigue crack pro-
pagating in the interface of the bond, which exposes the actual bonded
interface over the whole bonding area. This is in most cases not com-
pletely visible after shear testing due to the remnant of the sheared
nailhead, as in Fig. 4. The fracture surface of the completely lifted
nailheads can be seen in Fig. 7. For CueAl the fracture occurs in the Al
metallization layer as in Fig. 7b where a layer of Al remains on the lifted
nailhead.
The fracture surfaces for Cu-Cu
coarse
and Cu-Cu
fine
are shown in
comparison in Fig. 8. The magnified images of the interface of Cu-Cu
fine
show the actual bonded areas and some smooth areas where the me-
tallization grain structure in the backscattered electron image can be
distinguished (Fig. 8b). This indicates that the actual bonded interface
of the Cu-Cu is not bonded completely over the whole interface area.
120µm 184µm
Fig. 5. Tool tip geometry and design to reach the nailhead between the bond connections.
185µm
100µm
35µm
0-15µm
III III
IV V
Fig. 6. Schematic illustration of the gripping and testing process with the
BAMFIT method.
Fig. 7. Lifted Cu nailhead of a) Cu-Cu and b) CueAl thermo-sonic bond connection.
B. Czerny and G. Khatibi Microelectronics Reliability 114 (2020) 113831
4
For Cu-Cu
coarse
the interface in Fig. 8a seem to be bonded over a larger
area. The large grain structure of the metallization is also visible.
The BAMFIT fatigue test detects the cycles to complete lift off failure
(N
f
) for an applied excitation amplitude (dx). The tests were conducted
in the range of 1e3 up to 1e8 loading cycles. The results shown in Fig. 9
depicting dx in nm against N
f
for CueCu and CueAl for each annealing
stage separately. More than 600 nailhead bonds were tested for all 3
types of ballbonds at 4 stress levels, each with an average of 16 bonds
per level, providing enough data for a statistical analysis and calcu-
lating the percentile S-N curves. In this plot the size differences of the
wires and interconnects are not taken into consideration, but will be
covered in the next FEM section. The 50% fracture probability (P
f
) and
endurance limit from the fatigue data of each bond at each state was
calculated using the profatigue software [15]. The software uses a
Weibull model for calculating the endurance limit (C) for N → ∞ and
the percentile curves are hyperbolas with the asymptote Δx = exp(C).
The larger CueCu bonds can withstand much higher amplitude until
failure compared to the smaller CueAl bonds with data for Cu-Cu
coarse
being slightly above Cu-Cu
fine
despite the smaller interface area of Cu-
Cu
coarse
. The endurance limit for all bonds drops with increasing an-
nealing time. That is not due to a reduction in N
f
, but due to a de-
creasing curvature of the calculated fatigue curves. The endurance limit
drops about 1/3 of its value for CueAl from 55 to 35 nm. BAMFIT test
of the annealing stages actually shown a slight increase in N
f
. This can
be seen in more detail in the Weibull plots in Fig. 10, where P
f
is dis-
played against N
f
for one overlapping excitation amplitude of 90 nm.
The shift in N
f
for 50% P
f
can be seen for all bonds. For both CueCu
bonds the scattering of the fatigue results increases as well, as evident
due to a rotation of the fit curve.
Fig. 8. Fracture surface after BAMFIT testing on the metallization side for (a) Cu-Cu
coarse
and (b) Cu-Cu
fine
, indicating in red the perimeter of the overall bond area
used for shear stress calculations and FEM models. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this
article.)
B. Czerny and G. Khatibi Microelectronics Reliability 114 (2020) 113831
5
3.1. FEM analysis
Finite element analysis of the BAMFIT test was conducted to convert
the applied mechanical vibrational loads into shear stresses in the in-
terface. The dynamic conditions were simplified by consecutive static
simulations of the clamping, the preload and the excitation phases.
