EL «0.. ..
m4 1rd»... ; ..

 

_ q
3.
fit

.I

.

.

. z

3%}...
_

x2. .

99:3...

...
3
.52... .2
.i. .
4.1;... 9W3
s .s. . z

.4 :ms 9.....lfin

L.
nun...

s so;

i. ._ n3

: ;

$.[yuril
2...

u . I}.
£7.19. r.

bani-.2; .: 3w.

tiff
.I..|

 

2th.. like"?

stint. . 1. 5L

.. .5 :3“. 2. .
:1 4..- .5

9 3. 23:5:33. ; .. ..
, rt 13:12.»...5- .3

1.2.1.5.... a .1

7.23.... danglvfinuirro.

 
 

 

 

       

 

.émmuwwnwmwu

. 191‘)” ~
rut- r/A...
3 :Svii . ,
. . . 1"... 20!!
. A
name.
2 I. . , . . ,

     

31",, I 90v.

in?
. ,, , . {2:4, lfivnhfiafifi‘ .. dump“.
. .kwur: §.§ . . z...» . I

  
   

  

        

    
 

  

v I. 5%.
28...... 10;. f1... . .
u’n . I . .

 
 

.

 

THESIS

2005K

 

I In". A nv
Mlcnigan mate
University

 

 

 

 

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:IClRC/DateDuo.p65-p.15

 

STUDIES ON IN-SI TU PARTICULATE REINFORCED
Sn-Ag COMPOSITE SOLDERS RELEVANT TO
THERMOMECHANICAL FATIGUE ISSUES

By

SUNGLAK CHOI

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY

Department of Chemical Engineering and Materials Science

2001

 

ABSTRACT

STUDIES ON IN-SI TU PARTICULATE REINFORCED
Sn-Ag COMPOSITE SOLDERS RELEVANT TO
THERMOMECHANICAL FATIGUE ISSUES

By

SUNGLAK CHOI

Global pressure based on environmental and health concerns regarding the use of
Pb-bearing solder has forced the electronics industry to develop Pb-free alternative
solders. Eutectic Sn-Ag solder has received much attention as a potential Pb-free
candidate to replace Sn-Pb solder. Since introduction of surface mount technology,
packaging density increased and the electronic devices became smaller. As a result,
solders in electronic modules are forced to function as a mechanical connection as well as
electrical contact. Solders are also exposed to very harsh service conditions such as
automotive under-the-hood and aerospace applications.

Solder joints experience thermomechanical fatigue, i.e. interaction of fatigue and
creep, during thermal cycling due to temperature fluctuation in service conditions.
Microstructural study on thermomechanical fatigue of the actual eutectic Sn-Ag and Sn-
4Ag-O.5Cu solder joints was performed to better understand deformation and damage
accumulation occurring during service. Incorporation of reinforcements has been
pursued to improve the mechanical and particularly thermomechanical behavior of
solders, and their service temperature capability. In-situ Sn-Ag composite solders were
developed by incorporating Cu68n5, Ni3Sn4, and FeSnz particulate reinforcements in the

eutectic Sn-Ag solder in an effort to enhance thermomechanical fatigue resistance. In-

situ composite solders were investigated on the growth of interfacial intermetallic layer
between solder and Cu substrate growth and creep properties.

Solder joints exhibited significant deformation and damage on free surface and
interior regions during thermomechanical fatigue. Cracks initiated on the free surface of
the solder joints and propagated toward interior regions near the substrate of the solder
joint. Crack grew along Sn grain boundaries by grain boundary sliding. There was
significant residual stress within the solder joint causing more damage. Presence of small
amount of Cu and Pb phase resulting from Sn—Pb finish on electronic component did not
affect thermomechanical fatigue deformation of the eutectic Sn-Ag solder joints.

In the aging study up to 5000 hours, solder matrix consisting of Sn phase and
Aggsn particles showed extensive coarsening. Intermetallic particles of Cu68n5 Ni3Sn4,
and FeSnz particles in the eutectic Sn-Ag composite solder joints were effective in
retarding the interfacial intermetallic layer growth at low temperature less than 150°C.
Ni3Sn4 and FeSnz particles were stable during aging, whereas Cu6Sn5 particles exhibited
coarsening.

In-situ composite solder joints exhibited different creep behavior depending on the
type of incorporated particle, compared to the eutectic Sn-Ag solder joints. Cu68n5 and
Ni3Sn4 reinforced composite solder joints showed weakening at room temperature creep,
whereas composite solder joints reinforced with FeSnz particles exhibited strengthening
at the same temperature. However, the creep strength of composite solder joints at high
temperature of 85 and 125°C was similar to that of eutectic Sn-Ag solder joints,
indicating that there is no substantial effect of particle at high temperatures. Cyclic creep

by loading/unloading was found to significantly increase creep strength of solder joints.

to my father and the memory of my mother

to my wife for her assistance, endurance,
and encouragement

to Anne and soon-to-be—born son
for their ever—bright future

iv

ACKNOWLEDGEMENTS

I would like to express my sincere appreciation to my academic advisor, Dr.
Thomas R. Bieler, and co-advisors, Dr. K.N. Subramanian and Dr. J .P. Lucas, for their

guidance and encouragement to complete this research.

I also wish to thank the members of my committee, Dr. Dahsin Liu, Department of
Mechanical Engineering, and Dr. Phillip Duxbury, Physics and Astronomy Department,

for their valuable suggestions and comments on this study.

Particularly, I would like to express my sincere appreciation to Composite Materials
and Structures Center at MSU, Visteon Electronic Technical Center, Dearbom, MI,

Mannesmann VDO Control System, Auburn, IN, and National Science Foundation.

I would like to express my deep gratitude to my wife for her patient support and
encouragement during entire graduate study. I am grateful to my parent, Eunsang Choi
and Oksun Lee, for giving me great talent and their never-ending love. I also thank to my

daughter, Anne S. Choi, for bearing her trouble without bothering me.

TABLE OF CONTENTS

ix

 

LIST OF TABLES

LIST OF FIGURES

 

CHAPTER 1

 

1.1 INTRODUCTION

 

1.2 RATIONALES AND GOAL

 

1.3 RESEARCH STRATEGY

 

CHAPTER 2

 

2.1 INTRODUCTION

 

2.3 CREEP

 

 

2.3.1 Overview of Creep deformation

2.4 FATIGUE

 

 

2.4.1 Overview of Fatigue

 

2.4.2 Fatigue Deformation of Solder Joints

 

 

 

 

 

2.4.3.2 Thermomechanical Fatigue

3
5
OVERVIEW OF EUTECTIC Sn-Ag SOLDER AND SOLDER JOINT

8

2.2 CRYSTAL STRUCT UR AND MICROSTRUCTURE 8
12

12

2.3.2 Creep Deformation of Eutectic Sn-Ag Solder and Solder Joints ----------- 26

36

36

43

2.4.2.1 Fatigue Failure Criteria of Solder Joints 45

2.4.2.2 Isothermal Mechanical Fatigue of Solder Joints 46

2.4.2.3 Thermomechanical Fatigue of Solder Joints 48

2.4.3 Fatigue Properties of Eutectic Sn-Ag Solder and Solder Joints ------------ 53
2.4.3.1 Isothermal Mechanical Fatigue 53

63
78

 

CHAPTER 3
THERMOMECHANICAL FATIGUE BEHAVIOR OF Sn-Ag SOLDER JOINTS

vi

3.1 INTRODUCTION

79

 

 

3.2 EXPERINIENT AL DETAILS

3.3 RESULTS

80

83

 

3.4 DISCUSSION

 

 

3.5 SUMMARY

CHAPTER 4

 

CHARACTERIZATION OF THE GROWTH OF INTERMETALLIC
INTERFACIAL LAYERS OF Sn-Ag AND Sn-Pb EUTECTIC SOLDERS
AND THEIR COMPOSITE SOLDERS ON Cu SUBSTRATE DURING
ISOTHERMAL LONG-TERM AGING

4.1 INTRODUCTION

 

104

110

112

113

 

4.2 EXPERIMENTAL PROCEDURES

4.3 RESULTS AND DISCUSSION

 

4.4 CONCLUSION

 

CHAPTER 5

 

FORMATION AND GROWTH OF INTERFACIAL
INTERMETALLIC LAYERS IN EUTECTIC Sn-Ag SOLDER
AND ITS COMPOSITE SOLDER JOINTS

5.1 INTRODUCTION

 

5.2 EXPERIMENTAL DETAILS

114

118

137

139

140

141

 

5.3 RESULTS AND DISCUSSION

 

5.4 SUMMARY

 

CHAPTER 6

144

155

156

 

CREEP PROPERTIES OF EUTECTIC Sn-Ag SOLDER JOINTS
AND Sn-Ag COMPOSITE SOLDER JOINTS CONTAINING
INTERMETALLIC PARTICLES INTRODUCED BY IN-SITU METHODS

6.1 INTRODUCTION

157

 

vii

 

6.2 EXPERIMENTAL PROCEDURES

6.3 RESULTS AND DISCUSSION

 

 

6.4 SUMMARY

 

CHAPTER 7
MECHANISTIC ISSUES ON CREEP STRENGTH OF SOLDER JOINTS

7.1EFFEC1‘S OF CYCLIC CREEP ON THE CREEP STRENGTH ...............

159

162

178

181

181

190

 

7.2 EFFECTS OF PARTICLES ON THE CREEP STRENGTH

 

CHAPTER 8
SUMMARY

191

194

 

REFERENCES

viii

.r " 12““.1‘" {'1

LIST OF TABLES

CHAPTER 2

Table 2.1 Physical properties and cost of potential Pb-free eutectic solders 9

 

Table 2.2 Values of n and QC related with different creep mechanisms
occurring with pure metals 22

 

Table 2.3 Summary of published data for creep deformation

 

 

 

of the eutectic Sn-Ag bulk solder and joint 27
Table 2.4 Room temperature physical properties of intermetallics 72
CHAPTER 4
Table 4.1 Intermetallic thicknesses (pm) during aging at 150°C 125
CHAPTER 5

Table 5.1 Compositions (vol %) of composite solders
(*The exact ratio of FeSn and FeSn; particles is not clear at present) ------- 143

Table 5.2 Calculated layer-growth coefficients for the growth of IMC layers ----------- 151

ix

LIST OF FIGURES

CHAPTER 2

Figure 2.1 Unit cell of B-Sn phase (lattice parameter a=5.832 A, c=3. 182 A) -------------- 9

Figure 2.2

Figure 2.3

Figure 2.4

Figure 2.5

Figure 2.6

Figure 2.7

Figure 2.8

Figure 2.9

Typical microstructure of the eutectic Sn-Ag solder matrix

11

 

in a joint with Cu substrate

Microstructure of the eutectic Sn-Ag solder matrix in a joint
with Cu substrate after aging for 100 hours at 150°C

11

 

Schematic representation of a creep curve illustrating
different creep deformation stages

13

 

Normalized steady-state behavior of eutectic Sn-Ag
BGA (Ball Grid Array) solder joints. Power-law breaks down at

stress levels above 10'3 76, while a sinh function fits all data well ------

A plot of shear strain-rate vs. applied shear stress for

the eutectic Sn-Pb solder alloy with a grain size of 5.8 um at 63°C.

 

Note that Q, in Pb = 101 kJ/mol and Q, in Sn = 94 kJ/mol

A plot of applied stress vs. strain-rate for
the eutectic Sn-Ag single shear lap joints illustrating the

influence of different Ag3Sn morphology on the creep behavior ---------

A plot of normalized stress vs. strain-rate
for pure Sn and the eutectic Sn-Ag solder illustrating
the effects of different rnicrostructure on creep behavior

----- 17

24

----- 3o

30

 

A comparison of the creep behavior of eutectic Sn-Ag solders
along with pure Sn data. In legends, B denotes bulk,

SLJ denotes schematic single shear lap joint, and A]

denotes actual joint. *95Sn-5Ag bulk solder was used
instead of the eutectic Sn-Ag solder

 

Figure 2.10 Effect of gold content on tensile fracture elongation

Figure 2.11 Fatigue parameters

Figure 2.12 Notation for a symmetric hysteresis loop

 

and tensile strength

 

 

31

34

37

37

Figure 2.13
Figure 2.14

Figure 2.15

Figure 2.16

Figure 2.17

Figure 2.18

Figure 2.19

Figure 2.20

Figure 2.21

Figure 2.22

Figure 2.23

Figure 2.24

Figure 2.25

Figure 2.26

Figure 2.27

Cyclic behavior of a material under stress control 39

 

 

 

Cyclic behavior of a material under strain control 39
Stress (S) - cycles to failure (N) curves. A is for ferrous metals

and B is for nonferrous metals. SL is the fatigue or endurance limit -------- 41
Fatigue life in terms of total strain obtained by .
superposition of elastic and plastic curves 44
Number of cycles versus strain for 95Sn-5Pb solder.

Torsion tests without hold time 25°C 44

 

Schematic illustration of the three time-temperature
profile in thermal cycling. In these cycles, the strain

 

 

and temperature follow one another 50
Simplified test specimen used by Tribula et al.

to impose shear strain on the solder joints 52
Comparison of the fatigue life behavior of Sn-3.5Ag

and Sn-40Pb solder joints at 35°C and 150°C using a
50% load drop failure criterion 56

 

Comparison of fatigue lives of Sn-4Ag and Sn—37Pb
solder joints at room temperature using

 

 

 

 

 

50% load drop failure criterion 57
Fatigue crack growth rate (dac/dN) versus stress intensity

factor ( Ara/7m ) as a function of (I) for Sn-37Pb solder joints 59
Comparison of the crack growth rate

at (I) = 50% for the Sn-4Ag and Sn-37Pb solder joints 59
Fatigue lives of Sn-3.5Ag and Sn-37Pb bulk solders at 25°C

and with 1 sec ramp time and no dwell time using 50% cm 61
Effect of dwell time on the fatigue lives using 50% cm failure

criterion during tests with A8, = 0.006 and 1 sec ramp time at 25°C --------- 61
Effect of ramp time (frequency) on the fatigue lives

during tests with A8, = 0.01 and no dwell time at 25°C 62
Effect of temperature on fatigue lives of Sn-3.5Ag solder

during tests with 1 sec ramp time and no dwell time 62

 

xi

Figure 2.28 Schematic illustration of geometry of and cracking

in the controlled-collapsed chip (flip chip) Sn-3.5Ag solder joint ----------- 66

Figure 2.29 Geometry of the surface mount leadless Sn-3.5Ag solder

 

joint and the crack propagation through the solder joint

Figure 2.30 (a) Crack length as a function of number of cycles, and (b)

68

crack growth rate in surface mount leadless Sn-3.5Ag solder joint ---------- 69

CHAPTER 3

Figure 3.1 A schematic drawing of solder joint used in this study.
The cross section in A and B direction is used to examine

 

as-soldered and thermally cycled joints, respectively

Figure 3.2 (a)Optical and (b)ESEM micrographs of surface

82

84

 

appearance of as-soldered 95.58n/4Ag/0.5Cu solder joint

Figure 3.3 ESEM and optical micrographs showing microstructures
of a cross-section of as-soldered 95.5Sn/4Ag/0.5Cu solder
joint in the direction shown in Figure 3.1b. (a) is the bulk
solder region, (b) is the region near Aggsn intermetallic
layer, (c) is the region near Ni coating layer on Cu lead

85

 

Figure 3.4 (a) & (b) Optical micrographs of surface appearances
of representative eutectic Sn-Ag. (a) and 95.5Sn/4Ag/0.5Cu
(b) solder joint after 250 thermal cycles

87

 

Figure 3.4 (c) & ((1) Optical micrographs of surface appearances
of representative eutectic Sn-Ag (c) and 95.58n/4Ag/0.5Cu

88

 

(d) solder joint after 1000 thermal cycles

Figure 3.5 ESEM micrographs of circumferential crack on the
surface of solder joint. (a) Eutectic Sn-Ag solder joint
after 1000 thermal cycles, (b) and (c) 95.58n/4Ag/0.5Cu
solder joint after 1000 thermal cycles

89

 

Figure 3.6 A schematic drawing showing the circumferential crack
on the free surface of the solder joint and striations on
the bottom surface in the open crack

 

Figure 3.7 (a) & (b) ESEM micrographs of the microstructure of
solder joint after 250 thermal cycles. (a)and (b) are
eutectic Sn-Ag and 95.5Sn/4Ag/0.5Cu, respectively

91

92

 

xii

Figure 3.7 (c) & (d) ESEM micrographs of the microstructure of
solder joint after 1000 thermal cycles. (c) and (d) are
eutectic Sn-Ag and 95.SSn/4Ag/0.5Cu, respectively 93

 

Figure 3.8 ESEM micrographs of Ag3Sn intermetallic layer
between solder and 99Ag/1Pt metallization on alumina substrate.
(a) and (b) are eutectic Sn-Ag and 95.SSn/4Ag/0.5Cu after
250 thermal cycles, respectively. (c) and (d) are eutectic
Sn-Ag and 95.SSn/4Ag/0.5Cu after 1000 thermal cycles,
respectively 94

 

Figure 3.9 ESEM micrographs of cross-section of solder joint
after 1000 thermal cycles. (a) is the eutectic Sn-Ag solder joint
and (b) is the 95.5Sn/4Ag/0.5Cu solder joint 96

 

Figure 3.10 ESEM micrographs of a region near free surface
of the solder joint after 100 thermal cycles.
(a) is the eutectic Sn-Ag solder joint and
(b) is the 95.SSn/4Ag/0.5Cu solder joint 97

 

Figure 3.11 ESEM micrographs showing the crack within bulk region of
the solder joint after 1000 thermal cycles. (a) is the eutectic
Sn-Ag solder joint and (b) is the 95.5Sn/4Ag/0.5Cu solder joint ------------- 98

Figure 3.12 ESEM micrographs of (a) crack in the thin region
between Cu lead and 99Ag/1Pt metallized alumina substrate
and (b) microstructure in the region around the corner of Cu lead --------- 100

Figure 3.13 (a) & (b) ESEM micrographs of microstructures of two same
regions of the eutectic Sn-Ag solder joint in first and second
examination. (a) and (b) represent the interior region away
from the free surface in first and second examination, respectively -------- 101

Figure 3.13 (c) & (d) ESEM micrographs of microstructures of two same
regions of the eutectic Sn-Ag solder joint in first and second
examination. (c) and (d) represent the region near
Ag3Sn intermetallic layer, respectively 102

 

Figure 3.14 ESEM micrographs of microstructures of regions
beneath Cu lead of the eutectic Sn-Ag solder joint. (a) is
the region around the comer of the Cu lead and (b) is the
region beneath Cu lead 103

 

Figure 3.15 ESEM micrograph of a region near Ag3Sn intermetallic in
as-soldered 9SSn/4Ag/0.5Cu solder joint in the second examination ------ 105

xiii

CHAPTER 4

Figure 4.1 Schematic diagram of the assembly of solder joints
and aluminum fixture to hold solder joints (scales in mm)

116

 

Figure 4.2 Schematic diagram of geometry and dimensions of solder joints ----------

Figure 4.3 (a) & (b) SEM micrographs showing the microstructure
of as-fabricated solder joints. (a) eutectic Sn-Pb solder joint
and (b) eutectic Sn-Ag solder joint

 

Figure 4.3 (c) & ((1) SEM micrographs showing the microstructure
of as-fabricated solder joints. (c) eutectic Sn-Pb composite
solder joint and (d) eutectic Sn-Ag composite solder joint

 

Figure 4.4 (a) & (b) SEM micrographs showing the microstructure
of the solder joints after aging at 150°C for 5000 hours.

(a) eutectic Sn-Pb solder joint and (b) eutectic Sn-Ag solder joint ---------

Figure 4.4 (c) & ((1) SEM micrographs showing the microstructure
of the solder joints after aging at 150°C for 5000 hours.
(c) eutectic Sn-Pb composite solder joint, (d) eutectic
Sn-Ag composite solder joint

——117

119

120

—-122

123

 

Figure 4.5 Total intermetallic layer (Cu68n5 + CU3Sn) versus aging time for
the eutectic Sn-Pb and Sn-Ag solder joints, and their corresponding
composite solder joints after aging at 150°C for 5000 hours

127

 

Figure 4.6 Optical micrographs showing the accumulation of the
Pb-rich phases adjacent to the Cu68n5 intermetallic layer
in (a) the eutectic Sn-Pb solder joint and (b) the eutectic

130

 

Sn-Pb composite solder joint aged at 150°C for 350 hours

Figure 4.7 Thickness of the Cu68n5 and Cu3Sn intermetallic layers versus
square root of aging time at 150°C for the eutectic Sn-Pb and

Sn-Ag solder joints, and their corresponding composite solder joints----

Figure 4.8 Arrhenius plots for the growth of Cu68n5 intermetallic layer
in the eutectic Sn-Pb and Sn-Ag solder joints, and their corresponding
composite solder joints based on the interfacial intermetallic layer
thickness at 4000 hours of aging at different temperatures

 

CHAPTER 5

Figure 5.1 SEM micrographs showing the microstructures of as-made solder

xiv

133

136

joints, (3) eutectic Sn-Ag solder joint, (b) composite solder joint

containing FeSn/FeSn; particles, (c) composite solder joint containing
Ni3Sn4 particles, ((1) optical micrograph showing the layer containing

Ni element in composite solder joint containing Ni3Sn4 particles ----------- 142

Figure 5.2 SEM micrographs showing the microstructures of solder joints
after aging at 180°C for 1400 hours, (a) eutectic Sn-Ag solder joint,
(b) composite solder joint containing FeSn/FeSnz particles,
(c) composite solder joint containing N13SII4 particles 146

 

Figure 5.3 Initial thicknesses of IMC layers in
eutectic Sn-Ag and composite solder joints 148

 

Figure 5.4 The growth of total IMC layer (both Cu68n5 and CU3Sn)
during aging up to 1400 hours, (a) eutectic Sn-Ag solder joints,
(b) composite solder joint with FeSn/FeSn; particles,
(c) composite solder joints with Ni3Sn4 particles 149

 

Figure 5.5 Arrhenius plot for the growth of Cu6Sn5 IMC layer
in eutectic Sn-Ag and composite solder joints 154

 

Figure 5.6 Arrhenius plot for the growth of Cu3Sn IMC layer

 

in eutectic Sn-Ag and composite solder joints 154
CHAPTER 6
Figure 6.1 Schematic drawing of creep testing method 161

 

Figure 6.2 (a) & (b) SEM micrographs showing initial microstructure
of the solder joint. (a) eutectic Sn-Ag solder,
(b) Cu(,Sn5 particle reinforced Sn-3.5Ag solder 163

 

Figure 6.2 (c) & ((1) SEM micrographs showing initial microstructure
of the solder joint. (c) Ni38n4 particle reinforced
Sn-3.5Ag solder, (d) FeSn; particle reinforced Sn-3.5Ag solder -------------- 164

Figure 6.3 (a) & (b) Steady state strain-rate vs. applied stress.
(a) eutectic Sn-Ag solder joints, (b) Cu68n5 composite solder
joints along with data from Choi et al. 165

 

Figure 6.3 (c) & (d) Steady state strain-rate vs. applied stress.
(c) Ni38n4 composite solder joints, ((1) FeSn; composite solder joints ------- 166

Figure 6.4 (a) & (b) Two representative creep curves and schematic diagrams
illustrating different damage accumulation across the solder joint.

XV

 

(a) and (b) represent inhomogeneous deformation 168

Figure 6.4 (c) & ((1) Two representative creep curves and schematic diagrams
illustrating different damage accumulation across the solder joint.
(c) and ((1) represent homogeneous deformation 169

 

Figure 6.5(a) The normalized steady state behavior of eutectic Sn-Ag
and Cu6Sn5 composite solder joints along with data
of Choi et al. and Darveaux & Banerji’s 171

 

Figure 6.5(b) The normalized steady state behavior of eutectic Sn-Ag and
Ni3Sn4 composite solder joints along with data of Darveaux & Banerji’s---l72

Figure 6.5(c) The normalized steady state behavior of eutectic Sn-Ag and
FeSn; composite solder joints along with data of Darveaux & Banerji’s----l73

Figure 6.6 Representative fracture surface of the crept solder joint.
(a) Cu6Sn5 particle reinforced Sn-3.5Ag solder joint with
10.5% porosity, (b) smeared surface due to fracture in pure shear ------------ 176

Figure 6.7 (a) & (b) SEM micrographs showing fracture processes.
(a) cracks developed along Sn-Sn grain boundaries and
(b) dimples indicating ductile fracture 177

 

Figure 6.7 (c) & ((1) SEM micrographs showing fracture processes.
(c) complete debonding of Cu68n5 interfacial layer and
solder matrix, (d) partial debonding 179

 

CHAPTER 7

Figure 7.1 Comparison of steady state strain-rate of the Cu6Sn5
composite Sn-Ag solder joints for continuous and
cyclic creep deformation 182

 

Figure 7.2 Strain vs. time curve showing primary and secondary stages
of the datum point marked by an arrow in Figure 7.1 185

 

Figure 7.3 Simulated cyclic creep curve with unloading event every 10 ksec
and 30% strain drop at each unloading. Curve A is one
from continuous creep, curve B is the simulated cyclic creep,
and curve C is the experimentally observed one 187

 

xvi

CHAPTER 1

1.1 INTRODUCTION

Sn-Pb solders have been extensively used as joining materials in electronic
applications over the past several decades [1-5]. The widespread use of Sn-Pb solders is
due to several advantages such as low cost, good manufacturability, good wettability on
common substrate such as Cu and Ni used in electronics [1,3,4].

However, pending regulations regarding the toxic elements threaten to ban the use
of Pb-containing solders in electronic applications due to the environmental and health
concerns [1]. Another concern is that Sn-Pb solders are reaching the limit of
performance capability since the adoption of surface mount technology. Increasing
demands for greater packaging density and high performance are being realized with the
advent of surface mount technology, resulting in increase of the numbers of solder joints
per package and reduction of joint dimensions [6]. However, these changes demand the
solder joint to be a mechanical connection as well as an electrical contact. Reliability
losses in many electronic systems were identified with mechanical failure of the solder
joint rather than device malfunctions [3]. These concerns have stimulated substantial
efforts in the research and development of alternative Pb-free solder alloys with
reinforcements.

