effect of direction of ultrasonic vibration on flip-chip … · nation problems when making the...
TRANSCRIPT
38
Transactions of The Japan Institute of Electronics Packaging Vol. 6, No. 1, 2013
1. IntroductionThe development of packaging technologies such as
system-in-a-package (SiP) has greatly contributed to mak-
ing electronic devices smaller and more multifunctional.
Smaller, high-density packages require micro-interconnec-
tions between the die and substrate. Wire bonding and
flip-chip bonding are the main techniques used to make
these interconnections. Flip-chip bonding is a process in
which bumps are fabricated on a die and interconnected to
the package substrate, with the active die face down. Flip-
chip bonding has many advantages, such as high preci-
sion, high density, short interconnections, and small para-
sitic elements. However, the reliability of the
interconnections is crucial for flip-chip bonding.[1] Low-k
materials are used for insulating layers for large-scale inte-
gration, in order to reduce the signal delay; however,
because low-k materials are mechanically fragile, the bond-
ing force must be reduced. Ultrasonic vibration is com-
monly applied to the bonding head to decrease the force.
It has been reported that the structure and power of the
ultrasonic vibration horn affects the bonding state.[2]
However, failure analysis of flip-chip bonding is not easy,
because it is difficult to observe the connected area under
the face-down condition. Arai et al.[3] have reported on a
method to evaluate the bonding status using a die-pull tes-
ter, which can be used to evaluate whether the bonding
conditions are suitable for the production environment.[4]
It is known that the directions of the leads on the substrate
also affect the bonding state,[5] i.e., the bonding behavior
of longitudinal and lateral leads is different under the hori-
zontal vibration condition.
In this paper, the bonding states are investigated using
test element group (TEG) dies with gold stud-bumps.
Ultrasonic bonding is applied to substrates with longitudi-
nal and lateral leads, and the die-pull mode is investigated
systematically. The die is physically removed from the sub-
strate, and the failure mode is analyzed. The bonding
behavior with a vibrational head rotated by 45 degrees
with respect to the die configuration is compared with that
of a conventional head, and the process margin of flip-chip
bonding for both heads is discussed along with the failure
modes.
2. Experimental ProcedureThe conventional flip-chip bonding equipment has a
fixed vibration direction with respect to the die configura-
tion, as shown in Fig. 1(a). The flip-chip bonder we devel-
oped has an ultrasonic vibration head that can be rotated
in advance and operated with constant weight and fre-
quency, as shown in Fig. 1(b).
[Technical Paper]
Effect of Direction of Ultrasonic Vibration on Flip-Chip BondingMutsumi Masumoto*, Yoshiyuki Arai*,**, and Hajime Tomokage*
*Department of Electronics Engineering and Computer Science, Fukuoka University, 8-19-1 Nanakuma, Jonan-ku, Fukuoka 814-0180, Japan
**Research and Development Div., Toray Engineering Co., Ltd., 1-1-45 Oe, Otsu, Shiga 520-2141, Japan
(Received May 14, 2013; accepted October 29, 2013)
Abstract
Flip-chip bonding has several advantages, such as high precision, high density, short interconnections, and small parasitic
elements. However, creating reliable interconnections between chips and substrates is the key issue in flip-chip bonding.
In this work, bonding states are investigated using test element group (TEG) dies with gold stud bumps. Ultrasonic
bonding is applied to a substrate with longitudinal and lateral leads, and the die-pull mode is investigated systematically.
For conventional flip-chip bonding equipment, the die-pull test shows different bonding states for longitudinal and lateral
leads. However, we have developed a flip-chip bonder with a rotational vibration head, where the direction of the angle
of the vibration with respect to the die configuration can be changed. For a head rotated to 45 degrees, uniform bonding
is established on both the longitudinal and the lateral leads. A wide process margin for flip-chip bonding is obtained, with
a high yield.