Nonlinear kinematic hardening models were used for the Cu nailhead,
Cu and Al metallization, obtained from previous investigations on
25 μm Cu wires with different microstructures [16,17,19]. The data
from tensile tests of fine and coarse grained polycrystalline wires were
chosen for the fine metallization and the nailhead. For the coarse
metallization, where the grains exceed the layer thickness, data from a
bamboo structured wire were selected. The mesh size at the contacts
and interface was refined to 5 μm.
The experimental test conditions were replicated by a displacement
movement of the tip of the tweezers in X-direction against the nailhead
on a Si chip metallization. As boundary conditions, frictionless contacts
between the nailhead and the tweezers tool and a fixed support at the
bottom of the silicon chip were defined. The movements of the tweezers
tip are indicated in Fig. 11. The clamping of the tweezers at 5 μm
distance to the chip metallization resulted in a local plastic deformation
of the nailhead, then the vertical preload of 5 g was applied and finally
a displacement in the range of 50–150 nm in y direction of the tweezers
tip induces shear stresses in the interface.
Since in the experiments the movement at the bottom of the Si chip
can't be restrained to absolute 0 as in the FEA, the differential dis-
placement (dx) was measured at the bottom of the nailhead neck and on
the chip metallization using a LDV with a laser spot size of ~3 μm, as
indicated in Fig. 11. The same positions were evaluated in the FEA to
match the excitation displacement amplitudes from the experiments, at
which point the stresses in the interface were calculated. In Fig. 12a the
von Mises stress distribution of the wire and in 12b the equivalent
plastic strain distribution of the interface for Cu-Cu
fine
at 90 nm ex-
citation (dx) is displayed. The interfacial shear stress distributions (XZ-
plane) for all wires at 90 nm excitation are displayed in Fig. 13,
showing the highest gradient from center to the edge for Cu-Cu
coarse
.
Evaluating the average shear stresses in the interface for each ex-
citation amplitude and wire, as is plotted Fig. 14, provides a linear
conversion of the displacement to the shear stress. By converting the
excitation amplitudes from Fig. 9 with this FEA method to shear stress
levels, the BAMFIT results can then be plotted as conventional SeN
curves, in Fig. 15. In the SeN plot the difference between Cu
fine
and
Cu
coarse
increases, since due to the smaller bonding interface of Cu-
Cu
coarse
the shear stress increases for the same excitation level. The
trend of the SeN curves and the endurance limit don't differ much from
the excitation plot (Fig. 9), but the larger difference between Cu-Cu
fine
and Cu-Cu
coarse
brings the endurance limit for Cu-Cu
coarse
above the
others for all conditions. More investigations at higher and lower stress
amplitudes would be necessary for a better prediction of the curve
trends and endurance limits at very high cycles. Although the CueAl
bonds had the smallest interface, due to the much softer Al the shear
Fig. 9. BAMFIT results plotted dx against N
f
for Cu-Cu
fine
, Cu-Cu
coarse
and
CueAl for (a) the initial state and after annealing for (b) 100 h and (c) 200 h at
200 °C.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08
Fracture probability P
f
Loading cycles to failure Nf(cycles)
Cu-Al ini�al state
Cu-Al 100h @200°C
Cu-Al 200h @200°C
Cu fine ini�al state
Cu fine 100h 200°C
Cu fine 200h 200°C
Cu coarse ini�al state
Cu coarse 100h 200°C
Cu coarse 200h 200°C
Fig. 10. Comparison of BAMFIT fatigue results at 90 nm excitation plotting the
fracture probability.