Eutectic Sn-Ag solder has received much attention as a potential Pb-free candidate
for electronic and automotive applications. The eutectic Sn-Ag solder is non-toxic, has a
high melting temperature of 221°C, and generally exhibits better mechanical properties

than the eutectic Sn—Pb solder. This solder also show comparable wetting characteristics

to eutectic Sn-Pb solder. Owing to its high melting temperature, it is considered as a Pb-
free replacement in high temperature applications such as automotive under-the-hood and
step soldering process. In contrast, its high melting point is an obstacle to its use in
current manufacturing processes utilizing Sn-Pb solders.

Since the introduction of surface mounting technology that requires solders as a
mechanical as well as electrical connection, demands for strong as well as compliant
solder in soldering application have increased. Incorporation of reinforcements in Pb-
free solders has been pursued to improve the mechanical and particularly
thermomechanical behavior of solders, and their service temperature capability.
Incorporation of reinforcements has been also attempted to match Pb—free candidates to
current manufacturing processes based on Sn-Pb solder such as melting temperature,
wettability.

Extensive database of Sn-Pb solder on mechanical and physical properties, and on
deformation behavior and damage characterization has been established during past
decades. However, those for Pb-free candidate solders are still far from complete,
compared to Sn-Pb solder. The lack of a database for Pb-free candidates makes
difficulties in developing composite solder with appropriate reinforcements.

Mechanical properties such as creep and fatigue have been studied on eutectic Sn-
Ag solder. Attempts have been made to match melting temperature to that of Sn-Pb
solder as well as to improve mechanical properties by incorporating more alloying
elements. However, mechanical behavior of this solder is still not well understood.

Microstructural characterization on deformation and damage has not been detailed yet.

Furthermore, limited work has been conducted on the effects of reinforcements on

mechanical and aging properties of this solder.

1.2 RATIONALES AND GOAL

It is important to understand the deformation modes that the solder joint undergoes
during service before the Pb-free candidates are taken into account to replace Pb-bearing
solders. Thermomechanical fatigue (TMF) is the major deformation mode that the solder
joint experiences during service. As the electronic device is turned on and off, or as the
temperature of its environment changes, the solder joint is thermally cycled. During
thermal cycling, the strains are developed in the solder joint due to the mismatches in the
coefficients of thermal expansion (CTE) of the solder, substrate, and component. Thus,
the solder joint experiences thermal fatigue that causes the thermal strains in cyclic nature
[7-9]. Another important deformation mode is creep that is time dependent and operative
at high temperature above 0.5Tm. The solders have quite a low melting temperature,
resulting in a higher homologous temperature than 0.5Tm even at room temperature. The
temperature ranges that the solder joints are operated are at 50 to 80°C for typical
electronic products such as computer and home appliances [10], and are raised up to
150°C for automotive application [2]. Since the solders are used at high homologous
temperatures in service, the creep deformation is significant [7—9]. Consequently, the
solder joints suffer creep-fatigue interaction, i.e. thermomechanical fatigue (TMF), which
is cyclic, time dependent, and predominantly shear in nature [7]. The deformation during
TMF is described in the form of the hysteresis loop of stress and strain. During high

temperature regimes, the stress relaxation also takes place, introducing a further factor in

the deformation behavior [8,11]. In addition, the substrate affects the deformation of the
solder joint by imposing the mechanical constraint and causing residual stress [12,13].

The deformation behavior of the solder joint becomes more complicate due to the
complexity of microstructure of the solder joint in service. The solder is used in an as-
solidified microstructure depending upon the processing history [7]. The microstructure
of the solder is also altered due to the dissolution of substrate materials in the solder
matrix during reflow [14,15]. Because of the high homologous temperatures during
service condition, the as-solidified microstructure of the solder changes continuously
[15,16]. As a result, the mechanical properties of the solder joint during service are
different from those of the as-made solder joint. The microstructural evolution of the
solder causes the deformation complicate and reliability analysis of the solder joint
difficult.

A Pb-free solder candidate should have a steady microstructure exhibiting improved
thermal fatigue resistance and creep strength. There are also other factors that should be
considered such as melting temperature, solderability, thermal expansion coefficient,
interfacial IMC layer growth, cost, etc for a Pb-free solder candidate to be acceptable in
the industry-wide application. The shortcomings in the required properties can increase
the possibility of solder joint failures during manufacturing and service, causing open
circuits and/or mechanical failure of electronic modules [1].

The Pb-free candidates must be tested for various requirements discussed above to
determine whether they are suitable for a Pb-free replacement. Since the solder joint
undergo the creep-fatigue interaction, i.e. TMF, in actual service condition, it is important

to identify the deformation behavior during creep and fatigue and responsible

I.“

mechanisms. And then appropriate models should be developed to assess the reliability
of the solder joint based on lifetime estimation in the development of a superior Pb-free
alternate. Since the solders are susceptible to changes in microstructure affecting
mechanical properties during service, the microstructural evolutions should be taken into
account in the analysis of solder joint reliability.

Although several Pb-free solder candidates have been considered over the last few
years, there is no drop-in replacement for Pb-bearing solders yet. The database required
to develop suitable Pb-free solder replacement is still far from complete. The goal of this
research is to identify deformation behavior of the eutectic Sn-Ag and particulate
reinforced Sn-Ag solder joints. Understanding of the deformation of the solder joint
warrants how to improve the reliability of the solder joint. Another important factor
affecting the reliability of the solder joint, interfacial IMC layer growth, was also

conducted.

1.3 RESEARCH STRATEGY

This research aims to improve understanding of deformation behavior of the
eutectic Sn-Ag and its in-situ composite solder joints. In order to understand deformation
behavior of actual Sn-Ag solder joints in service conditions, thermomechanical fatigue
(TMF) utilizing actual solder joints was carried out. Microstructural characterization of
deformation behavior and damage accumulation occurring during thermal cycling was
conducted. Using the eutectic Sn-Ag solder as a model system, the eutectic Sn-Ag
solders reinforced with Cu68n5, NigSm, and FeSnz, known as composite solders, were

examined in the form of single shear lap solder joint made with two Cu substrates. Creep

If‘m‘v‘ I

deformation and interfacial IMC layer growth were investigated on eutectic Sn-Ag and its
in-situ composite solder joints. The effects of intermetallic reinforcements on creep
deformation and interfacial IMC layer growth were characterized. Properties of in-situ
Sn-Ag composite solders were compared to the eutectic Sn-Ag solder.

In chapter two, the eutectic Sn-Ag solder is reviewed in terms of crystal structure,
microstructure, creep, fatigue, interfacial intermetallic layers, and effects of
reinforcement. In the review of creep, creep data for the eutectic Sn-Ag solder are
summarized showing discrepancy in experimental results. Different viewpoints in the
analysis of creep data are revealed. It is also shown that microstructural characterization
of creep deformation did not support the measured creep data. Fatigue of the eutectic Sn-
Ag solder is reviewed in terms of isothermal mechanical fatigue and thermomechanical
fatigue. Important factors affecting fatigue behavior are summarized. Several features
about thermomechanical fatigue of actual solder joints are emphasized to make a contrast
to isothermal fatigue of bulk specimen and simplified solder joints.

In chapter three, detailed microstructural characterizations of thermomechanical
fatigue deformation and damage occurring in actual Sn-Ag solder joints are presented.
Several factors affecting thermomechanical fatigue are analyzed and discussed. Eutectic
Sn-Ag and Sn-4Ag-0.5Cu solder joints are compared to identify the effect of the small
amount of Cu on thermomechanical behavior.

Chapters four and five present effects of intermetallic reinforcements on interfacial
intermetallic compound (IMC) layer growth during aging of solder joints. Growth of
interfacial IMC layers is characterized during aging for time scale up to 5000 hours.

Chapter four compares the IMC layer growth behaviors of eutectic Sn-Ag and eutectic

Sn-Pb solder joints, and their corresponding composite solder joints with Cu6Sn5
particulate reinforcements. Chapter five evaluates the effect of different types of
reinforcement (Ni3Sn4 and FeSnz) on IMC layer growth in the eutectic Sn-Ag solder
joints.

Chapter six presents creep behavior of the eutectic Sn-Ag and three different Sn-Ag
composite solder joints. Effects of different type of reinforcement on creep behavior are
discussed. Difference between cyclic creep and monotonic creep are illustrated.

In chapter seven, possible mechanistic issues that could account for the observed
creep behavior are discussed. Future works are suggested to verify hypotheses regarding
mechanistic issues that determine creep properties of eutectic Sn-Ag and composite
solder joints.

Finally, important observations on thermomechanical fatigue, interfacial IMC layer
growth, and creep behavior of eutectic Sn-Ag and composite solder joints are

summarized.

 

CHAPTER 2

OVERVIEW OF EUTECTIC Sn-Ag SOLDER and SOLDER JOINT

2.1 INTRODUCTION

Several Pb-free solders have been considered for the replacement of Sn—Pb solders.
They include eutectic Sn-Ag, Sn-In, Sn-Zn, Sn-Bi, Sn-Sb, and Sn-Cu solders etc., and
have Sn in common as a base metal [10,17]. Table 2.1 summarizes physical properties
and cost of potential Pb-free solders [1,18-20]. The eutectic Sn-Ag solder is one of the
leading Pb-free candidates. It generally exhibits better wetting characteristics than the
eutectic Sn—Pb solder [4]. It has a higher melting temperature of 221°C in comparison to
the eutectic Sn-Pb solder as highlighted in Table 2.1 and is suitable for high temperature
applications such as automotive under-the-hood. In contrast, its high melting point is an
obstacle to its use in other electronic assembly applications [10,21]. In this chapter, a
literature review of eutectic Sn-Ag solder and solder joints will be given in terms of
crystal structure, microstructure, creep, fatigue, interfacial IMC layer, and reinforcement

strategies.

2.2 CRYSTAL STRUCTURE AND MICROSTRUCTURE

The eutectic Sn—Ag solder (96.5Sn—3.5Ag in wt%) alloy consists of two phases, [3-
Sn phase and Ag3Sn intermetallic compound, with a melting temperature 221°C. As
shown in Figure 2.1, B-Sn phase has a body—centered tetragonal structure with 4 Sn atoms
in the unit cell at 000;0 1/2 M: ; 1/2 ‘/2 1/2 ; 1/2 0 1% and c/a ratio of 0.546 [22]. Aggsn

intermetallic compound is known as orthorhombic structure with lattice constants of a =

Table 2.1 Physical properties and cost of potential Pb-free eutectic solders [1,18-20].

 

 

 

 

 

 

 

 

 

 

 

MP. = Mngelti Point, CTE = Coefficient of thermal expansion.
*Near-peritectic composition.

 

 

 

 

 

 

 

 

Solders M.P. (°C) CTE (10'°/°C) Density (g/cm3) Cost ($/kg) ~

63Sn-37Pb 183 21 8.4 5.87
96.58n-3.5Ag 221 22 7.36 13.73

48Sn-52In 1 18 20 7.3

428n-583i 138 15 8.7 7.79

9 l Sn-9Zn 199 7.99
99.3Sn-O.7Cu 227 8.62

9SSn-SSb* 245 27 8.36

Figure 2.1 Unit cell of B-Sn phase (lattice parameter a=5.832 A, c=3.182 A) [22].

5.968 A, b = 4.7802 A, and c = 5.1843 A, and 8 atoms in the unit cell, six atoms of Ag
and two of Sn [23]. However, a complete configuration of atoms in the Ag3Sn unit cell
has not been reported yet.

The eutectic Sn-Ag solder has an equilibrium microstructure consisting of Ag3Sn
intermetallic aciculae dispersed in an almost pure B-Sn phase with less than 0.04 wt% Ag
[24]. However, depending on the cooling rate during solidification process, it shows
either essentially pure B-Sn with rod-like Ag3Sn interrnetallics or primary B-Sn dendrites
with interdendritic eutectic microstructure, which is the mixture of Ag3Sn particles and [3-
Sn phase [5,10,14,16,25]. The solder joint commonly contains the microstructure of the
primary B-Sn dendrites with interdendritic eutectic mixture resulting from the faster
cooling rate. Figure 2.2 illustrates a typical microstructure of the eutectic Sn-Ag solder
matrix in a joint with Cu substrate [25].

The eutectic Sn-Ag microstructure was found to change due to aging at high
temperature. During aging, the Ag3Sn interrnetallics coarsened, showing a decrease in
number and increase in size [16,25]. The dendritic morphology of primary Sn phase also
disappeared, resulting in the microstructure of Sn phase with coarsened Ag3Sn
interrnetallics [25]. The coarsening was more significant at higher temperature, while it
was very limited at room temperature. Figure 2.3 illustrates the microstructure of the
eutectic Sn-Ag solder matrix in a joint with Cu substrate after aging for 100 hours at
150°C [25].

The microstructure and composition of the solder are altered when it is soldered to a
substrate due to the dissolution of the substrate. During soldering of the eutectic Sn-Ag

solder with a Cu substrate, Cu dissolves in the liquid solder and reacts with Sn, forming

10

 

Figure 2.2 Typical microstructure of the eutectic
Sn-Ag solder matrix in a joint with Cu substrate [25].

 

Figure 2.3 Microstructure of the eutectic Sn-Ag solder matrix in a joint
with Cu substrate after aging for 100 hours at 150°C [25].

11

also

the

Cu68n5 intermetallic dendrites or particles in solder matrix during solidification [14-
16,26]. The amount and size of the Cu6Sn5 dendrites in a solder joint were dependent on
the maximum soldering temperature [15]. The overall amount of the Cu6Sn5 particles
also depended on the cooling rate [14]. The Cu6Sn5 dendrites exhibited a coarsening
behavior similar to Ag3Sn particles [16].

The microstructure of the solder joint significantly affected the mechanical
properties. The microstructure containing fine Ag3Sn particles showed higher
rnicrohardness, higher maximum shear strength, and slightly reduced ductility compared
to that with coarse Ag3Sn rods [14-16]. The large Cu68n5 dendrites considerably reduced
the ductility and slightly increased the maximum strength [15]. It is apparent that the
microstructure produced by fast cooling rate and low soldering temperature would give
better mechanical properties. However, the solder joint may not preserve the mechanical

properties during service because of the changes in the microstructure such as coarsening.

2.3 CREEP
2.3.1 Overview of Creep Deformation

Creep is a time dependent deformation of a material when it is subjected to a
constant load at constant temperature and is important at temperatures greater than 0.5Tm.
Creep strain is measured as a function of time and the slope of the creep strain-time curve
is referred as a creep strain-rate. The creep curve is commonly described by reference to
three different stages as shown in Figure 2.4, i.e. primary, secondary, and tertiary stages
[27]. With a constant load applied to a specimen, the strain-rate decreases during the

primary stage after an instantaneous strain on loading, indicating that the material

12

 

 

 

Creep Strain, e

 

 

 

 

 

Rupture
<
Secondary strain-rate
8 = AE/AI
Primary As Tertiary
At
Secondary
Instantaneous strain

 

 

 

 

Time, t

Figure 2.4 Schematic representation of a creep curve
illustrating different creep deformation stages [27].

13

experiences an increase in creep resistance or strain hardening. After primary stage,
strain-rate reaches a secondary stage where the strain-rate is constant. The constant
strain-rate in the secondary stage is explained on the basis of a balance between the
competing processes of strain hardening and recovery, and in some sample alloys a
steady state microstructure develops. The secondary stage often has the longest duration
in the lifetime of a material during creep deformation. During the tertiary stage, the
strain-rate rapidly increases with time leading to fracture. An increase in strain-rate in
the tertiary stage is related to microstructural changes such as grain boundary separation,
the formation of cracks, cavities, voids, and neck formation, resulting in a decrease in the
effective load-carrying area.

The creep resistance of a material is often described by the minimum strain—rate in
secondary (or steady—state) stage. Both temperature and the level of applied stress
influence the creep characteristics. With either increasing stress or temperature, the
steady-state strain-rate is increased [29]. The stress dependence of strain-rate is
expressed using an empirical relationship,

75 = K17"

9

where K; is the constant and n is the material constant denoted as stress exponent. This
equation provides the basis for the "power-law" relationship between strain—rate and
stress, which have been used to describe high temperature creep behavior. The stress
exponent, n, can be obtained from the slope of a straight line in the plot of log 7: versus
log T . The influence of temperature on strain-rate is described by an Arrhenius

relationship,

14

 

. "Q
: K C
rs Zexp( kT ),

where K2 is the constant, k is Boltzmann's constant, and Q. is the activation energy for
creep. This equation shows that the strain-rate increases exponentially with increasing

temperature. The activation energy, Q, can be evaluated from the slope of log 75 versus

(1/1') plot. Combining the above two equations gives a phenomenological power-law

equation,

- I n _Q)
=A2' ex —-i
,5 .(k, ,

where A' is a constant. Like the Mukherjee-Bird—Dorn power-law equation, this
phenomenological power-law equation also shows the dependence of the steady-state
strain-rate on both stress and temperature. The value of n and Q in both power-law
equations varies depending on the applied stress and temperature.

Models for creep deformation based on steady-state are commonly expressed in the

form of a generalized, dimensionless rate equation called the Mukherjee-Bird-Dom

power-law equation [28],

—”""=AlillP-l “oeuvre
DGb G d °’ 0,012 G d kT ’

 

 

15

where 7, is the steady-state shear strain-rate, k is Boltzmann's constant, T is absolute

temperature, G is shear modulus, b is the magnitude of Burgers vector, A is a material-
related constant, Tis the applied shear stress, n is the stress exponent, d is the grain size
of a material, and p is a constant related to grain size. This equation determines both the
minimum strain-rate exhibited by a sample at given stress and maximum stress that can
be borne by a sample tested at given strain-rate [7]. D is the diffusion coefficient related
to rate-controlling diffusional processes during creep deformation and expressed as
following,

D = Do exp(:l-§—‘—],

where D0 is the pre-exponential constant and Q is the activation energy of the rate-
lirniting diffusion process for creep.

While the power law generally fits data at low and intermediate stresses, it breaks
down at very high stress level. Figure 2.5 illustrates normalized steady-state behavior of
eutectic Sn-Ag BGA (Ball Grid Array) solder joints where the power law breaks down at
high stress levels [30]. The Dom or phenomenological power law equation fits well up to
stress levels of 10'30’G, however, it breaks down at stress levels above 10'31/G due to
non-linear increase of normalized shear strain-rate. This non-linear increase in strain-rate
with increasing stress is often referred to as "power-law breakdown". The power-law

breakdown regions can be described by a single expression using sinh function,

 

 

 

 

 

 

l l l l I I I l I l l i T l l T l:

A 3 E C1 = 2.035-4 (Klseclpsl) E
'5 1° 50 - 38 5 (kJ/mol) """"""""""""""" “i
O. E c — . I ’3
B E """""""""""" I "E
: I I

g 101 g ----------------- , f ------ 1g
I I 2

£ 5' ““7" """"""" 1
A E ’ E
E; 10'1 g:- """"""""""""""""""""" i """"""""""" 1
\ E E
O a """""""""""""""" 1.:
z E 5

-3 ...........
g 10 :E 27°C 1;
A 5' El 77cc ------------ é
o .
E 10-5 :1 131°C --------- g
°>~ : , slnh function;
E l ----- power law
10.7 L 1 1 1 1 1 1 1 1 1 L 1 J 1 1
10" 10'3 10'2

(tIG)

Figure 2.5 Normalized steady-state behavior of eutectic Sn-Ag
BGA (Ball Grid Array) solder joints. Power-law breaks down at
stress levels above 10'31/6, while a sinh function fits all data well [30].

17

where adescribes the stress level at which the power—law dependence breaks down, and
C, is a constant.

Several mechanisms can be responsible for creep. The rate-controlling mechanism
depends both on stress level and temperature. The mechanisms responsible for creep can
be conveniently described as a function of the applied stress and be divided into four
major groups ; Diffusion creep, Dislocation creep, Dislocation glide, and Grain boundary

sliding [31,32].

Diffusion creep involves the flow of vacancies and interstitials through a crystal
under the influence of applied stress. It becomes the controlling mechanism at high
temperatures and relatively low stresses, 070 < 104. Two mechanisms are considered
important in this region ; Nabarro—Herring creep and Coble creep. Nabarro-Herring creep
involves the stress-directed vacancy flow inside the grain. The vacancies move from
1’ egions experiencing tensile stresses to those which have compressive stresses under the
applied stress. Simultaneously, there is a corresponding flow of atoms in the opposite
direction, leading to elongation of the grain. Coble creep is based on diffusion along the

grain boundaries instead of inside the grain. Coble creep predominates at relatively lower
temperatures than those at which Nabarro—Herring creep is operative. Diffusion creep
pt"EsCIicts a stress exponent close to unity, and activation energy (QC) close to that of lattice
Self—diffusion (Q) for Nabarro-Herring creep and grain boundary diffusion (ng) for
Coble creep. When diffusion creep processes dominate, the resistance to creep increases
with increasing grain size.
Dislocation creep involves the movement of dislocations which overcome barriers

by the diffusion of vacancies or interstitials. It occurs for stress ranges of 10“1 < 076 <

18

102. The treatment given by Weertman [32] constitutes the fundamental theory for
dislocation creep. It is based on a mechanism where dislocation climb is the rate-
controlling step. At high temperature, if a gliding dislocation is held up by an obstacle, a
small amount of climb may permit it to surmount the obstacle, allowing it to glide to the
next set of obstacles where the process is repeated. In the sequential process of gliding
and climbing, the gliding step produces almost all of the strain but the climbing step
controls the velocity. Since dislocation climb requires diffusion of vacancies or
interstitials, the rate—controlling step depends on the rate of atomic diffusion. This model
predicts the stress exponent of 3, however, creep experiments with a range of metals
show the stress exponent varying from 3 to 8, with a value of 5 most common. It also
can be applied for both high temperature creep where activation energy for creep (Q) is
that of lattice self-diffusion (Q1), and for low temperature creep where activation energy
for creep (QC) is that of pipe diffusion along dislocation cores (de).
Dislocation glide occurs at stress levels which are high relative to those normally
Considered in creep deformation, 07’G > 102. At a certain stress level the power-law
breaks down. It was observed in such a stress level that dislocation climb was replaced
by dislocation glide, which does not depend on atomic diffusion. The creep-rate resulting
from a dislocation glide mechanism is established by the ease with which dislocations are
itnl>eded by obstacles such as precipitates, solute atoms, and other dislocations. The
thet‘rnally activated dislocation glide is considered to be the rate-controlling step in creep
at Such high stress levels.
Grain boundary sliding usually does not contribute significantly to secondary creep.

However, it contributes in tertiary creep to the initiation and propagation of

19

‘1.)

T2

intercrystalline cracks. Another deformation process where grain boundary sliding plays
a significant role is superplasticity [33]. Most of superplastic deformation takes place by
grain boundary sliding. Superplastic deformation requires a fine grain size, a relatively

slow strain-rate (about 104 to 10'2 s'1

for many alloys), and a high homologous
temperature. A characteristic of the superplastic deformation is that the strain-rate
sensitivity m, which is the reciprocal of stress exponent n, is high (about 0.3 to 0.9). In
addition to grain boundary sliding, some accommodation process is required to maintain
material continuity during the sliding process such as diffusion and boundary migration.
It has been known that grain boundary sliding must be also present during diffusion creep
process to maintain material continuity.

More than one creep mechanism can operate at the same time [34,35]. If several

mechanisms are operating independently of each other, then the overall creep rate is the

sum of all the rates of the independent processes involved, i.e.
7: = Z 71' 9

where 7,. is the creep rate for the ith mechanism. For mechanisms operating parallel the

fastest mechanism will dominate the creep behavior. When the deformation depends on a
series of events occurring sequentially, with each event necessary for deformation to

continue, the creep rate is

1

7.=———

21%.)”

20

If there are i mechanisms operating in series, the slowest mechanism will control the
creep deformation.

Although two independent processes can contribute to the overall creep rate, one of
these processes is defined as dominant when the contribution made by other process is so
small that it can be ignored. The dominant creep mechanism can be identified by
comparing the experimentally measured values of n and Q with the values predicted
theoretically for different creep mechanisms, since each creep mechanism predicts unique
values of n and Q [36].

For instance, dislocation creep models generally lead to an n value of about 5,
whereas diffusion creep theories give n values of unity. Thus, a transition from n = 5 to n
= 1 with decreasing stress is usually taken as evidence for a change in the dominant
mechanism from dislocation to diffusion creep processes. Then, in the n = 5 regime,
when a change from Q 5 Q, to Q < Q, occurs, it is interpreted that lattice self-diffusion
is becoming more difficult with decreasing temperature until the deformation behavior is
controlled by preferential diffusion along dislocation core. Again, when diffusion creep
dominates creep deformation exhibiting n values close to unity, a transition from Q E Q,
to Q < Q, is related with a change in the dominant mechanism from Nabarro-Herring to
Coble creep. In the diffusion creep regime, atomic flow occurs predominantly through
the lattice at high temperatures giving QC 5 Q1, but predominantly along grain boundaries
as the temperature decreases below 0.6Tm giving QC .2. ng. The n and Q; values related
to different creep mechanisms for pure metals are listed in Table 2.2 [37], together with

the general stress and temperature conditions for each dominant creep process.

21

Table 2.2 Values of n and QC related with different
creep mechanisms occurring with pure metals [37].