Keywords: Flip-chip bonding, Die-pull test, Bump shear strength, Ultrasonic vibration
Copyright © The Japan Institute of Electronics Packaging
39
Masumoto et al.: Effect of Direction of Ultrasonic Vibration on Flip-Chip Bonding (2/5)
The test element group (TEG) dies used in this mea-
surement were 8.22 × 8.22 mm with 500 gold stud-bumps
per die. The pad size and pad pitch were 45 μm and 50 μm,
respectively. The substrates to be connected were FR-4
type with copper leads plated with nickel and gold. The
thickness of the nickel and gold were 0.08 ± 0.04 μm and
0.50 ± 0.25 μm, respectively. The lateral and longitudinal
lead patterns were placed equally on the substrates. The
conditions for flip-chip bonding were as follows: trigger
force 12 N (or 2.5 g per bump), bonding force 34.5 N,
bonding time 0.7 s, and amplitude of ultrasonic vibration
1.3 μm.
The reflow test was performed at 260°C, three times.
The direct-current resistance of a daisy-chain connection
between a die and a substrate was measured, and any sam-
ple which increased in resistance by more than 20% after a
reflow test was classified as not good (NG).
The die-pull test was performed using the mode count-
ing system.[3] The die was physically removed from the
substrate in the vertical direction. The pull speed was 1
mm/s, and the maximum force was 300 N. Then the pull-
off images were observed, and classified into four modes.
The different failure modes are shown schematically in
Fig. 2, along with the pull-off marks observed on the pad
and substrate sides for each mode. Mode A indicates
weaker bonding between the chip pad and stud-bump com-
pared with that between the stud-bump and lead. Mode B
is fracture of the bump, and implies strong bonding, while
mode C indicates incomplete bonding between bump and
lead. Finally, mode D is fracture between the lead and sub-
strate material. This also implies strong bonding, but
sometimes mode D also occurs because of copper delami-
nation problems when making the substrates.
The shear test was performed with a conventional shear
tester. To measure the shear strength of the bonds
between bumps and leads, the aluminum pads on a TEG
Fig. 1 Ultrasonic bonding head and die configuration: (a) conventional flip-chip equip-ment and (b) equipment with head rotated with respect to the die.
(a) (b)
Ultrasonic vibration
Fig. 2 Die-pull mode and typical photographs at pad side and substrate side: (a) mode A, (b) mode B, (c) mode C and (d) mode D.
(a) (b) (c) (d)
Pad-side
Substrate-side
(a)
TEG
(b) (c) (d)
Substrate
40
Transactions of The Japan Institute of Electronics Packaging Vol. 6, No. 1, 2013
die were chemically etched off, as shown in Fig. 3. The
etching was performed by immersing the die into a KOH
23% solution at 50°C for 40 min. We ascertained that this
chemical treatment did not affect the shear strength of the
stud-bump.
3. Results and DiscussionThe typical failures observed after the reflow tests are
shown in Fig. 4. From the resistance measurements, an
individual connection was determined as “NG”, and then
cross-sectioning was performed. The cracks usually
occurred between the die pad and bump, and between the
bump and lead. In order to obtain the distribution of bond-
failure modes on a wafer, the die-pull test was performed
with conventional flip-chip equipment. Figure 5 shows the
distribution of the bond-failure modes obtained for five
dies. The results for the longitudinal lead and the lateral
lead are shown in Fig. 5(a) and (b), respectively. Mode B
is dominant for the longitudinal lead (the proportions of
modes A, B, C and D were 2%, 96%, 0% and 2%, respec-
tively). On the other hand, the lateral lead connection usu-
ally failed in either mode A or B (the proportions of modes
A, B, C and D were 46 %, 53%, 0% and 1%, respectively), as
shown in Fig. 5(b). Mode B indicates strong bonding
between a bump and a lead. However, for the lateral lead
under this condition, there is a significant proportion of
mode A failures, meaning that excess vibrational energy
was applied to the pad during the lateral-direction
bonding.[3] Mode A is supposed to correspond to the
presence of a crack between the pad and bump after the
reflow test.