B. Czerny and G. Khatibi Microelectronics Reliability 114 (2020) 113831
6
10
2
10
3
10
4
10
5
10
6
10
7
10
8
10
9
30
40
50
60
70
80
90
100
110
120
130
140
Cu coarse
Cu fine
Cu-Al
Pf = 50%
Endurance limit
Displacement amplitude Δx [nm]
Inial state
(a)
(b)
(c)
10
2
10
3
10
4
10
5
10
6
10
7
10
8
10
9
30
40
50
60
70
80
90
100
110
120
130
140
Cu coarse
Cu fine
Cu-Al
Pf = 50%
Endurance limit
Loading cycles to failure Nf [cycles]
Displacement amplitude Δx [nm]
200h 200°C
10
2
10
3
10
4
10
5
10
6
10
7
10
8
10
9
30
40
50
60
70
80
90
100
110
120
130
140
Cu coarse
Cu fine
Cu-Al
Pf = 50%
Endurance limit
Displacement amplitude Δx [nm]
100h 200°C
stresses remain comparatively low between 20 and 24 MPa. The dif-
ferences in the lifetime curves between CueCu and CueAl can be re-
lated to the lower fatigue resistance of the softer Al metallization and an
effect of the intermetallic phase formation (IMC) in a CueAl bond
[2,7,8].
These SeN curve can be directly compared to the fatigue results of
Lassnig et al., since in this inertia based method they used the mass and
applied acceleration to calculate the shear stresses over the interface
area [3]. In that study the same CueAl ballbond specimen reached N
f
of
1e5-1e6 at 10–15 MPa for single ball test and 30–35 MPa for multiple
bond method. In comparison, the BAMFIT results for CueAl reach
5e3–1e6 N
f
at a calculated 20–24 MPa stress level, which are right
between the stress levels of the single and multiple test setups of Lassnig
et al. Despite the different approach to induce the mechanical stresses,
the testing frequency, the shelf life and the stress calculation, both re-
sults are in a good agreement.
differenal
LDV
stac
preload
displace-
ment
clamping
force
Fig. 11. (a) FEM model, (b) plastic deformation after tweezers gripping and preload of the nailhead, with indicated load directions and X displacement measurement
points.
Fig. 12. (a) Von Mises stress distribution of a Cu-Cu fine nailhead (bottom side), (b) equivalent plastic strain in the interface at 90 nm displacement.
Fig. 13. The XZ shear stress in the interface of a (a) CueCu fine (b) CueCu coarse (c) CueAl at 90 nm displacement.
B. Czerny and G. Khatibi Microelectronics Reliability 114 (2020) 113831
7
The calculated max von Mises stresses in the interface of the
BAMFIT test reaches ~107 MPa which due to the clamping method are
not as localized directly at the periphery of the bonded interface as in
the study of Lassnig, where peak values of 138 MPa were calculated.
4. Conclusion
Achieving results with the BAMFIT method for the long term fatigue
behaviour and revealing the actual bonding interface of Cu ball bonds is
unique for this type of interconnect. Due to its flexibility, no need for
intricate sample preparations and extremely fast test procedure, this
new method is most suitable for rapid screening and qualification at the
early stages of development and during production for lifetime eva-
luations and can reveal additional information to static shear or pull
tests.
As the results have shown, the shear test can deviate from the fa-
tigue behaviour. Comparing the average shear stress CueAl shows a
lower fatigue life despite reaching similar values as Cu-Cu
fine
in the
shear test. The fatigue resistance of the Cu-Cu
coarse
is higher than Cu-
Cu
fine
bonds and both Cu-Cu bonds show an increasing scattering after
annealing. Despite an increase in N
f
for all bond types for increasing
annealing time up to 100 h at 200 °C the curvature of the data de-
creased, resulting in a lower endurance limit.
The fracture surface for Cu-Cu
fine
revealed ribbon shaped areas in
bond direction of well bonded material as well as smooth surfaces,
which can't be seen by conducting static shear tests. This and the fact
that a much larger Cu-Cu
fine
bond compared to a Cu-Cu
coarse
still feature
a lower shear strength in the static shear test and a lower fatigue life in
the BMAFIT test, indicates that the actual bonded area of the Cu-Cu
fine
is not bonded uniformly over the whole interface.
