IEEE TRANSACTIONS ON ELECTRONICS PACKAGING MANUFACTURING, VOL. 31, NO. 3, JULY 2008 221
Study of Temperature Parameter
in Au Ag Wire Bonding
Zhili Long, Lei Han, Yunxin Wu, and Jue Zhong
Abstract The effect of the temperature on bondability and parameters and creating the empirical model [2] [5]. However,
bonding process for wire bonding are investigated. Bondability is
some necessary and valuable work still needs to be improved to
characterized by shear bonding strength and bonding process is
better understand bonding phenomena.
represented by input and output power of ultrasonic transducer.
Process monitoring of the bonding interface for wire bonds
A laser Doppler vibrometer and Labview software were used to
is difficult because the gold ball is small 20 m 60 m , and
record the velocity, voltage and current of transducer at different
the process is short (20 80 ms). Theories and attempts to solve
temperature settings. A K-type thermocouple sensor was used
to measure the bonding temperature. Experimental results show these problems have been developed [6] [9]. In the age of zero
that unsuccessful bonding happens at low temperature, and over
defect manufacturing, electronics manufacturing engineers are
bonding appears if the temperature is too high. Only when the
seeking for the best solution based on more reliable statistical
temperature is at appropriate settings, can a stable and satis-
data and analysis of the bonding process.
fied bondability be attained. The reason for this experimental
Temperature is a critical and important parameter for semi-
observation is analyzed. By using a high resolution transmission
conductor packaging. For example, current practice is to carry
electron microscope, the atom diffusion depth of Au-Ag bonding
interface was measured and the result is about 200 nm. By using out soldering or solder plating processes of the leads. If the en-
joint time-frequency analysis, the instantaneous characteristics
vironmental temperature is too high, the solder plated on the
of bonding process were observed completely and clearly. It is
copper frame will melt and contaminate the bonding surface.
found that input and output ultrasonic power vs. time-frequency
Moreover, high temperature will affect the equipment, such as
in a bonding process, including resonance frequency, harmonic
piezoceramic material in transducers.
components and amplitude of ultrasonic energy, vary along with
Although there are lots of studies about temperature effect
the change of temperature settings.
in wire bonding [10] [12], this topic is still worth investigating
Index Terms Bondability, bonding process, bonding tempera-
because that area is very complicated. Studies about tempera-
ture, joint time frequency analysis, wire bonding.
ture effects on input/output power of ultrasonic transducers are
scarce yet. Compared to other measurements of bonding quality
I. INTRODUCTION
such as pull strength, destructive shear measurement is more
controllable and stable, which avoids the disturbance from oper-
N recent years, as IC packaging development has moved to-
ator and environment. Therefore, the destructive shear strength
wards higher power, smaller size, thinner dimensions, and
I
between the gold ball and substrate is a common judgment for
denser circuits, wire bonding is still the most commonly used in-
bondability [7], [16]. The objective of this research is to under-
terconnection technique in first-level microelectronic packages
stand the effect of temperature on bondability and the bonding
[1]. The number of wire bonds made exceeds three trillion marks
process. Seventeen groups of bonding data at different tempera-
on an annual basis. Such a high volume has driven conventional
ture settings were compared to establish a relationship between
wire bonding to new levels of performance and reliability. While
bondability windows and input/output power of ultrasonic trans-
the wire bond failure rate for individually packaged parts is typ-
ducers. A laser Doppler vibrometer and Labview software were
ically in the low parts per million range, the failure rates for wire
used to monitor the bonding process.
bonds in multichip modules (MCMs), chip on-board (COB) as-
semblies, and other advanced packaging structures are consid-
erably higher [2].
II. EXPERIMENT
It is evident that bonding parameters are the influencing fac-
tors on bonding strength. In order to improve wire bond bond-
The wire-bond machine used in the study is a TS2100 ther-
ability, lots of effort has been spent on optimizing the bonding
mosonic gold wire bonder (Wei Tianxing Co., China). The di-
ameter of gold wire (99.99% gold) is 25 m. In order to easily
Manuscript received September 2, 2007; revised January 23, 2008. Pub- detect the microstructure and the diffusion area of the bonding
lished July 7, 2008 (projected). This work was supported by the National
interface, the copper substrate plated with Ag of 1 m thick-
Science Foundation of China under Contracts 50390064 and 50605064, the
ness is used in our experiment. Moreover, The Au Ag system
Ph.D. Program Foundation of Ministry of Education of China under Contract
20060533068, the China Department of Science and Technology Program 973
has been proven to be more reliable since it does not form exces-
under Contract 2003CB716202, and the Hunan Technology Project under Con-
sive intermetallic compounds, nor is it vulnerable to corrosion
tract 2007FJ3098. This work was recommended for publication by Associate
issues [17]. In this experiment, the bonding temperature was
Editor I. Fidan upon evaluation of the reviewers comments.