 

 

 

 

Creep mechanism Stress Temperature n value QC value '
low temperature 7 . i A
diffusion creep low 0.4 ~ 0.7Tm ~ 1 ng
(Coble creep)
high temperature
diffusion creep low above 0.7Tm ~ 1 Q,
(Nabarro-Herring creep)
low temperature . . .
dislocation creep intermediate/high 0.4 ~ 0.7T... > 3 de
h‘gh temperature intermediate/high above 0.7T... > 3 Q,

 

dislocation creep

Q31, = the activation energy for grain boundary diffusion, de = the activation energy

 

 

 

 

for diffusion along dislocation cores (often termed pipe-diffusion, de E Q81)

22

ll

Sn-Pb solder is a good example showing the variations in n and Q values
depending on stress and temperature. Figure 2.6 illustrates the different set of n and Q.
values in three distinct regions that have been identified in the eutectic Sn-Pb solder with
a superplastic microstructure of 5.8 pm grain size [38]. The low stress regime called
Region I exhibited a stress exponent of 3 and activation energy for creep close to that for
lattice self-diffusion. The mechanism that controls the creep process in Region I is
unknown. It was speculated that Region I is a distinct deformation mode intermediate
between the superplastic region H and the diffusion creep processes [40]. In the middle
range of strain-rates called Region II, a stress exponent in the range from 1.6 to 2.4 was
reported in literature. Most of reported activation energies for Region II lie between 44
and 57 lemol [4]]. These values are comparable to the activation energies for grain
boundary diffusion, which is approximately one-half that for lattice self-diffusion [39].
The mechanism controlling creep deformation in Region II is grain boundary sliding
accommodated by grain boundary diffusion, and is responsible for superplastic
deformation. At high strain-rates, the reported stress exponents are high, ranging from 4
to 7, and the activation energies are nearly equal to that for lattice self-diffusion [7].
Dislocation climb controlled by lattice self-diffusion is known as the deformation
mechanism for Region II.

Solid solutions with a single phase have been investigated for stress ranges of 10'4 <
070 < 102. Two very distinct responses were observed in solid solutions, referring to
class M alloys (also called class H alloys) and class A alloys (also called class I alloys)
[42]. Class M alloys show creep characteristics controlled by climb of dislocations, as in

the case of pure metals, giving n E 5. Class A alloys show creep behavior where the

23

 

1 0'1

    
    
 
  
    

 

 

 

E i llllllTT l ITITIIT] fl lllllll i TTIIIIE
E Region III 5
F , 56 ; T
102‘ “““““ Gc=81lemoI" """"""" 1
'1‘ E I I ' Z
' - 62$n-38Pb -
a 10.»... ............................ ..
5d=5.8 um : i
0 I : -
H " o ' . —1
I 104— --.Regionll m3
.5 E : n52 j
E “ Qc=45 kJ/mol -
9" 10'5? """""""""""""""""""""""""" ; """""""" "1:
(D E . 5
10.5,: ................... Reglonl .................... 1
E "=3 5
: Qc=84 kJ/mol I
1 .7 1 1111111 #1111111] 1 11111111 I 1111
10'2 10'1 10° ‘101 102

Stress (M Pa)

Figure 2.6 A plot of shear strain-rate vs. applied shear stress for
the eutectic Sn-Pb solder alloy with a grain size of 5.8 um at 63°C.
Note that Q, in Pb = 101 kJ/mol and Q, in Sn = 94 kJ/mol [38,39].

24

dislocation movement is controlled by the drag of solutes, producing n E 3.

Comparing solid solutions with pure metals gives a good understanding of creep
behavior of solid solutions [43]. In a pure metal showing a creep curve given in Figure
2.4, the initial strain—rate is very high because the existing and newly generated
dislocations can freely move. As the creep deformation proceeds to the secondary stage,
the dislocation density increases and the strain-rate decreases. The decrease in the strain-
rate is due to an increased density of barriers, resulting in strain hardening. For class A
alloys, each dislocation is pinned by a great number of solute atoms because the initial
density of dislocations is low, leading to a low initial strain-rate. As the deformation
proceeds, the dislocation density increases and the number of pinning solute atoms per
unit length of dislocation decreases. Hence, the strain-rate increases due to an increase in
mobile dislocation density with deformation. For class M alloys, dislocations are initially
controlled by drag of solute atoms, resulting in a low initial strain-rate. However,
dislocation creep becomes the rate-controlling mechanism after some deformation. Alloy
additions may not alter the creep characteristics in class M alloys and there is no change
in n value and creep mechanism, although a great number of solute atoms present in the
parent lattice increases creep resistance by impeding dislocation glide and recovery
processes.

Particle-hardened alloys (also called dispersion-strengthened alloys) are another
form of material which can significantly increase the creep strength, compared with pure
metals and solid solutions [44]. The improvements in creep resistance are achieved by
introducing a fine dispersion of precipitates or insoluble particles which provide effective

obstacles to dislocation movement. For particle-hardened alloys, dislocations can rarely

25

cut through the particles and consequently, dislocations attempt to surmount the obstacles
by climb and cross slip. When dislocation creep is dominant, the creep properties are
generally improved by increasing the volume fraction of particles. When diffusion creep
is rate-controlling process, particles can also improve creep properties by pinning grain
boundaries. Particles pinning grain boundaries can impede grain boundary sliding and
affect the ease that vacancies can be generated at grain boundaries. Particle-hardened
alloys are also characterized by a stress exponent higher than 7, considerably greater than
the values expected for the appropriate matrix alloys. For applications requiring high
creep resistance, it is essential that the particles do not coarsen rapidly during long
exposure at the service temperatures. Particle coarsening increases both the average
particle size and the mean interparticle spacing, so that it causes a decrease in the creep

resistance due to fewer particles being present to impede dislocation movement.

2.3.2 Creep Deformation of Eutectic Sn-Ag Solder and Solder Joints

The creep properties of the eutectic Sn-3.5Ag solder have been investigated by
several researchers in the form of a bulk solder specimen or a solder joint. Table 2.3 lists
all reported experimental creep data for the eutectic Sn-Ag solder. From the review of
the published data, it becomes apparent that there is no clear agreement in the
experimental data, resulting in the discrepancy for the rate—controlling mechanism. The
eutectic Sn-Ag solder was regarded as either pure Sn or a dispersion-strengthened alloy
in the analysis of the creep properties.

Mathew et al. [45] reported that the eutectic Sn-Ag bulk with 20 um average grain

size exhibited the stress exponent of 5 and 60.7 kJ/mol activation energy in the constant

26

Anna): .0 98 Ave: .2085: em Ba 3:5 2E.
.528 w<-=m £883 05 Co 93%.: wow: mm? 528 fine w<mémmm*

‘Il

 

 

 

 

 

 

 

 

 

:3 858888 812 as we .8 e as» em a: _ £3
: : eoses No.2 8 .2 2w 2 _ memeaowflwowamm 8:3
8 an em c8 vows
Ea eoees cmim cs 1 8 ode oaowmomfim no .ewmmw mm... 8:3
WE eoaes on; 9: .E .8 eommaofiee hoe m seem em a: 8 £3
63 eoaes 82m ca .9: .mm eomaaofiee nae m: Wfiwmwmwm 8:3
a: .85 aim we .5 ohm—mow Mm 8558 use; Hemmmmowwes
an son... aim 5 .R .5 eomwowwe new on a; owaww flees Amfiwmwfie
Eva/om 28 name
Co: 32: b H mafia—~88 em 2 .8228 5862?

 

 

 

 

 

 

 

45% Ba 628 fine w<-=m 282:0 05 no :ocegofiu queue e8 See 3533 no bagsm md 033.

27

 

load creep tests. In their tests, the method of increasing stress in steps when the steady
state was reached on a single specimen was used until fracture. They compared with the
creep data of pure Sn with 150 um grain size obtained from the same tests, which are n =
7.6, Q = 60.3 kJ/mol, and the dislocation climb controlled by lattice self-diffusion as the
deformation mechanism. And it was concluded that creep deformation of the eutectic Sn-
Ag solder took place by the same mechanism as that for pure Sn. It should be noted that
similar activation energy of about 60 kJ/mol was observed for the eutectic Sn~Ag solder
and pure Sn despite the large difference in the grain size. The 60 kJ/mol activation
energy was assumed to be that for lattice self-diffusion in pure Sn although it is more
comparable to one half that for lattice self-diffusion in pure Sn, 94 kJ/mol [39]. In their
analysis, the creep deformation controlled by grain boundary diffusion or dislocation pipe
diffusion was excluded due to high homologous testing temperature up to 0.8Tm.

Yang et al. [12] and Liang et al. [46] regarded the Sn-3.5Ag solder as a particle-
hardened alloy. The Sn-Ag bulk specimens were aged to stabilize microstructures prior
to testing in both studies as shown in Table 2.3. Yang et al. found n = 10 and Q = 50
kJ/mol in constant load creep tests of the eutectic Sn-Ag bulk. The activation energy
value of 50 kJ/mol was approximately one half that for lattice self—diffusion in pure Sn,
94 kJ/mol. Along with a high stress exponent, dislocation climb over the Ag3Sn particle,
controlled by either grain boundary or dislocation pipe diffusion, was interpreted as the
creep mechanism despite high homologous testing temperatures ranging from 0.6Tm to
0.92Tm. Liang et al. observed n = 7.83 and Q = 69.5 kJ/mol in the creep tests of 958n-
5Ag bulk. Dislocation climb was considered as the prevailing deformation mechanism

with a stress exponent of 7.83.

28

 

The creep properties of the eutectic Sn-Ag solder greatly depended on the
microstructure. Yang et al. [15] reported that the morphology of the Ag3Sn particles
resulting from different cooling rate greatly affected the creep behavior of the eutectic
Sn-Ag single shear lap joints as shown in Figure 2.7. At room temperature, the water-
quenched joints with fine Ag3Sn particles exhibited almost three magnitudes lower strain-
rate than the fumace-cooled ones with rod-like Ag3Sn interrnetallics. At 158°C, the
water-quenched joints showed a lower strain-rate at high stresses and a higher strain-rate
at low stresses than the fumace-cooled ones. On the other hand, the fumace-cooled joints
had a higher total strain-to-failure compared to water-quenched joints at both
temperatures.

It has been typically found that eutectic Sn-Ag solder is more creep resistant than
pure Sn due to the additional Ag3Sn particles [12, 47]. However, in the study of Mathew
et al. [45], the eutectic Sn-Ag solder was less creep resistant as shown in Figure 2.8 than
pure Sn. They used pure Sn with a 150m grain size, and the eutectic Sn-Ag with fine
rod-shaped Aggsn and a grain size of 20 um. The observations in Mathew et al.'s study
indicate that rod-shaped Ag3Sn particles do not play any role in the creep deformation,
and deformation along grain boundary might be significant, resulting in higher strain-
rate.

Solder joints are more creep resistant than the bulk solder. Figure 2.9 shows a
comparison of the stress vs. strain-rate data of the eutectic Sn-Ag bulk, and actual or
schematic shear lap solder joints along with additional pure Sn data [12,15,30,46-50].
Since the solder joints were tested in shear and bulk in tension or compression, the

appropriate comparison was made using following two equations,

29

 

 

 

 

 

 

1o-1 ea ~ va I _
. l:
E—e— Furnace,158°c ; {5
10., :5“ Water, 158°C_,_,-§. ,,,,,,,,,,,,,,, 1_:
A E—O—Furnace, RT 1 -' E
0 I I Z

0 3 "I" Water, RT ..........
s ‘° x a
o C ,I 2
‘5 104— -------------------------------------------- .’- ------ a
a: 5 ' 3
E I j
.5 ~ I -
'5 1o-5 — -------- 1
= fl
a) - -
P- , -i
10":— """"""" ," """""""""""""" ’ """""" a
E o' i

10_7 l 1 l 1

1 10

Stress (M Pa)

Figure 2.7 A plot of applied stress vs. strain-rate for the
eutectic Sn-Ag single shear lap joints illustrating the influence
of different Agasn morphology on the creep behavior [15].

 

 

 

 

10-9 : I I r I T V l 'I Y I 1 t I I 1 I:

E : D E
10'1°g_.8_ -: ------------- U ---------- 1

: SH (150 LII“); g

,0... Sn 3&9 (20 um) ................ E

: 3

fl 1042? """""""""""""""""""""""""" a:
e : 2
l- 10'”? """"""""""""""" o """"""""""" “-2:
X E E
.0) "‘ —4
10'“? """""""""""" r ““““““““““““ *5

E 0' E

1045 E— """""""""""" B'. """"""""""""""" ”—5!
10'16 E— """"""""""""""""""""""""""""" "g

E [:1 . g

1o_‘7 1 l l l l l I ll 1 l l l l 1 ll

10'5 10" 10'3

Figure 2.8 A plot of normalized stress vs. strain-rate
for pure Sn and the eutectic Sn-Ag solder illustrating
the effects of different microstructure on creep behavior [45].

3O

.5200 9.5m 0:00.30 05 .0 0020:. 000: 002. .0200 0:3 93-5mm.
.E_0_ .0300 «20:00 ...< 05... .25.. 03 50:0 0650 0:95:00 020ch 1.5 130 00.2.0.0 m .3502 c.
.Homévdmdtmz 200 cm 050 5.2. 9.2m 0.0200 9.15 0.82:0 .0 .0323 000.0 0.... .0 50:00.50 < QN 0.39”.

.35.. 32.0

 

 

 

 

cow OF P 6..
a a a 44 W. . a n 4 q . . 1 + . II. . O-

_..< 0:: m<1cm 550350 . ”2.0.3415 03003000020 _ C H
1<.Ohh.u<.cm.x:uo>.un O m 0.30 u<-cw 030050.500 o O D m
fi<.OhN.m<ucm.x:u0>huD o m. 0:39 cm mhsau+ax ........ .H .......................................... I“ 0.6—.

a n a H a .0 O D H
....m .5. 9.2.0 3.8.... I m a o m
. .u 1... .mca . .1 ............................................. O .................................... 1 .o_.

30 0mm..< m. >...> O- . so 0 +.. S

30 E 3&0 05> .s . m I ad _ o H n...
m.02.0<-:m.32.. d .1 ............................... c ....... woa ...... D++ ..... 1.9.: .01..
5.822-505... O H . d. 909 + H u
m.0S..a<.=m..ouoam D . o D” +++ .1: w
1| ............................. 0.-.. ... ..................................... .h n
m 68.0.1.0 .533. D - .l w h w... + + . 1.
. . . n o 9 Q .. u a
m 03 m<-:m .2000: u m I. D x m \H
w. ................. D ............... . .......................................... 1n.
m.0omp.0<-:m.0§>.... O . . x 1...: s
. a . I . X VA 1 a
m 00.: 3.5 93>... o m .. 9 @ w. XWM . a
m.0m~.m<-:m.m:u>.: o .1. ........ c n ........... O ...................... x .............................. .1. ”.3. (
0.3250632; + m o o h D x W
90035060520: + 1. I ............. D .................... . .......................................... 1.1:
. . £00.. . .
m.08.:m. m xW . o .9 D m
m 0““ cm cwwbm X UT» . L r _ . . _. . _ . _ e _ . m Pro—v

31

0'=T\/§ andé=%§

neglecting microstructural differences. The Following reasons were attributed for
difference in creep behavior between bulk and the solder joint [12,30,47] ; (i) change in
solder composition due to the dissolution of substrate in the solder matrix during
soldering, (ii) formation of interrnetallics in the solder matrix due to reaction of solder
and dissolved substrate element, resulting in dispersion-strengthening effect, (iii) the
mechanical constraint imposed by substrate and/or interfacial intermetallic layer between
solder/substrate, (iv) different microstructure of the solder joint depending on the
soldering parameters compared to that of bulk solder.

The creep data of actual solder joints are very limited at present. The creep
properties of actual solder joints are greatly affected by the specification of solder joint
such as geometry, materials to be attached, and metallization on attached materials.
Darveaux et al. [30] utilized the actual eutectic Sn-Ag solder joints used in flip chip
applications joining two ceramic substrates with a metallization consisting of Cu-Ni-Au.
They tested double lap shear specimens comprised of 9 joints in each interface. Prior to
testing, the solder joints were aged for several months at room temperature or 100 hours
at 100°C.

Results of Darveaux et al. study are shown in Figure 2.5 in a plot of normalized
strain-rate vs. normalized stress. In their study, the stress exponent of 5.5 was found,
indicative of dislocation climb controlled deformation. However, the activation energy

of 38.5 kJ/mol was considerably lower than the expected values of lattice self—diffusion

32

(106 kJ/mol), or grain boundary or pipe diffusion (63 kJ/mol) in pure Sn. They
interpreted that the activation energy might be stress dependent since the data in the stress
range more than 10'3I/G lies in the power law break down regime, causing lower
activation energy.

One important microstructural feature in the solder joints used by Darveaux et al. is
rod-like gold-tin intermetallic compounds formed by the reaction of Sn with Au from
gold metallization during soldering. Gold was found to significantly affect the
mechanical properties of the eutectic Sn-Ag solder in the study of Harada et al. [51]. As
given in Figure 2.10, when more than 6 mass% Au was added to the Sn-3.5Ag solder, the
fracture elongation was reduced to less than 5%, while the tensile strength of Sn—3.5Ag
solder increased gradually with the addition of gold. The gold-tin intermetallic
compounds at grain boundaries may provide sites for the stress concentration and crack
nucleation through grain boundary sliding process even with deformation as small as
0.2% [52]. In the study of Darveaux et al., if the gold-tin compounds were susceptible
sites for crack nucleation, an accumulation of damage may begin in early stage of creep
deformation and this may be the reason for low activation energy.

Igoshev et al. [52] have shown the discrepancy between the experimental results
and expected ones using the transition homologous temperature (11.) from lattice diffusion
controlled creep to dislocation pipe diffusion controlled creep. It is commonly observed
that as the homologous temperature (T/I‘m) is decreased, there is a transition from lattice
self-diffusion controlled creep to dislocation pipe diffusion controlled creep. The
transition homologous temperature (11.) is proportional to the applied stress and is usually

about 0.5 ~ 0.6 for most metals. According to the discussion given by Igoshev et al., the

33

 

Tensile fracture elongation (%)

 

 

 

 

60 _ 1 . r 1 1 1 . 1 ‘ rfi I ' ' ' l ‘ I I - 60
5o:— ---------- Q """ ....................... L 50
. .

: O 5 j
4oC_---O ----------------- QW'Q' """"""" E """""" ‘- 40
30 :_.....-. --‘-‘----"'""“"°"".-”: ............ . ........... {j 30
20 - """"""""""""""""""""""""""""" O” O ----------- j 20

E ' I
10 [— I I : """"""""""" L 10

E +Tenslle strength (MP8) . I

E —e—Tenslle fracture elongatlon (%) 1

0 1 l 1 l 1 1 1 l i l l l 1 l l l l 1 It 0

Au (mass%)

Figure 2.10 Effect of gold content on tensile
fracture elongation and tensile strength [51].

34

(can) uifiuans ensue;

dislocation pipe diffusion with QM = 44 kJ/mol. Igoshev er al. predicted that the
activation energy for creep of Sn-3.5Ag solder should be slightly or considerably higher
than for pure Sn due to Ag3Sn interrnetallics, depending on the microstructure. As listed
in Table 2.3, the activation energy measured by Darveaux et al. is slightly lower than the
expected one for pure Sn, while others have observed a much higher activation energy.
Igoshev er al. [52,53] suggested that in the analysis of the creep of Sn-3.5Ag solder,
some other mechanisms such as grain boundary sliding should be considered as well as
dislocation movement to account for the discrepancy in the creep data. The
microstructural studies of creep deformation appear to support this suggestion. Frear [48]
reported that the creep strain was accommodated through the Sn matrix with a grain size
of approximately 1 pm, at Sn—Sn grain boundaries, in compressive creep testing on the
eutectic Sn-Ag right-circular bulk sample. Igoshev et al. [53] noticed that microcracks
accumulated at grain boundaries due to grain boundary sliding, leading to intercrystalline
fracture during creep deformation of the eutectic Sn-Ag bulk specimen. The
accumulation of defects along grain boundary started at an earlier stage of creep and
lasted for at least 70 ~ 80% of the lifetime. In the study of Mathew et al. [45], the
eutectic Sn-Ag solder with 20 um grain size exhibited a higher strain-rate than pure Sn
with 150 um grain size. Even with additional Ag3Sn intermetallic phase, the fact that the
eutectic Sn-Ag solder was less creep resistant indicates that the deformation along grain
boundary significantly reduces the creep strength. However, such processes as grain
boundary sliding largely depend on microstructural features such as grain size. The Sn
grain size of Sn-3.5 Ag bulk used in the study of Frear [48] was about 1 mm, comparable

to the size of Aggsn particle formed by rapid cooling during soldering [15,16,25].

35

Although a great deal of efforts has been performed to understand creep
deformation of Sn-Ag solder, large discrepancies between experimental results still
remain. While the analysis of creep deformation mostly relies on the measured creep
data, there is no substantial study on the characterization of microstructures occurring
during creep deformation. Observed microstructural feature such as grain boundary
sliding does not support the measured creep data. In order to clearly understand creep
deformation of the eutectic Sn-Ag solder, systematic studies should be conducted by
using a consistent microstructure and specimen geometry. Study on microstructural

characterization must be conducted to verify measured creep data.

2.4 FATIGUE
2.4.1 Overview of Fatigue

Fatigue is defined as the phenomenon of failure of a material under cyclic loading
[54]. It is known that failure under cyclic stress or strain occurs at a stress level much
lower than that under conditions on monotonic loading. A fatigue failure is also insidious
because it occurs without any obvious warning, resulting in a brittle-appearing fracture
with no gross deformation at the fracture. Three basic factors are necessary to cause
fatigue failure, which are a maximum tensile stress of sufficiently high value, a large
enough variation or fluctuation in the applied stress, and a sufficiently large number of
cycles of the applied stress. There are some important parameters, as shown in Figure

2.11, which is useful in analysis of the fatigue phenomenon [55]. These parameters are

Cyclic stress range : A0 = amax "' 0'

min i

36

 

WM 6A
0.... U1 Time

Figure 2.11 Fatigue parameters [55].

 

stress

0

 

 

 

 

 

 

 

 

 

 

 

 

 

 

G A
A0 jg}:
i=5 0. ‘ 2
C . 1,
0 A0
Ag
215
BL , .1
Asp _A9
A8. — E
As

 

 

 

Figure 2.12 Notation for a symmetric hysteresis loop [56].

37

 

. . . _ max min _
Cyclic stress amplitude or altematrng stress : 0'0 - " ,

0' +0'

max min

 

Mean stress : 0'... =

In the cyclic loading in elastic regimes, stress and strain are related through the
elastic modulus. In such case, for any controlling factor, 0' or 8, the signal goes from O to
(o, e) to (—O', —8) and back to zero. For any cyclic loading that produces plastic strains,
however, the responses from a material are more complex. Figure 2.12 illustrates the
results schematically [56]. During initial loading in tension, the stress-strain curve is O to
A. On unloading and yielding in compression, we get to point B. On reloading in
tension from B and returning to point A, a hysteresis loop develops. The dimensions of
the hysteresis loop are described by the total stress range (A0) and the total strain range
(A8). The total strain range (A8) consists of elastic and plastic strains. The elastic
component is A8,, = Ao/E, where E is the elastic modulus, and the plastic component is
expressed as Asp = A8 — A8... The area of the hysteresis loop is equal to the work done or
energy loss per cycle.

_ There are two types of cyclic loading as shown in Figure 2.13 and 2.14
schematically, stress-controlled cycling and strain-controlled cycling [57]. In the case of
stress-controlled cycling, the controlling factor is the stress that oscillates between two
extremes, that is, the cyclic stress amplitude O'a is constant. However, the strain
amplitude is not constant. In the strain-controlled cycling, the strain range is constant and
the stress changes.

Since the plastic deformation produced during cyclic loading is not completely

38

 

 

 

 

 

l-s (DUI

 

 

 

 

 

Figure 2.14 Cyclic behavior of a material under strain control [57].

39

reversible, the structural modifications occur during cyclic loading and these can result in
changes in the stress-strain responses. Depending on the initial state a metal may
undergo cyclic hardening, cyclic softening, or remain cyclically stable. In the stress-
controlled cycling, the cyclic hardening occurs when a material increases in resistance to
further deformation with cycling. The cyclic strain then becomes smaller under the same
stress amplitude. On the other hand, cyclic softening occurs when a material shows an
increasing deformation with cycling under stress control. Figure 2.13 depicts the
phenomenon of cyclic hardening and cyclic softening under stress control. Figure 2.14
shows the cases of cyclic hardening and cyclic softening under strain control. The cyclic
hardening under strain control implies that resistance of materials to deformation
increases with cycling causing increase in stress to deform the material to the imposed
strain limits. The cyclic softening under strain control indicates that the material deforms
easily, resulting in a decrease in stress to deform to the imposed strain limits.

The behavior of a material under conditions of fatigue has been studied by means of
the S-N curve, a plot of stress S versus the number of cycles to failure N [58]. The stress
value in the S-N plot can be Ga, cm”, or 0mm, and the SN relationship is determined for a
specified value of cm. The S-N curve is mainly involved with fatigue failure at high
number of cycles (N > 105 cycles) where the nominal stress and strain are elastic. Figure
2.15 gives typical S-N curves for ferrous and non-ferrous metals [59]. The number of
cycle to failure increases with decreasing stress. There is a fatigue limit or endurance
limit given by a horizontal line which represents a stress level. Below this stress, the
material can be cycled without failure. The ferrous metals show a fatigue limit, while

non-ferrous metal does not have a true fatigue limit. The S-N curve in the high cycle

40

Cl)
r.

 

Stress Amplitude (S)

 

B
1 1 1 1

104 105 106 107 108
Cycles to failure (N.)

 

 

Figure 2.15 Stress (S) - cycles to failure (N) curves. A is for ferrous metals
and B is for nonferrous metals. SL is the fatigue or endurance limit [59].

41

region is described by the Basquin equation [60],

Nasz

,

where 0'a is the stress amplitude, and p and C are empirical constants.

When fatigue life or fatigue strength is considered, it is convenient to consider the
elastic and the plastic components of strain separately. It is also common to distinguish
by high cycle fatigue (N > 105 cycles) and low cycle fatigue (N <104 or 105 cycles). For
high cycle regime, the stress and strain on a gross scale are elastic. However, the
material deforms plastically in a highly localized region along persisted slip bands. The
conditions of low cycle fatigue are frequently caused by a thermal origin. Since the
thermal stress arise from the thermal expansion of the material, in this case the fatigue
results from cyclic strain rather than from cyclic stress.