Figure 6 shows the distribution of bump shear strengths
obtained from the shear tests on the dies bonded using
conventional flip-chip bonding equipment. The mean value
and standard deviation are tabulated in Table 1. The mean
shear strength and variation in shear strength are smaller
for the longitudinal leads than for the lateral leads. The
small variation in shear strength for the longitudinal lead is
consistent with the dominance of die-pull Mode B in Fig.
5(a). For the lateral leads, on the other hand, almost half of
failures correspond to Mode A, which accounts for the
larger value of the mean bump strength, and also the
larger variation in bump strength. Although Mode A cor-
responds to a crack between a pad and a bump, the result
of Fig. 6 was obtained with the bumps etched off from the
substrate. That might be the reason why the shear
strength for the lateral leads was larger than for the longi-
tudinal ones.
We consider that the ultrasonic vibration first induces a
low bonding force; the bonding strength then reaches its
maximum value, followed by fracture, as the vibration con-
tinues. Figure 7 shows the flip-chip bonding process as a
function of bonding time. When the bonding time is short,
lead open failures occur, because of the low bonding force.
Fig. 3 Process for carrying out a bump shear test: (a) after bonding, (b) alu-minum pad etched off chemically and (c) shear test.
Substrate
TEG chip
BumpLead
)c()b()a(
Substrate
TEG chip
BumpLead
)c()b()a(
Fig. 4 Typical defects observed after reflow test: (a) crack between a pad and a bump, (b) crack between a bump and a lead.
)b()a(
20 mµ
Table 1 Shear test for conventional and 45-degree-rotated bonders.
Conventional 45 degree rotated
Longitudinal Lateral Longitudinal Lateral
Sample number
36 36 36 36
Mean value (g)
6.86 10.4 8.24 8.48
Standard deviation (g)
1.17 2.12 0.89 0.96
(a) (b)
41
Masumoto et al.: Effect of Direction of Ultrasonic Vibration on Flip-Chip Bonding (4/5)
On the other hand, excess bonding energy causes cracks
when the bonding time is too long. For the lateral leads,
the ultrasonic energy can be absorbed more easily than for
the longitudinal leads. Therefore, the bonding states
obtained for the longitudinal and lateral leads can be
explained as shown in Fig. 7. The bonding energy, which
is the product of vibration power and bonding time, is dif-
ferent between the longitudinal and the lateral leads,
which can cause a time-lag in forming firm bonds, result-
ing in a reduced process margin for effective ultrasonic
bonding.
To change the bonding energy applied by the ultrasonic
vibration, the bonding head was rotated with respect to the
die configuration. Figure 8 shows the results for die-pull
tests and Fig. 9 shows the bump shear strength distribu-
tion for a head rotated 45 degrees. The mean value and
Fig. 8 Proportions of mode A, B, C and D failures for 45-degree-rotaed flip-chip bonding: (a) longitudinal lead, (b) lateral lead.
Fig. 6 Distribution of bump shear strengths.
10
20
30
40
50
60
0 5 10 15 20 25
Inci
denc
era
tio (%
)
Bump shear strength (g)
Lateral
Longitudinal
Fig. 7 Diagram showing bonding states and times for longi-tudinal and lateral leads bonded using a conventional flip-chip bonder. Note the small process margin.
Lead open
Lead open
CrackGood
CrackGood
Longitudinal lead
Lateral lead
Bonding time (s)
Process margin
0.5 1.0 1.5 1.05
Fig. 5 Proportions of mode A, B, C and D failures for conventional flip-chip bonding: (a) longitudinal lead, (b) lateral lead.