The authors are with the Department of Mechatronics, Central South Univer- tuned regularly from 40 C to 360 C with a step of 20 C, while
sity, Changsha 410083, China (e-mail: [e-mail]).
all other parameters such as bonding power, bonding time, and
Color versions of one or more of the figures in this paper are available online
normal pressure, were kept in an unvaried setting. The bonding
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TEPM.2008.926278 parameters are listed in Table I. During the bonding process,
1521-334X/$25.00 © 2008 IEEE
Authorized licensed use limited to: Imperial College London. Downloaded on June 07,2010 at 19:35:32 UTC from IEEE Xplore. Restrictions apply.
222 IEEE TRANSACTIONS ON ELECTRONICS PACKAGING MANUFACTURING, VOL. 31, NO. 3, JULY 2008
TABLE I
BONDING PARAMETERS
Fig. 2. Shear strength versus temperature.
TABLE II
MEAN AND STANDARD DEVIATION OF SHEAR
STRENGTH VERSUS TEMPERATURE
Fig. 1. Experimental setup.
the instantaneous driving voltage, current of the PZT (lead zir-
conate titanate) transducer, and velocity of the capillary tip were
recorded synchronously. After bonding, destructed shear tests
were used to evaluate the bonding strength of the wire-to-sub-
strate bond. Full statistical data involving repeating 50 times for
each bond and a total of 750 experiments were attained.
The driving voltage and current of the PZT trans-
ducer were sampled by using an NI (National Instrument Co.)
data acquisition card and Labview software. Bonding tempera-
or unsatisfied bondability happens when temperature as low
ture was measured by a K-type thermocouple sensor. The tem-
40 C 80 C . However, the over bonding phenomenon, in
perature sensor contacted directly the substrate. A shear tester
which bonding strength decreases as temperature increases, is
(Dage-4000) was used to evaluate the shear bonding strength.
found if the temperature is too high 320 C 360 C . Only
The velocity measurement of the capillary tip was carried
when the temperature is in a moderate 220 C 240 C set-
out using a Polytec laser Doppler vibrometer. The sample rate
ting, can stable and satisfied bondability be attained. Statistical
was 1 MHz. Therefore, the input and output ultrasonic powers
mean and standard deviations of bonding strength for each
of the transducer system, which indicates the dynamics of ultra-
experimental temperature are listed in Table II.
sonic energy in a bonding process, were recorded. The experi-
In a bonding process with a Au and Ag system of thermosonic
mental setup is shown in Fig. 1.
wire bond, temperature plays an important role in accelerating
The input ultrasonic power is expressed as
metal flow and softening the bonding surface. It is found that
poor bondability happens with low temperature. On the other
(1)
hand, high temperature deforms the gold ball badly, which
causes unsatisfied bonding strength. Therefore, a favorable
The output power is expressed by the root mean square (rms)
bonding process is only attained at moderate temperature. This
calculation of the ultrasonic velocity. That is
experimental phenomenon may be understood by the following
theories and experiments [11] [13].
(2)
1) When bonding temperature is low, the oxides and other
forms of contamination cannot be easily removed from the
surface of the Ag and Au metal, and the clean underlying
where is the vibration velocity of the transducer horn, and
metal cannot be exposed, even though the ultrasonic en-
is one vibration period.
ergy and pressure have loaded. Low bonding temperature,
which means that there is not enough energy from the out-
III. EFFECT OF TEMPERATURE ON BONDABILITY
side, cannot provide the environment for interfusion of the
The experimental results that show the effect of temper- materials. The microstructure of gold ball that was bonded
ature on bondability (shear bonding strength) are shown in at low temperature is shown in Fig. 3, where it is clearly
Fig. 2. From the experimental graph, unsuccessful bonding shown that there is a lot of contamination plated on the gold
Authorized licensed use limited to: Imperial College London. Downloaded on June 07,2010 at 19:35:32 UTC from IEEE Xplore. Restrictions apply.