The elastic component is readily described using the true stress amplitude and the

number of reversal (i.e. twice the number of cycles) [61],

 

0"
AgezgiL: __f_ 2'ny
2 E' E

where A6, is the elastic strain amplitude, 0;, the true stress amplitude, 0'; the fatigue

strength coefficient (equal to stress intercept at 2N; = 1), Nf the number of cycles to
failure, and b the fatigue strength exponent. This relation is an empirical representation
of the S-N curve above the fatigue limit. A smaller value of b results in longer fatigue

lives. The plastic strain component is described by Coffin-Manson relation [62],

42

A .
—:—p=8f(2ny

where A6,, is the plastic strain amplitude, e} is the ductility coefficient (equal to strain

intercept at 2N, =1), 2N; is the number of reversals to failure, and c the ductility exponent.
A smaller value of c results in a longer fatigue life. A valid equation for the entire range
(elastic plus plastic strain) of fatigue life is obtained by superposition of the elastic and

plastic strain component, which is,

Ae,_A.e, 3:... g ,
2 ——2—+ 2”— E 2Nf)”+ef(2Nf)‘.

 

Figure 2.16 schematically illustrates that the superposition of elastic and plastic
curves gives the fatigue life in terms of total strain [63]. The fatigue life curve in terms
of total strain tends toward the plastic curve at large total strain amplitudes, whereas it
will tend to the elastic curve at low total strain amplitudes. Figure 2.17 shows the
application of this equation to 95Pb-5Sn solder when tested in torsion from 0.05% to

2.0% at 25°C [64].

2.4.2 Fatigue Deformation of Solder Joints
The fatigue failure of solder joints in an electronic device results from the
imposition of strain due to different coefficients of thermal expansion of the joining

materials. The fatigue process in a device is controlled by the total strain, i.e. a solder

43

 

A28 (Log Scale)

 

 

01’ 1
Elastic 1 Plastic
2N1 (Log Scale)

Figure 2.16 Fatigue life in terms of total strain obtained
by superposition of elastic and plastic curves [63].

 

  

 

 

 

1o: 1 111nm] T ITIITTTI r 1 rrrnr] 1 1111111;

: 8 3

PE. 1:— r:

4 g 3

>33 3 :

Q _

g. 0.1“:— +AY E

5 +AY. 3

0.01 1 1 1111111 1 11111111 1 11111111 1 1111111
102 103 104 105 1O6

Cycles to failure (N,)

Figure 2.17 Number of cycles versus strain for 95$n-5Pb solder.
Torsion tests without hold time 25°C. [64].

joint experiences constant strain-controlled fatigue with dwell times at the strain extremes
[65]. The strain levels are determined by thermal expansion mismatch of joining
materials, thermal gradients in a package, temperature excursions during service, the
geometry of the solder joint, and compliance of the joint system [66].

The fatigue properties of solder materials used in electronic packaging are measured
by two different modes, isothermal fatigue and thermomechanical fatigue [67].
Isothermal fatigue is performed by mechanically cycling under a constant total or plastic
strain control at a constant temperature since a solder joint in a package undergo strain-
controlled fatigue. Thermomechanical fatigue is conducted by thermally cycling between
two temperature extremes, which is closer to the actual service conditions of electronic
packages and assemblies. However, the isothermal fatigue is easier to control, conduct,

and interpret, and less costly than thermomechanical fatigue tests.

2.4.2.1 Fatigue Failure Criteria of Solder Joints

The criteria of fatigue failure of a solder joint vary, causing difficulty in making a
comparison between data from studies that employed different variables. They can be
categorized as follows : catastrophic failure, surface cracks, load drop, and electrical
failure [67]. The catastrophic failure is the case when the specimen is separated into two
parts. The surface cracks are related to earlier processes in the failure sequence. In the
evolution toward a catastrophic failure, microcracks initiate at flaws, prOpagate, coalesce
to form a well—behaved crack that advances a certain amount per cycle and grows to
failure. This definition of failure was found to correspond to 1 mm of surface crack trace

per 1 m2 of surface area in 96.5Pb-3.SSn solder. Load drop is defined as the number of

45

cycles to reduce the total load range to 75%, 50%, or 25% of that of the first cycle in
hysteresis loop. When a different value of load drop is used as a criterion of fatigue
failure, it leads to a different number of cycles to failure. In industry, devices are
thermally cycled and solder joint integrity is monitored by mean of electrical resistance
measurements [68]. The electric failure criterion does not correlate well with load drop
because the load can drop 100% while the fracture surfaces are still in intimate electrical
contact, showing little or no change in resistance. Studies on the relationship between
load drop and resistance changes have shown that fatigue failure by load drop preceded
failure by an electrical criterion [68]. It is desirable to use isothermal fatigue for
estimates of thermomechanical fatigue life due to its convenience. However, studies
attempting to correlate isothermal fatigue data to thermomechanical fatigue data should

use the same failure criteria so that the comparison can be meaningful.

2.4.2.2 Isothermal Mechanical Fatigue of Solder Joints

To understand fundamental fatigue properties of solders, and to form a basis for
understanding thermomechanical fatigue properties, isothermal fatigue studies have been
performed in the strain and temperature ranges of interest. The most important factors
affecting fatigue properties include strain range, ramp time, hold time, temperature, and
aging [69].

Increasing strain range during cycling of materials leads to a decrease in the number
of cycles to failure. It is also obvious in the Coffin-Manson relation to have an inverse
relationship between strain range and number of cycles to failure. Effect of ramp time to

extremes of strain or stress in isothermal fatigue is considered by means of frequency in

46

the cycle. Frequency is usually taken as the number of cycles per unit time. However,
this definition of frequency needs a correction because the duration of a cycle consists of
both the ramp time and the hold time. Thus frequency is defined as the reciprocal of
twice the ramp time, separating frequency from the number of cycles per unit time [65].
This definition also makes frequency as a strain-rate variable. The reduction in frequency
is found to decrease the number of cycles to failure in isothermal fatigue. The effect of
frequency becomes more significant at high temperatures (> 0.5 MP.) because at high
temperatures, time dependent processes such as creep play important roles in deformation
and/or damage accumulation.

Hold times introduced in the cycle at maximum, minimum, or both strains, stresses,
or temperatures affect the fatigue life of materials. The number of cycles to failure
decreases with increasing hold time [65]. The reduction in fatigue life due to hold time is
attributed mainly to time-dependent creep deformation occurring during hold time.
When the effects of ramp time and hold time on fatigue life are compared, the effect of
hold time is much more dramatic than that of ramp time. The difference in the effects of
ramp time and hold time on fatigue life is attributed to differences in strain-rates
operating during ramp time and hold time. Strain-rates during very short hold times are
much slower than strain-rates during even very long ramp times. Since the fatigue failure
for solders are very sensitive to strain-rates, the effects of hold time and ramp time on
fatigue of solders thus are found to be dramatically different.

As temperature increases, the isothermal fatigue life is considered to decrease.
However, the degree to which fatigue life is affected depends on the material and testing

conditions. For near eutectic Sn-Pb solders, the effect of temperature is insignificant for

47

all strain ranges employed. The temperature effect on fatigue of low Sn Sn-Pb solders
such as 95Pb-SSn and 96.SSn-3.SSn is observed to be more pronounced than for near
eutectic Sn-Pb solders. The effect of temperature on fatigue life is also a function of
frequency and hold time. It is observed to be more significant for fatigue under low
strain-rate conditions, which correspond to long ramp time (or low frequency) and hold
time in the cycle.

The effect of aging on solder behavior is a function of composition and the initial
microstructure of the solder. Solders are generally two-phase materials with a low
melting point, where the solubility in each phase decreases with a decrease in
temperature. Precipitation of the second phase out of solid solution and phase coarsening
occur even at room temperature after the solder joint is solidified. When there is a
precipitation of the second phase during aging, the mechanical and fatigue strength
increase, while phase coarsening due to aging decreases mechanical properties of solders.
The fine microstructure of rapidly cooled solders leads to higher initial strength.
However, aging leads to a coarsened microstructure, causing reduction in strength. Thus,
aging is considered to neutralize the effects of initial microstructure on mechanical

properties of solders.

2.4.2.3 Thermomechanical Fatigue of Solder Joints

Thermomechanical fatigue of a solder joint arises when temperature fluctuations are
encountered in service, imposing cyclic strains in the solder joint due to the mismatch of
CTE. In contrast, isothermal fatigue involves imposition of mechanical strains at a

constant temperature. The testing methods of thermomechanical fatigue are divided into

48

four types depending on how the temperature is cycled and how the strain is imposed
[70].

0 Thermal cycling of actual component

0 Thermal cycling of simplified specimen

0 Thermomechanical fatigue of bulk solder

0 Thermomechanical fatigue of solder joint

Thermal cycling of actual component uses circuit boards as test vehicles for solder
joints. The whole electronic assembly is thermally cycled to impose strains on the solder
joints by the difference in CTE between the joined materials. The great advantage of this
technique is that the resistance to fatigue damage can be directly evaluated for a given
assembly system by monitoring the electrical and thermal continuity as a function of the
number of cycles. A difficulty with this method is that the strain and stress states in the
solder joints are very complex due to the specific geometry of the solder joints, which
makes the measurement and analysis of mechanical properties very difficult.

The time-temperature profile used in thermal cycling of actual component is usually
one of the three patterns illustrated in Figure 2.18 ; sine wave, sawtooth, and sawtooth
with hold times [71]. There are also three ways to apply temperature to the electronic
assembly ; thermal shock, thermal cycling, and power cycling. In thermal shock mode,
the solder joints are cycled with extremely short ramp times and some hold time at the
temperature extremes, resulting in very rapid deformation rate due to the rapid transfer of
heat. This could damage some electronic components. However, this is an easy test to
perform and accelerated results are achievable. In thermal cycling mode, the temperature

is cycled with a moderate ramp time to temperature extremes. Heat is transferred less

49

 

m -
U U11

Sine Wave

 

 

/\ -
V V

Sawtooth Wave

 

 

 

V V Time
Sawtooth Wave
with Dwell Time

Temperature/Strain Temperature/Strain Temperature/Strain

Figure 2.18 Schematic illustration of the three time-temperature profile in thermal
cycling. In these cycles, the strain and temperature follow one another [71].

50

rapidly and the deformation rate is slower than in thermal shock mode, resulting in
reduced possibility of damage of the solder joints during testing. Power cycling is
conducted by turning the electronic device on and off. The high temperature portion of
the cycle is generated by dissipating heat out of the package and into the solder joints
when the device is turned on. The low temperature portion is typically room temperature
or the temperature of the device when it is turned off. Power cycling is considered as the
primary source of thermomechanical fatigue, especially for consumer electronics [72].

Thermal cycling of a simplified specimen avoids the problem of complex strain
distributions in actual solder joints by employing a specimen with simplified state of
strain. Prototype solder joints are made by soldering materials of different thermal
expansion and thermally cycled between two temperature extremes. Figure 2.19 shows a
simplified test specimen to impose shear strain on the solder joint [73]. Normally the
strain imposed is either in simple shear or in a tension-compression orientation, as
defined by sample geometry. The advantage of this type of testing is the simplified strain
state in the solder joint such that microstructural evolution and failure characterization
can be studied easily. However, characterization of mechanical properties and the
number of cycles to failure is practically difficult in this method.

Thermomechanical fatigue of bulk solder is performed by externally applying
strains and temperature cycle simultaneously. Lawson applied this method to bulk
solders in flat dogbone design [74]. The temperature was cycled by generating heat by
radio-frequency heating coils in the grips and by forcing refrigerated freon through
machined chambers in the grips that hold the specimen. A tension-tension strain was

imposed by a load-frame. The advantages of this method are that a simplified tension-

51

   
 

  

Copper, (1 = 16x10’6/°C

‘ 11111111111110 (NI/01115111111111) Solder
' " 0.25mm

-a'='~“'25“x 101/10*" " -.

    
   
 

Copper

    

Figure 2.19 Simplified test specimen used by Tribula ef al.
to impose shear strain on the solder joints [73].

52

tension strain state can be achieved and mechanical data can be obtained during testing.
This method is similar to isothermal fatigue testing method so that a correlation between
thermal and isothermal fatigue can be made. However, the specimen used in this method
is fairly massive compared to a small solder joint, and it is most easily tested in tension-
tension mode, and not in shear mode.

Thermomechanical fatigue of solder joints attempts to collect mechanical data from
smaller scale solder joint having a simplified strain state. Frear proposed a testing
method using a double shear specimen where a servo-hydraulic load-frame imposes
cyclic strain under strain control and temperature is cycled by blowing resistively heated
air and liquid nitrogen vapor over the sample [48]. The temperature and strain were
controlled digitally. This method can achieve cycling from —55°C to 150°C. The
advantage of this method is that the strains are in simple shear and mechanical data are
easily measured with the load-frame. The drawback of this method is that the test is very
difficult because it is time consuming and needs special gas handling equipment and

control systems to be coordinated with the load-frame system.

2.4.3 Fatigue Properties of Eutectic Sn-Ag Solder and Solder Joints
2.4.3.1 Isothermal Mechanical Fatigue

The database of mechanical and fatigue properties of solder materials have been
established based on Sn-Pb solders during the past decade. However, those for lead-free
solders are still far from complete. Coffin-Manson low cycle fatigue law was often used
in the analysis of fatigue properties of bulk solders or solder joints. The form of Coffin-

Manson relation,

53

N?A£=C,

where N f is the number of cycles to failure, A6 total or plastic strain range, and aand

C are materials constants, has been frequently used to predict solder fatigue lifetimes
[75]. The load drop in the hysteresis loop was considered to be correlated with the
development and growth of fatigue cracks and used as a measure of the fatigue failure

[75-76]. The load drop was characterized by the load drop parameter (1) given by,

11117.1

where AP is the load range at any number of cycle and AP", is the maximum load range.
Solomon [75] performed isothermal mechanical fatigue testing of single shear lap
Sn-3.5Ag solder joints, which were made by soldering two brass or Cu blocks together, at
35°C and 150°C under plastic strain control. The microstructure of the solder joints
consisted of primary Sn dendrites with interdendritic eutectic phases. All the loading was
done with equal positive/negative plastic strain limit and loading rates (fully reversed
ramp loading), and a cycling frequency of 0.3Hz with no dwell time at strain extremes.
He examined load drop behavior from the hysteresis loop during fatigue testing and
analyzed fatigue lifetime on the basis of a 50% load drop as a fatigue failure criterion.
Solomon observed the load drop after the first cycle, or at most, after a few cycles. It

was found that the load drop followed a relationship of

<I>=ANB,

54

 

where N is the number of cycles, A and B are constants. B was measured to be 0.5 for
Sn-3.5Ag solder joints, so it was predicted that the crack growth rate will decrease as
cycling progresses and cracks grow larger. Figure 2.20 shows the comparison of fatigue
lives of Sn-3.5Ag and Sn-40Pb solder joints at 35°C and 150°C using 50% load drop
failure criterion. The number of cycles to failure was decreased with increasing plastic
strain range, following a Coffin-Manson low cycle fatigue law. The fatigue life of both
solders was lower at 150°C. At all strain ranges, the Sn-3.5Ag solder exhibited longer
fatigue life at both temperatures except that Sn-40Pb solder showed longer fatigue life at
the lowest strain range at 150°C. The fracture of the Sn-3.5Ag solder joints was observed
to occur by ductile hole growth and coalescence, which were more prevalent at 150°C.

Guo et al. [76] conducted isothermal mechanical cycling testing at room
temperature on the Sn—4Ag and Sn-37Pb single lap solder joint specimens. Testing was
performed under constant total shear strain control (Ax = 0.04 ~ 0.79), fully reversed
ramp cycling, and 0.1 Hz cycling frequency without dwell time at strain extremes. In
their study, the fatigue life was estimated using a 50% load drop criterion and the crack
growth rate was examined using load drop behavior in hysteresis loop during fatigue
testing.

Guo et al. compared the fatigue lives of the Sn-4Ag and Sn-37Pb solder joints at
50% load drop failure criterion as shown in Figure 2.21. The data followed the Coffin-
Manson type relation. The Sn-4Ag solder joint exhibited a higher fatigue resistance for
high strain ranges (A); 2 0.1) while it had lower fatigue resistance at low strain ranges
compared to Sn-37Pb solder joint. The relative fatigue resistance given for the load drop

(I) = 50% also occurred at other values of (I).

55

101

 

  

 

 

 

 

 

C 1 I 111ml 1 1 111111] T rrrnnl r r 1”an r r rnml I T rtrr_

C ' ' ' ' :

I g I
g, 1 g \ I

100.— ------------ \----; -------------------------------------- ~—_

C E ‘s I E
a . ‘ . s -
m “ ‘za 3‘ ‘
.5 . ‘ ‘ \ -
E 104; ------------------- ----- g-s- ---------------------- —,
"" : 1 ' :
m r E :
O 1 3 .
z: - ' - -
m Sn-3.5Ag, 35°C
2 10-2.—-- -------- ----~—
g 5 """Sn-40Pb, 35°C 3 z 5

: Sn-3.5Ag, 150°C 2

' """ Sn-40Pb,150°c 7

1 0.3 1 11111111 144111111 1 11111111 1 1 1111111 1 1 1111111 1 11111
10'1 101 103 105

Cycles to Failure

Figure 2.20 Comparison of the fatigue life behavior of Sn-3.5Ag and Sn-40Pb
solder joints at 35°C and 150°C using a 50% load drop failure criterion [75].

56

 

100 I I IIIIII] I I IIIIIII I I IIIIII] I I IIIIIL

  

 

 

A : C :
s: : :1 :
3 1 13115113 -
.5 ’ 1:1. ‘
CU — -
L-
6'5
1°":— 1
a ; a 3
g 1 ‘. El -
a, : +Sn-4A9 L; ‘35 1
To - "El-'Sn-37Pb .
fl
,2 ’ 0 = 50 °/. .
'2 I 1 1111111 I 1 1111111 I I 1111111 1 1111111

 

10° 101 1O2 103 10‘
Cycles to Failure

Figure 2.21 Comparison of fatigue lives of Sn-4Ag and Sn-37Pb solder
joints at room temperature using 50% load drop failure criterion [76].

57

Guo et al. proposed that the fraction of the fatigue—cracked area is directly

proportional to the load drop as follows

_A.
“A.

where A. is the cracked area and A. the original area. It indicated that the shear stress
(A1) acting on the solder joints remains essentially constant throughout a fatigue test with
a given Ax. Further, an estimation of the crack growth rate was obtained from the
estimate of the rate of crack area growth, which then can be obtained from the change in
the load drop with number of cycles. By taking an approximation for an average crack

length (ac) given by,
1
2ac = A/2 ,
the crack growth rate was estimated by

d_ac_ (A0¢)V2 d logo

dN 4N dlogN'

 

Figure 2.22 shows a plot of the crack growth rate (daC/dN ) versus stress intensity
parameter (AK ,, = [Mg/frat ) as 5 function of (I) for the Sn-37Pb solder joints. The Sn-
4Ag exhibited the similar behavior in crack growth rate. For a constant A71, dac/dN

decreased with increasing (D, i.e. the crack growth rate became smaller as the total crack
area (AC) increased. This finding corresponds to Solomon’s prediction on the crack

growth rate [75]. On the other hand, for a constant (I) (i.e. constant A), dac/dN

increased with increasing A74, i.e. with increase in Aror AK”. A comparison of the crack

growth rate at (I) = 50% is given in Figure 2.23 for the Sn-4Ag and Sn-37Pb solder joints.

58

1 0‘1

 

  
 

 

 

 

 

a ' ‘ "1ft? 1......5

E [,’_-' E

A : . III-i :
'g 104-:1- fl'l'; E
>0 E I: ." f
g _ 1.: q
a - ['5 Sn-37Pb -
1; 1°"; 1: '05: T: 300K 15
§ : ' +0: -5% 3
w, 10.4__ - O-o: 25% _
g I,"4 ,0 -El-<1>= -50%§
; flag}; :' --A--o= 75% j

. . men-1p: -90% ~

-5 1 1-115131111 1111111

10100 101 ‘102

A't(1ca)"2 (M Pa mm)

Figure 2.22 Fatigue crack growth rate (dao/dN) versus stress intensity
factor (A135; ) as a function of (I) for Sn-37Pb solder joints [76].

1o-1

? 1o-2
0
>1
0
E

-3

15, 10
12.

To 10-4
'U

10-5

1

 

 

 

 

More)"2 (M Pa m1/2)

j +Sn-4Ag j
:_ -e— Sn-37Pb o:
E (D = 50 % 0':
1- T = 300 K O / -
.— / -.
E 0 g '5
I ’ i
_ o / _
. /
E 06 f
1 O 1
0° 101

Figure 2.23 Comparison of the crack growth rate
at (I) = 50% for the Sn-4Ag and Sn-37Pb solder joints [76].

59

10'1 1 1

 

  
 

 

 

E III I ': l I 'IIIIE
: ’l’o :
” I". d
A "’ . .0 ‘
2 10-2; [211's .1
0 : l ' 2
> : ’0': j
3 »- '0... _j
E .3" I‘ .0. Sn-37Pb ‘
l— l I
£10; £'<?£> T=3OOK§
% : . +¢=5% 1
w 4 ’3', ' .—¢=25%
a 10' " l ‘ .B— _. o .1
u E ’0 no ¢-50/°§
f [a ; --A--o=75%3
_ ' :1' "‘9‘" tb = 90% 4
10-5 1 . 111'L1LL L 1 1 11111

 

10° 101 102
Away/2 (M Pa mm)

Figure 2.22 Fatigue crack growth rate (dao/dN) versus stress intensity
factor (Ah/7; ) as a function of (I) for Sn-37Pb solder joints [76].

 

 

 

 

10.15 i T I I j T T r
+Sn-4Ag
2 1o-2__ '63— Sn-37Pb Q
0 E E
f; : o = so °/. 0:;
E 1 T = 300 K o / -
-3__ _
g, 10 . gI i
s : 1 1
To 10-4r 0 ’ —_
1: E . / 5
E 06 f
1. . .1

10-5 1 1 a 1 1 1 1 1

10° 101

A1(1ta)1’2 (M Pa mm)

Figure 2.23 Comparison of the crack growth rate
at (I) = 50% for the Sn-4Ag and Sn-37Pb solder joints [76].

59

Apparently, the Sn-4Ag solder joints exhibited a higher crack growth rate for all strain
ranges studied. Thus it was suggested that the superior fatigue resistance of Sn-4Ag
solder joint at large strain ranges shown in Figure 2.21 is due to a higher resistance to
crack initiation.

Mavoori et al. [65, 77-78] tested the dogbone shaped bulk specimen of Sn-3.5Ag
and Sn-37Pb solders which were heat treated at 150°C for 24 hours followed by aging at
room temperature for a week. The specimens were tested for isothermal fatigue under
uniaxial loading and total strain control conditions (As, = 0.002 ~ 0.02) at 25°C and 80°C
in sawtooth pattern. The ramp time of 1 sec to strain extremes (i.e. 0.5 Hz cycling
frequency) was employed for tests with/without dwell time at maximum strain except for
tests where the ramp time was a variable. In their study, the number of cycles prior to
50% reduction in cm was used as a failure criterion (same fully reversed ramp cycling
as that used by Solomon [75] and Guo et al. [76] because Mavoori et al. used initial cross
section for stress calculation.

As shown in Figure 2.24, Mavoori et al. observed that the fatigue lives of the Sn-
3.5Ag solder was longer than that of Sn-37Pb solder in the test without dwell time at
25°C, following the Coffin-Manson type relation. However, the fatigue lives were
comparable at very low strain ranges. Figure 2.25 shows the effects of dwell time on the
fatigue lives during tests with As, = 0.006 at 25°C. An increase in tensile dwell time
dramatically reduced the number of cycles to failure. Beyond a certain dwell time, there
was no further reduction in the number of cycles to failure. The effect of dwell time is
greater for the Sn-3.5Ag solder. Figure 2.26 shows the effect of ramp time on the fatigue

lives during tests with As, = 0.01 and no dwell time at 25°C. Increasing time per cycle

60

 

101 l lllllll'l' T T llTlllI l llllllll lTrTllTL

T IITTTT

+ Sn-3.5Ag Z

:1? o --e--Sn-37Pb i
L

r ‘ ‘ O C 1
c
'- S‘
a \
1..“- 10°: 1
(D Z I
— F -1
5 : :
O
I— " T = 25°C, ‘

  

1 sec ramp tlme
no dwell time

.1 1 lllllll] 1 11111111 1 Llllllll l J_Llllll

1O2 103 104 105 10'5
Cycles to Failure

1

 

 

 

Figure 2.24 Fatigue lives of Sn-3.5Ag and Sn-37Pb bulk solders at 25°C
and with 1 sec ramp time and no dwell time using 50% omax [77]

1 5000

 

ITIIIIllITTTTTIHTTIITTTIllTTIllTTrTTTT

    
  
 
  
 

l

 

+Sn-3.5Ag ;

m --©--Sn-37Pb 1
h 1— -l
3 t = . ° :
= 10000__ total strain 0 6 A .1
If C T = 25°C, 50 % 0m: I
O : 1 sec ramp time 3
I." l- —
3 i i
E 5000:? ?

o

0 E“\ 21 /o ‘
E ‘6‘". -6. ------ 2.6-0/2 -@ g
r—llllllllllllllllJLlLlJLllllllllllllllll‘

 

 

 

100 200 300 400
Dwell Time (sec)

DO

Figure 2.25 Effect of dwell time on the fatigue lives using 50% omax failure
criterion during tests with A8,: 0.006 and 1 sec ramp time at 25°C [77].

6]

 

104 T T T TT—ml T F T TTTrTfi T T T TT7_TT

1 111111

TiTfirTTrTT

 

 

 

 

o
h
2
E 80 .... A
g 103: o" —_
m C i
g i 1
o 1 -
5' 1 +Sn-3.5Ag 1
"9"Sn-37Pb 1
2 l L 1_.LlLlll 1 l l [1111] l l l llLll
1 0'3 10-2 1 0-1 1 00

Frequency (Isec)

Figure 2.26 Effect of ramp time (frequency) on the fatigue lives
during tests with A191: 0.01 and no dwell time at 25°C [78].

