THE 9t!

.\3

This is to certify that the
thesis entitled

Effect of coarsened microstructure on electromigration
behavior of eutectic Pb-Sn solder joints

presented by
Yi-Chih Lee

has been accepted towards fulfillment
of the requirements for the

MS. degree in Material Science and

Engineering
I // J
1/ 5:?

Major Professor’s Signature

 

   

 

77/21 2/26, 20/0
{/’ ,

 

Date

MSU is an Affirmative ActiorVEqual Opportunity Employer

 

LIBRARY
Michigan State
Unive. sity

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:IProj/Acc&Pres/ClRC/Dateoue mdd

 

Effect of coarsened microstructure on electromigration behavior
of eutectic Pb-Sn solder joints

By

Yi-Chih Lee

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE

Materials Science and Engineering

2010

ABSTRACT

Effect of Coarsened Microstructure on Electromigration behavior of
Eutectic Pb-Sn Solder Joints

By
Yi-Chih Lee
Electromigration (EM) induced atom/ion movements in two-phase alloys should
depend on the size, shape, and distribution of both phases present. In order to evaluate the
roles of such morphological features on hillock/valley formation, and details of
microstructure evolution resulting from EM, joints made with eutectic Sn—Pb solder were
isothermally aged to coarsen the microstructure prior to current stressing. Effects of
coarsened microstructure on the events resulting from EM, when compared to those
noted in as-reflowed solder joints, indicate the roles of initial morphological features of

the phases present on the electromigration in a two-phase alloy.

ACKNOWLEDGEMENT

I would like to express my gratitude for my advisor Dr. A. Lee for his support,
encouragement, and help throughout the project. I would like to thank my committee Dr.
K. N. Subramanian for his help, guidance and encouragement. I would like to thank Dr. C.
E. Ho for his guidance and help. I would like to thank G. C. Xu, and T. Kobayashi for
their help in getting acquainted with various equipments required for this project.

I would like to thank to Department of Material Science at Michigan State
University for providing me golden opportunity of graduate studies at the university.

I deeply thank my parents and family for their emotional and financial support, and
their never-ending encouragement throughout my education. Finally, I would like to

thank my girlfriend, J. Y. Yu for her patience, love and support.

iii

TABLE OF CONTENTS
LIST OF TABLES .............................................................................. vi

LIST OF FIGURE ................................................................................ vii

1. INTRODUCTION

1.1. Electromigration is a significant reliability concern .................................... 1
1.2. Electromigration behavior ................................................................ 2
i. Phenomenon .................................................................. ,, ........ 2

ii. Quantitative evaluation for mean time to failure ................................ 5
1.3. Issues in accelerated electromigration of solder bump .............................. 6
i. Effect of current density on electromigration ..................................... 6

ii. Effect of temperature on electromigration ........................................ 7

iii. Effect of microstructure on electromigration .................................... 9
1.4. Solder joint configuration designed to reduce current crowding and
therrnomigration issue ..................................................................... 11

i. Current crowding issue ............................................................. 11

ii. Therrnomigration issue ............................................................ 12

iii. New design of solder joint to reduce current crowding and thermomigration
issues ................................................................................. 13

1.5. Analysis methods ........................................................................ 15
1.6. Motivation ................................................................................... 16
1.7. Aim of this study .......................................................................... 18

2. EXPERIMENTAL PROCEDURE

2.1. General ........................................................................................ 19
2.2. Sample preparation ........................................................................ 21
i. Materials used in this study ....................................................... 21

ii. Equipments used in this study ..................................................... 22

iii. Joining protocol ...................................................................... 25
2.3. Data collection and data analysis ........................................................ 30
i. Experimental protocol .............................................................. 30

ii. Methods of characterization ....................................................... 32

3. RESULTS AND DISCUSSIONS

3.1. Aging studies .............................................................................. 39
i. Aging at 100 °C ....................................................................... 39

a. Thickness of intermetallic compounds at solder/Cu interface ....... 39

b. Size distribution of Pb-rich phase within the solder joint .............. 40

c. Area fraction of Pb-rich phase between the two Cu substrates ........ 40

ii. Aging at 150 oC ...................................................................... 47

a. Thickness of intermetallic compounds at solder/Cu interface ....... 47

b. Size distribution of Pb-rich phase within the solder joint ............ 47

c. Area fraction of Pb-rich phases between the two Cu substrates.....48

iii. Summary and discussion .......................................................... 53

3. 2. Electromigration studies ................................................................ 55
i. Current stressing at 100 °C for 20 days.... .54

a. Characteristics of intermetallic compound layer ........................ 56

b. Size distribution of Pb-rich phase within the solder joint ............. 56

c. Area fraction of Pb-rich phases between the two Cu substrates ....... 56

ii. Current stressing at 150 °C for 6 days ............................................ 61

3. Characteristics of intermetallic compound layer ....................... 61

b. Size distribution of Pb-rich phase within the solder joint ............. 62

c. Area fraction of Pb-rich phase between the two Cu substrates ...... 62

iii. Summary and discussion ........................................................... 64

3. 3. Effect of coarsened microstructure on electromigration .......................... 68
i. Current stressing on coarsened microstructure at 100 °C for 20 days ...... 68

a. Characteristics of intermetallic compounds ............................ 69

b. Size distribution of Pb-rich phase within the solder joint .............. 70

c. Area fraction of Pb-rich phase between the two Cu substrates ........ 70

ii. Current stressing on coarsened microstructure at 150 °C for 6 days ........ 80

a. Characteristics of intermetallic compounds ............................... 81

b. Size distribution of Pb-rich phase within the solder joint ............. 81

c. Area fraction of Pb-rich phase between the two Cu substrates ....... 82

iii. Summary and discussion ............................................................ 91

4. SUMMARY ................................................................................... 93
5. REFERENCES ............................................................................... 94

LIST OF TABLES

Table 1. Hungtington’s diffusion regimes ......... , ............................................ 8
Table 2. Thickness of IMC layer in specimen age at 100°C ................................. 76
Table 3. Thickness of IMC layer in specimen aged at 150°C ................................. 87

vi

LIST OF FIGURES

Figure 1. Schematics drawing showing the solder joint configuration use ................. 20
Figure 2. Schematics drawing showing the Cu substrate package ........................ 23
Figure 3. Equipment used for alignment the parts for the soldering operation ............ 24
Figure 4. Schematics drawing showing the fabrication of the solder joint ................... 26
Figure 5. The overview of the joint used in this study ......................................... 28
Figure 6. The microstructure of as—reflowed solder joint cooled by airflow ............... 29
Figure 7. The microstructure of as-reflowed solder joint cooled by water ................. 29
Figure 8. Pb-rich phases outlined by ImageJ ................................................... 31
Figure 9. F low-chart of experimental protocol ................................................ 33
Figure 10. Observed area of the solder joint for size distribution of Pb-rich phase ........ 36
Figure 11. Area fraction of Pb-rich phase analysis ............................................. 37
Figure 12. Area fraction of Pb-rich phase analysis near anode .............................. 38
Figure 13. The image of as-reflowed solder joint microstructure .......................... 42
Figure 14. The image of solder joint isothermally aged at 100°C for 12 days .............. 43
Figure 15. The image of solder joint isothermally aged at 100°C for 15 days .............. 44
Figure 16. Size distribution of Pb-rich phase within solder joints at 100°C ............... 45

vii

Figure 17. Area fraction of Pb-rich phase at 100°C .......................................... 46

Figure 18. The image of solder joint isothermally aged at 150 °C for 3 days ............ 49
Figure 19. The image of solder joint isothermally aged at 150 °C for 6 days ............ 50
Figure 20. Size distribution of Pb-rich phase within solder joints at 150°C ............... 51
Figure 21. Area fraction of Pb-rich phase at 150°C .......................................... 52

Figure 22. The overview SEM image of as-reflowed solder joint subjected to current
stressing for 20 days at 100°C ...................................................................... 58

Figure 23. The image of as-reflowed solder joint subjected to current stressing for 20
days at 100°C ......................................................................................... 59

Figure 24. Size distribution of Pb-rich phase in as-reflowed solder joint subjected to
current stressing for 20 days at 100 °C ........................................................... 60

Figure 25. Area fraction of Pb-rich phase for as-reflowed solder joint subjected to current
stressing for 20 days at 100°C ...................................................................... 60

Figure 26. The overview SEM image of as-reflowed solder joint subjected to current
stressing for 6 days at 150 °C ...................................................................... 63

Figure 27 The image of as-reflowed solder joint subjected to current stressing for 6 days
at 150 °C .............................................................................................. 64

Figure 28. Size distribution of Pb-rich phase for joint without coarsening subjected to
current stressing for 6 days at 150 °C .......................................................... 65

Figure 29. Area fraction of Pb-rich phase for as-reflowed joint subjected to current
stressing for 6 days at 150°C ..................................................................... 65

Figure 30. The overview SEM image for Joint coarsened for 15 days at 100°C and
subjected to current stressing for 20 days at 100°C ............................................. 73

Figure 31. The image of solder joint isothermally aged at 100 °C for 12 days and
subjected to current stressing for 20 days at 100 °C .......................................... 74

Figure 32. The image of solder joint isothermally aged at 100 °C for 15 days and
subjected to current stressing for 20 days at 100 °C .......................................... 75

Figure 33. Size distribution of Pb-rich phase for coarsened microstructure at 100°C...77

viii

Figure 34. Area fraction of Pb-rich phase for coarsened microstructure at 100°C ........ 78

Figure 35. Accumulation of Pb-rich phase near anode for coarsened microstructure at
100°C ................................................................................................ 79

Figure 36. The overview SEM image of joint coarsened for 6 days at 150°C and subjected
to current stressing for 6 days at 150°C ......................................................... 84

Figure 37. The image of solder joint isothermally aged at 150 °C for 3 days subjected to
current stressing for 6 days at 150°C ............................................................ 85

Figure 38. The image of solder joint isothermally aged at 150 °C for 6 days and subjected
to current stressing for 6 days at 150 °C ......................................................... 86

Figure 39. Size distribution of Pb-rich phase for coarsened microstructure at 150°C...88

Figure 40. Area fraction of Pb-rich phase for joint isothermally aged for coarsened
microstructure at 150 °C ........................................................................... 89

Figure 41. Accumulation of Pb-rich phase near anode for coarsened microstructure at
150°C ................................................................................................ 90

1. Introduction
1.1 Electromigration is a significant reliability concern

Electromigration is the movement of the conducting materials due to interactions
between conducting electrons and diffusing metallic atoms/ions. It was not a significant
issue before 19703 since the current density in these earlier electrical and electronic
systems were well below the electromigration limit [1]. Due to the continuous
miniaturization of electronic devices, the current density reaches a level that is large
enough to make electromigration a significant issue. In an example of a VLSI process,
where integrated circuits are created on a single chip, thousands of transistor circuits are
connected with Cu thin-film line in dimension that are about 0.5 pm wide and 0.2 pm
thick. In such a circuit, a current as small as 1m Amps can result in current density of

about 10° A/cmz.

In direct current (DC) stressing, the atomic movement due to electromigration will
cause void formation at the cathode and extrusion at the anode. These defects result in
failure of these thin film integrated circuits [2]. The continuous miniaturization of
electronic device results in smaller interconnects and a high current density imposed on
these interconnects can cause the electromigration, such an event affects the electrical

and mechanical reliability of solder joint interconnects.

The approach in quantifying effect of electromigration in electronic devise is the
mean time to failure [3]. However, for the material science points-of-views, the mean
time to failure due to electromigration depends on factors such as alloy composition,

current density, service temperature and grain structure [4]. In this study, to isolate the

relationship between coarsened microstructures and electromigration behavior, eutectic
Pb-Sn solder joints were isothermally aged to different extents, leading to different

microstructure prior to the application of current stressing.

In the following section, the mechanism and phenomena of electromigration will
be described briefly. Issues in accelerated electromigration of solder joint are presented
below in order to reveal the links to various factors influencing electromigration. Then
problems existing in the currently popular joint configuration will be discussed in order
to give a clear picture that justifies development of a new solder joint design. The
complexity of microstructure after electromigration needs a systematic analysis method
to quantify and compare the results. Finally, the motivation and aim of this study will be

presented.

1.2 Electromigration behavior

i. Phenomenon

Under high current density, electromigration-induced damages will occur. The
corresponding microstructural changes include certain phenomena, such as valley/hillock
formation in the solder region [5], significant phase segregation of Pb—rich and Sn-rich
domains [6], asymmetric growth of Cu-Sn intermetallic layers at the interfaces [7], and

excessive depletion of Cu at the cathode side [8].
a. Hillock and Valley formation due to electromigration

Electromigration is a serious concern in electronics under high current density
and results in manifestation of visible surface features such as hillock and valley
formation. This phenomenon occurs when electrons move to cathode while metallic ions

2

move to the anode during the passage of electric current. Then, such movements result in
extrusions at the anode and voids at the cathode, which one call hillock and valley
formation at the cathode and at the anode, respectively. In other words, electromigration
behavior will change a flat solder surface into a buckled surface, which indicates the

cathode side has sunk in and the anode side has bumped out.