Mode A Mode B Mode C Mode DMode A Mode B Mode C Mode D
Die pull mode
Inci
denc
e ra
tio(%
)
Inci
denc
e ra
tio(%
)
Die pull mode
(a) (b)
Mode A Mode B Mode C Mode DMode A Mode B Mode C Mode D
Die pull mode
Inci
denc
e ra
tio(%
)
Inci
denc
e ra
tio(%
)
Die pull mode
(a) (b)
Die pull mode Die pull mode
(a) (b)
Inci
denc
e ra
tio(%
)
Inci
denc
e ra
tio(%
)
100
80
60
40
20
0
100
80
60
40
20
0
Die pull mode Die pull mode
(a) (b)
Inci
denc
e ra
tio(%
)
Inci
denc
e ra
tio(%
)
100
80
60
40
20
0
100
80
60
40
20
0
(a) (b)
42
Transactions of The Japan Institute of Electronics Packaging Vol. 6, No. 1, 2013
standard deviation are tabulated in Table 1. The curves for
the longitudinal and lateral leads have the same mean
strength and variation, and the overall variation is small
compared with that obtained using conventional flip-chip
equipment. SEM images of the longitudinal and lateral
leads after the die-pull test are shown in Fig. 10. Mode B
pull-off patterns, with similar structures, were formed on
both the leads.
Figure 11 shows a schematic of the bonding status ver-
sus bonding time for a 45-degree head. The ultrasonic
energy was delivered to both longitudinal and lateral leads,
resulting in a much wider process margin for effective
bonding.
Using the 45-degree-rotated head improved the yield for
flip-chip bonding. For example, in one product, the yield
on 3,000 dies increased from 85.1% to 99.4%. Low-k Black
DiamondTM material was used for the insulating layer.
After open and function tests, dies bonded using the
45-degree-rotated head also showed high reliability in tests
such as preprocessing humidity resistance level 3, heat
cycle test 55/125°C, high temperature and humidity pres-
ervation test 110°C/85%, and high temperature self-test at
150°C.
We are still doing research on other factors such as flip
chip bonding temperature, and gold stud bump shape and
will publish the results in a next paper.
4. ConclusionUltrasonic flip-chip bonding of TEG dies with gold stud-
bumps was applied to substrates with longitudinal and lat-
eral leads, and the die-pull mode was investigated system-
atically. When conventional flip-chip bonding equipment
was used, the die-pull test showed different bonding states
for the longitudinal and lateral leads. A flip-chip bonder
with a vibration head that could be rotated with respect to
the die configuration was developed. For a 45-degree-
rotated head, uniform bonding was established on both
longitudinal and lateral leads. A wide process margin for
flip-chip bonding was obtained with a high yield.
References[1] Y. Jin, Z. Wang, and J. Chen, “Introduction to micro-
system packaging technology,” Science Press, pp.
73–83, 2011.
[2] A. Yamauchi, S. Kuwauchi, S. Sato, and S. Nakai,
“Enhancements in ultrasonic flip chip bonding to flex-
ible printed circuit substrate (in Japanese),” MES
2003 Symposium Proceedings, pp. 200–203, 2003.
[3] Y. Arai, W. Jimyung, S. Aoki, K. Imai, and Y.
Miyamoto, “Die pull tester for flip-chip bonding,”
ICEP2011 Proceedings, pp. 688–692, 2011.
[4] M. Masumoto, N. Nakanishi, and A. Okazaki,
Japanese Patent JP2010-258302A, 2011.
[5] Y. Arai, Y. Miyamoto, S. Aoki, and K. Shimatani,
“Mode counting system for die pull test,” ICEP2012
Proceedings, pp. 373–377, 2012.
Mutsumi MasumotoYoshiyuki AraiHajime Tomokage
Fig. 9 Distribution of bump shear strengths for bonds made with the 45-degree-rotated bonder.
10
20
30
40
50
60
0 5 10 15 20 25
Bump shear strength (g)
Inci
denc
era
tio (%
)
Lateral
Longitudinal
Fig. 10 Mode B pull-off patterns for the longitudinal (a) and lateral (b) leads.
Fig. 11 Diagram showing bonding states and times for longi-tudinal and lateral leads bonded using a 45-degree-rotated bonder. Note the increased process margin.
Lead open
Lead open
CrackGood
CrackGood
Longitudinal lead
Lateral lead
Bonding time(s)
Process margin
0.5 1.0 1.5 0.95 1.2
(a) (b)(a) (b)