 

 

 

 

 

101l— T T T T T T IT] I T T T T T T T_‘
Sn-3.5Ag + 25°C
A
o\° - 'A- - sooc ‘
V - .1
:13" 1 sec ramp time
<1 ” ' no dwell time
c O
I- o_ .—
2 10 C A :
.1- 1
‘0 : l
'5 _ -
H
o _ -
I" 1 _
10.1 L 1 1 J l 1 IL! 1 l l l l LLl
103 104 105

Cycles to Failure

Figure 2.27 Effect of temperature on fatigue lives of Sn-3.5Ag solder
during tests with 1 sec ramp time and no dwell time [78].

62

in tests without dwell time (i.e. decrease in cycling frequency and reduction of strain-
rate) led to a reduction in the umber of cycles to failure. The fatigue lives of the Sn-
3.5Ag solder were more sensitive to the ramp time variable than Sn-37Pb solder. The
effect of temperature was to reduce the number of cycles to failure with increasing
temperature as shown in Figure 2.27. There appeared to be a greater reduction at lower

total strain ranges.

2.4.3.2 Thermomechanical Fatigue

As with studies for Sn-Pb solders, the fatigue studies on the Sn-3.5Ag solder
concentrated on the assessment of the fatigue lifetime or resistance using Coffin-Manson
type relationships in the analyses. These approaches were criticized by Winterbottom [2]
in his work where the deformation process in the actual solder joint was studied using a
thermal cycling profile similar to that in real service condition and simulated using a
nonlinear finite element analysis. A number of geometric and microstructural features in
the actual solder joint complicate the deformation process. He has also pointed out that
the number of cycles to failure can be influenced by the frequency of temperature
cycling, temperature extremes, ramp times, dwell times, and the magnitude of the strain
ranges. It has become apparent that such a simple model as Coffin-Manson relation is an
over-simplification in assessment of fatigue life of the actual solder joints and a more
accurate model is needed for more accurate life predictions.

Several important results were obtained from the analysis of Winterbottom’s
simulation study on deformation process, which were verified in experimental results.

The first feature was that the distribution of the accumulated strain in the solder joint was

63

inhomogeneous and highly localized depending on the joint geometry. This does not
support the uniform strain assumption throughout the simplified solder joints or bulk
solders employed in application of the Coffin-Manson type relation. The localized strain
accumulation at certain locations suggested that the crack initiation and propagation
processes would be more appropriate for predicting fatigue lifetime than an approach
assuming uniform strain based on global displacements. Secondly, the deformation
produced during the heating and cooling regimes of the thermal cycling was not
reversible but produced continually changing distortion and strain distributions. As a
result, the joint shape and the stress distribution changed irreversibly as well. This
behavior implied that isothermal fatigue testing couldn’t reveal exactly the failure
mechanism, i.e. the cracking process, induced by thermal cycling. The nucleation and
propagation of fatigue cracks were observed to further alter the distribution of strain and
stress. Thirdly, the comparison of accumulated strain induced by time-independent
plastic deformation and time-dependent creep deformation provided insight into the
cracking process for fatigue failure. The crack trajectory in the solder joint was found to
pass through the regions having greatly accumulated plastic strain and very little creep
strain. With inputs from the study on deformation process, a failure criterion was taken
as follows. Once a critical strain density is reached at a location in the solder joint, a
crack is nucleated and extended. The solder joint will fail when the cracked area reduces
the joint strength below that required to support the imposed load.

In the earlier study using actual solder joints by Harada et al. [51], controlled-
collapsed chip joints (flip chip) of Sn-3.5Ag solder between two ceramic chips, whose

solderable area was plated with Ni/Au, were used. The solder joints were thermally

cycled between —50°C and +150°C in a thermal shock mode with 30 min dwell at each
temperature in air, i.e. a cycle/hour. Figure 2.28 schematically illustrates the geometry of
the solder joint and cracking in the solder joint. Fine fatigue striations with intervals of
0.04 ~ 0.2 pm were observed on the fracture surface. These striations were considered as
the crack length created during each thermal fatigue cycle, so the intervals of striations
were used to estimate fatigue crack growth rate. As the crack propagated during thermal
cycling, the striation interval increased. Thus, the crack growth rate increased with
increasing number of cycles, contrary to the results in the study using single shear lap
solder joints by Guo et al. [76] and Solomon [75]. This was attributed to reduced the
effective joint area that increased the stress during thermal cycling. It indicates that the
load did not decrease during thermal cycling, resulting in the increasing stress because of
the reducing joint area. This is contrary to load drop behavior observed in strain-
controlled isothermal mechanical fatigue that estimated a constant shear stress (AZ) acting
on the solder joints throughout a fatigue test. The striation intervals in the fracture
surface of Sn-3.5Ag solder joint were generally about a half of those of Sn—37Pb solder,
indicating the Sn-3.5Ag solder to have about twice fatigue life of Sn-37Pb solder for the
same joint geometry.

Frear [48] used the double lap shear Sn-3.5Ag solder joint in his TMF tests where
the cyclic strain and temperature were imposed simultaneously. The temperature range
was -55°C to +125°C and a total of 10% shear strain was imposed at a deformation rate
of 2.9 x 104/sec. The Sn-3.5Ag solder joint had a slightly longer fatigue lifetime than the
Sn-40Pb solder joint tested under the same condition. The Sn-3.5Ag solder joints failed

by intergranular cracking at Sn-Sn grain boundaries.

65

 

 

 

fSSn-BfiAgg} ———> W

Chip 2 .
Chip 2

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 2.28 Schematic illustration of geometry of and cracking
in the controlled-collapsed chip (flip chip) Sn-3.5Ag solder joint [51].

66

Shangguan [79-80] tested Sn-3.5Ag solder joints made by soldering surface mount
leadless ceramic chip resistors to Cu pad on printed wiring board substrates. The solder
joints were cycled in air under thermal shock mode between —40°C to +125°C with 18
minutes dwell time at each temperature extreme up to 1500 cycles. Figure 2.29 illustrates
the geometry of the solder joint and the crack propagation through the solder joint. The
cracks initiated at a location in the end of the component termination between 250 and
500 cycles. The cracks then propagated through the stand-off region towards the end
fillet region. Once the cracks reached the end of the component, they propagated either
upward along the component termination or along an approximately 45° direction in the
end fillet. The length of crack which propagated during every 250 cycles was measured
and the average crack growth rate was calculated by dividing the crack length increment
by the number of cycles (250 cycles). Figure 2.30a and b show the crack length as a
function of number of cycles and average crack growth rate, respectively. The crack
initiation took place between 250 and 500 cycles. Initial crack growth rate was rapid and
then became slow once the crack reached the end of the component. The crack growth
rate significantly decreased in the end fillet region to 0.18 jun/cycle between 1250 and
2500 cycles. It is thought to be due to the large volume of solder in end fillet region
compared to that in the stand-off region.

In the case of solder joints with leadless surface mount component, the crack
initiation and growth greatly depended on the geometry of the solder joint and volume of
the solder. The crack initiation did not appear to be the major portion of fatigue life. The
slow crack propagation rate in the end fillet region of the solder joints significantly

extended the total life of the solder joint. Since the crack initiated in the end of the

67

       

_ zaj termination
leadless ceramic

chip resistor

PWB substrate

Figure 2.29 Geometry of the surface mount leadless Sn-3.5Ag
solder joint and the crack propagation through the solder joint [79].

68

 

 

 

 

 

 

 

 

 

700 r—I FT T I T T l l I T T TT I T Tl T I T T TH] l T T T

A 500 E+ Sn-3.5Ag —
E : I
3 500 : j
5 : :
a, 400 1’ 71
g : 1
.l 300 r 1:
=1: 5 3
a 200 r j
5 L’ 2
_ -l
100 r t
o P— l l l l 1 Lil l l l l 1 1L l 14 1 l l 1 1+:

0 500 1000 1500

Number of Cycles
104 PT I T T T l T T l I T T T T T l l l T I r1 T T T TTT T_,
A ,_ a
a1 1 2 . -
‘2; - L‘ +Sn-3.5Ag :
3' 1 L .1
\ 1. ..
E : :
:1. 0 8 - —
V >- .1
3 C l
a: 0-5 r i
1: 1 -
.: 0.4 l L
III-0 - a
B : :
2 “-2 r 1
C5 : :
0 l 1 l l l I 1 l l 11 L 1 1 l L l l l l l l l l l l l l l
0 500 1000 1500

5'

Number of Cycles

Figure 2.30 (a) Crack length as a function of number of cycles, and (b)
crack growth rate in surface mount leadless Sn-3.5Ag solder joint [79].

69

component termination and rapidly propagated through thin stand-off region, it is critical
to increase the stand-off height to delay crack initiation and growth. The cracking and
fracture process of the solder joint in this study were identical to the criterion determined
in the work of Winterbottom [2].

Fatigue properties of solder joints have been analyzed using measured fatigue data
from bulk solder and simplified joint specimen. The fatigue deformation of those
specimens was characterized using a Coffin-Manson type relationship. However, it is
difficult to directly apply the fatigue data obtained from bulk solder and simplified
specimen to actual modules. This is mainly due to several other factors such as
geometry, complicate stress-strain state, microstructure, etc. Furthermore, there is no
microstructural characterization of fatigue deformation and damage. Fatigue properties
of solder joints should be examined incorporating distinct parameters related to each
specific type of solder joint to estimate the fatigue resistance or lifetime of solder joints.
The fatigue resistance of solder joints can be improved by optimizing geometry of solder
joints. Eventually, the performance of solder joints during thermomechanical fatigue can
be maximized by microstructural control of solder joints toward defem'ng fatigue
damage. Thus, as with studies of Sn-Pb solder joints, detailed microstructural study on
fatigue deformation and damage is required to better understand the fatigue properties of

Sn-Ag solder joints.

7O

 

2.5 INTERFACIAL INTERMETALLIC COMPOUND (IMC) LAYERS

The Sn-based solders form the interfacial intermetallic compound (IMC) layers at
the solder/substrate interface during soldering due to the reaction of Sn and the substrate
elements. The [MC layers grow by means of solid state diffusion and continued reaction
of Sn and substrate elements during service. In the case of the Cu substrates extensively
used in electronic applications, two IMC layers of Cu68n5 and Cu3Sn are formed at the
solder/substrate interface [81-85]. Only the Cu68n5 layer was found to form during
soldering and Cu3Sn layer is usually found only after aging at elevated temperatures [81-
86].

The physical properties of intermetallic compounds contrast significantly with those
of solder or the substrate materials. Interrnetallic compounds are less ductile, and less
thermally and electrically conductive. Table 2.4 provides physical properties of Cu-Sn
intermetallic compounds along with those of Cu and Sn for comparison [87-88]. The
intermetallic compounds used to generate this data were produced by rapid solidification
by supersonic, inert gas, metal atomization, and resulted in fine-grained single phase.
The hardness is greater for the intermetallic compounds, indicating their greater
brittleness. With these higher values of hardness for the interrnetallics, no plastic
deformation would be expected under normal levels of stress acting on the solder joints.
The elastic modulus is likewise greater for the interrnetallics, and interrnetallics usually
have a low ductility. However, the coefficients of thermal expansion for the
interrnetallics are similar to that of Cu and Sn. Thus, it appears that the intermetallics
would expand and contract much like their associated substrate material. Thermal and

electrical conductivities are generally lower for the interrnetallics, but they are considered

71

Table 2.4 Room temperature physical properties of interrnetallics [87-88].

 

 

 

 

 

 

 

 

 

 

 

Properties 0165115 Cu3Sn C“ S“
Vickers Hardness
(kg/mmz) 378 i 55 343 i 47 30 100
Toughness
(MPa-mm) 1.4 i 0.3 1.7 i 0.3
Youngs Modulus 85.56 i 1/65 108.3 i 4.4 117 41
(GPa)
Shear Modulus
(GPa) 50.21 42.41
Thermal Expansion
(x log/,0 16.3:l:0.03 19.0 :1: 0.3 17.1 23
Thermal ‘23“quin 0.145 i 0.015 0.24 i 0.024
(cm /sec)
Heat Capacity
( J lg ml deg) 0.286 1: 0.012 0.326 i 0.012 0.385 0.227
Res1st1v1ty 17.5 i 0.1 8.93 i 002 1,7 11,5
(uohm-cm)
Dens‘W 8.28 i 0.02 8.9 i 0.02 8.9 7-3
(gm/cc)
Thermal C°“d“c“v"y 0.341 i 0.051 0.704 t 0.098 3.98 0.67

(watt/cm-deg)

 

72

 

 

 

as minimal problems for a thin intermetallic layer.

While the formation of the IMC layers ensures the firm bonding of a solder joint,
the resulting thickness of the IMC layer is of significant concern to the reliability of the
solder joint. Two major concerns related to the IMC layer are the solderability loss and
the degradation of mechanical properties [81-85]. The solder pads of printed wiring
board and component termination are usually pre-coated with pure Sn or Sn-Pb solder as
a solderable metallization to improve their solderability during manufacturing. The
coating process also creates IMC layers, which in turn grow during storage.
Consumption of the solderable metallization due to the growth of the IMC layers in the
solid state during storage can strongly affect subsequent solderability. The degradation
of the mechanical properties is due to the embrittlement of the joint with excessive
growth of the [MC layers during service.

Frear et al. [48, 82] reported that the fracture toughness of the Sn-3.5Ag solder joint
was drastically decreased below a half of that of as-solidified one when the IMC
thickness increased up to 25 pm. They also found that the fracture of the Sn-3.5Ag
solder joint was dependent on the IMC thickness. In the as-solidified condition with
about 1.24 pm Cu6Sn5 IMC layer, the failure occurred via interfacial separation at the
solder/IMC interface, while the fracture propagated through the 45ij Cu3Sn layer after
aging 10 days at 205°C. After 205°C for 25 days, the fracture occurred through a 37.8p.m
Cu68n5 layer.

Yang et al. [15] showed the effects of IMC layer thickness on the mechanical
properties using the Sn-3.5Ag solder joints produced by two different reflow times of 20

and 1000 minutes during soldering. The solder joint reflowed for the longer time

73

exhibited no ductility in tensile test and significantly reduced strain at failure during
creep compared to that reflowed for short time. The solder joints with long reflow time
showed brittle fracture through IMC layer, while the solder joints reflowed for a short
time failed in the bulk solder.

The growth kinetics of the IMC layers in liquid and solid state was examined in
many studies. The growth of two IMC layers, Cu(,Sn5 and Cu3Sn, in solid state occurs in
planar manner by the diffusion of Sn and Cu in Opposite direction. As in Sn-Pb solder
[89-90], the solid state growth of IMC layers in the Sn-3.5Ag solder joint exhibited a

diffusion-controlled growth characterized by the model,

x=x0+J797,

where x0 is the initial thickness, x is the IMC layer thickness at time t, and D is IMC layer
growth rate constant [81, 82, 91]. However, Frear et al. [82] observed that the IMC
layers grew following 1°35 dependence instead of rm during aging at 205°C, suggesting a
different mechanism controlling IMC layer growth at high temperature.

For liquid state growth of IMC layers during soldering, Blair et al. [84] observed

”2 of IMC layer growth at temperature ranges of 228 ~ 326°C and for

the dependence of t
reflow times from 10 to 120 sec. However, London et al. [83] observed the 1m
dependence of IMC layer growth during soldering in the study using 240 ~ 360°C
temperature ranges and reflow time up to 216 hours. Choi et al. [86] observed that only

Cu68n5 layer was formed for reflow time less than 120 sec, while two layers of Cu6Sn5

and Cu3Sn were formed when the reflow time was more than 120 sec. The difference in

74

 

results between Blair et al. and London et al. appears to be due to the formation of Cu3Sn

layer when longer reflow time was used, affecting the grow rate of the total IMC layer.

2.6 EFFECTS OF REINFORCEMENTS

Several types of reinforcements have been used in attempts to improve mechanical
properties of the solder materials. The types of the reinforcement studied include carbon
fiber, metallic elements such as Cu, Ni, Au, Ag, Fe, In, and Sb, and intermetallic
compound particles such as Ni3Sn4, Cu68n5, and Cu3Sn. The intermetallic particles were
incorporated directly in the solder materials or by in-situ method [25, 26, 92—94]. While
the effects of reinforcements on the Sn-Pb solder were extensively studied [89-90, 93-
106], the studies using Pb-free solder are only a few. In the case of Sn-3.5Ag solder,
only some metallic elements and intermetallic particles by in-situ method were attempted
[5, 21, 25-26, 48, 92, 107-109]. For Pb-free solders, the reinforcements were
incorporated not only to improve mechanical properties but also to match the melting
temperature of the eutectic Sn-Pb solder.

For the eutectic Sn-3.5Ag solder, the metallic elements were incorporated to reduce
the 221°C melting temperature because it is too high to apply directly to manufacturing
processes. Miller et al. [107] found a ternary eutectic Sn-4.7Ag-l.7Cu which melts at
216.8°C. Further alloying with Sb resulted in a newly invented solder alloy, CASTIN,
Sn-2Ag-0.8Cu-0.6Sb, with melting range of 210 ~ 217°C [108]. Frear [48] extensively
studied the ternary Sn-4.7Ag-1.7Cu for creep, solid state IMC layer growth, TMF
deformation under the same condition as those used for Sn-3.5Ag solder. The

microstructure of the ternary Sn-4.7Ag-1.7Cu was similar to the eutectic Sn-Ag solder

75

except having additional Cu68n5 intermetallic particles. The ternary Sn-4.7Ag-1.7Cu
alloy exhibited very similar deformation behavior to the Sn-3.5Ag solder, with the stress
exponent of 6 in creep, decreasing fracture toughness with increasing IMC layer, and
similar number of cycles to failure during TMF. The small addition of Cu did not have
any substantial effects on mechanical properties of the Sn-3.5Ag solder.

McCormack et al. [21] examined two ternary alloys, Sn-3.5Ag-51n and Sn-3.5Ag-5
Bi, produced by incorporating In or Bi into Sn-3.5Ag solder. They obtained a reduced
melting temperature of 213°C and 212°C for the In and Bi additions, respectively. Both
types of additions strengthened the alloys at small strain ranges, but they severely
degraded the maximum strain limit that the alloy can withstand, showing a brittle
fracture. This was attributed to the segregation of In and Bi in Sn phase. They also
examined the ternary alloys Sn-3.5Ag-1Zn with a reduced melting temperature of 217°C
[5]. This alloy showed increased 0.2% offset yield stress, ultimate tensile strength, and
comparable elongation to the Sn-3.5Ag solder. It also exhibited significant creep
resistance with creep strain more than an order of magnitude smaller than Sn-3.5Ag
solder. The improvements of mechanical properties were interpreted in terms of
microstructural refinement. The added Zn suppressed the soft primary Sn phase and
refined Ag3Sn particles resulting in very fine and uniform two phase microstructure. Zn
was found exclusively within the Ag3Sn phase. It appears that Zn affected the
microstructure of primary Sn phase by refining Ag38n particles which may provide
nucleation sites of Sn phase.

Kariya et al. [109] investigated the effects of Cu and Bi on Sn-3.5Ag solder with up

to 1% Cu and 5% Bi addition during TMF using actual solder joint with Quad Flat Pack

76

 

(QFP) leads. They found that the increased addition of Bi drastically degraded the
thermal fatigue resistance of Sn-3.5Ag solder. They attributed the degraded performance
to the formation of Bi precipitates in Sn phase when more than the solubility limit of 2
mass% Bi in Sn was exceeded. The Cu addition did not change significantly the thermal
fatigue property of the Sn-3.5Ag solder.

The effects of reinforcements have been investigated on physical properties (e.g.
melting temperature), mechanical properties (e.g. creep, fatigue, and tensile), and
microstructure. However, there are very few studies about effects of reinforcements on
interfacial intermetallic layer growth. Furthermore, these studies were performed with
Sn-Pb solders and time scales ranging to a few hundred hours. It is necessary to examine
the effect of reinforcements on interfacial intermetallic layer growths during long-term

ranging to several thousand hours.

77

CHAPTER 3

THERMOMECHANICAL FATIGUE BEHAVIOR OF Sn-Ag SOLDER JOINTS

ABSTRACT

Microstructural studies of therrnomechanically fatigued actual electronic
components consisting of metallized alumina substrate and tinned copper lead, soldered
with Sn-Ag or 95.5Ag/4Ag/0.5Cu solder were carried out with an optical microscope and
environmental scanning electron microscope (ESEM). Damage characterization was
made on samples that underwent 250 and 1000 thermal shock cycles between —40°C and
+125°C, with a 20 minute hold time at each extreme. Surface toughening and grain
boundary cracking were evident even in samples thermally cycled for 250 times. The
cracks were found to originate on the free surface of the solder joint. With increased
thermal cycles these cracks grew by grain boundary decohesion. The crack that will
affect the integrity of the solder joint was found to originate from the free surface of the
solder very near the alumina substrate and progress towards and continue along the solder
region adjacent to the Ag3Sn intermetallic layer formed with the metallized alumina
substrate. Re-examination of these thermally fatigued samples that were stored at room
temperature after ten months revealed the effects of significant residual stress due to such
thermal cycles. Such observations include enhanced surface relief effects delineating the

grain boundaries and crack growth in regions inside the joint.

78

3.1 INTRODUCTION

The need for improving the structural performance of solder is increasing rapidly
since solder joints that connect surface mounted electronic components are structural as
well as electrical connections. The lack of detailed characterization of the realistic solder
joint microstructure, and its evolution in service, especially in lead-free solders,
represents a major hurdle to overcome before a solder microstructure that provides
optimal service performance can be engineered and introduced into manufacturing
settings. A highly recommended suitable alternative for Pb-bearing solders used in
electronic applications, especially for higher temperature service conditions, is eutectic
Sn-Ag solder. Such a solder under service environments like in automotive under-the-
hood, aerospace, and military application will experience severe thermal excursions that
will result in significant thermal stresses due to the thermal expansion mismatch between
soldered components and the substrate. Although there have been studies [75,80,110-
112,79,109] dealing with thermomechanical fatigue behavior of Sn-Ag solders, a
microstructural evaluation to develop a basic understanding of the processes involved is
far from complete.

Typically, manufactured solder joints have non-equilibrium casting microstructures
with considerable porosity. As the Pb-Sn solder joint undergoes long term
thermomechanical deformation, the microstructure coarsens heterogeneously and
develops well defined phase boundaries, especially in highly stressed areas near the
interface, and the coarsened regions are more susceptible to fatigue crack initiation
[113,114,73]. In addition, a continuous, hard intermetallic layer forms at the

solder/substrate interface. This layer and the constraints imposed by the substrate affect

79

 

the stress state and the deformation behavior in the solder joint [115,65]. When hold
times between temperature changes are imposed, the number of thermal cycles to failure
is substantially reduced [116-118]. There has been significant activity to model the
complex stress histories that occur in solder joints in order to predict the failure
processes. These analyses usually use mechanics based approaches for modeling the
complex creep-fatigue conditions imposed by thermal cycling [110,119-124]. To
develop suitable material models for reliability predictions, it is necessary to characterize
the processes associated with microstructural evolution, particularly generation of
microstructures that lead to plastic instability that stimulates the failure process to explore
ways to prevent or retard damaging microstructural changes such as coarsening.

It is important to note that thermomechanical fatigue in Sn-Pb solders have been
extensively investigated during the last decade. However, such a knowledge base cannot
be directly applied to Sn-Ag solders because the phase fractions and solubility behavior
of the two phases are distinctly different. Pb has high solubility for Sn but not vice versa,
and the Pb rich phase is ductile, has a volume fraction near 30% in Pb-Sn solder ; in
contrast to Ag3Sn, which is hard and comprises only about 4% of the volume in Sn-Ag
solder. The main aim of the proposed study is to investigate the thermomechanical
fatigue damage in an actual electronic component assembly consisting of tinned copper

lead soldered to metallized alumina substrate with Sn-Ag solders with or without Cu.

3.2 EXPERIMENTAL DETAILS
Several, as-made or thermally cycled electronic modules, containing chip

components soldered to a rigid alumina substrate circuit board were examined during the

80

 

course of this investigation. Two types of solders, 96.5Sn/3.5Ag (eutectic Sn-Ag) and
95.5Sn/4Ag/0.5Cu, were used to connect the lead of chip component to the alumina
substrate metallized with 99%Ag/1%Pt. The lead of the chip component was made of Cu
and was plated with Ni/Sn or Sn-Pb.

Solder paste was screened to the solder pad on the metallized alumina substrate to a
thickness of 150 ~ 200 um. The circuit boards were then reflowed with a maximum
temperature of 235 ~ 240°C for less than 20 sec in a nitrogen environment. Figure 3.1
shows a schematic drawing of the solder joint connecting the lead of the chip component
to the substrate. The solder joints were subjected to thermal shock by cycling between -
40°C and +125°C for 250 or 1000 cycles with a 20 minutes hold time at extreme
temperatures. The two temperature extremes were chosen to represent the harsh
conditions in the automotive under-the-hood environment [80, 2].

The initial examination of the surface appearance of the as-made and thermally
cycled solder joints was carried out using optical microscopy and environmental scanning
electronic microscopy (ESEM). Thereafter, the solder joints were mounted in epoxy
resin, cross-sectioned using a diamond saw, and wet polished to 0.05mm finish using a
Si02 colloidal suspension. Polished sections of the solder joints were examined with
ESEM and energy dispersive spectroscopy (EDS). ESEM facilitated direct examination
of the solder joint assemblies without additional preparation such as gold or carbon
conductive coating that will be required for analysis in conventional scanning electronic
microscopy [111, 125]. These solder joints were also reexamined using ESEM without

any further treatments after about 10 months.

81

 

  
   

Cu Lead

Solder

Alumina
Substrate

(a) A

 

 

 

 

 

 

 

 

 

Cross section in A direction Cross section in B direction used
used to examine as-soldered joint to examine thermal cycled joint
0)) (C)

 

Figure 3.1 A schematic drawing of solder joint used in this study.
The cross section in A and B direction is used to examine
as-soldered and thermally cycled joints, respectively.