According to Lee et a1. [10], by using confocal laser scanning microscopy
(CLSM), the height of the hillock and the depth of the valley can be measured. It is worth
noting that the valley appears to be deeper than the height of hillock. It is because the
valley at the cathode results from the migration of atoms away from the cathode. These
atoms movement will tend to push the materials toward the free surface available at the
anode, then the hillock forms. In other words, the hillock formation is a result from the

blockage of atomic movements.

b. Significant phase segregation of Pb-rich and Sn-rich domains

According to Ho et al. [5], after current stressing with high cm'rent density,
there will be a dense layer of a Pb-rich domain at the anode end of the solder. Adjacent to
the Pb-rich layer, towards the solder will be a Sn-rich layer and a random mixture of Sn-
rich, Pb-rich, and Cu6Sn5 phase, in sequence. Such segregation implies that the lowest
energy to make this two-equilibrium phases stable can be reached only when the Pb-rich
and Sn-rich domain exist as two separate domains with a single continuous interface
boundary. So, under high current density, the significant two-layer segregation of Pb-rich

and Sn-rich domain will be observed not only because the diffusivity of Pb and Sn under

high current density is different, but also the lowest energy can be reached only when this

two-layer segregation of Pb-rich and Sn-rich phase result.

c. Asymmetric growth of Cu-Sn intermetallic layers at the interfaces

The essential process in solder joining is the chemical reaction between Cu and

Sn to form intermetallic compounds:

6Cu+5 SIT-*CU6SI'15
9Cu+Cu68n5—>5Cu3Sn

After reflow, the scallop-type Cu6Sn5 compounds form at solder/Cu interface.
And, a very thin layer of Cu3Sn may also form between the Cu68n5 scallop/Cu interfaces
in the initial state. However, after current stressing both of Cu6$n5 and Cu3Sn transformed

into layer-typed.

It is worth noting that if the joint was subjected to current stressing, massive
transportation of electrons move from cathode to anode resulting in the enhancement of
IMC growth at the anode, and inhibition of the IMC grth at the cathode. Therefore, at
the anode, the total IMC layer always will be much thicker than that at the cathode. This

is called the polarity effect of electromigration on IMC growth [7].
(1. Excessive depletion of Cu at the cathode side

In addition, the transformation of 1 molecule of Cu68n5 into 2 molecules of
Cu3Sn will leave behind 3 molecules of Sn {Cu6Sn5——>2(Cu3$n)+3Sn}, which will attract
9 atoms of Cu to form 3 more molecules of Cu3Sn. Therefore, the vacancy flux needed to

transport the Cu atoms will accumulate at the Cu/Cu3Sn to form Kirkendall voids. In

other words, the Cu at the cathode side will be consumed excessively. Voids are
undesirable in device application; therefore it is of reliability interest to limit the growth

of Cu3Sn and reduce the depletion of Cu at the cathode.

In order to understand the fundamental failure mechanism resulting from
electromigration, in-situ electromigration tests were performed by Lee et al. [10]. At the
cathode side of the solder bump, the interfacial crack formed at the entry of the chip side
between IMC layer and the solder. As mentioned in the previous paragraph, due to
excessive depletion of Cu at the cathode, Cu was quickly consumed, followed by void
formation at the contact area. The void reduces the contact area and displaces the
electrical path, causing the current crowding and Joule heating inside the solder bump. A
large joule heating inside the solder bumps can cause melting of the solder bump and the
failure occurs quickly. Then, the interfacial crack propagates along the whole IMC
layer/solder interface with increasing time resulting in a void formation along IMC

layer/solder interface. Once a void is nucleated, electron flow will be interrupted.

ii. Quantitative evaluation for mean time to failure

Electromigration is associated with atomic migration resulting from electron wind. It
is a serious reliability problem in electronics encountering high current densities. Such
migration quite often results in manifestations of visible surface features and
microstructural evolution in the interior region. At certain length of time, such migration
will cause failure of the electronic device, which is called the mean time to failure
(MTTF). Since the late 19605, Black’s equation [11] provided to predict current density
and temperature for the mean time to failure of aluminum conductors due to

electromigration.

other words, the Cu at the cathode side will be consumed excessively. Voids are
undesirable in device application; therefore it is of reliability interest to limit the growth

of Cu3Sn and reduce the depletion of Cu at the cathode.

In order to understand the fundamental failure mechanism resulting from
electromigration, in-situ electromigration tests were performed by Lee et al. [10]. At the
cathode side of the solder bump, the interfacial crack formed at the entry of the chip side
between IMC layer and the solder. As mentioned in the previous paragraph, due to
excessive depletion of Cu at the cathode, Cu was quickly consumed, followed by void
formation at the contact area. The void reduces the contact area and displaces the
electrical path, causing the current crowding and Joule heating inside the solder bump. A
large joule heating inside the solder bumps can cause melting of the solder bump and the
failure occurs quickly. Then, the interfacial crack propagates along the whole IMC
layer/solder interface with increasing time resulting in a void formation along IMC

layer/solder interface. Once a void is nucleated, electron flow will be interrupted.

ii. Quantitative evaluation for mean time to failure

Electromigration is associated with atomic migration resulting from electron wind. It
is a serious reliability problem in electronics encountering high current densities. Such
migration quite often results in manifestations of visible surface features and
microstructural evolution in the interior region. At certain length of time, such migration
will cause failure of the electronic device, which is called the mean time to failure
(MTTF). Since the late 19605, Black’s equation [11] provided to predict current density
and temperature for the mean time to failure of aluminum conductors due to

electromigration.

MTTF=A - 1'" - exp(EA/ KT)
where,
A is the proportionality constant, J is the current density, n is the exponent on current
density, E A is the activation energy of the electromigration failure mechanism, T is
temperature and K is Boltzmann constant.

It is noteworthy that the MTTF of solder bump decreases exponentially as the
stressing temperature increase. Wu et al. [12] conducted electromigration tests on eutectic
PbSn solder bumps, and found that the MTTF decreased from 711 hours to 84 hours
when the testing temperature increase from 125°C to 150°C with an applied current
density of 5.0><103 Amp/cmz.

Additionally, according to Nah et al.[13], the MTTF of solder bump decreased as
the current density increased. The composite solder joint used in their study did not fail
after one month of current stressing at 4.07 X104 Amp/cmz, but they failed after 10 hrs of
current stressing at 4.58 X104 Amp/cmz. At a slightly higher current stressing of 5x 104
Amp/cmz, these joints failed after only 0.6 hrs by the melting of the composite solder
bumps.

These findings indicate that electromigration must be strongly dependent on the
temperature and current density.
1.3 Issues in accelerated electromigration of solder bump

i. Effect of current density on electromigration

The current density is required to be larger than a critical current density to initiate
and advance the atom/ion migration within a reasonable time. Besides, higher current

density leads to faster electromigration. As discussed in the previous section, atomic

migration may not only occur by bulk diffusion, but also through dislocations, grain
boundaries, and external surface. Although these “short circuit” diffusions are faster than
bulk diffusion, they are insignificant in most situations because the cross-sectional areas
of these paths are extremely small. However, as the continuous miniaturization of solder
joint and the associated increases in current density, the effect of short circuit diffusion
cannot be neglected. It is because the path for electromigration in these miniaturized

devices is not only through lattice diffusion, but also along grain boundary and/or surface

diffusion.

Chan et. al [15] classify the current density into three ranges: low, moderate, and
high current density. Low current density refers to 10°-10l Amp/cmz. In the moderate
range, 102-103 Amp/cm2 is the experimental condition. High current density refers to 104-
105 Amp/cmz. Most of the electromigration studies are conducted in an accelerated
manner where a high current density (104 to 105 Amp/cmz) and a high temperature more
than 100°C. On the other hand, with a moderate current density at 125°C, no significant

electromigration was noted.

Thus, it was confirmed that no obvious electromigration occurred with the moderate
current density. However, above a critical current density of 104 Amp/cmz, significant

electromigration behavior occurs.
ii. Effect of temperature on electromigration

Temperature plays an important role in electromigration. It is because
electromigration occurs by an atomic diffusion mechanism. Atomic migration may not

only occur by bulk diffusion, but also through dislocations, grain boundaries, and

external surface. These are sometimes called “short circuit” diffusion paths inasmuch
with rates much faster than for bulk diffusion. Based on the mode of diffusion,
Huntington [15] classifies the electromigration phenomenon into three types, A-C,
depending on the homologous temperature (THET/TM where TM is the melting
temperature of alloy) as shown Table 1. If the conducting line is kept at a very low
temperature (e.g., liquid nitrogen temperature), electromigration cannot occur because
there is no atomic mobility of diffusion, even though there is a driving force. However, if
the temperature increases at homogenous temperature below 0.5 for example, only grain
boundary is the principal electromigration path. Moreover, if temperature increases to
423K (150°C), solder (TM=456°K) typically operates well above TH=0.5 in Type A
diffusion where lattice and grain boundary diffusion domains. In other words, the

diffusion rate increases causing significant electromigration.

Thus, higher temperature accelerates the electromigration behavior. On the other

hand, it is difficult to investigate electromigration behavior at lower temperatures.

Table 1. Hungtington’s diffusion regimes [15].

 

 

 

Homologous . .
Type Diffuswn mode
temperature (TH)
A TH>O.5 Lattice and grain boundary
B 0.3311605 Grain boundary with some trans-boundary diffusion

 

C 0.3<TH Grain boundary only

 

 

 

 

 

iii. Effect of microstructure on electromigration

Numerous studies have been carried out to investigate the effect of microstructure on
electromigration in Al interconnects and Cu interconnects. However, only few studies
focus electromigration in solder joint, such as those using eutectic Pb-Sn solder. The
eutectic PbSn solder is a two-phase alloy. The microstructure of the alloy present in the
solder joint region depends on the alloy composition and the fabrication of process used
to make the solder joints.

a. Alloy composition

According to Liu et. al, [16] at ambient temperature imposition of a current
density of 105 Amp/cmz, eutectic Pb-Sn solder can result in very fast failure rate due to
electromigration in this alloy composition. It is because this eutectic alloy has the lowest
melting point and a highest density of lamella structure. Changing the composition of the
alloy by increasing percentage of Sn to produce hyper or hypo condition reduce the
failure rate due to electromigration. In other words, the highest density of lamella
structure with the highest density of interfacial boundaries results in the highest rate of
electromigration behavior. Thus, it is believed that grain boundary diffusion is the main
kinetic path of atomic transport in electromigration at ambient temperature. Moreover,
as the alloy composition changes, the electromigration rate changes. Starting from pure
Sn, the electromigration rate decreases when Pb is added to it; however, it increases as
the alloy approaches the eutectic composition. At the eutectic point, solder joint has the
highest electromigration rate. Then it decreases again when alloy composition moves
towards Pb. Therefore, electromigration rate is strongly dependent on the microstructure

of solder joint.

b. Microstructure coarsening during annealing

As the solder joint is isothermally aged, the microstructure of solder joint will
be coarsened. According to Jung et al. [17], the eutectic coarsening kinetics for eutectic

PbSn solder joints annealed at 50°C to 150°C was determined to be of the form:
D" — D—o"= K0 0 t - exp(-Q/RT)
where,

D is the mean linear intercept phase size, D—o is the as-reflowed phase size, n=4.] 1:015,
Q=39.8:O.8kJ/mole, K0=1*10'23-4.5*10'23,t is time, T is temperature, and R is
Boltzmann constant.

From the above equation, the phase size is strongly dependent on the time and
temperature. In other words, by controlling the temperature and time, the phase size can
also be controlled. Microstructural coarsening not only decreases the grain boundary of
solder alloy joints, but also reduces the diffusion path available in the solder joint, and
slows down the hillock and valley formation.

According to Yang et al. [18], it was found that pre-aging at 170°C was able to
increase the electromigration lifetime of flip-chip Sn-Ag solder joints. It was believed
that solid-state aging may form thicker intermetallic compound (IMC) layers and this

may be able to increase the electromigration resistance.

However, in order to gain a better understanding of electromigration behavior as
well as diffusion paths during electromigration, this study was focused on the effect of

coarsened microstructure by using our newly designed solder joint to reduce unnecessary

10

issues, such as thermomigration and evaluate the electromigration behavior without

additional complications.

1.4 Solder joint to reduce current crowding and thermomigration issue

In order to study electromigration behavior, there is a need to develop a new
design for solder joint. It is because using the currently popular joint configuration it is
hard to carry out studies that isolate critical events that occur mainly due to the
electromigration. This joint configuration minimized the roles of other issues, such as

“current crowding” and thermomigration.

1. Current crowding issue

Current crowding is serious issue in the failure mode of electromigration in flip
chip solder joints. As the electron flows in the thin film and approaches the edge of a
solder bump, electric current will take lowest resistance path and jam at the entrance into
the solder bump, resulting in current crowding. In other words, the electric field, E rotates
to accomplish this resulting in current crowding. Especially in the line to bump geometry
of a flip-chip solder joint, as the current changes direction from line to bump, there is a
very large current density change at the contact. It is because the cross-section of the line
on the chip side is two orders of magnitude smaller than that in solder bump. Thus, as the
same current passing through them, the current density on the chip will be at least two
orders of magnitude lager than that of solder bump [19]. Since the rate of
electromigration damage is roughly proportional to the square of the current density, void

nucleation will occur at the via edge [1].

11

According to Tu et al. [20], there are two significant effects due to the current
crowding. First, there is an abrupt change in current density before and beyond the point
where current turns into the bump. Second, the current density in the solder bump near
the entrance point will be about one order magnitude higher than the average current
density in the middle of the bump. It will be 105 Amp/cm2 near the entrance when the
average current density in the middle of the bump is 104 Amp/cmz.

Thus, it is current crowding or the high current density at the entrance point of
electric current that will lead to electromigration, not the average current density in the
bulk of the joint. 80, it is our intent to investigate if the role of current density in the
entire solder joint where the average current density influences the electromigration
behavior.

ii. Thermomigration issue

Due to Joule heating, electromigration may cause a non-uniform temperature
distribution in a flip chip solder joint. So, there may be a component of thermomigration
in any electromigration experiment. In other words, electromigration in a flip chip solder
joint is accompanied by thermomigration when a large current density is applied or when

the current distribution is non-uniform because of current crowding.

Thermomigration is a mass transport caused by a large temperature gradient. If an
initially homogeneous two-component alloy is placed within a temperature gradient,
above a critical level, mass diffusion can cause disintegration of the components. Then,
the components will move from the hot side to the cold side, and as a result the hot region
becomes depleted in that component. This effect is known as the Soret effect in fluids.