The fatigue results, converted into a SeN plot by FEA, showed a
good correlation to the fatigue investigation of Lassnig et al. [3] under
consideration of the different method of inducing the shear stresses and
different shelf life. The FEM analysis provides a better comparison to
other loading conditions, since the BAMFIT test uses a unique way of
inducing the shear stresses.
The BAMFIT adaptation for ballbonds is a promising addition to the
conventional qualification tests, since it is a very fast testing method,
needs little specimen preparation, provides lifetime fatigue results for
lifetime evaluations and can reveal the actual bonding interface.
CRediT authorship contribution statement
B. Czerny: Conceptualization, Methodology, Data curation,
Investigation, Project administration, Software, Supervision,
Validation, Visualization, Writing - original draft, Writing - review &
editing. G. Khatibi: Conceptualization, Project administration, Funding
acquisition, Supervision, Writing - review & editing.
Fig. 14. Calculated average YZ shear stress of the interface against the differ-
ential displacement for all types of wire bonds.
Fig. 15. SeN curves for Cu fine Cu coarse and CueAl for the (a) initial state, (b)
after annealing for 100 h and (c) 200 h at 200 °C.
B. Czerny and G. Khatibi Microelectronics Reliability 114 (2020) 113831
8
10
2
10
3
10
4
10
5
10
6
10
7
10
8
10
9
Cu coarse
Cu fine
Cu-Al
P
f
= 50%
Endurance limit
Avg. shear stress Δτ
avg
[MPa]
100h 200°C
15
20
25
30
35
40
10
2
10
3
10
4
10
5
10
6
10
7
10
8
10
9
Cu coarse
Cu fine
Cu-Al
P
f
= 50%
Endurance limit
Loading cycles to failure N
f
[cycles]
Avg. shear stress Δτ
avg
[MPa]
200h 200°C
15
20
25
30
35
40
10
2
10
3
10
4
10
5
10
6
10
7
10
8
10
9
Cu coarse
Cu fine
Cu-Al
P
f
= 50%
Endurance limit
Avg. shear stress Δτ
avg
[MPa]
Inial state
15
20
25
30
35
40
(a)
(b)
(c)
Declaration of competing interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influ-
ence the work reported in this paper.
Acknowledgements
This work was conducted as part of the CD Laboratory RELAB. The
financial support by the Austrian Federal Ministry for Digital and
Economic Affairs and the National Foundation for Research,
Technology and Development is gratefully acknowledged. The authors
want to thank M. Nelhiebel from Infineon Technologies Austria for
providing test samples, M. Zareghomsheh for the nano indentation
tests, J. Pelzman and G. Akhlaghpoor for their assistance conducting the
experiments.
References
[1] G. Harman, Wire Bonding in Microelectronics, 3rd ed., McGraw-Hill, 2010.
[2] C.P. Liu, S.J. Chang, Y.F. Liu, W.S. Chen, Cu-Al interfacial formation and kinetic
growth behavior during HTS reliability test, J. Mater. Process. Technol. 267 (2019)
90–102, https://doi.org/10.1016/j.jmatprotec.2018.12.012 no. December 2018.
[3] A. Lassnig, M. Lederer, G. Khatibi, R. Pelzer, W. Robl, M. Nelhiebel, High cycle
fatigue testing of thermosonic ball bonds, Microelectron. Reliab. 71 (2017) 91–98,
https://doi.org/10.1016/j.microrel.2017.02.019.
[4] B. Czerny, G. Khatibi, Accelerated mechanical fatigue interconnect testing method
for electrical wire bonds, Tech. Mess. 85 (4) (2018) 213–220.
[5] B. Czerny, G. Khatibi, Interface reliability and lifetime prediction of heavy alu-
minum wire bonds, Microelectron. Reliab. 58 (2016), https://doi.org/10.1016/j.
microrel.2015.11.028.
[6] B. Czerny, G. Khatibi, Cyclic robustness of heavy wire bonds: Al, AlMg, Cu and
CucorAl, Microelectron. Reliab. 88–90 (July) (2018) 745–751, https://doi.org/10.
