Академический Документы
Профессиональный Документы
Культура Документы
1
Department of Materials Science and Engineering, University of Texas at Arlington
501 West First Street, Box 19031, Arlington, Texas 76019
2
present: Globalfoundries, 400 Stone Break Extension, Malta, New York 12020
3
Texas Instruments, 12500 TI Boulevard, Dallas, TX 75243
4
contact e-mail: choongun@uta.edu
978-1-4799-8609-5/15/$31.00 2015 IEEE 135 2015 Electronic Components & Technology Conference
joint because the failure may develop by two kinetically substantially reduced. There exists a few places where IMC
competing processes, which are the rate of voiding by Sn EM growth is found to be highly enhanced, which is a subject of
and IMC growth by Cu EM. This speculation is indeed proven our ongoing research; however, the resulting solder joint
in our study that investigates the kinetic mechanism of EM microstructure, consisting of narrow channels of Sn encased
damage in low-profile solder joint made of Cu-post, Sn--Ag by thick IMC, provides us with an ideal condition for studying
solder, and Ni coated Cu lead-frame. Our study finds that EM the EM voiding mechanism with limited amount Sn in the
damage stops when Sn solder layer is fully converted to Cu- joint.
Sn IMCs and that the joint is left with residual voids In order to confine the EM damage to the target joint, we
developed in Sn prior to its conversion to Cu-Sn IMCs. The placed a number of supporting posts near the EM post and
residual damage appears as voids surrounded by IMC phase, used them to provide distributed current exit paths to the post
and the degree of voiding is found to be affected by EM under EM testing. This arrangement is used to make the test
condition. This paper presents examples of our data with current from the LF to Cu-post, meaning that EM occurs from
weighted emphasis on the effect of EM on the rate of Sn the post to the LF direction. Also developed for this study was
conversion to IMC and the degree of EM damage appearing a customized EM test oven to enable effective joule heat
as residual voids. dissipation and maintenance of temperature uniformity and
stability for duration of EM testing. The developed oven
Experimental Method and Sample Configuration system uses a large surface heater with high speed blower and
The samples used in our investigation are based on a flip min-air conduits made of Al fin arrays. High speed forced air
chip on lead-frame structure with Pb-free Sn solder containing circulation over the array of samples enabled us to keep the
approximately 2.5% Ag. As presented in Figure 1(a), where joule heat induced temperature rise to under 6oC at 7.6A
cross-sectional view of the joint is shown, the Cu-post is (~27kA/cm2) test current and 150oC test temperature. EM
patterned on top of Si wafer. The dimension of Cu post is testing was conducted at five different conditions: 150oC (6.4,
approximately 300m in length, 110m in width 35m in 7.0, 7.6A), 140oC/7.0A, and 160oC/7.0A. The temperatures
height. The mating electrode of the bump is the Cu-lead frame used in our testing were determined by the on-chip
(LF) coated with 1m thick Ni. The solder layer was formed temperature sensor. An average of 16 samples were used per
by plating the solder alloy and by subjecting the assembly to a test condition. During EM testing, resistance of the joint was
standard reflow condition. The distance between the top of the closely tracked in order to monitor the level of damage
Cu-post and bottom of LF is approximately 15m. Due to resulted by EM.
formation of Cu6Sn5 at the post interface and (Cu,Ni)6Sn5 at
the LF interface, the actual solder layer thickness is reduced to Results
~12m after the first reflow. In order to simulate the reflow 80
condition that the chip may receive in its application, the t
P
70 6.4A
samples were subjected to three additional reflow treatments.
60 R
P
R(t)/R(0), %
7.0A
50
40 7.6A
30
20
10
0
0 500 1000 1500 2000 2500 3000
Time, hrs
Figure 2. A plot showing the representative change in the
sample resistance during EM testing at 150oC under three
different test currents.