This resulting diffusion process is known as thermomigration or thermal diffusion [21].

12

According to Basaran et al.[22], there isa large temperature gradient across nano-
electronics and power electronics solder joint, especially Si based devices that operate at
high temperature. It is because high current density in power electronic devices leads to
thermal gradient due to local Joule heating. These increased current density and
temperature gradients are new major reliability concerns, especially for the solder joints

of an electronic package.

Ye et al. [23] have shown that when the thermal gradient is large enough the
thermomigration can be the dominant migration process, even larger than
electromigration. According to Chen et al. [24], a hot spot was found in the vicinity of the
entrance point of the Al interconnect, which is detrimental to the electromigration life
time of the solder joints. According to Nah et al. [25], this hot spot was the most
vulnerable part in the solder joint, since it may experience much larger electron wind
force due to the higher current density and the higher diffusivity owing to the higher
temperature. Hence, instead of being influenced by electromigration, the formation of

voids and the cause of failure in solder joints may result from thermomigration.

iii. New design of solder joint to reduce current crowding issue and
thermomigration issue
To understand fundamentals of these complex interaction, and to investigate
failure mechanisms during electromigration, there is a need to developed a new design of
solder joint to focus on the atomic movement purely due to electromigration. Fortunately,
a new design has been developed by our group which will efficiently reduce current
crowding and thermomigration during electromigration study. The followings are the

advantages of this new design of solder joint. First, the geometry of solder joint is

13

symmetrical. Thus, the effect of current crowding will be efficiently reduced. Second,
although there are still some current crowding region in the solder joint, the extent of
current crowding regions can be controlled by appropriate minor modification in
dimensions. Third, there is no serious joule heating due to Cu pillars and substrate has an
excellent thermal conductivity. As a result, there was no thermal gradient from the
cathode to anode of the joint. In other words, this joint configuration will avoid the
influence of thermal migration. Last but not least, it is easy to adjust the size and change
the alloy composition of the solder joint [26].

By using this joint configuration, several factors will influence the
electromigration has already been identified, such as effect of current stressing time,
effect of thickness, and effect of direct current (DC) and alternating current (AC).

a. Effect of current stressing time

According to Lee et al.[27], after three days of continuous current stressing with a
current density of 104 Amp/cm2 at 150 °C, significant hillock/valley formation while Pb-
rich and Sn-rich start to segregate into two-layer morphology in the 25-um joint.
Additionally, prolonging the current stressing for four more days causes this two-layer
morphology formation completely.

b. Effect of thickness of solder joint [27]

The time required to achieve the two-layer morphology in the thicker (75-pin) joint
is longer than in the thinner (ZS-um) joint.

c. Effect of DC and AC

According to Ho et al.[28], if the joint was stressed by DC, electromigration induced

damage occurred, such as valley/hillock formation, two-layer segregation of Pb-rich and

14

Sn-rich domains, and asymmetric growth of Cu-Sn intermetallic layers. On the other
hand, no significant electromigration was observed after the AC treatments, especially for
the treatment with AC frequency higher than 1 hour". The dependence of the damage on
AC frequency suggests that electromigration in solder joints can be inhibited to a large
extent when a proper reverse current is delivered.
1.5 Analysis method

There are numerous analysis methods used to investigate electromigration behavior.
For example, Scanning Electron Microscopy (SEM), Optical Microscopy (OM),
Transmission Electron Microscopy (TEM) [30], Energy Dispersive X-ray spectroscopy
(EDX) [31] have already been widely used to evaluate the microstructure evolution of
solder joint, and to analyze the composition at various locations in the solder joints.

However, few studies were focused on quantitatively analyzing the result. One of
the currently popular ways to further discuss the electromigration behavior is to calculate
the effective charge number 2*. According to Huntington et al.,[20] the drift velocity is
taken as

VdED/KT-Z‘oq-j-p ,

where
Vd is drift velocity, D is diffusivity, Z* is effective charge number, T is temperature,
q is the charge of an electron, j is current density, and p is resistivity
This indicates that if considers the self-electromigration of the noble metal and
measure the drift velocity, knowing the diffusivity, D, the effective charge number can be

calculated.

15

However, for a two-phase alloy, one must first have a detailed study of the
microstructure before studying its electromigration. Without knowing the diffusion
species and paths in the microstructure, the calculation of effective charge number is not
necessarily accurate. [30]

Thus, instead of calculating the effective charge number, a systematic method
should be developed to characterize the result. In this study, due to the large contrast for
the Pb-rich and Sn—rich regions for the eutectic Pb—Sn solder joints under optical
microscope (OM), it is easy to distinguish these two regions. In this way, by investigating
the microstructure change of Pb-rich and Sn-rich regions after electromigration, the
diffusion paths and diffusion species can be inferred.

1.6 Motivation

i. Grain coarsening

Grain coarsening effects plays an important role in electromigration related
phenomenon. Isotherrnally aged sample joint will result in the change in the grain size of
solder joint. So, by isothermally aging the solder joints to different extents of time, one

can control the size of the grain as well as control the electromigration rate.

Besides, grain size is an important factor for mass diffusion rate. The smaller the
grain size is, the faster the diffusion is. It is because the net contribution of lattice and
grain boundary diffusion is the active electromigration paths for eutectic solder joint. As
the grain size changes, the effect of grain boundary diffusion on electromigration can be

evaluated.

16

ii. Diffusion mode and diffusion of atomic species at 100°C and 150°C

Diffusion mode and diffusion species under electromigration at 100°C and 150°C
are still unclear. It is generally believed that the mode of diffusion path at 100°C is lattice
diffusion and at 150°C is through grain boundary. If this saying is true, as we change the
size of grain boundary, the electromigration phenomenon should remain the same at
100°C and change at 150°C. So, we can operate at two different temperatures and
compare the result of changing the different coarsened structure to see the effect of

coarsened microstructure on electromigration behavior.

Besides, the temperature of 100°C and 150°C are both very critical temperatures
for electronic industry. It is because the working temperature of Si devices is around
100°C and an aging at 150°C for 1000 hour is a required test in reliability specification
[15]. So, if one operates experiment at these two temperatures, the behavior of the solder

joint at these two temperatures can be further discussed.
iii. Aging study

In order to compare different extent of the coarsened microstructure, how many
isothermally aging days is also very important to decide the size of grain area. At lower
temperature (100°C), the coarsening time should be longer than at higher temperature
(150°C) to reach the same size of grain area. So, it is also interesting to investigate the

growth of grain size for the different coarsening time at different temperature.

17

Aim of this study

Following were the objective of this study:

a. To develop a consistent method to fabricate new design of solder joint, that has
less current crowding and thermomigration issues.

b. To gain a better understanding of material movement on electromigration
behavior by isothermally aging the eutectic Pb-Sn solder joint to achieve
different extents of coarsened microstructure prior to current stressing.

c. To evaluate the diffusion mode and species at 100°C and 150°C by comparing
the results of different coarsened microstructure subjected to current stressing
for the same length of time.

d. To understand the effect of coarsened microstructure on electromigration
behavior of eutectic Pb-Sn solder joint.

e. To conduct systematic analysis to characterize the average grain area of Pb-
rich phase and the concentration of Pb-rich domains to evaluate the
accumulation Pb-rich layer, and to analyze the area distribution of Pb-rich

domains.

18

2. EXPERIMENTAL PROCEDURE

2.1 General

In order to study electromigration behavior, there is a need to develop a new
design of solder joint. It is because using the currently popular joint configuration is
difficult to clearly characterize the electromigration behavior without interference from
other influences. The currently used joint configuration is always involved with other
issues, such as “current crowding” and the thermomigration. Due to the thin-thick
divergence, this configuration exists a region where local density can vary by two or
more orders of magnitude over a very short distance [32, 33, 34]. Additionally, this
“current crowding” region will cause a significant local Joule heating and result in the
local temperature differences [35, 36]. This temperature difference between the chip side
and the board side will cause visible thermomigration from chip side to the board side. As
the result, the influence of thermomigration cannot be ignored.

Therefore, in order to minimize the current crowding, as well as the local Joule
heating, the current path in this joint needs to be straight. Moreover, in order to eliminate
the thermomigration issue, the heat generated in the solder joint should be dissipated as
soon as possible. Based on these two concepts, a new design of solder joint has been
developed and is shown in Figure 1. Instead of using the thin-thick divergence
configuration, the geometry of the joint in this new design is symmetrical and straight to
reduce “current crowding.” Also, the joint should be made by lots of copper to facilitate
good heat conductivity. Thus, the heat generated will be dissipated easily.

According to the results by Ho et al. [37], more than 80 volume percent of solder

in this new design of joint has current densities in-between a range of 15% of targeted

19

 

 

l
.l

 

 

Solder mask

Figure l. Schematics drawing showing the solder joint configuration used in
this study

Note: 6 indicates direction of current stressing

20

value. During the 5.9 Amps current stressing, the temperature in the solder joint is only
raised to about 2.5 (i2)°C above ambient environment temperature over a range of 30-
150°C.Unlike in circuit of solder bump flip-chip joints, no serious heat generation occurs
in the solder present in this new design of joint configuration. This suggests that there
was no significant thermal gradient during current stressing.

However, this new design of solder joint was not easily fabricated. During the
process of fabrication, any slight movement will result in significant uncertainty in the
solder joint, such as cracks or uneven thickness of the solder joint. As long as there are
cracks in the solder joint, these will significantly influence the events that occur during
current stressing. Moreover, uneven thickness of solder joint will result in different
current density in the solder joint. Therefore, the method of fabricating this new design of
solder joint needs to be carefully controlled.

2.2 Sample preparation
i. Material used in this study

a) Solder ball

The solder ball used in this study was eutectic 63 Sn3 7Pb prepared by Kinsus.
b) Solder mask

The solder mask used for mounting in this study was prepared by mixing the heat
hardening solder mask and the hardener with a weight percentage of 9:1 respectively.
c) Cu pillar

The Cu pillar is 1.2 mm diameter and the tips of Cu pillars were metallographically

polished before soldering.

21

(1) Cu substrate package

The Cu substrate package included a 1200 um diameter Cu pillar, a 500 um thickness
Cu pad, and a 70 um thickness solder mask. As shown in Figure 2a, a solder mask with
an opening of 460 um was attached on the topside of Cu pad, and a Cu pillar was
soldered on the backside to Cu pad by using eutectic SnAgCu paste in order to conduct
the electric current. The use of eutectic SnAgCu assured that backside of this package
will not fall off in the soldering process due to its higher melting point of 217°C.
e) Solder paste

A surface that is oily or oxidized will prevent the solder from spreading during
soldering. This soldering paste with brand named Taiyo Soldering Accessory was used to
remove such oil or oxides and to improve wettability while soldering.

ii. Equipment used in this study
a. Equipment for alignment
In order to recreate the new design proposed by Ho et al. [37], the equipment

set-up for alignment shown in Figure 3 was used. The Jack was used for controlling the
z-axis movement impositions between the inserting Cu pillar and the Cu substrate
package. X-Y axis controller was used for aligning the inserting Cu pillar with the solder
ball on the Cu substrate package. In addition, an aluminum stage with a 1.28 mm-
diameter hole was connected on the X-Y axis controller to ensure vertical alignment. In
order to heat up this aluminum stage as well as Cu pillar while soldering, a heat belt was

surrounded on the aluminum stage.

22

Solder mask

70pm
500pm

 

 

(b)

70pm
500nm

 

 

 

 

Figure 2. Schematics drawing showing the Cu substrate package (a) side
view and (b) top view.

23

x-y axis controller Heat belt Aluminum stage

 

 

Steal block
Heat plate
1 ~ . \
Temperature . || Temperature
controller 2 axus contro er controller

Figure 3. Equipment used for alignment the parts for the soldering
operation.

24

b. Heat plate

A steal block was placed on the heat plate to ensure constant temperature
during sample fabrication.

c. Temperature controller

Two parts were needed to maintain the temperature that was controlling by a
temperature controller. One is connected to the aluminum stage on the equipment for
alignment and the other, is connected to the hot plate. The temperature was maintained at
220°C while soldering.

d. Aluminum stage

For the purpose of placing Cu substrate package, an aluminum stage with a
hole of 1 cm-diarneter was used to hold the Cu substrate package while soldering. The
use of aluminum stage assured that the entire Cu substrate package was heated to 190°C
while the hot plate temperature was controlled at 220°C

e. Power supplier

For current stressing, a well regulated constant voltage/ constant current
supply with brand named Model 1796 high current regulated DC. power supply was
used. It delivered 0-16 V at 0-50 Amps and could be adjusted continuously throughout
the output range.

iii. Joining Protocol

Schematics are provided in Figure 4 to illustrate the details corresponding to the

steps described in the following methods.

Prior to the soldering, Cu pillars were cleaned by nitric acid Solution (30%) and

water, and then fluxed. First, the cleaned and fluxed Cu pillar was placed into the

25

.Sv waive—om “can can A8 waive—om 28cm
game.“ .528 05 mo nouaotnfl 2: mega? magma mouafionom .v 95“....—

 

 

 

 

 

 

 

 

 

 
   

 

 

 

 

 

 

 

..oEom
.3520
23-5
_<_:o__< 2 30—2

 

 

 

 

 

3v 3

26

equipment for alignment heated up to 220 (i5)°C, and then aligned with the opening of
solder mask. Second, Cu substrate package was then placed on the aluminum stage
heated to 220 (i5)°C. At this temperature, a fluxed solder ball was then placed in the
opening of solder mask. After these two steps, the aligned Cu pillar shown in Figure 4a
was then inserted and squeezed to the Cu substrate package to see the excessive solder
drained from the pad and wetted the inserted pillar. Last, after heating for 40 see, the as-

reflowed solder joint was then cooling down by airflow for another 40 see.