82

 

3.3 RESULTS
Surface Appearance and Microstructure of the As-Soldered Joint

Optical and ESEM micrographs of the surfaces of the as-soldered
95.5Sn/4Ag/0.5Cu solder joints are given in Figure 3.2. There was no observable
damage on these surfaces. The solder joint surface consisted of almost pure B-Sn, Ag3Sn,
Cu6$n5 interrnetallics, and Pb phase. The B-Sn phase exhibited an elliptical-shape
morphology. Ag3Sn interrnetallics existed in needle and disk shapes, and were embedded
within the B-Sn phase. Hexagonal shaped Cu6Sn5 interrnetallics were also prevalent on
the surface of the solder joint. Pb phase was found mostly between B-Sn grains at
various regions of the solder joint.

The microstructures of a cross-section (indicated in Figure 3.1b) of as-soldered
95.SSn/4Ag/0.5Cu solder joint are given in Figure 3.3. The solder comprised of pure B-
Sn phase matrix with eutectic Sn-Ag phase that is a mixture of B-Sn and small needle
shaped Ag3Sn interrnetallics. Pb phase (bright areas) was also identified and its existence
was attributed to the Sn/Pb solder tinning used for the lead of the chip component. Few
irregular shaped Cu6Sn5 interrnetallics were also observed. Ag3Sn intermetallic layer was
formed between solder and alumina substrate as a result of reaction between Sn from the
solder and Ag from the 99Ag/1Pt metallization as shown in Figure 3.3(b). This Ag also
reacted with Sn resulting in the formation of large rod shaped Ag3Sn interrnetallics,
which could have broken away from the Ag3Sn layer and entered into the bulk solder
region during reflow.

The microstructure of the regions between solder and Cu component lead is

provided in Figure 3.3(c). Sn-Pb layer tinning on the surface of the Cu component lead

83

 

Figure 3.2 (a)Optical and (b)ESEM micrographs of surface
appearance of as-soldered 95.5Sn/4Ag/0.5Cu solder joint.

84

\ luminzl
(l )

( u cornponcnt's lcml

30jun

pin
\ilapcr " §-——- holrs_, ‘

 

Figure 3.3 ESEM and optical micrographs showing microstructures of a
cross-section of as-soldered 95.5Sn/4Ag/0.5Cu solder joint in the direction
shown in Figure 3.1b. (a) is the bulk solder region, (b) is the region near
Agasn intermetallic layer, (0) is the region near Ni coating layer on Cu lead.

85

dissolved in molten solder during reflow resulting in the formation of Pb phase in bulk
solder region as shown in Figure 3.3(a). The Ni layer plating on a Cu component lead
exhibited discontinuity due to several pin-hole defects. Molten solder penetrated through
the pin-hole defect and reacted with the Cu lead resulting in the formation of Cu3Sn and
Cu6Sn5 interrnetallics inside the hole. Similar damage in the Ni plated leadless chip
resistor has been identified in a previous study [111]. Since the Ni protective layer was
breached, ternary intermetallic layer consisting of Sn, Ni, and Cu replaced Sn or Sn-Pb

layer tinning adjacent to the Ni layer.

Surface Appearance and Microstructure of the Thermally Cycled Joints

Surface appearances of representative eutectic Sn-Ag and 95.5Sn/4Ag/0.5Cu solder
joints after 250 and 1000 thermal cycles are shown in Figure 3.4. Thermally cycled
solder joints exhibited significant surface cracking that was circumferential to the
component lead, irrespective of which of the two solders was used to make the joint.
After 250 thermal cycles, the 95.5Sn/4Ag/0.5Cu solder joint showed more surface
cracking than the eutectic Sn-Ag solder joint. After 1000 cycles, the surface appearance
of the 95.SSn/4Ag/0.5Cu solder joint was slightly more grainy and rougher compared to
the eutectic Sn-Ag solder joint. However, the crack density on the joint surface was
apparently the same for both types of solder joints after 1000 cycles. The extent of the
opening of the cracks on the joint surface after 1000 thermal cycles was more than that
after 250 thermal cycles.

Closer examinations of the representative cracks on the surface of the solder joint

shown in Figure 3.4 are presented in Figure 3.5. Figure 3.5(a) shows that the cracking

86

8.96 35.9: 8“ 5%. 25.. 528 3V 8333693 Eu 3 3-5 2623
9:35.852 ho moocmcmoaam momtsm Co 23523:. .330 g a 3 in 95mm

3.98 cmn
war—.1. 9:93.:—

W 3.93:3
e emiufiéimma

 

87

$.93 .255 82 5% E2 828 A3 sonobgcmmmm Em A3 915 2523
02235859 Co moocmnmoaam course 8 23692:. 8230 A3 a. A8 vd 239“.

3.3... :2:
3.-.} 233.5

 

88

 

Figure 3.5 ESEM micrographs of circumferential crack on the surface
of solder joint. (a) Eutectic Sn-Ag solder joint after 1000 thermal cycles,
(b) and (c) 95.5Sn/4Ag/0.5Cu solder joint after 1000 thermal cycles.

89

occurred between B-Sn grains during thermal cycling. The cracking between B-Sn grains
indicates that the microcracks developed between elliptical Sn grains evolved as a
macrocrack that is circumferential to the component lead. Figures 3.5(b) and 3.5(c) also
show that the cracking along the grain boundaries between B-Sn grains at the bottom
surface of the open crack. The bottom surfaces of the open crack revealed striations that
were parallel to the direction of the crack opening and closing during thermal cycling as
shown in Figures 3.5(b) and 3.5(c). These striations were also found to be perpendicular
to the circumferential direction of cracking around the component lead. The schematic
drawing given in Figure 3.6 illustrates the direction of striations with respect to the
circumferential cracks. These striations were found at the bottom surface at every
location of the circumferential crack around the component lead as illustrated in Figure
3.6. These striations are attributed to rubbing of the harder Ag3Sn particles present at the
grain boundaries on the grain surface during opening and closing of the crack.

The microstructures of eutectic Sn-Ag and 95.SSn/4Ag/0.5Cu solder joint after 250
and 1000 thermal cycles, provided in Figure 3.7, illustrate the coarsening of the Ag3Sn
interrnetallics. Small needle shaped Ag3Sn interrnetallics present in as-joined condition
(shown in Figure 3.3(a)) have coarsened into more rounded particles and are found to be
more uniformly distributed. The details of the microstructural coarsening have been
discussed in other previous studies [25,26,92]. The coarsening behavior of the
constituent phases was prominent with increased number of thermal cycling. Ag3Sn
layer that formed between the solder and 99Ag/1Pt during reflow also exhibited
significant growth during thermal cycling, depending upon the number of cycles as

shown in Figure 3.8.

90

 

 

 

Figure 3.6 A schematic drawing showing the circumferential crack on the free
surface of the solder joint and striations on the bottom surface in the open crack.

91

$32888. 58.923698 new. 3.5 2.85.... an E Eu E
.865 .9255 0mm Eta E5. .528 8 2382622.: 05 Co 298992.: 2mmm 3v w 3 En 2:9".

 

92

562688. 58535598 Em grew 28%. as A3 Em §
.mo_o>o 3:55 ooow noun E9. .528 co 9202592,: 9: .0 2.3522,: $.me A3 w 3 hm 2:9“.

1.7.59 .
“Nil”: .. a
A .. .. ,

 

93

29:088. 6296 Echo...

82 5% 38.035598 as... 2.5 9628 2m :3 use §

262588. $.26 .255 SN 5% 38.035288 Em 2-5 2598 9a 3V Ea E

Elam

Suzanne mc_E:_m co co=m~=§oE E 35¢ 28 .5qu comics
.93. 2:53:25 cmmm< Co mzamcmocgE Ewmm Qm 2:9”.

.2: 5:7.

.5 i4?
emf

3v

5:537.

 

94

Studies on the Cross Section of the Solder Joint That Underwent Thermal Cycling

Microstructure observed in the cross-section (indicated in Figure 3.1(c)) of the
eutectic Sn-Ag and 95.5Sn/4Ag/0.5Cu solder joint after 1000 thermal cycles are given in
Figure 3.9. A major crack, which can lead to failure of the solder joint, was observed in
each solder joint after 1000 thermal cycles. An important microstructural illustration of
grain boundary (GB) sliding indicated that during the thermal cycling significant
deformation occurred only in few locations. Such features were observed in regions near
the free surface of the solder joint (denoted by A), regions around the major crack
(denoted by B), and regions around the comer of the Cu lead (denoted by C). However,
there was no such a microstructural indication associated with deformation in the solder
joints that underwent 250 thermal cycles

The regions near the free surface of the solder joint (denoted by A in Figure 3.9)
that exhibited deformed B-Sn grains due to thermal cycling are shown in Figures 3.10(a)
and 3.10(b). It appeared that B-Sn grains present in this region experienced GB sliding
due to thermal cycling. Such GB sliding was not observed in the interior regions of the
solder joint although it was pronounced in regions near the free surface of the solder
joint. Figures 3.11(a) and 3.11(b) reveal that similar GB sliding occurred in regions
around the major crack (denoted by B in Figure 3.9). This indicates that GB sliding was
accompanied by the formation and propagation of the crack along grain boundaries of B-
Sn. It is also apparent that the crack propagated through regions in the bulk solder near
the Ag3Sn layer.

Cracks observed in thin regions between the Cu lead and Ag3Sn intermetallic layer

on the metallized substrate of the 95.5Sn/4Ag/0.5Cu solder joint are illustrated with

95

Eutectic Sir-Ag

95.5., n/4Ag/0.5(‘u

 

Figure 3.9 ESEM micrographs of cross-section of solderjoint
after 1000 thermal cycles. (a) is the eutectic Sn-Ag solder joint
and (b) is the 95.5Sn/4Ag/O.5Cu solder joint.

96

 

Figure 3.10 ESEM micrographs of a region near free
surface of the solder joint after 100 thermal cycles.
(a) is the eutectic Sn-Ag solder joint and

(b) is the 95.5Sn/4Ag/0.5Cu solder joint.

97

26:83... sonob<scmnmm 9: a 3v Eu :3 528 2-5 0:023 as a E
.363 _m§m£ ooor Eta E_o_ $28 9: Co :23. x33 55:; x090 o5 9:265 2.35928 Ewmw Cd 9:9”.

«5:57.

55:52

.1 5V.

. ...,,_...E:w§o Ea =me<

 

micrographs given in Figure 3.12. The cracks propagated along the boundaries between
the Ag3Sn layer and bulk solder. In the region around the comer of Cu lead (marked by
C in Figure 3.9(b)), the GB sliding, which was identified in regions near the free surface
and around the major crack, was also prevalent. The eutectic Sn-Ag solder joint,
however, did not reveal any feature of deformation such as GB sliding and crack in the

same region (around the comer of the Cu lead).

Re-Examination of the Cross Section of the Solder Joint

The re-examination of the cross-section with ESEM of the solder joints shown in
Figure 3.9 were performed without any additional treatments after 10 months since the
first examination was carried out. Figure 3.13 compares the microstructures seen in the
first and the second examination of the interior region away from the free surface, and the
region near Ag3Sn intermetallic layer, in the cross-section of the eutectic Sn-Ag solder
joint. The microstructural characteristics of GB sliding were observed in both regions
during the second examination where no microstructural evidence of GB sliding was
found in the first examination. It is obvious that the cross-section of the solder joint
underwent further GB sliding during storage in regions where no microstructural
characteristics could be observed in the first examination. The regions beneath the Cu
lead in the eutectic Sn-Ag solder joint that did not exhibit cracking during the first
examination; however, the same region exhibited cracking during the second examination
as shown in Figure 3.14. The crack propagation was along the grain boundaries of B-Sn
in the region around the comer of the Cu lead where the GB sliding was prominent. The

crack path changed to the interface between Ag3Sn intermetallic layer and the solder in

99

(in lead

Alumina

 

Figure 3.12 ESEM micrographs of (a) crack in the thin region
between Cu lead and 99Ag/1 Pt metallized alumina substrate
and (b) microstructure in the region around the comer of Cu lead.

2026032 .coszmem 2508 new “we: :_ 83.3. m9: 9: :8: >955 .552 8:25 9:
E3952 3v ace 3 .cozngmxm 2003 use “2: E E9. .528 91% 0:023 on.
.o 9559 9:3 22 Co 8582392,: Co mcnmumofiE 2mmm fly a E mtm 2:9“.

 

101

5.6283me :98. 2:33:25 cmmm< Em: 859 65 29852 AB uca A3 .coszmeo 9.88 can “2: 5
$6.. 528 2.5 2823 9: Co 2059. mEmm 92 Co 3582592,: Co 29238? 2mmm A3 a g 9d 959”.

5.25:4 .
r. :3? w .5 :32.

.3? =5 ,. . . .,.
.. .. ._. yam/W _ ,, 3:: =mfl<

 

102

('u lczul

: A'ggSn: ‘ ‘
993:1] I l’t
Alum inn

(:1)

(‘u lead

Ni layer

" z

. a
n
\-

1,
Alumina

 

Figure 3.14 ESEM micrographs of microstructures of regions
beneath Cu lead of the eutectic Sn-Ag solder joint. (a) is the region
around the comer of the Cu lead and (b) is the region beneath Cu lead.

103

the regions beneath the Cu lead. In this region only small amount of solder is present
compared to that around the comer of Cu lead. Unfortunately, microstructural features in
the 95.SSn/4Ag/O.5Cu solder joint could not be observed in the second examination due
to severe contamination of the specimens during storage.

As-soldered joints also exhibited the microstructural characteristics of GB sliding in
regions near Ag3Sn intermetallic layer as Figure 3.15 reveals. However, the extent of GB
sliding in the as-soldered joints during storage appeared to be much less compared to the

solder joints subjected to 1000 thermal cycles.

3.4 DISCUSSION

The present microstructural study dealt with deformation and fracture aspects in
two Sn-Ag based solders during thermomechanical cycling of an actual electronic
component. This section addresses the implications of the findings and possible reasons

for the same.

Efl'ects of Pb Contamination

There have been concerns about the effects of tinning of Cu leads with Pb
containing solders [17,126,127]. Pb entering the Sn-Ag solder results in affecting its
higher temperature capabilities, due to the formation of low melting ternary eutectic
phase of 178°C [128]. During the course of this study we did observe the presence of Pb
containing phases within the solder, distributed throughout the joint. However, if such
phases were to affect the thermomechanical fatigue behavior of the solders studied, the

region where the detrimental crack could form and grow should be along the regions

104

 

Figure 3.15 ESEM micrograph of a region near Agasn intermetallic in
as-soldered 95Sn/4Ag/0.5Cu solder joint in the second examination.

105

where the Pb contamination should be a maximum. Contrary to this expectation the
detrimental crack formed near the ceramic substrate where the effects due to mismatch in
CTE will be a maximum. However it should be emphasized that the highest temperature
experienced by the solder joints in this study was only 125°C, reasonably below the
melting temperatures of low melting Pb-containing phases (178°C) that will be present in
the contaminated solder. If higher temperatures are involved in the thermal cycling, so as
to evaluate the higher temperature capabilities of the solders studied, problems due to Pb

contamination may arise.

Effects of Breached Ni Coating on Cu Lead

There have been significant concerns regarding the defects in the Ni coating,
especially the electroless coating, on the copper substrate [129-131]. Similar defects
were found in Ni coating on leadless chip resistor [111]. Defects such as pinholes present
in such a layer can facilitate the diffusion of copper and tin, and promote formation of
interrnetallics that can affect the integrity of the joint. During the course of this study
formation of such interrnetallics, especially mushroom-type growth within the Ni coating
and the Cu lead, have been observed. If these were potential sources for
thermomechanical fatigue failures, the detrimental crack that affects the integrity of the
joint should be in the region near the copper lead. However, this was not the case for the
temperature range and holding times used during thermal cycling in this study.
Obviously the CTE mismatch seems to be more of a contributing factor for the failure in

the component studied, rather than breached Ni coating. However, a breached Ni coating

106

on the substrate may significantly affect the thermomechanical fatigue behavior of the

solder joint.

Effects of Small Amount of Copper Addition to Sn-Ag Solder

During the course of this study no significant differences in thermomechanical
fatigue behavior due to the small addition Cu to Sn-Ag solder could be detected. Such an
observation tends to indicate that the breached Ni-coating on the copper lead, or other
means of impregnating the solder with small amounts of Cu will overshadow the effects
of small additions of Cu to Sn-Ag solder. However, it should be noted that the small
decrease in ternary eutectic temperature to 216.8°C [107] due to the addition of Cu to the
Sn-Ag solder still results in a melting temperature well above the highest temperature

used for thermal cycling of this study.

Thermomechanical Fatigue Crack Nucleation

In all the samples observed it was apparent that the thermomechanical crack
nucleation always started on the free surface. There have been studies indicating that
stress imposed decreases and strain imposed increases during thermomechanical fatigue
[75,123] Some other studies indicate that stress-strain hysteresis loop during thermal
cycling stabilizes after a few cycles [110]. In stress relaxation studies on single shear lap
specimens made with the eutectic Sn-Ag solder joints having similar joint thickness, the
stress relaxation after about an hour at room temperature was about 50% and that at

150°C was about 60%, for comparable imposed strains [132]. All these observations are

107

based on specific joint geometry and joint thickness. Such observations elucidate the role
of constraints on stress relaxation.

Unlike in controlled studies that use single shear-lap specimens or joint specimens
with controlled geometry, the free surface area of the solder joint to the interface area
between solder and substrate/component is much higher in the solder joints studied. As a
consequence, the strain accumulated due to thermomechanical fatigue could dissipate
from regions of high strain concentration to the free surface. This is more probable based
on hot-deformation condition that exists in solders due to the high homologous
temperatures experienced by the solders. Such strain dissipation at the free surface could
create surface undulations that provide potential sites for crack opening. This dissipation
of strain/stress appears to be accommodated by grain boundary sliding. Although such
crack nucleation could occur any where on the free surface of the solder, presence of the
copper lead, with its own CTE, seems to favor the circumferential surface cracking
observed during the course of this study. However, these circumferential cracks did not
prove to be detrimental due to the soft interior of the bulk solder joint that was able to
accommodate the strain gradients by plastic deformation, and deter the crack
propagation.

The location of the detrimental crack tends to indicate that the region that
experiences the maximum effects of CTE mismatch is the potential location for it.
Although studies based on Pb-Sn solders indicate that localized phase coarsening in such
a region contribute to the thermomechanical fatigue failure [113,114,73], no significant
coarsening of grains in such regions was observed in Sn—Ag solders used in this study.

However, such an observation does not preclude the possibility of dynamic

108

recovery/recrystallization in such a region. Such microstructural rearrangement will also
facilitate stress relaxation. The important finding is that there is no grain
coarsening/growth in these regions. Grain sizes in these regions are about the same as

that in the as-made joint.

Thermomechanical Fatigue Growth of the Catastrophic Crack

The catastrophic crack that contributes to the thermomechanical fatigue failure was
found to nucleate on the free surface near the interface between solder and intermetallic
layer on the metallized ceramic substrate. This crack propagated by grain boundary
decohesion/sliding within the solder in a region adjacent to this interface. This crack did
not reach the solder/intermetallic interface within the bulk of the solder joint. If and
when the intermetallic region grew due to aging, this crack tended to reach this interface.
Such a feature can be realized in the very thin region underneath the copper lead where
the amount of leftover solder was very little. This observation tends to indicate that grain
boundary sliding within the solder to be dominant mechanism of thermomechanical

fatigue failure in the Sn-Ag solder used in the electronic component studied.

Residual Stresses due to Thermomechanical Fatigue

Another important finding of this study is the existence of significant amount of
residual stress due to thermomechanical fatigue. Such an effect became obvious by the
enhanced relief effects revealing grain structures noted in specimens re-examined after
about ten months following the initial examination, with no additional specimen

preparation. The residual stresses that exist in thermomechanically fatigued samples are

109

much higher than those present in as-joined samples, as evidenced by the much enhanced

surface relief in the former compared to a bare minimal one in the latter.

3.5 SUMMARY

Following summarizes the findings based on studies with actual electronic

components containing metallized ceramic substrate and tinned copper leads soldered

with Sn-Ag or 95.SSnl4Ag/O.5Cu solder:

1.

Thermomechanical fatigue damage in Sn-Ag and 95.SSn/4Ag/0.5Cu solder joints
occurs by processes that are different from those observed in Pb-Sn solders. There
appears to be no grain coarsening in Sn-Ag solders as a consequence of thermal
cycling.

Cracks formed due to thermal cycling in Sn-Ag and 95.SSn/4Ag/O.5Cu solder joints
initiate on the free surface of the solder joint and progress towards the interior
regions of the joint.

Crack growth due to thermomechanical fatigue is along the grain boundaries of [3-
Sn phase. Opening and closing of these cracks during each thermal cycle results in
striations on the grain boundary surfaces that intersect the opening crack. These
striations can be due to the rubbing of Ag3Sn particles present at the grain
boundaries onto the grain boundary surfaces.

The crack that will contribute to the ultimate thermomechanical fatigue failure of
the joint occurs very near the ceramic substrate, initiating at the free surface of the
solder joint and progressing inwards and continuing along the solder region near the

interface with Ag3Sn intermetallic layer formed with metallized ceramic substrate.

110

Significant amount of residual stresses resulting from thermomechanical fatigue can

cause further damage to solder joints during storage at room temperature.

Presence of small amounts of copper does not significantly affect the
thermomechanical fatigue behavior of the Sn-Ag solder joints.

Presence of Pb phase in the solder, arising from tinning of Cu lead, also does not
appear to affect the thermomechanical fatigue behavior of the Sn-Ag solder joints

as long as the maximum temperature experienced is below 150°C.

111

CHAPTER 4

CHARACTERIZATION OF THE GROWTH OF
INTERMETALLIC INT ERFACIAL LAYERS OF Sn-Ag AND Sn-Pb
EUTECTIC SOLDERS AND THEIR COMPOSITE SOLDERS ON
Cu SUBSTRATE DURING ISOTHERMAL LONG-TERM AGING

ABSTRACT

Single shear lap joints were made with four different solders, Sn-Pb and Sn—Ag eutectic
solders, and their composites containing about 20 vol% in-situ Cu68n5 intermetallic
phases about 3-8 micrometers in diameter. Two sets of experiments were performed : In
the first set, all of the above four solder joints were aged at 150°C for periods ranging to
5000 hours and the intermetallic growth was monitored periodically. In the second set,
each of the above four solder joints was aged at five different temperatures for 4000
hours. The interfacial layers between solders and the Cu substrate were examined using
optical and scanning electron microscopy. The growth kinetics of intermetallic
interfacial layers formed between solder and Cu substrate was characterized. The effect
of in-situ Cu68n5 intermetallic phases on the growth rate is discussed. The growth rate of
the intermetallic layers in the eutectic Sn-Pb composite was slower for the first 150 hours
as compared to the eutectic Sn-Pb non-composite. The growth rate of the intermetallic
layers was similar for both the eutectic Sn-Ag and eutectic Sn-Ag composite throughout
the aging duration. The activation energies for Cu6Sn5 layer growth for the eutectic Sn-
Pb and Sn-Ag solder joints are evaluated to be 111 kJ/mol and 116 kJ/mol, respectively.
The eutectic Sn-Pb and Sn-Ag composite solder joints exhibit higher activation energies

of 161 kJ/mol and 203 kJ/mol.

112

4.1 INTRODUCTION

Eutectic Sn-Pb solder has been widely used in electronic industry as an interconnect
material between electronic components and printed circuit boards. Recently, the
environmental and mechanical concerns have warranted the electronic industry to
develop alternatives to Pb-free solders [26]. Eutectic Sn-Ag solder is being considered as
a possible Pb-free solder substitute for several reasons. The eutectic Sn-Ag solder is non-
toxic, has a higher melting temperature of 221°C, is suitable for higher temperature
applications, and offers better mechanical properties [2,5]. Its wetting characteristics are
comparable to that of eutectic Sn-Pb solder [4].

The Sn-Pb and Sn-Ag solders bond strongly with Cu substrate by forming
intermetallic layers between the solder and Cu substrate. Two intermetallic layers,
namely, Cu68n5 (n-phase) and Cu3Sn (re-phase), are produced at the interface between the
solder and Cu substrate. However, these layers tend to grow with time by solid state
diffusion even at ambient temperatures. These intermetallic layers may adversely affect
the reliability and solderability of solder joints due to excessive growth during storage
and service. They can be sources of mechanical weakness in solder joints due to the
brittle nature of the interrnetallics or can cause delamination at the interface [133,82].

Because of the importance of the intermetallic layer to bonding, and its influence on
the mechanical properties of solder joints, many studies have been carried out on the
crystal structures [134] on the growth kinetics [l35-137,90,16,91], and on their effects on
mechanical properties [83,138-140] While over the last few years the kinetics and

microstructural data published [82,137,90,l6,9l] on the intermetallic layers in the Pb-free

113

solder and composite solder joints have been considerable, a basic understanding is far
from complete.

This study investigates microstructural evolution and growth kinetics of the Cu58n5
and Cu3Sn intermetallic layers in the eutectic Sn-Pb and Sn-Ag solder, and their
corresponding composite solder joints during long-term solid state aging. The grth
kinetics of intermetallic interfacial layers formed between solders and Cu substrate was
measured as a function of time and temperature, and activation energies for the

intermetallic layer growth were calculated using a diffusion—controlled growth model.

4.2 EXPERIMENTAL PROCEDURES

Single shear lap solder joints were fabricated by soldering two dog-bone shaped Cu
strips together using the solder preforms of four different solders, eutectic Sn-Pb and
eutectic Sn-Ag solders, and their corresponding composite solders with 20 vol% Cu68n5
reinforcements. Eutectic Sn-Pb and Sn-Ag solders were provided in the form of cast
ingots by Alpha Metals. Composite solders were prepared by introducing about 20 vol%
Cu68n5 intermetallic particles into the eutectic Sn-Pb and Sn-Ag solders by in-situ
methods. These solders were rolled to produce sheets about 100 pm thick, and cut into

2 size.

the preform foils of about 1.0 mm

Copper substrates were machined from 0.5 mm-thick Cu sheet using
electrodischarge machining (EDM) to produce specimens 11.5 mm long with a gage
width of 1 mm. The Cu substrates were painted at the end of the gage section of the
substrate with a solder mask and baked for about 1 hour at 150°C to confine the molten

2

solder within a 1 mm area during joining operation. The Cu substrates were precleaned

114

with a 50% aqueous HN03 solution and rinsed in methyl alcohol just prior to soldering to
remove the thin tarnish layers.