Figure 1. Cross-section micrographs showing the overall view
of the Cu-post/LF solder joint (a), the joint microstructure
after the first reflow (b), and after additional reflow treatments EM Damage Measured by Resistance Change
(c). No complete EM failure was seen in our EM testing that
lasted well over 4000 hours in most test splits. There were a
This induces an additional growth of the Cu6Sn5 and few samples, less than 5%, that showed complete failure at
(Cu,Ni)6Sn5 IMCs but the growth is found to be more early stage of the testing but subsequent failure analysis
extensive in a few places along the long-axis direction. As indicated that these early failures were not the result of EM.
shown in Figures 1(b) and (c), where cross-sectional SEM Judging from the fact that they were more or less random in
view of the joint after the first reflow and additional reflow is terms of test conditions, we concluded that they were
compared, the growth of IMC phase on both interfaces, associated with the flaws introduced during assembly process,
post/solder and solder/LF, makes the solder layer to be such as uneven stand-off or residual stress distribution. Since
136
such flaws can be eliminated by improved process Cu6Sn5 is at its terminal stage where EM induced voiding is
optimization and they were not related to EM reliability, they disrupted. The amount of initial Sn in solder layer is likely to
were removed from our EM analysis. be similar among samples, making the contribution of Sn
While the majority of samples showed no clear sign of consumption by Cu6Sn5 on the joint resistance more or less
EM failure, we found a consistent trend that could be related the same irrespective of sample and EM test conditions.
to EM induced residual damage in the joint. The indicator for Therefore, it is reasonable to assume that the variation in the
the residual damage was found from the increase in the joint peak resistance reflects the total amount of EM induced
resistance. Figure 2 compares the most representative damage in the Sn phase prior to its complete conversion of
resistance traces taken from samples tested at 3 different Cu6Sn5 IMC phase. Also it is reasonable to assume the peak
stress currents at 150oC. Note that the resistance change is not time to be the time for completion of Sn to Cu6Sn5
a monotonic function of time. Rather, it shows an initial stage conversion.
of rapid increase followed by the slowly decaying stage. As a Post EM Joint Microstructure
result, there exists a peak point in the resistance change (RP). In order to substantiate our assumption that the phase
The existence of the resistance peak may appear odd because conversion and EM induced damage in Sn layer is responsible
both the conversion of Sn to Cu6Sn5 IMC and EM damage are for the unique change in the joint resistance shown in Figure
likely to increase the joint resistance. However, we find that 2, EM tested samples were subjected to microscopic
the peak resistance appears because conversion of Cu6Sn5 to inspection. This analysis, conducted on multiple samples per
Cu3Sn IMC phase reduces the joint resistance. test condition, revealed that the joint could be characterized
by two types of isolated voids trapped inside of Cu6Sn5 and
Cu3Sn formed at Cu6Sn5/Cu-post interface. An example of
such microstructure is shown in Figures 3, where SEM
micrographs of the joint taken after ~5200 hours of EM
testing at 150oC under 7.0A are presented. Note that there is
no Sn phase remaining in the joint. This means that Sn phase
is completely consumed and replaced by Cu6Sn5 phase. It also
indicates that the voids visible within the Cu6Sn5 phase are
developed during Sn consumption process because low
diffusivity of Sn and Cu in Cu6Sn5 phase would prevent EM
from occurring to a degree to induce voids of such a scale.
Our inspection on multiple samples suggests that there is no
clear preference for the voiding site.
137
growth extent of Cu3Sn phase is found to be influenced by 120
EM. An example evidence of the accelerated Cu3Sn phase
growth is shown in Figures 4. The joint microstructure shown 100
in Figure 4 was taken from a sample tested at 150oC under 7.6A 6.4A
7.6A stress current after ~4900 hours. It can be seen that the 80
thickness of the Cu3Sn phase formed in EM tested joint is
RP
considerably thicker than the one found in the reference joint 60
in the same sample that was subject to thermal aging only.