1016/j.microrel.2018.07.003.
[7] A. Lassnig, R. Pelzer, C. Gammer, G. Khatibi, Role of intermetallics on the me-
chanical fatigue behavior of Cu-Al ball bond interfaces, J. Alloys Compd. (2015),
https://doi.org/10.1016/j.jallcom.2015.05.282.
[8] R. Pelzer, M. Nelhiebel, R. Zink, S. Wöhlert, A. Lassnig, G. Khatibi, High tempera-
ture storage reliability investigation of the Al-Cu wire bond interface,
Microelectron. Reliab. 52 (9–10) (2012) 1966–1970, https://doi.org/10.1016/j.
microrel.2012.06.070.
[9] S. Manoharan, C. Patel, S. Hunter, P. McCluskey, Mechanism of wire bond shear
testing, Microelectron. Reliab. (2018), https://doi.org/10.1016/j.microrel.2018.06.
061.
[10] A. Lassnig, W. Trasischker, G. Khatibi, B. Weiss, M. Nelhiebel, R. Pelzer, Accelerated
lifetime estimation of thermosonic Cu ball bonds on Al metallization,
Microelectron. Eng. 106 (2013) 188–194, https://doi.org/10.1016/j.mee.2013.02.
015.
[11] EIA/JESD22-B116, Wire Bond Shear Test Method, EIA/JEDEC Standard, (1998).
[12] ASTM, F1269–89 Test Methods for Destructive Shear Testing of Ball Bonds, (2001).
[13] E.J. Jang, J.W. Kim, B. Kim, T. Matthias, Y.B. Park, Annealing temperature effect on
the Cu-Cu bonding energy for 3D-IC integration, Met. Mater. Int. 17 (1) (2011)
105–109, https://doi.org/10.1007/s12540-011-0214-0.
[14] M. Martinez, et al., Mechanisms of copper direct bonding observed by in-situ and
quantitative transmission electron microscopy, Thin Solid Films 530 (2013) 96–99,
https://doi.org/10.1016/j.tsf.2012.02.056.
[15] A. Fernández-Canteli, C. Przybilla, M. Nogal, M.L. Aenlle, E. Castillo, Profatigue: a
software program for probabilistic assessment of experimental fatigue data sets,
Procedia Eng. 74 (2014) 236–241, https://doi.org/10.1016/j.proeng.2014.06.255.
[16] A. Mazloum-Nejadari, G. Khatibi, B. Czerny, M. Lederer, J. Nicolics, L. Weiss,
Reliability of Cu wire bonds in microelectronic packages, Microelectron. Reliab. 74
(2017) 147–154, https://doi.org/10.1016/j.microrel.2017.04.014.
[17] B. Czerny, M. Lederer, B. Nagl, A. Trnka, G. Khatibi, M. Thoben, Thermo-me-
chanical analysis of bonding wires in IGBT modules under operating conditions,
Microelectron. Reliab. 52 (9–10) (2012) 2353–2357, https://doi.org/10.1016/j.
microrel.2012.06.081.
[18] V.N. Popok, S. Buhrkal-Donau, B. Czerny, G. Khatibi, H. Luo, F. Iannuzzo,
K.B. Pedersen, Comparative study of wire bond degradation under power and
mechanical accelerated tests, J. Mater. Sci. Mater. Electron. 30 (2019)
17040–17045, https://doi.org/10.1007/s10854-019-02050-0.
[19] B. Czerny, A. Mazloum-Nejadari, G. Khatibi, L. Weiss, M. Zehetbauer, Fatigue
testing method for fine bond wires in an LQFP package, Microelectron. Reliab. 64
(2016) 270–275, https://doi.org/10.1016/j.microrel.2016.07.068.
B. Czerny and G. Khatibi Microelectronics Reliability 114 (2020) 113831
9