In order to investigate interior microstructure of solder joints, the joints were heat-
mounted in the solder mask and metallographically polished to over one-half of the joint
diameter before applying current. Based on the known diameter of the solder joint, the
current stressing area was calculated. In other words, current density was able to calculate
by dividing the current passing through the joints by the contact area of current stressing.
Figure 5 shows the overview of the polished solder joint used in this study. The dot line
in the middle region indicated the region of observation. The image of initial
microstructure of as-reflowed solder joint is shown in Figure 6. However, if the solder
joint was cooling down by spraying water on the connection of Cu pillar and Cu substrate

package, the microstructure is different as shown in Figure 7

Prior to current stressing, in order to observe the effect of coarsened
microstructure during current stressing, solder joints were isothermally aged in a
temperature-controlled, convection air oven for a required condition to cause different

extents of coarsened microstructure.

27

 

Figure 5. The overview of the joint used in this study.

28

.583 ,3 838 Emom $28 326:2-8 mo ogogmocomfi 2:. .b «...—um.”—

r: .....fi
; 1.. .....al

kh.x..; ..c 3‘.“ an»

LMA» .fldwwu

 

29

These isothermally aged solder joints were then subjected to current stressing for a
required time by placing them into a temperature-controlled, convection air oven set at a
required temperature. The details of all required experimental conditions will be
described in the next section. For current stressing, a constant electrical current of 4.9
Amps was passed through the solder joints, which generated a global average current

density of 104 Amps/cm2 within the solder joint.

After current stressing for a known period of time, surface features of solder joints
were observed by using optical microscopy (OM) and a JEOL 6400 scanning electron
microscopy (SEM) operated at 20 keV. In order to characterize the image, image analysis
software ImageJ was used to automatically identify the phase region boundary, and
measure phase area and average diameter. By setting the threshold in ImageJ, the Pb-rich
region boundary could be detected, and the average area could be measured. The Pb-rich

phase region is outlined by the software as shown in Figure 8.
2.3 Date collection and data analysis
i. Experimental protocol
Basically, this experiment was performed at two temperatures, 100°C and 150°C.

At 100°C, the solder joints were isothermally aged for 12 days and 15 days. Then,
these isothermally aged solder joints were subjected to current stressing for 20 days at
100°C; and compared to those noted in as-reflowed solder joints subjected to current

stressing at 100°C.

30

3 a .52... Bass. 3 e858 .... 3 a .34.... SEE a 2%...—

 

 

f... at, 1...... fetal:
Q... ...mw .
0...... 4?... Q... $ng Km

 

 

$1$¢§>Lr.fly ..

....A

I.
. ....

I

. .. 2.3.2.... 4...... outwit. .11.... .. i. ...... ....
... fir fl... ...:
an iv...“- ahafit wiflrmm 5.5.3:

 

.W...

.....va

31

At 150°C, the solder joints were isothermally aged for 3 days and 6 days. Then, these
isothermally aged solder joints were subjected to current stressing for 6 days at 150°C;
and compared to those noted in as-reflowed solder joints subjected to current stressing at

150°C.

Figure 9 shows the flow chart of experimental protocol.

ii. Methods of characterization

In order to characterize the distribution of this two-phase structure, grain area for
each Pb-rich phase was measured in this experiment since digital image processing
sofiware ImageJ was employed. The area of individual Pb-rich domain, A5, on the solder

surface is measured for every Pb-rich phase region.

The average size of Pb-rich domain, <Apb>, was determined by:

 

Area occupied by Pb-rich phases _ 2111 A;

<APb>= . .
Number of Pb-r1ch domarn pr

where pr is the total number of Pb-rich domain within the chosen area of interest. The

area fraction of Pb-rich phase, Pb (%), is determined by:

Area occupied by Pb-rich of the specific layer

 

Pb(%)= x 100(%)

Observed specific layer

For the purpose of comparing the size distribution of Pb-rich phase, the segregation of
Pb-rich phase and Sn-rich phase, and the accumulation of Pb-rich layer, three types of

analyzed were developed to characterize these solder joints.

32

..ooouoa advantage mo incisofi .a 9:43,.—

 

 

comtmano w mzflmom _ _ comtmano w mzsmom

 

 

 

é , g

 

 

0.86 .523... .3 5E oooormo .5035... .2 55,
9.7.8me. Eotzo mEmmobw E950

 

 

 

 

 

: .
oooo.© -
nom< .mEEoEofl

 

 

Doom F©
nomad. 5:558.

 

 

 

 

 

 

E5. 523 poser—9-2

 

 

33

a. Size distribution of Pb-rich phase within solder joint

As shown in F igurelO, an area of constant width (350 um) and length (30pm)
within solder joint are chosen on the solder image. By plotting the size distribution of Pb-
rich phase, not only mean size of Pb-rich phases but also standard deviation for the size
distribution could be evaluated. In this way, the average size of Pb-rich phases resulting
from different experimental condition could be compared. In addition to compare the
mean size of Pb-rich phase, uniformity of the area distribution of Pb-rich phase could be
evaluated by comparing the standard deviation of size distribution of Pb-rich phase
within solder joint.

b. Area fraction of Pb-rich phase
a) Between two Cu substrate

In order to evaluate the distribution of Pb-rich area toward anode and cathode, a
constant width of 220 pm in the center of solder joint was chosen. As shown in Figure 11,
we increase the length from 1 to 35 um towards anode and cathode. In this way, a layer
by layer analysis of area fraction of Pb-rich domain from the cathode to anode can be
plotted.
b) Anode accumulation
Since the accumulation of Pb-rich layer was not flat, the value of the thickness of

Pb-rich accumulation layer was difficult to measure. As shown in Figure 12, in order to
focus on the accumulation of Pb-rich phase as well as the thickness of Pb-rich
accumulation, an enlarged image was used as shown in Fig. 8b. Besides, in order to see
the change in the fraction of Pb-rich phase while increasing the distance away from anode,

we fix a constant baseline with width of 120 um and change the length of chosen area

34

from 1 to 30 um. Comparing the fraction of Pb-rich phase from each chosen region and
drawing the chart of Pb-rich phase fraction vs. the distance away from the cathode, the

area fraction of Pb-rich phase near anode will be discussed.

35

omega scuba Mo 83356 cum com “50.4 $20... 2: no 33 33030 .3 «...—ur—

_ . . . . .
. l . I. '0 , \W. . u 01..
...1111n1.1.1.»1.1.~.9111flllw..

. $518.83

11 1.11-11.11.1‘H1.41.11_.n1- L
1.4:; vn9.~*0wdfl.1rcn-Caw 34.

.3...

 

36

. .568
3 2:8 30:30 2: 80¢ 33m 893mm. .9» 853 goth; mo 8.88m 05 8a Ego a 5.3. .3

5.5 :2on 58:0 :80 82m 03% £3.35 no 888... 05 3.39.80 .0858 98 3.2a 3.339
E: mm 8 fl 80¢ 83 58:... mo 59.2 2: 38.85. 98 E: omm mo 523 5.3 0:: “8:8 05

xi .88 33030 mo 830E womeEm SV .38 33030 A3 0.62 E 93w 0 8m mammmobm €0.58
8 3.8.3:... was was. m How 3:838 8.2. .mmmbanw 08% £2.25 mo 5on we}. .5 Paar—

...... .. ...N..1.1 3241‘; .J1.... 1&1 . s ... 3N9. de... ... ._ .. .. .... Yum-”.... ..
Elmm -. .

f . .

w .
. flu... ”...... cum...

.9.
.. ..9... firmUofimO

.u)

 

 

 

37

.8580 00 00.00 000500

80.0 830 008380 .m> 08:0 £0033 mo 800.000 05 .8.“ 8:0 0 0B .0088 800050

£000 80.0 0083 noflbm m0 00003 06 8880800 .X .3 000—88 .81 om 00 _ 80¢ 008
800000 .80 580— 05 0.8088 08 81 ofi .«0 5005 505 082089 08 in .008 003030
00 New 0082 00000.30 .3 0080:00 0W8: By .008 003030 m0 N cm 0080— 008030
.3 0080200 0w08_ A3 Doom_ 8 080 0 000 8700.3 808.0 00 00000350 08 080

m 000 0088800 883 .0008 80: 08308 085.0 500.00 .80 80008.“ 00?. . N~ 008E

.. .. . .1. .7 . 1'3. . . 1,. f v. u... . N MHQ‘ . ~.. ..5 F
. .. .. .. . . \. . ...‘w . . . 1 _ .1
1i... .. ...... $0. ...... .. .. . .....L a ......2. ...

     

......

    

      

    

  

.....w

   

5%nW fix ...“..Mdb. . . . 1 ....... .
.1 8&5. .& 911w .MEH1.E.F.: _. .....nyr .....N... in. $81.1. 0.11.1114?» mc__®mmm WWW”...

g.. .. 1.5 x. .. .. 4 . 1 . u ..m . 1 . : .101. . .. «4.1.11.1 4011.111 I'll ...... ...-W121hflb. 1‘1 x.
O 1 . . . . .
I. a 1 . .

Enema .

         

111..
$

.7 ..
.

o
. I
1‘

n .

    
 

\
s

 

...-H-

81. x:

 

I

 

  

.‘

 

 

38

3. RESULTS AND DISCUSSIONS
3.1. Aging studying

In this section, the results for 4 different as—reflowed solder joints that
experienced different experimental conditions will be reported. Among these 4 solder
joints, two were isothermally aged at 100°C for 12 and 15 days, respectively; the other
two were isothermally aged at 150°C for 3 and 6 days, respectively. The structure of the
chapter will be organized into 3 parts: first, the clear images of optical micrograph and
the thickness of IMC layers will be demonstrate for each experimental conditions;
followed by an analysis of the size distribution of Pb-rich phase within solder region; and

alternate analysis of Pb-rich phase will also be provided in the last part of the chapter.

i. Aging at 100°C
3. Thickness of intermetallic compounds for solder/Cu substrate

interface

Microstructural features of the as-reflowed solder joint are shown in Figure 1.
A scallop-shaped of Cu68n5 layer with an average thickness of about 1 um at both sides
is presented in this figure. The as-reflowed solder joint exhibits the classic interlamellar
eutectic structure of Pb-rich and Sn-rich phases that are clearly visible as dark and white

domains, respectively, in this optical micrograph.

Optical micrographs of thejoints coarsened for 12 and 15 days at 100°C are
shown in Figure 2 and Figure 3, respectively. Morphologies of Pb-rich and Sn-rich

regions in these micrographs indicate that both phases coarsen from isothermal aging.

39

 

Despite the extent of aging, the observation provided in Figure 2 and 3 show that the Cu-
Sn interface IMC layers at both sides in both joints (coarsened for 12 and 15 days) are

about 2 mm.
b. Size distribution of Pb-rich phase within the solder joint

For the purpose of comparing results of joints that experienced for different
isothermal aging conditions, the best way is to compare the size distribution of Pb-rich
phase. The size distribution of Pb-rich phases show not only the area distribution of Pb-
rich phases, but also the mean size of Pb-rich phase and standard deviation of these Pb-
rich phases. The size distribution of Pb-rich domain in intermediate region of solder joint
calculated by image analysis software ImageJ is shown in Figure 4. The results showed
that when the solder joint is isothermally aged for longer periods of time, the average size
of Pb-rich domain increased. In addition, for Pb-rich phases smaller than 2 pmz, their
normalized count of Pb-rich phase was obviously decreased after solder joints was
isothermal aged, indicating that some individual Pb—rich phase region was coarsened, not

the exchange of smaller ones.
c. Area fraction of 'Pb-rich phases between the two Cu substrate

In order to observe the area fraction of Pb-rich phases during isothermal aging,
a layer-by layer analysis from one end of the interface to the other end was carried out.
Observation of Pb-rich phase occupied in each selected region (220*2 umz) layer by
layer away from the interface is shown in Figure 5a and 5b. As the plot indicates, the area

fraction of Pb-rich phase forjoints isothermally aged 12 and 15 days are similar. The

40

percentage of each Pb-rich area was held constant around 30% except at the end regions
of joints. In other words, isothermally aging samples in general does not change the area

fraction of Pb-rich phase in a given region of the joint.

41

 

Figure 13a. The image of as-reflowed solder joint microstructure. The
enlarged picture of regions A and region B are shown in Figure lb.

 

Figure 13b. In regions A and B, IMC layer are both less than 1pm thick.

42

, navy-«..- ,. . _
, . Cu-S

'-~a— ‘-~‘

. _ r ‘ ‘ ' ‘ -
' 7' I,
, a . ¢
' . ' _ ” ' ..{‘§. 3”,.
, ,o - ‘
. ' ,, ,‘J‘ .- g3,
' e 47' . , .
.n - .
. v . ,. ‘ . ,‘, L
. 4 , . ‘ 'T. .
. - - ‘ _ . '- . '7 . L .3 y e y '0‘.
u - , .. n ' .
‘ 1 ‘p‘ . ' _ ‘ - I I f-f-J‘: 'v.‘ \

       

Figure 14a. The image of solder joint isothermally aged at 100 °C for 12
days. The enlarged picture of regions A and region B are shown in
Figure 2b.

 

1‘1 5pm

Figure 14b. In region A and B, IMC layer in both ends about 2 pm thick.

43

 

Figure 15a. The image of solder joint isothermally aged at 100 °C for 15
days. The enlarged picture of region A and B are shown in Figure 3b.

 

 
       

B 2A- ’45-; F. staff‘?

Figure 15b. In regions A and B, IMC layer in both ends about 2 pm thick.