LONG et al.: STUDY OF TEMPERATURE PARAMETER IN Au Ag WIRE BONDING 223
Fig. 3. Microstructure of wire-to-substrate at low temperature.
Fig. 6. EDXS measure of Ag Ag bonding interface.
Fig. 4. Microstructure of wire-to-substrate at moderate temperature.
Fig. 5. Microstructure of wire-to-substrate at high temperature.
Fig. 7. Typical velocity signal of capillary tip. (a) Time waveform and (b) fre-
quency spectrum.
ball. It is the difficulty in removal of the oxidation film of
Ag and Au that results in poor and unsatisfied bondability.
2) Appropriate and moderate temperature can soften the
metal material and accelerate the atomic interfusion be- interaction between Au and Ag. The microstructure of
tween Au and Ag, which results in favorable bonding. satisfied bondability is shown in Fig. 4. It is found that
According to expression of atom diffusion coefficient, there are no cracks or contamination in the wire bed root.
, where is diffusion coef- 3) From the microstructure of the wire bed at high tempera-
ficient, and is diffusion activation energy. It indicates ture settings shown in Fig. 5, it can be observed that there
that metal atoms tend to diffuse to each other when tem- are some cracks and cavities in the bonding interface. We
perature increases. Moreover, when temperature goes up, can deduce that it is the redundant temperature energy
it can soften the metal, reducing the stress required to that deforms the bonding interface and leads to unsatisfied
deform plastically a metal and generating of metallurgical bondability.
Authorized licensed use limited to: Imperial College London. Downloaded on June 07,2010 at 19:35:32 UTC from IEEE Xplore. Restrictions apply.
224 IEEE TRANSACTIONS ON ELECTRONICS PACKAGING MANUFACTURING, VOL. 31, NO. 3, JULY 2008
Fig. 8. STFT analysis result (at 80 C setting). (a) Input energy. (b) Output energy.
IV. INTERFACE CHARACTERS OF Au Ag BONDING transform (STFT), Gabor expansion, wavelets, Wigner Ville
distribution, and so on. In our paper, we apply STFT to the
The cross-sectional feature of the Au Ag bonding interface
input/output power of an ultrasonic transducer. The expression
was inspected by using a high-resolution transmission electron
of STFT is described as follows [14], [15]:
microscope (TEM). Because the standard sample for TEM is
3.0 mm in diameter, the Au Ag bonding sample must be dealt
by punching, grinding, and thinning. First, the sample was filled
(3)
with a special epoxy, and cured subsequently at 100 C for one
hour. Then, the sample was sawn to 3 mm diameter and 0.5 mm
thick. The specimen was ground manually and further thinned where is input or output velocity of the ultrasonic trans-
by a precision ion polishing system. Thus, a vertical cross sec- ducer.
tion of the Au Ag bonding interface was produced. Then, the Figs. 8 10 are the STFT results of the input and output en-
sample was measured by HRTEM (F30) with a line scanning ergy of the ultrasonic transducer at 80 C, 220 C, and 300 C,
energy dispersive X-ray spectroscopy (EDXS) at 300 kV. respectively. They show that 1) the dominant frequency of the
Fig. 6 is the scanning result of EDXS along the Au Ag ultrasonic transducer is about 60 kHz. It can be deduced that it
bonding interface, which was bonded at 220 C. It shows should be the working frequency because it is in good agree-
that the atom diffusion depth of the Au Ag is about 200 nm. ment with the manufacture s value of 59.0 61.0 kHz. It is also
According to the Au Ag phase diagram, this bonding interface observed that the working frequency changes slightly during
of Au Ag is a kind of solid solution structure, not a kind of bonding processes. 2) There are some low-frequency compo-
intermetallic compound, because both Au and Ag are unlimited nents mixed in input energy, which indicates that some ultra-
solid-solution metals. sonic energy is wasted in the other ways. There are not any ob-
vious harmonic components in output energy. 3) The bonding
V. EFFECT OF TEMPERATURE ON BONDING PROCESS
process can be divided into three periods. At an initial 0 6 ms,
In general, the signals are analyzed either in the time or the energy starts and increases gradually. This initial period may
frequency domain. There are some shortcomings for those represent the phase locking chaos. From 7 45 ms, the energy
conventional methods. For example, from a typical velocity maintains steadily, indicating the bonding strength formed by
signal waveform of the capillary tip in the time domain shown the generating of metallurgical interaction between Au and Ag
in Fig. 7(a), we cannot determine the frequency components metal. At an ending 46 50 ms, the energy attenuates gradually.