Since the reference joint was not subjected to EM, the Cu3Sn 40
phase must have been formed by pure inter-diffusion. Thus,
the net thickness difference in these two joints is attributed to 20
the growth acceleration by Cu EM while Sn phase was
surviving in the joint. It should be noted that the accelerated 0
0 500 1000 1500 2000 2500 3000 3500 4000
growth of Cu3Sn phase is not a result of Cu EM in Cu-post
tP, hrs
nor in Cu6Sn5 phase as low diffusivity of Cu in those phases
would make the Cu EM flux to be negligibly small. The Figure 6. A plot showing RP of samples tested at 150oC
growth of Cu3Sn is known to proceed when the Cu6Sn phase under 6.4 and 7.0A as a function of tP.
is sizable [20], and therefore thicker Cu3Sn phase seen in EM
stressed joint can be attributed to the fact that faster
conversion of Sn to Cu6Sn5 makes Cu3Sn growth process to Sn consumption, the result showing decrease in average tP
start faster. It then follows that Cu EM in Sn phase induces with increase in stress current is consistent with the
accelerated growth of both Cu6Sn5 and Cu3Sn IMC phases. expectation. What is unexpected at first glance, though, is the
This conclusion is in good agreement with the findings made result that the average RP is almost independent of current or
from the peak time analysis detailed below. slightly increases with decreasing current. There is a
considerable scatter in RP, but there exists an unmistakable
Dependence of RP and tP on EM Condition
trend that it is more or less independent of current. Since RP
Our attempts to characterize the peak time, tP, of EM is proportional to the amount of voids formed while Sn is
tested samples indicate that it is indeed affected by the EM remaining in the joint and Sn EM is a predominant source of
condition. We were not able to determine RP and tP from all the void, it is expected to increase with the current. In this
samples because some showed resistance with too much respect, the result shown in Figure 5 appears to be odd and
fluctuation when it approached to the peak point while the disagrees with the usual EM damage model. However, we
resistance of others, especially the ones tested at 140oC, did believe that our result may be explained when the effect of tP
not reach the condition to provide clear determination of the on EM voiding kinetics is considered. It is important to note
peak within the given EM testing time. However, we were here that the EM voiding stops when Sn phase is completely
able to collect RP and tP from a reasonably good number of consumed. In this circumstance, the extent of voiding may not
samples, which was sufficient to draw statistically meaningful be a function of the current but can be affected by tP. At lower
conclusions. current, it takes longer time for Sn to be consumed, providing
4 more time for Sn EM to continue, yet such effect is somewhat
1 10
100
counter-balanced by the smaller Sn-EM flux. This can result
in RP that is at similar or higher level than at higher current.
8000 While there may be other contributing factors, such as the
thermal voiding, our view is supported by the observation of
6000 the strong dependence of RP on tP within samples tested at
tP, hrs
R , %
138
within the known range of activation energy for Cu diffusion high power electronic devices because the current carrying
in Sn phase [21-23]. capability of the joint can be substantially increased.
While the benefit of low-profile solder joint is evident in
10
4
150 terms of EM resistance, it results in a challenge in conducting
EM reliability assessment. It is important to realize that the
critical amount of solder for EM immunity is not fixed but
varies with EM condition. For aggressive EM condition, the
1000 100 EM immunity would require small amount of solder, while it
R , %
t , hrs
P
that there is a possibility that EM failure kinetics do not scale
P
139
Sn due to the growth of Cu6Sn5 IMC. The interruption of 11. F.-Y. Ouyang, H. Hsu, Y.-P. Su, and T.-C. Chang,
voiding leaves EM damage in Cu6Sn5 phase in the form of Electromigration induced failure on lead-free micro
residual voids. The amount of the residual voids is found to bumps in three-dimensional integrated circuits
vary with EM condition because the Cu6Sn5 phase growth rate packaging, Journal of Applied Physics, vol. 112, p.
is affected by the Cu EM. The kinetic interplay between the 023505, 2012.