(a) As-reflowed

 

 

Normalizede-rich phase

 

 

- . 2 l
30% T—"‘* 22- 2.2“] Mean SIZE. l3.3(urn L _ ___J
50% I; ___,Sthev :10. 985 m ._
40% ,5 Total counts. E0212, g * |
3. ' l
20% _ <100 counts: P00“? 1
0% i 54:}: "'_.’_T_T'I7 TI I I I I I I IT I I I I FT IfTI I I T T] ITI
40 50 60 7O 80 90 100
size (umz)
(b) 12 days of isothermal aging at 100 °C

 

 

80% T~ —- , 3+

L Mean size: b.2(pm2)
50% It _ I ‘ I I- sIdDev: {12 .058

40% i . I Total counts F73

<1OOCounts: E372

 

 

20%

Normalizede-rich phase

0%

 

40 50 60 7O 8O 90 100
size (umz)

(c) 15 days of isothermal aging at 100 °C

 

 

 

 

 

 

 

Q)

3; 80% ""— miMean size: ‘14. 8(pm2) l?“
c WW ___2 z _z __________
Ig’ 50% _ I Sthev: 36.208

13 40% 1, ' i Total counts. 167 I
"Z“ 20% <100 counts 153 :I—
2 0% -t‘ { Mv—T T ,,,,,,,, T—rT—rT—r—f—T—rr-Hfi—g—I—m—fl—T—fi—fl

20 30 40 50 60 70 80 90 100
size (umz)

Figure 16. Size distribution of Pb-rich phase within solder joints (a) as-
reflowed (b) 12 days of isothermal aging at 100 °C (c) 15 days of
isothermal aging at 100 °C.

45

(a) 12 days of isothermal aging at 100 °C

1 00%
90%
80%
70% *—

60% ’
50% X /

40% k y

30% W

20% ‘9 ’ .

10%
0% I I I I I I I

0 10 20 30 40 50 60 70

 

 

 

 

 

 

 

 

 

 

Area fraction of Pb-rich phase(%)

Midpoint of given layer from the interface (pm)

(b) 15 days of isothermal aging at 100 °C
100%
90%
80%
70%
60% 9

50% g f
40% . Q

30% I . . .

20% ” ’

10%

0% I j— I I I r 1 fl

0 10 20 30 40 50 60 70 80

 

 

 

 

 

 

 

 

 

Area fraction of Pb-rlch phase(%)

 

Midpoint of given layer from the interface (pm)

Figure 17. Area fraction of Pb-rich phase (a) Joint isothermally aged for 12
days at 100°C(b) Joint isothermally aged for 15 days at 100°C.

46

ii. Aging at 150°C
a. Thickness of intermetallic compounds layer at solder/Cu substrate

interface

Optical micrographs of the joint coarsened for 3 and 6 days at 150°C are
shown in Figure 6 and Figure 7, respectively. Morphologies of Pb-rich and Sn-rich
regions in these micrographs indicate that both phases coarsen significantly from
isothermal aging. However, the thicknesses of Cu-Sn interfacial IMC layers at both sides
are significantly thicker than in joints coarsening at 100°C. Additionally, the total Cu-Sn
interfacial IMC layer can be divided into two regions (Cu68n5+Cu3Sn). The thicknesses
of Cu-Sn interfacial IMC layers at both sides of the joints coarsened for 3 days and 6
days at 150°C are about Sum and 6 pm thick, respectively. Besides, the thickness of

Cu68n5 layer is greater than the thickness of Cu3Sn layer.
b. Size distribution of Pb-rich phase within the solder joint

The size distribution of Pb-rich domains in interior region of solder joint
isothermal aged at 150°C is shown in F igure8. The results indicate that the aging
temperature will have positive influence on the mean size of Pb-rich domain. Compared
to joints isothermal aged at 100°C, coarsening of the phases is more significant at higher
temperature. In addition, while comparing the size distribution of Pb-rich phase for as-
reflowed, 3-days isothermal aged, and 6-days isothermal aged, one can note that
isothermally aging of solder joints caused the normalized counts of Pb-rich phases

smaller than 10 pm2 decreased, while larger ones coarsened.

47

c. Area fraction of Pb-rich phases in the interior region of the joint

The observation of Pb-rich phase occupied in each selected region (220*2 umz)
layer by layer away from the interface is shown in Figure 9. The concentration of Pb-rich
phase is about 30% in the middle region of solder joints and increases at both ends of
solder joints. It indicates that there is an accumulation of Pb-rich layer at both ends of

solder joints. However, this accumulation layer is not a continuous layer.

48

1 l. a. ,. .‘1‘ ‘
.. ‘ 3 .

V . '4. - l
. ‘I" ‘ H" '2' , . _ . s ‘
u - .

e I I ‘. " 1‘" r . - (o I
.- . .-

.-fs ,1! at ,9"? 7, .- I- ‘ .
. ' ' ‘ . I v 1 u

I . I .3 ,

- . ' i. '\ . v. . .

' ‘ Q I " "' "-'~ a

t .

II
”I

 

Figure 18a. The image of solder joint isothermally aged at 150 °C for 3
days. The enlarged picture of regions A and B are shown in Figure
6b.

 

Figure 18b. In regions A and B, IMC layers are both about 5 pm thick.

49

 

Figure 19a. The image of solder joint isothermally aged at 150 °C for 6
days. The enlarged picture of region s A and B are shown in
Figure 7b.

 

Figure 19b. ln regions A and B, IMC layers are both about 6 pm thick.

50

(a) As reflowed

 

 

 

 

 

 

 

 

 

 

 

3i .. .........................................
N l
‘5. 609’ IIIIIIII _ I I I Mean size: p.9(um2) Ii—
;§ 7.1 ; I
{I 40% A, d Sthev. 7.212 __
g g Total counts: 661 1
E 20% <100 counts: 560 [L—
2 ; - ..........................................
0% nITIrTrTrrTTlTII—rT—IIIIIIIITIIII
0 10 20 30 40 50 60 70 80 90 100
size (umz)
3 (b) 3 days isothermal aged at 150 °C
" I
g 50% I ' Mean size: (A 114. .3(um2) ”J“
-= ' l
a 40% : -— ~ ~ ~—~ ——v3___Sf‘?'D?Y _ I223- ._ z;—
E Total counts: 136 §
" 20% I ”I " I—
; <100 counts: i134 J
z 0% "' " ' “ . -H-HH-I-r"I—r-I—I-rI-H—I-I-I-I-w-I—rw—I-H—r—m—r'fifi—I
0 10 20 30 40 50 60 70 80 90 100
size (umz)
(c) 6 days isothermal aged at 150 °C
3 - __ fl Hm- _...____.. L
f: 60% 7 :Mean size. __03. 2(pm2) ;
g 40% . _ , ,, , _._ sthev: 60. 559 L
1. ~+ ——~:
g Total counts 1 L
,5 20% . III III III ‘III I
TE. <100 counts: 36 I
g 0% « -..-J34. ,
0 10 20 30 40 50 60 70 80 90 100
size (umz)

Figure 20. Size distribution of Pb-rich phase within solder joints (a) as-
reflowed (b) 3 days of isothermal aged at 150 °C (c) 6 days of
isothermal aged at 150 °C.

51

(a) As reflowed

 

 

 

 

 

 

 

 

 

 

 

31 - ..........................................
~ I
'5. 60% IIIIIIIII I III Mean size: 33.9mm?) 'I——
'5 i I ‘
t J ‘ ' I
Q 40% _‘l Sthev. 17'212 7*
g Total counts: I661
E 20% i ' <100 counts: 660 9'—
o . - ..........................................
z 0%;(‘I - ' I.{TlIjrilllTrTllIrrTlT1llIliTllTl
0 10 20 30 4O 50 60 70 80 90 100
size (umz)
a: (b) 3 days isothermal aged at 150 °C
.. I
:‘1 60% ‘ Mean size: I14. 3(um2) IL‘
0 " ‘ “ “— “'1
g 40% 7 a _ .- Stdoev: {22.3 _ I"
E .... Total counts: I136 I
To 20% " “ i—
g <100 counts: 5134 ;
z 0% —-I-n-
40 50 60 7O 80 90 100
size (umz)
(c) 6 days isothermal aged at 150 °C
3 1 - -11--- “I.
E 60% Mean size 33. 2(um2) .
E 40% . I Sthev: 50.559 I__
'8 Total counts: 91 '
,5 20% ---~ II —
E I» <100 counts: 86 J,
g l‘ ‘ . . *-H%PI-W
0 10 20 30 40 SO 60 70 80 90 100
size (umz)

Figure 20. Size distribution of Pb-rich phase within solder joints (a) as-
reflowed (b) 3 days of isothermal aged at 150 °C (c) 6 days of
isothermal aged at 150 0C.

51

(a) 3 days of isothermal aged at 150 °C

100%
90%
80%
70%

50%

Area fraction of Pb-rich phase(%)

100%

80%
70%
60%
50%

10%

Area fraction of Pb-rich phase(%)

60% ‘ _

40%
30%
20% -
10%
0% .

 

90% -

40% I
30% g
20% .

0% ‘

 

 

 

 

 

 

 

 

 

 

 

 

10 20 30 40 SO 60 7O
Midpoint of given layer from the interface (um)

(b) 6 days of isothermal aged at 150 °C

.z‘,

 

 

 

 

 

 

 

 

 

 

 

 

Midpoint of given layer from the interface (pm)

Figure 21. Area fraction of Pb-rich phases in the interior region of

solder(a) Joint isothermal aged for 3 days at 150°C(b) Joint
isothermal aged for 6 days at 150°C

52

d. Summary and discussion

Experimental research on the Pb phase coarsening in eutectic Pb-Sn solder
joint under isothermal aging is reported. Phase growth is observed under different
temperatures and times. Three interesting points should be discussed. First, longer aging
time results in larger grain size. Second, isothermal aging does not change the area
fraction of Pb. In other words, the distribution of Pb-rich phase and Sn-rich phase is
uniform. Third, higher temperature leads to faster grain coarsening and thicker

solder/substrate interface IMC layers.
a) More aging days results in larger grain size

Comparing the results of isothermally aging of solder joints with increasing days, it is

obvious to find out the grain size is strongly dependent on the time. The experimental

results can be explained by the equation 5" —- D—o"=Kt= K0 exp(-Q/RT)t, provided by
Jung et. al [1 6] where, 5 is the mean linear intercept phase size, D—o is the as-reflowed
phase size, n=4.] 10.15, Q=39.8i0.8kJ/mole, K0=1*10'23-4.5*10'23,t is time, T is
temperature, and R is Boltzmann constant.

As the initial phase size Do is fixed, the mean linear intercept of the grain size

depends on time. In other words, more aging days results in larger grain size.
b) Isothermal aging does not change the area fraction of Pb-rich phase

While comparing the results of solder joints aged for six days and eletromigrated for
six days, one can easily find out that isothermal aging does not change the area fraction
of Pb-rich phase. Isothermal aging only causes the enhancement of grain size. Unlike

53

current stressing, there is no driving force during isothermally aging. Thus, each

individual of Pb-rich phase coarsens due to by the self diffusion within solid.
c) Higher temperature leads to faster grain coarsening and thicker IMC layers

Comparing the results of isothermally aging of solder joints at lower temperature

(100°C) and at higher temperature (150°C), higher temperature leads to faster grain

coarsening and thicker IMC layers. According to Jung et al., 5 n — D—o"=Ko exp(-Q/RT)t,
where n=4.], Q=39.8kJ/mole, 140:1":10'23-4.5*10'23 m4/s R=8.31 J/K-mole, we can insert
T=100°C (373K) and 150 (423K). For constant time and constant initial phase size (Do),
the linear intercept phase size D at higher temperature is always higher than lower
temperature. Moreover, in solid-state aging, it grows a thicker layer of CU3SI] and Cu68n5
at solder/substrate interface at higher temperature. It is because higher temperature has
higher diffusion rate resulting in higher rate of Cu-Sn reactions. In addition, under such

condition the thickness of Cu6Sn5 is greater than the thickness of Cu3Sn.

54

3.2. Electromigration study

In this chapter, as-reflowed joints without subjected to current stressing with a current
density of 104 Amp/cm2 for 20 days at 100°C and for 6 days at 150°C, respectively will
be provided. The structure of this chapter will be organized into 3 parts: first, SEM
observations will be given to show the surface features, such as valley/hillock formation.
Second, for the purpose of revealing detailed microstructures at regions below the
valley/hillock formation, the specimens were metallographically polished to remove the
surface manifestations. Thus, a clear image of optical micrograph and the characteristic
of IMC layers will be demonstrated; followed by an analysis of the size distribution of
Pb-rich phase within solder region; and then another analysis of the area fraction of Pb-

rich phase will also be provided in the last part of this chapter.
i. Current stressing at 100°C for 20 days

Solder joints without isothermal aging exposed to current stressing with current
density of 104 Amp/cm2 for 20 days exhibited electromigration behavior result in the
surface features shown in Figure 10a and 10b. SEM observations illustrate that the valley
formed near the cathode and extruded a gentle hillock at the anode. Figure 10b, which is
a manifestation of the same feature by tilting the specimen by 45°, shows the contrast of
valley/hillock more clearly. The microstructure noted in interior regions below the hillock
and valley are presented in Figure 11. These orientation show that the morphologies of
both Pb-rich and Sn-rich regions were coarsened from current stressing. As the result
indicates, there was a dense layer of Pb-rich phase accumulated at the anode end of the

solder. The accumulation of Pb-rich layer was about 5 pm thick.