Figure 4.1 schematically illustrates fabrication of solder joints along with
dimensions of Cu substrate. Solder joints were fabricated in an aluminum fixture
designed to hold the assemblies of solder joints in proper alignment during soldering
operation. Liquid acid flux (a200-L) was applied to the area to be soldered in each Cu
substrates prior to sandwiching the solder preforms between them. They were placed on
an aluminum fixture. The temperatures of solder joints during joining operation were
monitored by a thermocouple attached to the aluminum fixture. Since aluminum has very
high thermal conductivity, it was assumed that the temperature of the aluminum fixture
was nearly the same as that of the solder joints. The aluminum fixture with an assembly
of 10 solder joints was placed on a preheated hot plate maintained at about 500°C, and
allowed to reach about 230°C for eutectic Sn-Pb and its composite solder joints, and
270°C for eutectic Sn-Ag and its composite solder joints. When the solder joints reached
the temperature allowed, the aluminum fixture was moved to a cooling surface. Figure
4.2 illustrates geometry and dimensions of fabricated single shear lap solder joints. The
solder joints possess about 1mm2 joining area and about 100 um thickness. During
soldering, the intermetallic particulate reinforcements in composite solders remained in a
solid state during reflow because of their higher melting temperature, while the solder
matrices in them were reflowed and solidified during soldering process.

Excess solder present on both sides of the joints were ground off, and one of its
sides was polished using 400 and 2000 grit sand papers followed by a finish using 0.05

um SiOz colloidal suspension. The solder joints were etched with a solution of 40%

115

.EE 5 mofiomv 252 .528 20... 2 93x: E::_E:_m ucm £50.. 528 ho 29:8me 9: .0 89030 25625 fie 2:9".

233.... E::_E:_< vogue:

      

$55 mazooofiach

5:". .52. 23.2
__o..._ E35..."— 320m

116

.250.— ..oEom ho mco_mcoE_u can 395% ho ESmmfi ozmsmzow N6 959".

new m:_c_o_ NEs. 7.
32:32. E1 2::

 

 

E3 328

 

 

 

 

117

H20/40% NH40H/20%H202 for about 10 seconds to reveal microstructural features.

Two sets of aging treatments were performed. In the first set, the joints made with
all of the four solders were aged at 150°C for various periods of time ranging to 5000
hours. The second set consisted of aging each of the four solder joints at five different
temperatures for 4000 hours. Sn-Pb solder joints were aged at 50, 70, 100, 120 and
150°C, Sn-Ag solder joints were aged at 70, 100, 120, 150 and 180°C. Aging was
conducted by placing solder joints on an aluminum plate maintained at a steady
temperature by means of a hot plate. The solder joints were slightly repolished and
etched periodically after each aging treatment to reveal the resultant microstructure. This
procedure helped to monitor the progress of the aging process on the same specimen at
various stages.

The solder joints were examined with an optical microscope and a scanning
electron microscope operated at 25 kV. Energy dispersive X-ray spectra (EDS) were
used to identify elements present in phases and intermetallic layers. Optical micrographs
of solder/substrate interfacial intermetallic layers were used to measure its thickness.
These micrographs were scanned with HG Scanet 3C scanner and scanned images were
then analyzed using Adobe Photoshop software to measure the thickness of intermetallic
layers. The intermetallic layer thicknesses reported here are averages of about 10

measurements per micrograph for each solder joint.

4.3 RESULTS AND DISCUSSION
As-joined microstructures of the eutectic Sn-Pb and Sn-Ag solder, and their

composite solder joints are presented in Figure 4.3. The Sn-Pb solder joint consisted of

118

.E_o_ 520m m<-cm 2523 g new E5. .628 3.5 0:023 E
$50.. 528 09853.3 “—0 93026225 9: 9.36% 2359.28 2mm g a E 9v 239”.

E1 .: 3.2.3.; :9 :3 E1 c— .32.52.; :9
£5er rims... .
mamxuaEtm

. s

, 1. . q a . .
«firemeaefizw \ r

. a;
, ¢ .
t .1 v
. e
4

 

119

.33 528 £8qu0 2.5 282.5 A3 new E_o_ 5.28 26368 ndém 2823 3
.mE_o_ $23 88038.3 8 9.2026228 9: 9.56% 2929.28 Eww A3 a A3 m4. 659“.

92:227. :9

ease

0:“.31237. =.v :—1

E: :—

IS.

2..

£33.23 £5.59

1

 

120

three regions namely, the Cu substrate, the Cu68n5 intermetallic layer, and the eutectic
Sn-Pb phase. Sn-Ag solder joint revealed four distinct regions, the Cu substrate, the
Cu68n5 intermetallic layer, primary Sn-rich phase and the eutectic Sn-Ag phase. The
composite solder joints possessed microstructures similar to their corresponding non-
composites except for the uniformly distributed 3 ~ 8 um size intentionally incorporated
Cu68n5 intermetallic particles. Prior to aging only the Cu68n5 intermetallic layer at the
solder/substrate interface was observed for all four solder joints.

Figure 4.4 reveals the microstructures of all the four solder joints after aging for
5000 hours at 150°C. The aged eutectic Sn—Pb solder joint consists of the Cu substrate,
the Cu38n intermetallic layer, the Cu68n5 intermetallic layer, the Pb-rich (white) phase,
and the Sn rich (dark) phase. The Pb-rich and Sn-rich phases exhibit extensive
coarsening due to aging. The growth of the intermetallic layers depleted the Sn-rich
phases adjacent to the Cu68n5 intermetallic layer and led to the formation of a layer of
Pb-rich phase across the entire intermetallic layer/solder interface. Such a feature has
been observed in other studies [133,82,90].

The aged eutectic Sn-Ag solder joint comprises the Cu substrate, the Cu3Sn
intermetallic layer, the Cu68n5 intermetallic layer, the Ag3Sn particles embedded within
the Cu68n5 intermetallic layer, the Aggsn particles in the solder, and the Sn-rich phase.
The Aggsn particles present in the solder and the Cu68n5 intermetallic layer exhibited a
significant amount of coarsening due to aging. With increased aging, the Ag3Sn particles
became more prevalent in the Cu68n5 intermetallic layer. This observation is consistent
with a previous study [91] on the intermetallic growth in a eutectic Sn-Ag/Cu diffusion

couple. The existence of Aggsn particles within the Cu6Sn5 intermetallic layer is related

121

 

.22 528 915 2.85.... 3. as... :2 528 £20 2.8.3 E .

050; ooom .2 9.02 8
92mm 5.8 2.22 3200 05 .0 22022092.: 05 @2305 05369.2... 2mm 3v a 3 v6 0.59“.
E: o. 0:221; Pa :3 :3 :—

.w_=.:wm_=m =...v

=m...=D
:6: v .

.. , u:méu
”ems—Q _

swag. RV. .._

.

3.52.50.

 

122

.E_0_ .0200 0:000:50 m<-:w 020050 AB 420.. .0200 02000E00 anew 262:0 A8 .050; ooom .2
ooomr “a 9200 5:0 0:20.. .0200 05 *0 0.2020022... 05 023050 02006920.. 2mm A3 .0 8V v6 052“.

:3 c.

J

30.52.; :0
:mEU
25...: v

r

‘

..L......:rAl.. :wfi4
8.03.50 . )\ ,
, .. E mag—U .

00.02.30 Ewe: u.

 

0.5.3.; =U
.555

£0.50

v
r

123

to the depletion of Sn close to intermetallic layer.

Composite solder joints retained characteristics similar to corresponding non-
composite solder joints with the presence of the Cu3Sn intermetallic layer, the Cu6$n5
intermetallic layer, coarsened Pb-rich and Sn-rich solder phases, and Ag3Sn particles.
The Cu68n5 particles present in the composite solder joints also showed coarsening after
long-term aging of 5000 hours. In addition, the Cu68n5 particles present in the region
near the intermetallic layer tended to attach with the Cu6Sn5 intermetallic layer as shown
in Figure 4.4 (c) and ((1), resulting in “abnormal” growth of such layers. This undesirable
feature is less pronounced in the Sn-Pb composite solder joint than in the Sn-Ag
composite solder joint because of the Pb phase present in the Sn-Pb solder. Since this
phenomenon is not the direct cause for the intermetallic layer growth, it was not included
in the quantitative analysis of the layer growth measurement. It has been observed in a
35 um thick Sn-Ag composite solder joint that the joining of the Cu68n5 particles with the
Cu68n5 intermetallic layer led an excessive intermetallic layer growth and finally caused
two Cu58n5 intermetallic layers at opposing sides to bridge together. Q

Table 4.1 lists the thicknesses of the Cu68n5 and Cu3Sn, intermetallic layers for all
four solder joints for the aging times up to 5000 hours at 150°C. Presence of these two
intermetallic layers was common to all of four solder joints aged in a solid state. The
thicknesses of both intermetallic layers increased with time for all four solder joints, and
in general the thickness of the Cu6Sn5 intermetallic layer was always larger than that of
Cu3Sn layer.

The Cu3Sn intermetallic layer was not observed for short aging times for all of four

solder joints and only Cu6$n5 intermetallic layer was observed. A Transmission Electron

124

Table 4.1 Interrnetallic thicknesses (pm) during aging at 150°C.

 

10 50 150 350 700 1200 22005000

 

 

 

 

 

 

 

Solder Joint/'1‘ ime (hours) 0
Eutectic Sn-Pb Cu68n5 1.85 2.3 2.8 3.8 4.67 5.15 6.5 7.0 7.8
S°l°°r J°mt Cugsn 2.4 2.6 3.0 3.5 3.9 3.5 4.8
Eutectic Sn-Pb Cu68n5 1.9 1.35 1.6 2.9 3.9 5.7 6.9 6.8 8.0
Composite
Solder Joint Cu3Sn 0.99 1.0 2.35 2.7 2.6 3.3 3.6 4.9
Eutectic Sn- Ag Cu68n5 3.5 3.8 4.3 3.8 4.2 4.8 6.4 7.3 6.5
S°l°°r J°mt Cu3Sn 1.3 2.1 2.5 3.1 4.0 5.6
Eutectic Sn-Ag Cu68n5 2.3 2.33 2.62 2.5 2.75 3.0 3.7 5.5 5.5
Composite
Solder Joint Cu3Sn 1.2 1.9 2.1 2.6 4.0 4.6 .

 

 

 

 

 

125

 

 

 

 

 

 

Microscopic study by Y. Wu et a1. [90] reported the presence of about 0.2 ~ 0.3 um
thickness Cu3Sn layer in an as-fabricated eutectic Sn-Pb solder/Cu joint. Since the solder
joints in the current study were investigated by the optical and SEM microscopy, it
should be noted that formation of such a thin Cugsn layer during soldering could not be
observed.

Sn—Ag and its composite solders formed thicker Cu68n5 intermetallic layer than the
Sn-Pb and its composite solder as a result of the joining operation. Presence of higher
amount of Sn present in eutectic Sn-Ag solder (96.5Sn wt%) composition appears to
promote a thicker Cu68n5 intermetallic layer as compared to the eutectic Sn-Pb solder
(63Sn wt%). The eutectic Sn-Pb composite solder produced a Cu68n5 intermetallic layer
of comparable thickness to that of the eutectic Sn-Pb non-composite. On the other hand,
the eutectic Sn—Ag composite solder formed a less thick Cu68n5 intermetallic layer than
the eutectic Sn-Ag solder due to joining operation. It is not clear how the presence of the
Cu68n5 particles in the solder affects the intermetallic layer formation during with Sn-Ag
solder.

Figure 4.5 shows plots of the total intermetallic layer (Cu68n5 + Cu3Sn) thickness
versus aging times up to 5000 hours using the data in Table 4.1. For clarity, the same plot
is shown in three different aging time scales, namely two expanded short time scales and
an overall time scale incorporating all aging times investigated.

During the initial stages of aging up to 50 hours, the eutectic Sn-Pb solder joint
showed faster growth of the intermetallic layers than did its composite solder joint, and
the growth of the intermetallic layers slowed after 50 hours. As shown in Table 4.1, the

eutectic Sn-Pb solder joint exhibited no visible Cu3Sn intermetallic layer growth up to 10

126

j—i
A

 

 

 

 

 

   
 
  

 

 

 

 

 

 

 

 

      

 

 

 

 

 

 

 

 

 

PIIIIIIIIIIFITTTIIIIIjWFd

12:— _v

A t :

L I j

E s- _

.3 - _

u . _

IE 6 1

E-4 . -

... , EutectrcSn/Pb _

g 4 ----- CompositeSn/Pb:

E" 2 EutecticSn/Ag _‘

----- CompositeSn/Ag:

0 1111111llllllllllllllllld

0 1000 2000 3000 4000 5000

AgingTime(hrs)
14prITTrIIIIIIIIIIIIIIIIIIIIIIIIITIFF_ l4L-lllIIIjIIlIIIIIITfiIIII‘IIHIII—
: Eutectic Sanb ; : Eutectic Sanb 3
12 :- ----- Composite Sn/Ph -: 12 :- ----- Composite Sn/Pb j
A — Eutectic SnlAg - E r- Eutectic SnlAg ~
g 10:. -----CompositeSn/Ag .1 3' lol— -'--'CompositeSn/Ag L
g 82— ‘1‘ g 8E- i
.0 i 3 .2 3 ‘
BE 6: “ 4:" 6? '1
_ Jl [— L 2:1—4
34 —. 2‘7
H I' : h E, -—-—-—‘;—--- :
2 — 2 -
I C I
0IlllILllllLlIllllllllllLllllll1111‘ ()lnlllLlIlIlllllllllillllllllLll-4
0 100 200 300 400 500 600 700 0 50 100 150
Aging Time (hrs) Aging Time (hrs)

Figure 4.5 Total intermetallic layer (Cu68n5 + Cuasn) versus aging time
for the eutectic Sn-Pb and Sn-Ag solder joints, and their corresponding
composite solder joints after aging at 150°C for 5000 hours.

127

 

 

 

hours. The Cu3Sn intermetallic layer was observed after 50 hours of aging. After the
formation of Cu3Sn layer, both Cu6Sn5 and Cu3Sn intermetallic layers grew in a
progressive manner. These observations for the intermetallic layers in the eutectic Sn-Pb
solder joint imply that the Pb-rich phases began to accumulate between solder/Cu6Sn5
intermetallic layer interface due to the depletion of Sn-rich phases. They consequently
affected the growth of the intermetallic layers by reducing the Sn concentration at the
Cu6Sn5 intermetallic layer/Cu interface, resulting in the formation of the Cu-rich Cu3Sn
intermetallic layer. The enriched Pb-rich phases adjacent to the Cu68n5 intermetallic
layer, once formed, appears to significantly affect the growth of the intermetallic layers
throughout the rest of the aging process.

Sn-Pb composite solder joint exhibited no significant growth of the intermetallic
layers initially and after 50 hours, the intermetallic layer growth in this solder showed
faster growth than corresponding non—composite solder that finally decelerates after 150
hours as shown in Figure 4.5. After 700 hours, the eutectic Sn-Pb composite solder joint
showed growth of the intermetallic layers similar to its non-composite counterpart. The
Cu3Sn intermetallic layer in the eutectic Sn-Pb composite solder joint was noticed even
after 10 hours aging, a time period shorter than in its non-composite counterpart, and
occupied a considerable portion of the total intermetallic layer.

Since the accumulation of the Pb-rich phases in the eutectic Sn-Pb composite solder
joint was not significant prior to an aging period of 50 hours, it appears that the Cu68n5
particles affected the intermetallic layer growth by delaying Sn diffusion toward the
Cu68n5 intermetallic layer, resulting in the earlier formation of the Cu3Sn intermetallic

layer of similar total thickness up to 50 hours. The Sn depletion at Cu6Sn5 intermetallic

128

 

 

 

 

 

layer/solder interface occurred after 50 hours causing the accumulation of the Pb-rich
phases. The intermetallic layer growth appears to be affected by the accumulated Pb-rich
phase throughout the rest of the aging process beyond 150 hours.

Accumulation of the Pb-rich phases adjacent to the Cu68n5 intermetallic layer in the
eutectic Sn-Pb and its composite solder joints after 350 hour aging at 150°C can be
observed in Figure 4.6. It is evident that the Cu68n5 intermetallic layer in the eutectic Sn-
Pb solder joint was completely covered with Pb-rich phase. However, in Sn-Pb
composite solder joint only about 87 % of the interface layer was covered with the Pb-
rich phase. The comparable enrichment of the Pb—rich phases after aging for 350 hours
and the similar growth rate of the intermetallic layers after 700 hours in the eutectic Sn-
Pb and its composite solder joints suggest that the growth of the intermetallic layers in
the eutectic Sn-Pb composite solder joint was moderated by the enhanced concentrations
of Pb-rich phases rather than the Cu6Sn5 particles for most of the later stage of aging.
The Cu6Sn5 particles were effective in retarding the intermetallic layer growth during the
initial stages of aging in this solder.

It was considered as a possible mechanism in a previous study [137] that the Cu68n5
particles can act as Sn sinks so that they reduce the amount of Sn available for the
reaction at the substrate/solder interface. However, since the Cu68n5 particles do not
react with the solder and can be considered as an inert phase with respect to the solder, it
is unlikely that the CU6SII5 particles can be sinks for Sn. One possible mechanism is that
the Cu(,Sn5 particles act as partial obstacles to Sn diffusion toward the substrate/solder
interface resulting in an increase of its total diffusion distance and delay the Sn supply

required for the intermetallic growth. A similar suggestion on the role of the Cu6Sn5

129

 

.052. 000 .2 0000. a 800 as. .028 20858 #30 2.850 05 3. 0:0
.20.. .0200 00.5 2.0050 05 .0. 2 .0>0_ 2:93:22 05.50 0.... 0. 2000.00
0000.... 52.00 05 .0 02.03.0080 05 02.5050 0500592.: 02.00 0... 052“.

.01.: 0.0.50 .530 .5
l «tying :0

«0.3.2.0 is... $20.; 3.. ..

 

130

particles has been made in a previous study [90].

The eutectic Sn-Ag and its composite solder joints revealed similar growth rates
throughout aging. The difference of the intermetallic layer thickness produced during
soldering between the eutectic Sn-Ag solder joint and its composite solder joint was
maintained for the entire duration of this aging study as shown in Figure 4.5. It can be
seen from Table 4.1 that the Cu3Sn layer in the eutectic Sn-Ag and its composite solder
joints exhibited similar thickness values and only the CUfiSUS layer retained the initial
difference of the intermetallic layer thickness portion in the total thickness of the
intermetallic layer during aging, resulting in a similar growth rate. This maintenance of
the difference in the initial thickness of the intermetallic layers by the Cu68n5
intermetallic layer indicates that there is no substantial effect of the Cu68n5 particles on
the intermetallic layer growth, although it affects the formation of this layer during
joining.

An interesting feature about Cu3Sn intermetallic layer formation in Sn-Ag solder
joint occurs far later than in Sn-Pb solder joint as shown in Table 4.1. This suggests that
lack of diffusion of tin from the solder is a dominant factor in determining the rate of
growth of the Cu3Sn layer. Since Sn-Ag solder can provide more tin than Sn-Pb solder
matrix, tin deficiency that will occur at Cu68n5 intermetallic layer/Cu interface with Sn-
Pb solder results in early formation of Cu3Sn layer (in the time range of 0-50 hours as
compared to 50-150 hours).

The four solder joints investigated in this study exhibited a saturation in the layer
growth after they reached a certain thickness of the intermetallic layers at different aging

times. The slow saturated layer growth suggests a possible change in growth rate

131

 

controlling factor during aging. A possible explanation may rest with that a certain
intermetallic layer thickness may cause this stabilization. If the intermetallic layer is
thick enough to cause diffusion of Cu and Sn through it to become much slower, the
solder joint may exhibit the saturation in the layer growth. Thus, the thickness of
intermetallic layers themselves can be a controlling factor for the layer growth during
long-term aging in this solder.

The growth of the intermetallic layers is influenced by factors such as (a) the Pb—
rich phase accumulation at the intermetallic layer/solder interface, (b) the intermetallic
layer thickness, (c) the Cu6Sn5 particles, and (d) the number of interface boundaries
encountered by the diffusing species. The influence of each factor on the growth of the
intermetallic layers is complicated.

A simple layer growth model was used to determine the layer-growth coefficients K

for Cu68n5 and Cu3Sn intermetallic layers, which is expressed as follows.

d=d0+fié

where K is strongly related to the diffusion coefficient of atomic elements of the
intermetallic layer. d is the intermetallic thickness (pm) at time I, do is the initial
thickness (pm) of the intermetallic layer, K is the layer-growth coefficient (cm2/sec), and
t is the aging time (sec). The layer-growth coefficients were determined from a linear
regression analysis of (d-do) versus I”, where the slope is K V 2.

Figure 4.7 shows values of the layer-growth coefficients for Cubsns and Cu3Sn
layer growth obtained from plotting intermetallic layer thickness versus square root of

time at 150°C. The linear correlation coefficients were generally greater than 0.9 for the

132

 

Thickness (um)

Thickness (jun)

Eutectic Sn-Pb solder joint

 

   
     
   
 

 

 

 

 

 
      
    

 

 

 

 

9" I I I I I I I I I I I I I I I I I I I I I Ii
3 “03305) = . a
7 48 x 10'15 cmzls i
/ A
6 / *3 E
5 / I 5 3
/ O 3 g
4 —: E
3 ° 5 0
3 ""
2 °’ K(Cu,Sn) = 9.2 x 10 ~15 cm2/s 3 5'5
1 i
() l 1 J l l l L,1 L41 1 l l l l J,l l l I l l l l:
0 1000 2000 3000 4000 5000
(Time in sec)“2
Eutectic Sn-Ag solder joint
S’V INT III I I I I I I I I I I I I I I I I I I3
8 “(3068115) = —:
7 17 x 10 '15 cmzls _j
C .4
6 ,, Li a
5 / g 3
4 .3 g
3 Q .0
2 0’ é a
1 / K K(Cu38n) = 14.3 x 10-15 cmzls a
() 1 l l l l l 141 1 I l l l l l l LAL,1 l l l 1 1d

010002000300040005000

(Time in sec)"2

Eutectic Sn-Pb composite solder joint

'IIIIIIIITIfII

K(Cu‘Sn5) =
75 x 10 '15 cmzls

 

III I I I I I

l

   
      
    
  

111111111111

0
\

O

/ K(Cu,Sn) = 15 x 10 -15 cmzls
O

QHNUAUIGQfl‘O

 

t-
111111111lllLlllllllllllilllJJl

l l l l l l l J L,1 l l l 1 1,L,1 l l l l l l

010002000300040005000

('l‘imeinsec)"2

Eutectic Sn-Ag cmite solrbr joint

I I I I I I I I I I I I I I I I I I I I I I I I

 

   

II

  
 
 

K(Cu‘S r3) = 10 x 10 "5 cmzls

IIIIIIIIIIIIIIIIIIIIIIIIIIIIII

A

 

 

¢HNU&UIG\\IN@

t:
0 1000 2000 3000 4000 5000
('I‘inninsec)"2

Figure 4.7 Thickness of the Cu68n5 and Cuasn intermetallic layers
versus square root of aging time at 150°C for the eutectic Sn-Pb
and Sn-Ag solder joints, and their corresponding composite solder joints.

133

 

 

 

four types of solder joints used except the K value of 0.83 for the Cu3Sn layer in the Sn-
Pb composite and 0.84 for the Cu68n5 layer in the Sn-Ag composite. This good linear
correlation suggests that a thermally activated diffusion process is valid approach for
assessing the intermetallic layer growth.

Eutectic Sn-Pb composite solder joint showed higher layer-growth coefficients for
the growth of Cu6Sn5 and Cu3Sn layers than the corresponding non-composite solder
joint. The high layer-growth coefficient values observed for Sn—Pb composite tend to
suggest that the Cu5Sn5 particles, added in-situ, did not promote layer growth retardation,
particularly in the later stage of aging. In contrast, both eutectic Sn-Ag solder joint and
its composite counterpart exhibited comparable layer-growth coefficients for the Cu5Sn5
and Cu3Sn intermetallic layers.

The eutectic Sn-Ag and its composite solder joints exhibited lower layer-growth
coefficients for the Cu6Sn5 layer, and higher layer-growth coefficients for the Cu3Sn layer
than the eutectic Sn-Pb and its composite solder joints. The data in the Figure 4.7
resulted from the aging at the same temperature of 150°C. The same aging temperature
of 150°C represents 0.85 for the eutectic Sn-Ag solder and 0.9 for the eutectic Sn-Pb
solder in terms of their melting temperatures. Sn diffusion toward the intermetallic layers
in the eutectic Sn-Ag and its composite solder joints was slower than in the eutectic Sn-
Pb and its composite solder joints due to the aging at lower homologous temperature.
The slower Sn diffusion resulted in lower layer-growth coefficients for the Sn-rich
Cu6Sn5 intermetallic layer, and higher layer-growth coefficients for Cu-rich Cu3Sn

intermetallic layer.

134

 

 

The activation energies for the intermetallic layer growth was calculated using the

Arrhenius relationship for the diffusion dominant process,

K = Aexp(—E* /RT)

where K is the layer-growth coefficient (cm2/sec), A is the layer-growth constant
(cmzlsec), E. is the activation energy (J/mol) for layer growth, R is the ideal gas constant
(8.314 J/mol K), and T is the absolute temperature (°K). The activation energy is
obtained from the slope of the In K versus 1/T plot. Figure 4.8 is the Arrhenius plots from
which the activation energies for the growth of the Cu68n5 intermetallic layer in the four
solder joints can be obtained. The data reported here are in good agreement with the
results from the previous studies [82,137,90,16,9l]. Unfortunately, the analysis for the
Cu38n layer is not shown in this study because of lack of relevant data.