Cu EM, responsible for Cu6Sn5 growth, and Sn EM, 12. Y. C. Chan, and D. Yang, Failure mechanisms of solder
responsible for voiding, is found to be the factor that interconnects under current stressing in advanced
determines the degree of EM damage. This kinetic interplay electronic packages, Progress in Materials Science, vol.
between these two factors makes EM voiding kinetics of the 55, pp. 428-475, 2010.
low-profile solder joint to be more complex than what is 13. D.-Q. Yu, T. C. Chai, M. L. Thew, Y. Y. Ong, V. S. Rao,
found from the more conventional solder joint. L. C. Wai, and J. H. Lau, Electromigration study of 50
m pitch micro solder bumps using four-point kelvin
References
structure, in Proc. 2009 Electronic Components and
1. K. N. Tu, H. -Y. Hsiao, and C. Chen, Transition from flip
Technology Conference (ECTC), IEEE, 2009, pp. 930-
chip solder joint to 3D IC microbump: Its effect on
935.
microstructure anisotropy, Microelectronics Reliability,
14. K.-W. Moon, W. J. Boettinger, U. R. Kattner, F. S.
vol. 53, pp. 2-6, 2013.
Biancaniello, and C. A. Handwerker, Experimental and
2. H. Huebner , S. Penka, B. Barchmann, M. Eigner, W.
thermodynamic assessment of Sn-Ag-Cu solder alloys,
Gruber, M. Nobis, S. Janka, G. Kristen, and M.
Journal of Electronic Materials, vol. 29, pp. 1122-1136,
Schneegans, Microcontacts with sub-30 m pitch for 3D
2000.
chip-on-chip integration, Microelectronic Engineering,
15. M. Murugesan, Y. Ohara, T. Fukushima, T. Tanaka, and
vol. 83, pp. 21552162, 2006.
M. Koyanagi, Low-resistance Cu-Sn electroplated
3. C.-C. Lee, K.-S. Kao, R.-S. Cheng, C.-J. Zhan, and T.-C.
evaporated microbumps for 3D chip stacking, Journal of
Chang. Reliability enhancements of chip-on-chip
Electronic Materials, vol. 41, pp. 720-729, 2012.
package with layout designs of microbumps,
16. B. Chao, S.-H. Chae, X. Zhang, K.-H. Lu, J. Im, P.S. Ho,
Microelectronic Engineering, vol. 120, pp. 138-145, 2014.
Investigation of diffusion and electromigration parameters
4. O. M. Abdelhadi, and L. Ladani, Effect of joint size on
for CuSn intermetallic compounds in Pb-free solders
microstructure and growth kinetics of intermetallic
using simulated annealing, Acta Mater. vol. 55, pp. 2805-
compounds in solidliquid interdiffusion Sn-3.5Ag/Cu-
2814, (2009).
substrate solder joints, Journal of Electronic Packaging,
17. B. Chao, X. Zhang, S.-H. Chae, P.S. Ho, Recent
vol. 135, p. 021004. 2013.
advances on kinetic analysis of electromigration enhanced
5. C.-B. Lee, I.-Y. Lee, S.-B. Jung, and C.-C. Shur, Effect
intermetallic growth and damage formation in Pb-free
of surface finishes on ball shear strength in BGA joints
solder joints, Microelectronics Reliability vol. 49, pp.
with Sn-3.5 mass% Ag solder, Materials Transactions,
253-263, 2009.
vol. 43, pp. 751-756, 2002.
18. M. Li, Z. Li, Y. Xiao, and C. Wang, Rapid formation of
6. I. Shohji, F. Mori, and K. F. Kobayashi, Thermal fatigue
Cu/Cu3Sn/Cu joints using ultrasonic bonding process at
behavior of flip-chip joints with lead-free solders,
ambient temperature, Applied Physics Letters, vol. 102,
Materials Transactions, vol. 42, pp.790-793, 2001.
p. 094104, 2013.
7. M. Lee, M. Yoo, J. Cho, S. Lee, J. Kim, C. Lee, D. Kang,
20. K. Zeng, R. Stierman, T.-C. Chiu, D. Edwards, K. Ano,
C. Zwenger, and R. Lanzone, Study of interconnection
and K. N. Tu, Kirkendall void formation in eutectic SnPb
process for fine pitch flip chip, in Proc. 2009 Electronic
solder joints on bare Cu and its effect on joint reliability,
Components and Technology Conference (ECTC), IEEE,
Journal of Applied Physics, vol. 97, p. 024508, 2005.
2009, pp. 720-723.