55

a. Characteristics of solder/substrate intermetallic compound layer

Optical micrographs of as-reflowed solder joints subjected to current stressing
for 20 days at 100°C are shown in Figure 11a and b. The thicknesses of Cu-Sn interfacial
IMC layers at both sides are thicker than initial microstructure. The thicknesses of Cu-Sn
interfacial IMC layers at both sides are about 2 pm. Besides, the thickness of Cu68n5 is
greater than the thickness of Cu3Sn. In the interior regions of the solder, far away from

the solder/substrate interface, Cu6$n5 particles were found near anode.

b. Size distribution of Pb-rich phase within the solder joint

At 100°C, the size distribution of Pb-rich domain in the interior region for the
joint that was not aged and subjected to current stressing for 20 days is shown in Figure
12. While comparing the size distribution of Pb-rich phase of as-reflowed solder joint
(Figure 4a) with joint subjected to current stressing for 20 days (Figure 12), one can be
observe that the mean size of Pb-rich area were coarsened after current stressing for 20
days. In addition, the standard deviation is significantly increased, indicating the sizes of

individual Pb-rich phase regions are significantly different from each other.

e. Area fraction of Pb-rich phase between two Cu substrates

Area fraction of Pb-rich phase in the interior region between two Cu
substrates for as-reflowed solder joint subjected to current stressing at 100°C is shown in
Figure 13. The trendline remains at about 30% until a distance of 55 pm away from

cathode is reached, where it increased significantly. Hence, two-layer segregation of Pb-

56

rich and Sn-rich becomes obvious. Besides, there is a continuous Pb-rich layer at distance

of 63 um away from cathode.

57

30pmx

 

Hillock

Figure 22a. The overview SEM image of as-reflowed solder joint subjected
to cun‘cnt stressing for 20 days at 100 0C.

 

Figure 22b. As-reflowed solder joint subjected to current stressing for 20 days
at 100 “C. Inside the dotted area is the tilted image of Figure 10a.

58

WW“;

.7 ' .e . ., ' ' 3.. . I VI!” -1" 3.3““, . h" I ‘.'u§:?*€u,‘ .75". .5. III-I‘ _‘
. ' 1' I ‘ \ 1. . ’n’ilffv'. I ' -_; I ;‘;' . , .
m s- «lever-- B. '.‘ was '~ w r: . . -
’. 2 '. _ . . . _- - | . ~ I. ' I
. ~43“! , - 2 ' _ . '2: gsofiitjljgje. i'» ti 'r .7 +' ”e. "

 

Figure 23a. The image of as-reflowed solder joint subjected to current
stressing for 20 days at 100 0C. The Pb-rich accumulation layer is
about 5 pm thick. The enlarged picture of regions A and B are shown
in Figure llb.

A . ...Mf ;‘

 

A - IleI all“!

..,f Swn f.

l2pm

 

Figure 23b. In region s A and B, IMC layers are both about 2 pm thick.

59

 

50%

 

40%

 

Mean size: 10.2(pm2)

 

30%

20%

10%

Normalized Pb-rich phase

0%

‘7 .—J.

Sthev: 20.69

 

 

 

 

LTotal counts: 212

 

 

 

 

L <100 counts: ,208

_l_ _ __—

 

10 20 30 40 50 60 70 80 90 100
size (umz)

Figure 24. Size distribution of Pb-rich phase in as-reflowed solder joint
subjected to current stressing for 20 days at 100 °C.

100%
§ 90%
8 80%
I:
g 70%
‘9’ 60%
Q
9; 50%
2 40%
O
'3 30%
«E 20%
g 10%
0%

 

 

 

 

 

 

10 20 30 40 SO 60 7O 80
Midpolnt of glven layer from the cathode (um)

Figure 25. Area fraction of Pb-rich phase for as-reflowed solder joint

subjected to current stressing for 20 days at 100°C.

50%

 

 

40%

 

FMean size: I 10.2(pm2l

 

30%

20%

10%

Normalized Pb-rich phase

0%

Figure 24. Size distribution of Pb-rich phase in as-reflowed solder joint

 

 

Jam _ ”29.-:59.
Total counts: 212

 

 

 

 

R9? _

 

 

 

10 20 30 40 SO 60 7O 80
size (umz)

subjected to current stressing for 20 days at 100 °C.

100%
.\° 90%
a
a 80%
S
f; 70%
0
'c 60%
A
a. 50%
3
c 40%
0
=3 30%
g 20%
(I
g 10%

0%

Figure 25. Area fraction of Pb—rich phase for as-reflowed solder joint

00 000

 

0 10 20 30 40 50 60
Midpoint of given layer from the cathode (um)

subjected to current stressing for 20 days at 100°C.

60

70

90

100

80

ii. Current stressing at 150 °C for 6 days

Solder joints without isothermal aging exposed to current stressing with a current
density of 104 Amp/cm2 for 6 days at 150°C exhibited the surface feature shown in
Figures 14a and 14b. These SEM observations illustrate the valley formed near the
cathode with a gentle hillock formed near the anode. Figure 14b is view tilted by 45° for
illustrate the contrast of valley/hillock more clearly. Compared to those stressed at 100°C
for 20 days, the valley/hillock formation is more significant when current stressing at
150°C is carried out. The microstructure in interior regions of this specimen from current
stressing is presented in Figure 15. It illustrates the coarsening of Pb-rich and Sn-rich
regions from current stressing. As can be seen, there is a dense layer of Pb-rich phase
accumulated at the anode end of the solder joint. The accumulation of Pb-rich layer was
about 10 pm thick. In the interior regions of the solder, far away from the solder/substrate

interface, Cu68n5 particles were also found.
a. Characteristic of intermetallic compound

Optical micrographs of as-reflowed solder joints subjected to current stressing
for 6 days at 150°C are given in Figure 15b and 150. The thicknesses of Cu—Sn interfacial
lMC layers at both sides are thicker than initial microstructure and are asymmetric. The
thickness of llMC layer at anode is 2 um greater than at cathode. Besides, the total Cu-Sn
interfacial IMC layer can be divided into two regions (Cu68n5+Cu38n). At anode, the
thickness of Cu68n5 is greater than the thickness of Cu3Sn. However, at cathode, the

thickness of CU3Sn is greater than the thickness of Cu6Sn5.

61

b. Size distribution of Pb-rich phase within the solder joint

Size distribution of Pb-rich phase for as-reflowed solder joint subjected to
current stressing for 6 days at 150°C is shown in Figure 16. The mean size of Pb-rich
phase was increased from 3.6 pm2 to 55.7 pmz, which indicates the Pb-rich phase has
significantly coarsened from current stressing. Besides, the standard deviation of Pb-rich
phase was increased from 10 to 100, which indicates the size of each individual Pb-rich
phase is different from each other. In other words, with the driving force resulting from
current stressing, each individual Pb-rich phase was not only coarsened but also

combined together.
c. Area fraction of Pb-rich phase between two Cu substrate

Area fraction of Pb-rich phase between two Cu substrates is shown in Figure
17. There are two interesting points that should be mentioned. First, the accumulation of
Pb-rich layer near anode is significant. A continuous layer of Pb-rich phase is observed at
a distance of 55 pm to 60 pm away from the cathode. Second, the trendline of area
fraction of Pb-rich phase increases significantly, which indicates the segregation of Pb-
rich and Sn-rich is significant. However, the situation would become distinctly different
near the cathode. No continuous layer of Pb-rich phase or segregation of Pb-rich phase

and Sn-rich phase were found near the cathode.

62

,9.

¥

 

Figure 26a. The overview SEM image of as-reflowed solder joint
subjected to current stressing for 6 days at 150 °C.

‘r ‘c .I I "J '

     

Figure 26b. As-reflowed solder joint subjected to current stressing for 6
days at 150 °C. Inside the dotted area is the tilted image of Figure
10a.

63

IM C layer Pb-rlch layer

 

Figure 27a. The image of as-reflowed solder joint subjected to current
stressing for 6 days at 150 °C. The Pb-rich accumulation layer is
about 10 pm thick. The enlarged picture of regions A and B are
shown in Figure 15b.

CU3SI’1 > cussns

‘I . Q
.... _ ' _b I
a .. ‘ .

   
 
     

’ "W. . - 14%

\

f

as "

   

 

   

. . swim-.1
. , 5pm
‘3. :‘

CU3 S“ < CU58n5
Figure 27b. ln region 3,A the [MC layer 18 about 4 pm thick. The thickness
of Cu3Sn layer 18 thicker than Cu6Sn5 layer. In region B, the IMC
layer is about 6 pm thick. The thickness of Cu3Sn layer is thinner
than C u(,SnS layer.

 

64

Normalizede-rich phase

Area fraction of Pb-rich phase(%)

60%

40%

20%

As-reflowed solder joint subjected to 6 days of current stressing at 150 °C

 

 

Mean size: 55.7(pm2)

 

 

 

   

Std Dev: 100.007 _

Total counts: 49

>—- _ '_ __—._ __ *—__._—_—._ __—

 

 

0%

   
  

<100 counts:

 

 

 

 

 

llI1IlIIlIII

10 20 30 40 50 60 70 80 90 100
size (umz)

Figure 28. Size distribution of Pb-rich phase for joint without coarsening

subjected to current stressing for 6 days at 150 °C.

As-reflowed solder joint subjected to 6 days of

1 00%
90%
80%
70%
60%
50%
40%
30%
20%
1 0%

0%

current stressing at 150 °C

0

 

0 10 20 30 40 50 60 70
Midpoint of given layer from the cathode (um)

Figure 29. Area fraction of Pb-rich phase for as-reflowed joint subjected to

current stressing for 6 days at 150°C

65

d. Summary and discussion

Electromigration study on as-reflowed solder joints under two different
temperature of 100°C and 150°C was reported in this section. Regardless of temperature,
three interesting points were noted to occur due to electromigration. First, current
stressing results in large accumulation of Pb-rich layer near the anode. Second, the area
fraction of Pb-rich phase near the anode is more than near the cathode. Third,
solder/substrate interface IMC layer is thicker at the anode than at the cathode. While
comparing two different current stressing temperatures (100°C and 150°C), the
segregation of Pb-rich phase and Sn-rich phase under higher temperature is more

significant.
a) Current stressing results in large accumulation of Pb-rich phase at the anode

After electromigration with current density of 104 Amp/cmz, the interwoven lamellar
eutectic structure nearly translated into a two layer structure of Pb-rich phase and Sn-rich
phase near anode. Because Pb does not react with Cu, this segregation implies that with a
driving force of current stressing, Pb is driven in the same direction of the electron flow

towards anode resulting in accumulation at the anode.
b) The percentage of Pb-rich domain at the anode is higher than the cathode

Afier electromigration with a current density of 104 Amp/cmz, Pb is driven in the
same direction of electron flow from the cathode to the anode. So, unlike in the aging

study, the area fraction of Pb-rich phase near the anode after electromigration is always

66

more than that near the cathode. In other words, atomic movement resulted from

electromigration causes the uneven distribution of Pb-rich phases.
c) IMC layer is thicker at the anode than at the cathode

After electromigration with a current density of IO4 Amp/cmz, asymmetry in the
grth of Cu-Sn interface IMC layer is observed. The Cu-Sn IMC layer is thicker at the
anode than at the cathode. It is because Sn is driven in the same direction of electron flow
and reacts with Cu. So, the growth of both Cu3Sn and Cu68n5 at the anode would be
enhanced. Other excessive Sn atoms are squeezed out from the anode by compressive
stresses as a consequence of electromigration [8]. Besides, the thickness of Cu3Sn is
greater than Cu68n5 near cathode. It is because Sn is driven away from the cathode. Thus,
lack of Sn to react with Cu. In other words, present of more Cu than Sn results in the

thicker Cu3Sn layer than the Cu6Sn5 layer.

(1) Higher temperature leads to more significant electromigration behavior

Electromigration is a diffusion-control mechanism. So, it follows the rule of D=Do

exp(-Q/ RT). As the temperature increases, the diffusivity increases. Thus, the atomic
movement during electromigration is faster at higher temperature. However, diffusion
mode and diffusion species under electromigration at 100°C and 150°C are still unclear. It
is generally believed that the mode of diffusion path at 100°C is lattice diffusion and at
150°C is through grain boundary. If this saying is true, as we change the size of grain
boundary, the electromigration phenomenon should remain the same at 100°C and change
at 150°C. So, in the next chapter, the results of coarser microstructure after current

stressing will be provided to see the effect of coarsened microstructure.