in the vibration signal and how the vibration energy changes This final attenuation may reflect the remains of kinetic energy
along time. Similarly, from its corresponding frequency spec- in the transducer structure.
trum shown in Fig. 7(b), we cannot tell when these harmonic The main difference of input/output ultrasonic energy at dif-
components happen and determine how the energy of those ferent temperature settings is the amplitude of the SFTF result.
signals evolve over time. When at 80 C, the input energy for a steady period is 12 (di-
However, joint time frequency analysis, which covers si- mensionless parameter), while it decreases to 8 when at 220 C
multaneously the time and frequency domain, is suitable to and goes down to 4 when at 300 C. A similar trend happens to
analyze the signal of the bonding process. The results of joint the output energy. Moreover, there are some differences for the
time frequency analysis can clearly and completely reveal bonding process at different temperature settings. When tem-
all information evolving over time, including the shift of fre- perature is 80 C, the initial period of input energy is longer
quency, change of harmonic components, and its corresponding compared with the one for 220 C and 300 C, which means
energy. Therefore, joint time frequency can help to under- that it needs more time to lock phase. At 220 C, the input and
stand the bonding process. There are quite a number of joint output ultrasonic energy is very steady in the middle stable pe-
time frequency analysis methods, such as short-time Fourier riod, which causes it to be satisfied and with stable bondability.
Authorized licensed use limited to: Imperial College London. Downloaded on June 07,2010 at 19:35:32 UTC from IEEE Xplore. Restrictions apply.
LONG et al.: STUDY OF TEMPERATURE PARAMETER IN Au Ag WIRE BONDING 225
Fig. 9. STFT analysis result (at 220 C setting). (a) Input energy. (b) Output energy.
Fig. 10. STFT analysis result (at 300 C setting). (a) Input energy. (b) Output energy.
TABLE III
BONDING PROCESS COMPARISONS AT DIFFERENT TEMPERATURE SETTINGS
When the temperature is high at 300 C, the input/output ul- energy versus time frequency in a bonding process, including
trasonic energy decreases gradually during the middle period. the shift of frequency, harmonic components, and amplitude
Table III lists the comparisons of STFT result of input/output of corresponding energy, vary along with the change of tem-
energy, working frequency in the middle period, and bonding perature. It is concluded that temperature not only affects the
strength at these three temperature settings. bonding strength, but it also affects the ultrasonic energy of the
transducer system.
VI. CONCLUSION
Based on the experimental measurements, the effect of
ACKNOWLEDGMENT
temperature on bondability and the bonding process for wire
The authors would like to thank the technical reviewers and
bonding was investigated. It is found that unsuccessful bonding
Associate Editor Dr. Ismail Fidan for their constructive reviews.
happens with low temperature, and over-bonding happens if
the temperature is too high. Only when the temperature is
REFERENCES
moderate can a stable and satisfied bondability be attained.
[1] G. G. Harman, Wire Bonding in Microelectronics: Materials, Processes
The atom diffusion depth of the Au Ag bonding interface is
Reliability and Yield, 2nd ed. New York: McGraw-Hill, 1997.
about 200 nm. By using the joint time frequency analysis, the
[2] H. K. Charles, K. J. Mach, and S. J. Lehtonen,  Wire bonding at higher
instantaneous characteristics of the bonding process can be
ultrasonic frequencies: Reliability and process implications, Micro-
observed completely and clearly. The input/output ultrasonic electron. Rel., vol. 43, pp. 141 153, 2003.
Authorized licensed use limited to: Imperial College London. Downloaded on June 07,2010 at 19:35:32 UTC from IEEE Xplore. Restrictions apply.
226 IEEE TRANSACTIONS ON ELECTRONICS PACKAGING MANUFACTURING, VOL. 31, NO. 3, JULY 2008
[3] F. Oldervoll and F. Strisland,  Wire-bond failure mechanisms in plastic Zhili Long received the M.S. and Ph.D. degrees in
encapsulated microcircuits and ceramic hybrids at high temperatures, mechatronics engineering from Central South Uni-
versity, Changsha, China, in 2002 and 2007, respec-
Microelectron. Rel., vol. 44, pp. 1009 1015, 2004.
tively.
[4] Y. F. Yao, B. Njoman, and K. H. Chua,  New encapsulation devel-
He is a Lecturer in the Department of Mecha-
opment for fine pitch IC devices, Microelectron. Rel., vol. 45, pp.
tronics, Central South University, Changsha, China.