21. Y. C. Chan, A. C. K. So, and J. K. L. Lai, Growth kinetic
8. H.-Y. Son, S.-K. Noh, H.-H.ee Jung, W.-S. Lee, J.-S. Oh,
studies of CuSn intermetallic compound and its effect on
and N.-S. Kim, Reliability studies on micro-bumps for 3-
shear strength of LCCC SMT solder joints, Materials
D TSV integration, in Proc. 2013 Electronic Components
Science and Engineering B, vol. 55, pp. 5-13, 1998.
and Technology Conference (ECTC), IEEE,2013 , pp.29-
22. N. Mookam, and K. Kanlayasiri, Evolution of
34.
intermetallic compounds between Sn0.3Ag0.7Cu low-
9. Y.-M. Lin, C.-J. Zhan, J.-Y. Juang, J. H. Lau, T.-H. Chen,
silver lead-free solder and Cu substrate during thermal
R. Lo, M. Kao, T. Tian, and K.-N. Tu, Electromigration
aging, Journal of Materials Science and Technology, vol.
in Ni/Sn intermetallic micro bump joint for 3D IC chip
28, pp. 53-59, 2012.
stacking, in Proc. 2011 Electronic Components and
23. Z. Mei, A. J. Sunwoo, J. W. Morris, Analysis of low-
Technology Conference (ECTC), IEEE, 2011, pp. 351-
temperature intermetallic growth in copper-tin diffusion
357.
couples, Metallurgical Transactions A, vol. 23, pp. 857-
10. J. Kim, J.A. Noquil, T. K. Tan, C.-L. Wu, and S.-Y. Choi,
864, 1992.
Multi-flip chip on lead frame overmolded IC package: a
24. R. Labie, W. Ruythooren, K. Baert, E. Beyne, and B.
novel packaging design to achieve high performance and
Swinnen, Resistance to electromigration of purely
cost effective module package, in Proc. 2005 Electronic
intermetallic micro-bump interconnections for 3D-device
Components and Technology Conference (ECTC), IEEE,
2005, pp. 1819-1821
140
stacking, in Proc. 2008 Interconnect Technology
Conference (IITC), 2008, pp. 19-21.
25. Y. Wang, S.-H. Chae, R. Dunne, Y. Takahashi, K.
Mawatari, P. Steinmann, T. Bonifield, T. Jiang, J. Im, and
P. S. Ho, Effect of intermetallic formation on
electromigration reliability of TSV-microbump joints in
3D interconnect, in Proc. 2012 Electronic Components
and Technology Conference (ECTC), IEEE, 2012, pp.
319325.
26. C. C. Wei, C. H. Yu, C. H. Tung, R. Y. Huang, C. C.
Hsieh, C. C. Chiu, H. Y. Hsiao, Y. W. Chang, C. K. Lin,
Y. C. Liang, C. Chen, and T. C. Yeh, "Comparison of the
electromigration behaviors between micro-bumps and C4
solder bumps in Proc. 2011 Electronic Components and
Technology Conference (ECTC), IEEE, 2011, pp.706-710.
27. D.-Q. Yu, T.C. Chai, M.L. Thew, Y.Y. Ong, V.S. Rao,
L.C. Wai, and J.H. Lau, Electromigration study of 50 m
pitch micro solder bumps using four-point Kelvin
structure, in Proc. 2009 Electronic Components and
Technology Conference (ECTC), IEEE, 2009, pp. 930
935.
28.. Y. Wang, K. H. Lu, V. Gupta, L. Stiborek, D. Shirley, J.
Im, and P. S. Ho, Effect of Sn grain structure on
electromigration reliability of Pb-free solders, in Proc.
2011 Electronic Components and Technology Conference
(ECTC), 2011, pp. 711-716.
29. K. Lee, K.-S. Kim, Y. Tsukada, K. Suganuma, K.
Yamanaka, S. Kuritani, and M. Ueshima, Effects of the
crystallographic orientation of Sn on the electromigration
of Cu/SnAgCu/Cu ball joints, Journal of Materials
Research, vol. 26, pp. 467-474, 2011.
141