67

3.3. Effect of coarsened microstructure on electromigration

In this section, joints isothermally aged to different extents of microstructure
coarsening were subjected to current stressing with current density of 104 Amp/cm2 for 20
days at 100°C and for 6 days at 150°C. The structure of the chapter will be organized into
3 parts: first, SEM observations will be given to show surface features, such as the
valley/hillock formation resulting from electromigration. Second, optical micrographs
and the characteristic of IMC layers resulting from electromigration will be presented;
followed by an analysis of the size distribution of Pb-rich phase within solder region.
Another analysis of area fraction of Pb-rich phase will also be provided to compare with

those noted in as-reflowed solder joints in the last section of this chapter.
i. Current stressing on coarsened microstructure for 20 days at 100°C

Through SEM observations, the passage of high current density is given in Figure
18. It can be noted that no significant valley/hillock formation in the solder joint
isothermally aged for 15 days in Figure 18b. Figure 18b is tilted 45° view for showing the
contrast of the image. The dotted area is the same region as shown in Fig 18a. However, a
comparison with the surface geometry observation on as-reflowed specimen, the

undulation of the surface was not significant in the joint coarsened for 15 days.
a) Joint coarsened for 12 days at 100°C subjected to current stressing for 20 days

Feature in a solder joint coarsened for 12 days at 100°C followed by current stressing
for 20 days at 100°C is shown in Figure 19. There are several microstructural features

that are different from the observation on as-jointed specimen subjected to current

68

stressing. First, the size distribution of Pb-rich phases was more random. Second, the
accumulation of Pb-rich layer is about 3 pm in thickness, which is thinner than that
observed in as-reflowed solder joint. Third, the grain area size of Pb-rich domains were
more coarsened than as-reflowed joints subjected to current stressing due to the fact that
they had been coarsened for 12 days prior to current stressing. Last, fewer Cu68n5 were
found in the interior of the solder as compared to that in as-reflowed solder joint
subjected to current stressing, and the thickness of interfacial Cu-Sn layer at both

interface is about 2 pm thick.

b) Joint coarsened for 15 days at 100°C subjected to current stressing for 20 days

The interior microstructures evolution of the solder joint coarsened for 15 days
subjected to current stressing for 20 days at 100°C is shown in Figure 20. The results
indicate that the two-layer segregation of Pb—rich and Sn-rich region was hardly observed
because the Pb-rich and Sn-rich phases were randomly distributed. and the accumulation
of Pb-rich layer near anode was not as much as joints without coarsening subjected to
current stressing (Figure 11). The thickness of Pb-rich layer was about 2 pm. In addition,
there were less Cu68n5 phase observed inside the solder matrix and a fewer amount of

IMCs nucleated at both interfaces.

a. Characteristic of intermetallic compound

Thickness of IMC layer for different experimental conditions at 100°C was
shown in Table 1. Two interesting points should be mentioned. First, the thickness of

IMC layer is symmetric for the joint current stressed at 100°C. Whether the joints were

69

coarsened or not, the thickness of IMC layer at both sides are the same. Second, for as-
reflowed solder joints subjected to current stressing, the thickness of Cu68n5 layer near
cathode is smaller than the thickness of Cu3Sn layer. However, for solder joints aging
prior subjected to current stressing, the thickness of Cub-Sns layer near cathode is greater

than the thickness of Cu3Sn layer.
b. Size distribution of Pb-rich phase within the solder joint

Size distribution of Pb-rich phase within the solder joint for coarser
microstructure subjected to current stressing for 20 days at 100°C is shown in Figure 21.
Compared with the as-reflowed solder joint subjected to current stressing (Figure 11), the
average area of Pb-rich domain is more coarsened since the solder joint had experienced

a thermal annealing treatment for more days before current stressing.

c. Area fraction of Pb-rich phase in the interior region between two Cu

substrates

The observation of Pb-rich phase occupied in each selected region (220*2 pmz)
layer by layer away from the interface is shown in Figure 22a and 22b. While comparing
with samples that reacted isothermal aging only (Figure 5a-b), one can find that current
stressing does influence the area fraction of Pb-rich phase. There are few interesting
things to note. First, from the center towards cathode, the distribution of Pb-rich phase
remains constant at 30%. However, from the center towards anode, the distribution of Pb-
rich phase depends on whether the microstructure is coarsened or not. Isothermally aging

for longer period of time condition, more uniform distribution of Pb-rich phase is. For

70

example, in Figure 13 (joint without coarsening), there was a drop in Pb—rich phase
around 55 pm away from the interface, and a pile-up at 62 pm away fi'om the interface.
In other words, there is a significant segregation of Pb-rich phase and Sn-rich phase near
the anode. On the other hand, in Figure 22c (joint coarsened for 15 days), it is relatively
constant in the distribution of Pb-rich phase. There is no such segregation of Pb-rich

phase and Sn-rich phase since the joint was coarsened.

Additionally, in order to quantify the accumulation of P'b-rich phase near
anode, method described in experimental procedure 3.2 p.25 was used. By changing the
thickness of a selective region and comparing the area fraction of Pb-rich phase with each
selected region (120* 2 pm 2), one can draw a chart of the percentage of Pb-rich phases vs.
distance away from the cathode as shown in Figure 23. From this chart, the thickness of
Pb-rich layers and the segregation of Pb-rich phase and Sn-rich phase near anode can be
examined.

The percentage of Pb-rich phase for the joint without coarsening subjected
to current stressing remains at 30% to a distance of 55 um away from anode and this
exhibits an increasing trend. Because the trendline increased significantly, we know the
two-layer segregation of Pb-rich and Sn-rich phase is significant. On the other hand, as
the result indicates, the accumulation of Pb-rich layer for the joint coarsened for 15 days
subjected to current stressing, the trendline remains at 30% and gradually increased
without dropping at distance away from cathode about 60 pm. So, the two-layer
segregation is not significant in joint aged for 15 days because the slope of the trendline
is relatively small compared to the joint without coarsening. In addition, as the result

indicates, for fine microstructure, the area faction of Pb-rich phase almost reaches to
71

100%, which implies that there is a continuous layer of Pb—rich layer near anode. On the
other hand, for coarser microstructure, the maximum area fraction of Pb-rich phase is

around 80%. Thus, there is no continuous layer of Pb-rich layer near anode.

72

 

30pm

Figure 30a. The overview SEM image for Joint coarsened for 15 days at
100°C and subjected to current stressing for 20 days at 100°C.

 

Figure 30b. Joint coarsened for 15 days at 100°C subjected to current
stressing for 20 days at 100°C. Joint was tilted by 45°. The dot
area is the same region as Figure 18a No hillock/valley
formation was noted.

73

 

Figure 31a. The image of solder joint isothermally aged at 100 °C for 12 days
and subjected to current stressing for 20 days at 100 °C. The enlarged
picture of regions A and B are shown in Figure 19b.

 
 
 

i.
" d

. 1"; .. p .. s) _ .' . . .'
)il g i I lTi I 31 $1.19
0 F . / . ‘ '

0'

Figure 31b. In region s A and B, solder/substrate IMC layers are both
about 2 pm thick.

74

 

Figure 32a. The image of solder joint isothermally aged at 100 °C for 15 days
and subjected to current stressing for 20 days at 100 °C. The enlarged
picture of regions A and B are given in Figure 20b.

 

Figure 32b. In region s A and B, solder/substrate IMC layers are both
about 2 pm thick.

75

amnauxamg.

 

 

 

an...“ EA .uz_ ...u... 9.5358»?va

on...” onzu ..Nz— BA @5308 “KEYNF
....

L 45.an mag Lv . _v tog—.912

—

 

 

 

892 89:8
Ea... 8.8.. Aee<§§ E... mini €352...

 

 

§8.@e€eou AQSFQER.
5.22 5288

 

 

 

.0o 2: “a coma 586on E no.3— 02 mo 305—029 .N «San.

76

 

Normalized Pb-rich phase

Normalized Pb-rich phase

(a) 12 days of isothermal aging followed by 20 days
of current stress at 100 °C

 

45%
40% _ _

35% Mean size: 217.8(pm2)
309‘ Sthev: 134.418

25% 1
Total counts: 154
20% .

15% L <1OO counts: 145

10%
5%

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0% _ — _
Ij1lIII[JFIIIIWfiITIT71JIIIIYIIllj

0 10 20 30 40 50 60 70 80 90 100
size (umz)

(b) 15 days of isothermal aging followed by 20 days
of current stress at 100 °C
50%

:3: A] Mean size: j 1332(pm2)
35% ‘7 Sthev: 45.689
30% F T— ”_'i ’_“
25% — .
20% L <100 counts: 1106
15% --- _ ___ —
10%
5%
0%

 

 

 

 

 

 

——-‘~%

 

 

 

1 Total counts: 1111

 

 

 

 

 

 

 

 

0 10 20 30 4O 50 60 70 80 90 100
size (umz)

Figure 33. Size distribution of Pb-rich phase for (a) joint isothermally
aged for 12 days at 100 °C and subjected to current stressing for 20
days at 100 °C (b) joint isothermally aged for 15 days at 100 °C
and subjected to current stressing for 20 days at 100 °C.

77

90%

 

 

 

 

 

 

 

 

 

 

 

A 0
if; 80% f
D.
g 50% O /
“a 40% <> <>
5 0 AA 1; W l
3 3°” 00”) V O ‘70 ‘0 [>0
g 20% 0 00,.
u M
g 10% O
0% I T I I I I I l
0 10 20 30 40 SO 60 7O 8O

Midpoint of given layer from the cathode (um)

Figure 34a. Area fraction of Pb-rich phase for joint isothermally aged
for 12 days at 100 °C and subjected to current stressing for 20
days at 100 °C.

 

100%
=3, 90%
§ 80%
J: <>
5; 70%
1‘7.’ 60%
.0
9; 50%
2 40%
g° 30% 00

O
E . 000 o
I; 20/o
g 10%

0%

O 10 20 30 40 50 60 7O 80

Midpoint of given layer from the cathode (um)

Figure 34b. Area fraction of Pb-rich phase for joint isothermally aged for
15 days at 100 °C and subjected to current stressing for 20 days at
100 °C.

78

Area fraction of Pb-rich phase(%)

 

100% CD-

[A
90% ‘ ‘
9‘ Oil
80%

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I
/ “i i
70% [i
1"] A
60% ..I 4.
50% '
.. 73 I ‘
40% \ ; T
30% ° £\ {3.2. "T <7- }0’ : l
a
20% *2, lfizqeoe O ’ A C A
:\ D.. ‘0 I
10% \q} 1723
0% T VT D D T T fi
30 4o 50 60 7o 80

Midpoint of given layer away from the cathode (pm)

 

Poly. (As-reflowed) - - Poly. (12-days isothermal aging)
- o o - Poly. (15-days isothermal aging)

Figure 35. Accumulation of Pb-rich phase near anode using image taken
by 50X objective lenses . Comparison of specimens with extents
of coarsened microstructures subjected to current stressing for 20
days at 100 °C.

79

ii. Current stressing on specimen aged for 6 days at 150°C

SEM observations, the joint coarsened for 6 days subjected to current stressing for
6 days at 150°C is shown in Figure 24. As compared with the joint without isothermal
aging subjected to similar current stressing conditions (Figure 14), no serious
valley/hillock formation was found in Figure 24. This suggests that joint coarsening for 6

days at 150°C inhibits the formation of hillock/valley formation.

a) Joint coarsened for 3 days at 150°C subjected to current stressing for 6 days

The results of joint coarsened for 3 days at 150°C and subjected to current stressing
for 6 days at 150°C is shown in Figure 25. An accumulation of Pb-rich layer was clearly
observed near anode from current stressing. The thickness of Pb-rich layer was about 6
pm. As the result indicates, adjacent to the Pb-rich layer, towards the solder, was an Sn-
rich layer followed by a random mixture of Sn-rich and Pb-rich. The two-layer
segregation of Pb-rich and Sn-rich phases is not as obvious as Figure 16 (joint without .
coarsening). This suggests that area distribution of Pb-rich and Sn-rich phases is more
random if the joint is coarsened for 3 days at 150°C prior to current stressing.
Additionally, fewer Cubsns regions were found in the coarsened joint than in the joint
without coarsening. Last but not least, the interfacial Cu-Sn IMC layer at both interfaces

had a thickness of about 6 pm.

b) Joint coarsened for 6 days at 150°C subjected to current stressing for 6 days
The interior microstructures evolution of the solder joint coarsened for 6 days

subjected to current stressing for 6 days is shown in Figure 26. Comparing to the as-

80

reflowed solder joint subjected to current stressing (as shown in Figure 16),one can note
that the Pb-rich phases was more randomly distributed on the surface of solder. The
accumulation of Pb—rich layer in thickness is reduced to 5 pm in joint coarsened for 6
days at 150°C. As the result indicates, no significant two-layer segregation of Pb-rich and
Sn-rich layers was observed. Additionally, fewer Cu6Snj particles were observed within
the solder matrix and the interfacial Cu-Sn IMC layer at both interfaces had a thickness
of about 6 pm.

a. Characteristic of intermetallic compound

The thickness of solder/substrate interface IMC layers for different
experimental conditions at 150°C is shown in Table 2. Three interesting points should be
mentioned. First, the thickness of IMC layer is asymmetric for joint without isothermal
aging subjected to current stressing at 150°C. However, it is symmetric for pre-aged
solder joints. Second, for as-reflowed solder joints subjected to current stressing, the
thickness of Cu68n5 near cathode is smaller than the thickness of Cu3Sn near the cathode.
However, in the specimen with coarser microstructure subjected to current stressing, the
thickness of Cu68n5 layer near cathode is larger than the thickness of Cu3Sn layer near
the cathode. Last but not least, one can note that the overall thickness of solder/substrate

interface IMC layer result in 150°C is greater than in 100°C.
b. Size distribution of Pb-rich phase within the solder joint

Size distribution of Pb-rich phase within the solder joint for coarser
microstructure subjected to current stressing for 6 days at 150°C is shown in Figure 27.

Two interesting points should be noted here. First, the mean size of Pb-rich phases within

81

solder joint is coarsened due to current stressing 'and the extent of coarsening is reduced if
the joint was pre-aged. For example, in the as-reflowed solder joint the size of this phase
increased from 3.9 pm2 to 55.7 um2 and in sample that had 6-days isothermal aging
solder joint the increase is from 33.2 pm2 to 40.7 umz. In other words, the effect of
isothermal aging prior to current stressing hinders the grth of mean size of Pb-rich
phase while current stressing. Second, the difference in sized of each individual Pb—rich

phase after 6-days of current stressing is reduced with increased extents of coarsening.
c. Area fraction of Pb-rich phase between two Cu substrate

The results of area fraction of Pb-rich phase occupied in each selected region
(220*2 umz) layer by layer for joints exposed to current stressing at 150°C is shown
Figure17 and Figure 28a-b. The results indicate that after electromigration for 6 days at
150°C, whether the microstructure was coarsened or not, the concentration of Pb—rich
phase near anode is higher than near the cathode, which indicates Pb-rich phase was
accumulated at the anode. In other words, electromigration results in Pb—rich
accumulation near the anode. However, the segregation of Pb-rich phase and Sn-rich
phase is more significant for as-reflowed solder joint than isothermally aged. In Figure 17,
there is a drop at the position of 43 pm away from the cathode, which indicates Pb-rich
phase is hardly found in this layer. In other words, this layer is Sn-rich layer. On the other
hand, one cannot find any drop in coarsened microstructure as shown in Figure 28b. This
observation indicates the distribution of Pb-rich phase and Sn-rich phase is more random

after current stressing starting with coarser microstructure.