1222 1229, 2005.
His research interests are ultrasonic packaging,
[5] R. Warrant and Anderson,  Reliability considerations for ESD protec-
design, and modeling of piezoelectric actuators.
tion under wire bonding pads, Microelectron. J., vol. 42, pp. 367 373,
2001.
[6] Z. L. Long, Y. X. Wu, and L. Han,  Measurement of driving electrical
signal and input impedance analysis of PZT transducer in thermosonic
bonding, in Proc. 6th IEEE CPMT Conf. High Density Microsyst. De-
sign Packag. Compon. Failure Anal., 2004, pp. 322 325.
[7] L. Han, F. L. Wang, and W. H. Xu et al.,  Bondability window
Lei Han received the B.S., M.S., and Ph.D. degrees
and power input for wire bonding, Microelectron. Rel., vol. 46, pp.
from the University of Science and Technology of
610 615, 2006.
China, Hefei, in 1982, 1984, and 1989, respectively.
[8] P. W. Chu, H. Li, and H. L. Chan et al.,  Smart ultrasonic transducer for
From 1991 to 1995 and 2000 to 2003, he was
wire-bonding applications, Mater. Chem. Phys., vol. 75, pp. 95 100,
with Oregon State University, Corvallis, Lehigh
2002.
University, Bethlehem, PA, the State University
[9] Z. Khatir and S. Lefebvre,  Boundary element analysis of thermal fa-
of New York at Stony Brook, and Case Western
tigue effects on high power IGBT modules, Microelectron. Rel., vol.
Reserve University, Cleveland, OH, as a Research
44, pp. 929 938, 2004. Associate. He is currently a Professor at Central
[10] S. J. Hu, G. E. Lim, and K. P. Foong et al.,  Study of temperature pa- South University, Changsha, China. His research
rameter on the thermosonic gold wire bonding of high-speed CMOS, interests include experimental mechanics, smart
IEEE Trans. Compon., Hybrids, Manuf. Technol., vol. 14, no. 4, pp. structures, wavelet analysis, and electronics packaging reliability.
855 858, Dec. 1991.
[11] M. Y. Li, H. J. Ji, and C. Wang,  Interdiffusion of Al-Ni system en-
hanced by ultrasonic vibration at ambient temperature, Ultrasonics,
vol. 45, pp. 61 65, 2006.
Yunxin Wu received the B.S. and M.S. degrees in
[12] D. S. Liu and C. Y. Ni,  A thermo-mechanical study on the electrical
mechanical engineering from Central South Univer-
resistance of aluminum wire conductors, Microelectron. Rel. , vol. 42,
sity, Changsha, China, and the Ph.D. degree from the
pp. 367 374, 2002. Faculte Polytechnique de Mons, Mons, Belgium, in
[13] S. Murali,  Formation and growth of intermetallics in thermosonic wire 1999.
bonds: Significance of vacancy-solute binding energy, J. Alloys Com- He is a Professor at Central South University. His
pounds, vol. 426, pp. 200 207, 2006. main research interests are dynamical model and con-
trol of mechanical engineering.
[14] L. Cohen, Time Frequency Analysis. Englewood Cliffs, NJ: Pren-
tice-Hall, 1995.
[15] D. Zhang and S. Ling,  Monitoring wire bonding via time-frequency
analysis of horn vibration, IEEE Trans. Electron. Packag. Manuf., vol.
26, no. 3, pp. 216 220, Jul. 2003.
[16] N. Srikanth, S. Murali, and Y. M. Wong,  Critical study of thermosonic
copper ball bonding, Thin Solid Films, vol. 462, pp. 339 345, 2004.
Jue Zhong graduated from the Beijing University of Science and Technology,
[17] J. Lee, M. Mayer, Y. Zhou, and S. J. Hong,  Iterative optimization of
Beijing, China, in 1958.
tail breaking force of 1 mil wire thermosonic ball bonding processes
She is currently a Professor at Central South University, Changsha, China,
and the influence of plasma cleaning, Microelectron. J., vol. 38, pp.
and a Fellow of the Chinese Academy of Engineering. Her research interests
842 847, 2007.
include industrial and microelectronics packaging equipment.
Authorized licensed use limited to: Imperial College London. Downloaded on June 07,2010 at 19:35:32 UTC from IEEE Xplore. Restrictions apply.