82

Additionally, for the purpose of focusing on quantification of Pb-rich phase
accumulation near anode, methods described in experimental procedure 3.2 p.25 was
used. The area fraction of Pb-rich phase vs. distance away from the cathode is shown in
Figure 29. Few interesting points should be noted. First, isothermal aging of solder joints
after current stressing shows less segregation of Pb-rich and Sn-rich than in as-reflowed
solder joints. As the result indicates, the area fraction of Pb-rich phase for specimens with
fine microstructure after current stressing is around 10 % in the intermediate region of
solder joints and 30% for coarser microstructure. This shows the area fiaction of Pb-rich
phase is relatively uniform for coarser microstructure subjected to current stressing.
Second, the thickness of Pb-rich accumulation layer for as-reflowed solder joints is
thicker than pre-aged solder joints. As the result indicates, at 50 pm away from the
cathode, the percentage of Pb-rich phases for the joint without coarsening remains at 95 %
for about 10 pm, which indicates that the thickness of Pb-rich accumulation is about 10
pm. On the other hand, for the joints coarsened for 6 days and subjected to current
stressing for 6 days, area fraction of Pb-rich phase is more than 95%. Besides, the
distance of high area fraction of Pb-rich is relatively shorter than as-reflowed solder
joints, indicating the Pb-rich accumulation layer is thinner. In addition, the area fraction
of Pb-rich phase reach 100% for as reflowed solder joints subjected to current stressing,
indicating that there is a continuous layer of Pb-rich phase near anode. However, no such

continuous layer was found for pre-aging solder joints.

83

1‘

Q-
. .‘Jv' ' .

«mm» m. ,werscmrtzenv» .r.~»~.«6»- «my...

a —’ .v \—
-

 

Figure 36a. The overview SEM image of joint coarsened for 6 days at 150°C
and subjected to current stressing for 6 days at 150°C.

 

Figure 36b. Joint coarsened for 6 days at 150°C and subjected to current
stressing for 6 days at 150°C. Joint was tilted by 45°. The dotted
area is the same region as in Figure 24a. No hillock/valley
formation noted.

84

em: ,3“; I}. 4i
9‘ '5

ii- iii Jr was

r:

 

Figure 37a. The image of solder joint isothermally aged at 150 °C for 3 days
subjected to current stressing for 6 days at 150 °C. The enlarged picture
of regions A and B are shown in Figure 25b.

Cu3Sn < Cu68n5

   

:}6um

Cu3Sn < Cu68n5

Figure 37b. In regions A and B, IMC layers are both about 6 pm thick. The
thickness of Cu3Sn layer is thinner than Cu6Sn5 layer in both
regions.

85

 

Figure 38a. The image of solder joint isothermally aged at 150 °C for 6
days and subjected to current stressing for 6 days at 150 °C. The
enlarged picture of regions A and B are shown in Figure 26b.

Cu3Sn < CUSSn5

 

i Q “ ‘7» 7 __ ,* - ” ‘

- . , y a l n
a
. ,

 

Cu3Sn < Cu6$n5

Figure 38b. In region s A and B, IMC layer are about 6 pm thick. The
thickness of Cu3Sn layer is thinner than Cu6Sn5 layer in both
regions.

86

£90930...

 

 

 

 

 

 

 

 

 

 

.o .o .m .m 3:338 at:
.o ..o .v .v 3.338%“:
.c Emnaovwocmsév .pv . ...v 39.2
Asawmwcewuom Egg Ea mcoaom €5< Sana
Accomp©a€vs §8.@m§é
5m 35‘ im— 980m

 

 

.9. ca 3 Ban SE89. 5 as. Us: do $0505 .m 2.3.

87

Normalized Pb-rich phase

Normalized Pb-rlch phase

(a) 3 days isothermal aging followed by 6 days current

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

    

stressing at 150 °C
60% ,
1 ' - T ' 2
50% ., Mean suze. E 8_1(pm)
Sthev: 67.472
40% ;
Total counts: 65
30% <100 counts: 55
20%
10%
0% r , er HEW: 1 1
0 10 20 30 40 SO 60 7O 80 90 100
size (umz)
(b) 6 days isothermal aging followed by 6 days current
stressing at 150 °C
60%
50%
40%
r- _______ _ . _ .
30% : Mean suze: 0.7(um2)
‘ Sthev: 1.237
20% , _
Total counts: 3 3
L.i.1.99..°.2"'?£?= F1

 

 

 

0 10 20 30 40 50 60 70 80 90 100
size (umz)

Figure 39. Size distribution of Pb-rich phase for (a) joint isothermally aged for

3 days at 150 °C and subjected to current stressing for 6 days at 150 °C
(b) joint isothermally aged for 6 days at 150 °C and subjected to current
stressing for 6 days at 150 °C.

88

 

100%

 

 

 

 

 

90% ()9
80%

70%

60% //\\
50% <>

 

40°/ A v
‘309; ‘(’;yL—7§<7<7ZS <><>C> <> //Vy <>
10% L0

0% l I I l I l I
10 20 30 40 SO 60 70

 

Area fraction of Pb-rich phase('/o)

 

 

0K

Midpoint of given layer from the cathode (um)

Figure 40a. Area fraction of Pb-rich phase for joint isothermally aged for 3
days at 150 °C and subjected to current stressing for 6 days at 150

 

 

 

 

 

 

 

 

 

 

 

 

°C.
100% (‘2
a? 90%
O
E 80%
g 70% 0
g 50% /4:zfi> / \
2 40% o
s 30% / <\O°<>m [)0 V
g 4; Q \ 40 Y
5* 10% H
0% , , I I T do s
0 10 20 30 4O 50 60 70

Midpoint of given layer from the cathode (pm)

Figure 40b. Area fraction of Pb—rich phase for joint isothermally aged for 6
days at 150°C and subjected to current stressing for 6 days at 150°C.

89

Area fraction of Pb-rich phase(%)

 

 

 

 

 

 

 

 

 

 

 

 

Midpoint of given layer away from the cathode

 

Poly. (As-reflowed) - - Poly. (3-days isothermal aging)
- 0 0 0 Poly. (6—days isothermal aging)

Figure 41. Pb-rich phase accumulation at anode region using image taken
by 50X objective lenses . Comparison of specimen with different

extents of coarsening subjected to current stressing for 6 days at
1 50°C.

90

(1. Summary and discussion

Most important of all, the effect of coarsened microstructure on
electromigration is investigated in this study. By controlling the annealing conditions,
solder joints with various extent of coarsened microstructures can be fabricated. As
compared with as-reflowed solder joint subjected to current stressing, several interesting
points are discussed. First, coarsened microstructure results in less Pb-rich accumulation
at the anode. Second, coarsened microstructure results in more uniform distribution of
Pb-rich phases between the cathode and the anode. Last but not least, coarsened
microstructure leads to more symmetric growth of Cu—Sn IMC layers at the interfaces.

a) Coarsened microstructure results in less Pb-rich accumulation at the anode

After electromigration with current density of 104 Amp/cm2 on the coarsened
microstructure, Pb is driven in the direction of electron flow from the cathode to the
anode. However, unlike the as-reflowed solder joints, coarsened Pb-rich phase has less

grain boundary between phases. So, it results in less Pb-rich accumulation at the anode.

b) Coarsened microstructure results in more uniform distribution of Pb-rich

phases between the cathode and the anode

The domains of Pb-rich and Sn-rich phase were still distributed randomly, except for
the fact that they had coarsened to different extents with time. The microstructural
evolution in the solder region looks as if the solder joints experience an isothermal aging

treatment without current stressing.

91

c) Coarsened microstructure leads to more symmetric growth of Cu-Sn IMC layers

at the interfaces

After electromigration with a current density of 104 Amp/cm2 on the specimen with
coarsened microstructure, not only Pb, but also Sn is driven in the same direction of
electron flow from the cathode to the anode. However, coarsened microstructure has less
grain boundary. So, less Sn is pushed to the anode as compared to the as-reflowed solder
joint. As a result, the growth of interface IMC layers at the anode is almost the same as
the growth of IMC layers at the cathode. The microstructural evolution at the interface
looks the same when the joints are subjected to isothermal aging treatment only. In other
words, the atomic movements during electromigration in specimens with coarser
microstructure in the solder region are not as fast as that in specimen with fine

microstructure.

92

4. SUMMARY

In summary, isothermal aging solder joints results in coarsened microstructure.
Coarsened microstructure has less interfacial area between phases. Regardless of
temperature, electromigration on specimen with more coarsened microstructure shows
less accumulation of Pb-rich phase at the anode. This suggests that the atomic movement
during current stressing at 100 °C and 150 °C is not only through lattice diffitsion but also
through grain boundary diffusion. Moreover, this finding also suggests that the
electromigration of Cu/eutectic Pb-Sn/Cu solder joints can be inhibited significantly by

appropriate pre-aging treatment of solder joints.

93

5. REFERENCES

1. Glenn A. Rinne, Microelectronics Reliability, 43, 2003; 1975
2. Tu K. N., Journal of applied physics, 94, 2003, 5451

3. Huntington H. B., Diffusion in solid: Recent Development, Nowick A. S. and
Bm'ton J. J. (Eds), Academic Press, New York, 1974, chapter 6

4. Glenn A. Rinne, Microelectronics Reliability, 43, 2003; 1980

5. Lee A., Ho C. E., Subramanian K. N., Journal of material research, 22(11),
2007;3268

6. Wu W. H., Peng S. P., Lin C. 8., Ho C. E., Journal of electronic material, 38,
10,2009; 2184

7. Gan H., Tu K. N., Journal of applied physics 97, 2005; 63514
8. Zhang L., Ou S., Huang J ., Tu K. N., Applied physics letters, 88, 2006; 12106

9. Lee A., Liu W., Ho C. E., Subramanian K. N., Journal of applied physics 102,
2007; 53507

10. Lee J. H., Lee Y. D., Park Y. B., Electronic components and Technology

conference, IEEE components, Packaging, and Manufacturing Technology
Society, 2007; 1436

11. Black J. R, Mass transport of aluminum by momentum exchange with conducting
electrons. In: Proc. 6th Ann. Int. Rel. Phys. Symp., 1967; 148

12. Wu J. D., Zeng P. J., Lee K., Chiu C. T., and Lee J. J., Proceeding of the 52th
Electronic components and Technology conference, IEEE components, Packaging,
and Manufacturing Technology Society, San Diego, 2002; 452

13. Nah J. W., Suh J. 0., Paik K. W., Tu K. N., IEEE 2005

14. Gupta D., Oberschmidt J. ML, Diffusion processes in lead based solders used in
microelectronic industry. In: Presented at Proceedings of Design and Reliability

of Solders and Solder Interconnection Symposium, Orlando, FL, USA, 1997; 59

15. Wu B. Y., Zhong H. W., Chan Chan Y. C., Alam M. 0., Journal of material
science, 17, 2006; 943

16. Liu C. Y., Chen C., and Tu K. N., Journal of applied physics, 88,10, 2000; 5703

94

17. Jung K., Conard H., Journal of electronic materials, 30, 10,2001; 1303

18. Yang P. C., Kuo C. C., Chen C., Journal of minerals, metals, and material society,
60, 6, 2008;77

19. Yeh Everett C. C., Choi W. J ., Tu K. N., Elenius P., Balkan H., Applied physic
letters, 80(4) 2002; 580

20. Tu K. N. , Solder joint technology, Springer, 2006; 13

21 . Huntington H. B., Difiitsion in solid: Recent Development, Nowick A. S. and
Burton J. J. (Eds), Academic Press, New York, 1974, chapter 6

22. Basaran C., Li Shidong, Abdulhamid M. F., Journal of applied physics, 103, 2008;
23520

23. Basaran C., Lin M., Ye H., Int. J. Solid struct. 40, 2003; 7315
24. Chiu S. H., Liang S. W., Chen C., Yao D. J. Liu Y. C., Chen K. H., Lin S. H.,
Electronic components and Technology conference, IEEE components, Packaging,

and Manufacturing Technology Society, 2006 ; 663

25. Nah J. W., Paik K. W., Suh J. 0., Tu K. N., Journal of applied physics, 94, 12,
2003; 7560

26. Ho C. E., Lee A., Subramanian K. N., Journal of materials science: Material in
electronics 18 (6)2007; 569

27. Lee A., Ho C. E., Subramanian K. N., Journal of material research, 22(11), 2007;
3265

28. Wu W. H., Peng S. P., Lin C. S., Ho C. B, Journal of electronic material, 38,
10,2009; 2184

29. Lee T. Y., Tu K. N., F rear D. R., Journal of applied physics, 90(9),2001; 4502

30. Chen K. C., Wu W. W., Liao C. N., Chen L. J ., Tu K. N., Science 321, 2008;
1066

31. C. Y. Liu, Chen C., Tu K. N., Journal of applied physics 88(10), 2000; 5708
32. Tu K. N., Journal of applied physics. 94, 2003; 5458

33. Yeh E.C., Choi W.J., Tu K.N., Elenius P., Balkan H., Applied physic letters,
80,2002; 580

95

34. Zhang L., Ou S., Huang J ., Tu K. N., Gee S., Nguyen L., Applied physio letters,
88, 2006; 12106
35. Ye H., Basaran C., Hopkin D. S., Applied physic letters, 82, 2003; 1045

36. Liang S. W., Chang Y. W., Chen C., Liu Y. C., Chen K. H., Lin S. H., Journal of
electronic materials, 35, 2006; 1647

37. Ho C. E., Lee A., Subramanian K. N., Journal of material science: Materials in
Electronics 18, 2007, 569

96