The two composite solder joints exhibited higher activation energies than their
corresponding non-composite solder joints. Higher activation energies in the two
composite solder joints appear to indicate that the Cu6Sn5 particles in the composite
solders were effective in retarding the intermetallic layer growth. The higher activation
energies, however, resulted from higher layer—growth coefficients at high aging
temperature and lower layer-growth coefficients at low aging temperature. Therefore, it
is evident that the Cu6Sn5 particles were not efficient diffusion barriers at high
temperature, while they were effective barriers at low temperatures.

The activation energy values for the CueSns intermetallic layer in this study are
slightly higher than those reported from the previous studies, which are 77 kJ/mol for

eutectic Sn-Pb solder [90], 118 kJ/mol for the eutectic Sn-Pb composite solder containing

135

 

 

In K (cmzlsec)

In K (cmzlsec)

Eutectic Sn-Pb solder joint

 

   
     

 

   

 

 

 

   
     
 

 

 

 

' I I I I I I I I I I r I I I I I I I I q
' - I
.30 E - lll kJ/mol fl
A = 2.22 cmzlsec :
.32 j A
3 a
4. s
i E
-36 q 3
a M
-38 a 5
.40 Y. Wu et al. [90] -3
D Pinizzotto et al. [137] d
.42 L1 1 1 1 1 1 I 1 1 1 L44 1 l 1 1 1 _,
2.2 2.4 2.6 2.8 3 3.2
l/T (K'1)x 1000
Eutectic Sn-Ag solder joint
'26 .. r I I I I I I I I I I I I I I I I I I _,
.23 E’ = 116 lemol _‘
A = 2.75 cmzlsec :
'30 .1 A
-32 T: «g
: 5
-34 j E
-36, —: E
C 0 Flanders et al. [91 1
'33 :‘ u Frear & Vianco [82] 1
I O Yang et al. [MI I
'40 1 l l l L L 1 l l l l l l l l L
2 2.2 2.4 2.6 2.8 3

HT (K'l) x 1000

. I I T I T T I I I—I I I I I T _
E o o 11
-32 :- E = 161 kJ/mol 1
Z A = 2.7 x 106 cmzlsec ‘
-34 _— .21
-36 E— _:
-38 {— .3
.42 3- _:
.44 i—F- Present Study 3
C 0 Pinizzotto at al. [137] j
.46 h #1 l l l l l l l l l l l l l l l l l
2.2 2.4 2.6 2.8 3

Eutectic Sn-Pb composite solder joint

 

        
 
  
 

 

 

 

3.2
III (K'l) x 1000

Eutectic Sn-Ag composite solder joint

 

 

 

 

 

I I I I I I I I I I I I I I I I I I I d
: E‘ = 203 lrJ/mol f
_— 0 A =712x 109cm1/sec -_
t :
E E
h l l 1 J l l l l l 1 J J l l l l 1 l _
2 2.2 2.4 2.6 2.8

III‘ (1(4)): 1000

Figure 4.8 Arrhenius plots for the growth of Cu68n5 intermetallic layer
in the eutectic Sn-Pb and Sn-Ag solder joints, and their corresponding
composite solder joints based on the interfacial intermetallic layer
thickness at 4000 hours of aging at different temperatures.

136

 

 

20 wt% Cu6Sn5 particles [90], and 107 kJ/mol for the eutectic Sn-Ag solder [91]. The
previous studies [133,91] reported that the aging at low temperatures less than 110°C up
to about 2184 hours did not cause significant growth of the intermetallic layers.
However, the four solder joints studied here revealed slight increases of the Cu68n5
intermetallic layer at low temperature of 70 and 50°C during the long-term aging up to
5000 hours. The slightly higher activation energy values for the intermetallic layer
growth in this study appear to result from the low layer-growth coefficients produced at

low aging temperature during long-term aging up to 5000 hours.

4.4 CONCLUSION

This study detailed the microstructural evolution and growth kinetics of the Cu68n5
and Cu3Sn intermetallic layers in the eutectic Sn-Pb and Sn-Ag solders, and their
corresponding composites during aging at temperatures ranging 50°C to 180°C up to
5000 hours. During aging, the eutectic Sn-Pb and Sn-Ag solder phases revealed
extensive coarsening and the intermetallic layers also grew significantly. The
intermetallic layers in the composite solder joints exhibited undesirable features,
specifically, (a) coarsening of the Cu6Sn5 particles, (b) the ‘abnormal’ growth of the
intermetallic layer.

Factors such as Pb-rich phase accumulation, the layer thickness itself, the aging
temperature, and Cu6Sn5 particles play significant roles in the growth of the intermetallic
layer. The Cu68n5 particles affected the growth of the intermetallic layers at the initial
stage of aging for the eutectic Sn-Pb composite solder joint. However, there was no

effect of the Cu6Sn5 particles for the intermetallic layer growth in the Sn-Ag composite

137

 

 

solder joint aged at 150°C for long periods of time, even though there was an apparent
effect on the intermetallic layer formation during soldering. The Cu6Sn5 particles were
effective in reducing the layer growth in Sn-Pb and Sn-Ag composite solder joints for
low temperature aging, but at high temperatures above 150°C significant coarsening of
the Cu6Sn5 particles was observed and caused the enhanced intermetallic layer growth.

A possible mechanism which can explain the effect of the Cu6$n5 particles on the
kinetics of intermetallic growth is based on assuming that the particles act as barriers to
the Sn-diffusion toward intermetallic layer and reduce the amount of Sn available at the

interface for reaction.

138

 

 

h-

CHAPTER 5

FORMATION AND GROWTH OF INTERFACIAL INTERMETALLIC LAYERS
IN EUTECTIC Sn-Ag SOLDER AND ITS COMPOSITE SOLDER JOINTS

ABSTRACT

Single shear lap joints were made by soldering two Cu substrates with eutectic Sn-Ag
solder, and its composite solders containing FeSn/FeSn; or Ni3Sn4 intermetallic particles
introduced by in-situ method. Aging of solder joints was performed at 70, 100, 120, 150,
180°C for 1400 hours. The growth of the interfacial intermetallic compound (IMC) layers
was characterized assuming diffusion-controlled growth kinetics. Effects of such
FeSn/FeSn; and Ni3Sn4 particulates on the IMC layer growth rate were extensively
characterized. Composite solder joints in the fabricated condition formed thinner IMC
layers compared to the corresponding non-composite solder joints. The Cu6Sn5 IMC layer
grew faster at temperatures above 120°C(T/Tm = 0.8), while growing slower at
temperatures below 120°C in composite solder joints. In-situ introduced FeSn/FeSnz and
Ni3Sn4 particle reinforcements in composite solder joints proved effective in reducing the
overall growth of the interfacial Cu6Sn5 IMC layer only at lower temperatures.
Composite solder joints exhibited slower growth of the Cu38n layer during aging at all
temperatures used in this study. Two different regions having different activation energies
depending on the temperature were identified for the growth of Cugsns and Cu38n IMC

layers.

139

 

,Q.

5.1 INTRODUCTION

Solders will bond with a Cu substrate through the formation of a dual Cu-Sn
intermetallic compound (IMC) layer consisting of CuGSns and Cu3Sn that exist between
the solder and Cu substrate. The Cu68n5 layer forms adjacent to the solder side and the
Cu3Sn layer next to the Cu substrate. Over time in service, these layers grow in the solid
state and occupy a significant portion of the solder joint. Because IMC layers are more
brittle than the solder matrix, they can be a site of mechanical weakness causing failure of
the joint with the fracture in the IMC layer itself or along the interface between solder
and IMC layer [133,82].

Several studies have been carried out on the growth kinetics of IMC layers
[135,136,89,90,16,91] and their subsequent effects on the mechanical properties of solder
joints [l38,83,139,140]. Most of these studies deal with IMC layer in Pb-Sn solder
joints, and complementary Pb-Sn composite solder joints [133,135,89,90,138-140]. In
contrast, fewer studies exist on the formation and growth of IMC layers in eutectic Sn-Ag
solder and eutectic Sn-Ag based composite solder joints [82,16,91].

This study investigates formation and growth of IMC layers in eutectic Sn-Ag
solder and its composite solder joints during long-term isothermal aging up to 1400
hours. Our previous study [141] investigated the effects of in-situ introduced Cu68n5
particles on the growth of IMC layers in both Sn-Ag and Pb-Sn solder joints. In this
study, the effects of in-situ introduced Fe-Sn and Ni3Sn4 particle reinforcements on the
growth of IMC layers in eutectic Sn-Ag solder were characterized. A comparison was

made between composite solders reinforced with Fe-Sn, Ni3Sn4, and Cu68n5 particles.

140

 

 

The growth kinetics of IMC layers was determined as a function of time and temperature

assuming a diffusion-controlled growth model.

5.2 EXPERIMENTAL DETAILS

Three types of solders were used in this study, eutectic Sn-Ag and its composite
solders containing FeSn/FeSnz or Ni3Sn4 intermetallic reinforcement particles. Eutectic
Sn-Ag solder was provided in the form of ingots cast by Alpha Metals. Composite
solders were prepared by incorporating additional Ni and Sn or Fe and Sn to introduce
about 20 vol.% FeSnz or Ni3Sn4 intermetallic particulate reinforcements into the eutectic
Sn-Ag solder using in-situ method. However, the composite solder reinforced with Fe
and Sn produced two different compositions of particles, FeSn and FeSnz, as illustrated in
Figure 5.1. The ratio of FeSn and FeSnz in particulate reinforcements is not clear at
present. The compositions of composite solders are listed in Table 5.1.

Single shear lap joints were made using solder preform foils with two dog-bone
shaped Cu strips together. Fabrication and preparation methods of solder joint specimens
are described in detail in Chapter 4. Excess solder present on both sides of the joints
were ground off, and one of its sides was polished using 400 and 2000 grit sand papers
followed by a finish using 0.05 um SiOz colloidal suspension. The solder joints were
etched with a solution of 40% H20/40% NH40H/20%H202 for about 10 seconds to
reveal microstructural features.

Solder joints were then isothermally aged at five temperatures of 70, 100, 120, 150,
180°C for long time periods up to 1400 hours by placing them at different locations on an

aluminum plate heated by a hot plate at the end so that is maintained at these steady

141

 

 

'“t-CuttSns

(‘u substrate _
lllum b

(_ "(.8115 layer

 

(u substrate 101““

 
   

Ni
containing
Layer

. '.-".2-,.«w. .4 . .~. .. ,
\.,__,o..4. VEIIgSns ‘ ' * . Cu Sn

 

C (‘u substrate 1 layer

Figure 5.1 SEM micrographs showing the microstructures of as-made
solder joints, (a) eutectic Sn-Ag solder joint, (b) composite solder joint
containing FeSn/FeSnz particles, (0) composite solder joint containing
Ni3$n4 particles, (d) optical micrograph showing the layer containing
Ni element in composite solder joint containing Ni38n4 particles.

142

 

Table 5.1 Compositions (vol %) of composite solders
(*The exact ratio of FeSn and FeSn; particles is not clear at present).

 

 

Eutectic Ni Fe Sn Introduced
Sn-Ag interrnetallics
Composite solder
reinforced with 80 5.1 16.84 20 vol% Ni3Sn4

Ni3Sn4 particles

 

 

Composite solder
reinforced with FeSn 80 4.12 18.98 FeSn & FeSnz
& FeSnz particles

 

 

 

 

 

143

 

temperatures. The solder joints were removed from the set-up periodically, and were
slightly repolished to reveal the resultant microstructure using an optical microscope. In
order to measure the IMC layer thickness, optical micrographs of [MC layers between
solder and Cu substrate were scanned with HG Scanet 3C scanner and then analyzed
using Adobe Photoshop software. The average value of 10 measurements per micrograph

was used to analyze the growth kinetics of IMC layers.

5.3 RESULTS AND DISCUSSION

Figure 5.1 shows the as-joined microstructures of solder joints prior to aging.
Typically, eutectic Sn-Ag joints consisted of Sn dendrites and eutectic Sn-Ag phases
having Sn phase peppered with fine Ag3Sn particles as shown in Figure 5.1a. The
microstructures of composite solder joints containing FeSn/FeSn; or NigSn4 intermetallic
particles in a eutectic Sn-Ag solder matrix are shown in Figure 5.1b and 5.1c,
respectively. In-situ introduced Fe-Sn particles were found to be of two types, FeSn and
FeSnz, as evidenced by the different contrast shown in Figure 5.1b. According to Fe-Sn
phase diagram [142], these two intermetallic compositions are possible in this system.
The Fe-rich FeSn is shown by the dark contrast in the inside region of Fe-Sn particles,
and Sn-rich FeSnz is exhibited by the light contrast in the outside region of Fe-Sn
particle.

The solder joints have been found to form two IMC layers between solder and Cu
substrate, Sn-rich Cu6Sn5 layer adjacent to the solder and Cu-rich Cu38n layer adjacent to
the Cu substrate, as identified in phase diagram of Cu-Sn binary system [16,82,83,89-

9l,133,135,136,138-141]. The non-composite and composite solder joints in as-

144

 

 

 

fabricated condition contained a continuous Cu6$n5 IMC layer between solder and Cu
substrate formed by the reaction of Sn and Cu during reflow. The CU3Sn IMC layer
which has been identified to form during reflow [90] was not clearly visible in the as-
joined solder joints of this study. Few Cu6$n5 particles were found in the solder matrix of
as-fabricated joint as shown in Figure 5.1a. Cu58n5 particles appear to exist in solder
matrix by breaking of Cu68n5 scallop from the Cu6Sn5 IMC layer and entering the solder
matrix during reflow. Composite solder joint having Ni3Sn4 particles exhibited an
additional component to the Cu6Sn5 layer which contained Ni as shown in Figure 5.1d.
This observation suggests that Ni3Sn4 particles in composite solder participated in the
formation of IMC layer during reflow, resulting in Ni containing layer adjacent to Cu68n5
IMC layer. The exact composition of the Ni containing layer is not clear at present. The
Ni containing layer was not included in the quantitative analysis of Cu6Sn5 and Cu3Sn
IMC layer growth, but was considered in computing the total thickness of IMC layers.
Figure 5.2 shows the microstructures of solder joints after aging at 180°C for 1400
hours. The Ag3Sn particles present in the solder exhibited a significant amount of
coarsening as has been observed in previous studies [16,25]. Few Cu6Sn5 particles that
were present in as-joined non-composite eutectic Sn-Ag solder joint also showed
considerable coarsening during aging [16]. The FeSn/FeSnz and Ni3Sn4 particles in
composite solder joints were stable during aging in this study, as compared to Cu68n5
reinforcing particles that exhibited significant coarsening during aging for 5000 hours at
this temperature [141]. Ag3Sn particles were also found to be embedded within the
Cu68n5 layer and in Cu6Sn5 particles of the solder matrix. Such an embedding of Ag3Sn

particles is attributed to the depletion of Sn close to the Cu6Sn5 layer and Cu6Sn5 particles

145

 

 

C1165 n;

{7115811
( u substrate

   

 

,-_
I

,. 4“ a” A \I; ' —.
(habits
(7113811 _——
(7u substrate 151”"

Figure 5.2 SEM micrographs showing the microstmctures of solder joints
after aging at 180°C for 1400 hours, (a) eutectic Sn-Ag solder joint,

(b) composite solder joint containing FeSn/FeSnz particles,

(0) composite solder joint containing Niasm particles.

146

 

 

within solder matrix [141,91]. The Cu68n5 and C0380 layers exhibited significant growth
during aging. There were voids present only in Cugsn layer apparently due to Kirkendall
effect in diffusion [15]. It has been proposed that void coarsening during aging can cause
the failure of a solder joint at the interface between Cu3Sn IMC layer and Cu substrate
[16].

Initial thicknesses of interfacial INIC layers of solder joints in the as-joined
condition are provided in Figure 5.3. Composite solder joints exhibited thinner IMC
layer, with average thickness of about 2.0 mm, as compared to an average thickness of
about 3.8 um observed in eutectic Sn-Ag solder joints. The IMC layer thickness
differences are readily observed in Figure 5.1. In our previous study [141], composite
solder joints reinforced with Cu58n5 particles in the as-joined condition also produced a
thinner IMC layer than did the non-composite eutectic Sn-Ag solder joint in the as-joined
condition. These observations indicate that reinforcing particles affect the formation and
its initial thickness of the IMC layer during the joining operation.

Plots illustrating the growth of total IMC layers (including both Cu6Sn5 and Cugsn)
at five temperatures up to 1400 hours are presented in Figure 5.4. Both nonocomposite
and composite Sn-Ag solder joints exhibited a similar behavior. The growth of total IMC
layer was considerably faster at higher temperatures of 180 and 150°C, as compared to
that at lower temperatures below 120°C.

The [MC layers, Cu6Sn5 and Cu3Sn layers, present in solder joints are formed by
the reaction of Sn in solder matrix and Cu from Cu substrate during reflow and grow
upon aging. The growth of IMC layers occurs through diffusion of Cu and Sn in opposite

directions. In this study, due to the complexity of two diffusing atomic species in

147

 

 

Initial IMC layer thickness (um)

 

 

I I I I I I I I I I I I I I I I I I I I I l I I I I j I
, .

I I I I

Average . :
""“3.8'um """" """"""""""""

............................................................

1111 [ML ILLLI L1

2-0 um 2.0 um

......................................................
............

 

IIIII‘IiITIIIIIIIIIIIIITTII
. . .

l l I I I l l l l I l I l l

 

 

    

 

. artectllc Sin-' 1 i
Eutectic Sn-Ag w, Fesmssf Euta we Sn-Ag
partlcles WI "'3an particles

Figure 5.3 Initial thicknesses of IMC layers in
eutectic Sn-Ag and composite solder joints.

148

 

I

st—ITIIIIIIIIIIIIIIFIIj—TII
O

—0— 1.0-c Eutectic Sn-Agf

   
 

 

15

1O

IIIWIIIIIIIIIIIIII

I

I

h

I

l

a

I

I

t

'

a

I

I

I

I

I

I
11141111111111].ng

 

 

 

Total IMC layer growth (um)

5 - 7315-59770", $117.19- . ——

L114

 

I I
I I
111111111111111111111111

o 500 1000 1500 2000 2500
a) (Time in sec)"2

A

 

 

   

 

25», I I I I Ifijj I FT VI I I I I I I rT I Id
A ——9-—180-c :EutecticSn-Ag o a
g :- a— 1m 5 wlFoSnIFaSn, :
V 20 .— —(>— 1m"? particles “' """ j
1: ~ ' : . -
1; --a--1oo-c: . ' _
0 E I - I r
._ 15 . ...... _
a j . : , :
E, 1 : - : 1
g 10 ------- .0 ------ f --------- : ------- 3
0 ' : f : ,9
2 i :43” - 3
Z 5 a! ...... _
8 I q. w -1
O ' Isl-Liana :
P ' -1

111 111 1 1 111 1 1 111 1 141111 1

 

 

0o 500 1000 1500 2000 2500
b) (Timeinsecw2

A

 

 

 

 

 

 

20 """"" I I I I I I I I I I I I I I I I
A _—9— 1m Eutectic Sn-Ag , :
g .- 9— 1506 w! "1,511, particles : _
v " -0 — 1m : : ' “
i 3
c I
h
a _
3 —t
>. i
2 -l
o I
g E
5 -
O a
I- I

o 1 1 1 1 1 1 1 1 1 1 1 I l I 1 I l 1 I 1 1 1 1 1
o 500 1000 1500 2000 2500

(C) (T lme in sec)"2

Figure 5.4 The growth of total IMC layer (both CUGSns and Cuasn) during aging
up to 1400 hours, (a) eutectic Sn-Ag solder joints, (b) composite solder joint
with FeSn/FeSn; particles, (c) composite solder joints with Ni38n4 particles.

149

opposite directions, the thickness measurements of IMC layers were used to address the
growth kinetics of the [M layers during aging, assuming diffusion-controlled growth of
IMC layers by Sn and Cu in opposite directions. As a result, layer growth coefficient is
used for analysis in solder studies, rather than traditional diffusion coefficient based
equations [141].

The growth kinetics of the CU68I15 and Cu3Sn IMC layers was determined using a

simple growth model,

d=¢+fi6

The layer-growth coefficient, K, is strongly related to the diffusion coefficient of
diffusing atomic species determining the growth of the [MC layers. d is the IMC layer
thickness at aging time t, and do is the initial thickness of the IMC layers. The layer-
growth coefficient, K, was detemrined from a linear regression analysis of d versuss rm,
where the slope is K”. K is then related to activation energy for the growth of IMC

layers by Arrhenius relationship describing the temperature dependence of K as follows :

K = A exp(- EX”)

where A is the temperature-independent pre-exponential factor, E * is the activation
energy for the IMC layer growth, R is the ideal gas constant, and T is the absolute
temperature. The activation energy was calculated from the slope of In K vs 1/1‘ plot.

The layer-growth coefficients, K, for the growth of Cu6$n5 and Cu38n IMC layers

150

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

awed find on
wwod 3.6 09
. . . . 85 us:
.03 o cm o 35 o 2 m CS .530
£06 36 mod 5.5m cw“
mmmd five 03.0 2: mad 3; of
Sod mvod wad _ me 966 end on
$06 _ _.o 8.9 3 .0 mad mwd o3
. . . . . . .05 us:
Goo N~_ :mo SL 030 mm_ OS £530
506 h. _ 5 $06 w.m~ ammd @0— of
mwmd mmm v36 hmm 306 3 N 9:
NM A000\~Eom_b: M mm C0$Eo£b$ M Na A000\~Eon_-o$ M G v
O
020an Ems Z 53, 00.0303 ~=m0n<=m0m FEB w<-:m 0:033“ 008809009
w<-:m ouoousm m<ém 300:5

 

 

 

 

 

 

000.?— US: 0.0 530% 05 .8.— 00:0_05000 fiaouwaomfl 0000—00—00 Nd 030m.

151

in non-composite and composite solder joints at 180, 150, 120, 100, and 70°C are
provided in Table 5.2. Most of the linear correlation factor, R2, were greater than 0.9.
This study and other previous studies [89,16,91] confirm that the growth of IMC layers is
diffusion-controlled in both non-composite and composite solder joints over the
temperature range studied.

At higher aging temperatures of 180 and 150°C, composite solder joints had higher
layer-growth coefficients for the Cu68n5 HVIC layer growth than non-composite solder
joints. At an intermediate aging temperature of 120°C, both non-composite and
composite solder joints exhibited similar layer—growth coefficients. These observations
suggest that FeSn/FeSn; and N13Sl’l4 particles have no substantial effects on reducing the
growth of Cu6Sn5 layer during aging at temperatures above 120°C. Therefore,
FeSn/FeSnz and Ni3Sn4 particles are not expected to reduce the layer growth in
conditions encountered in automotive under-the-hood environment, where the
temperature can reach as high as 150°C [2].

On the contrary, composite solder joints exhibited lower layer-growth coefficients
for the growth of Cu6Sn5 layer than non-composite solder joints at lower aging
temperatures of 100 and 70°C. This indicates that the reinforcement particles in
composite solder are effective in reducing the growth of Cu68n5 layer at temperatures
below 120°C. Thus, composite solders containing FeSn/FeSn; or Ni38n4 particles would
be beneficial in stabilizing the Cu6Sn5 layer in relatively low-temperature applications
such as in microelectronic components in typical office/home environment.

The growth of Cu3Sn layer was slower in composite solder joints than in the non-

composite Sn-Ag solder joint at all aging temperatures used in this study as shown in

152

 

 

Table 5.2. This observation is consistent with that in a previous study [90] in which Ni
retarded the growth of Cu3Sn layer while increasing the growth of (21.168115 layer.
Moreover, the Cu3Sn IMC layer became visible only at a far later aging time in
composite solder joints aged at lower temperatures than in the non-composite solder
joints. The composite solder joints thus were efficient in reducing the growth of CU3Sn
layer at temperatures used in this study than the non-composite Sn-Ag solder.

Figures 5.5 and 5.6 are the Arrhenius plots from which activation energies for the
growth of Cu6Sn5 and Cu3Sn layers were obtained. The best fit to the K vs T data shows
that the activation energies for Cu6$n5 and Cu3Sn layer were temperature dependent.
Such results suggest that the controlling mechanisms for CuGSns and CU3Sn layer growth
are different at high and low temperatures. The activation energies for the growth of
Cu68n5 layer in the non-composite and composite solder joints with FeSn/FeSnz or
Ni3Sn4 particles at higher temperature region (TITm ~ 0.8 — 0.92) were 1.3, 1.38, and 1.46
in eV/atom, respectively. Higher activation energies in composite solder joints were due
to the higher layer-growth coefficients at higher aging temperatures of 180 and 150°C.
The activation energy of 0.12 eV/atom in the non-composite solder joint in low
temperature region was quite lower than that at higher temperatures. The activation
energy in composite solder joints at lower temperatures was not determined because of
insufficient data. The activation energy for Cu3Sn IMC layer growth in the composite
solder joint with FeSn/FeSn; particles was higher than that in non-composite solder joint.
The activation energy for Cu3Sn IMC layer growth in composite solder joints could not

be determined due to the inability to visually resolve the Cu3Sn layer.

153

 

 

In K (cmzlsec)

 

 
  
       

 

 

 

I I I j I I I I I I I I fT I I ffi I .1
' ' ——e— Eutectlc Sn-Ag
I E’ = 1.3 eV/atom
'30 " """"""""""""""""""""""" j n! = 0.999
3 -El — Eutectic stt-Ag
'32 T """""""""""""""""""""""" ’2 wl FeSn/FeSq pertlclee
C . Eutectic Sn,” E' = 1.33 eVIatom
.34 _ --.§--.E*=o.129v - 82:0.998
: \k E 3 --<>-- Eutectic Sn-Ag
.35:— ....................... \ .......... _ wINgSn‘partlclea
: 1 I ~;- -_ a j 5* = 1.46 eVIatom
r - : “+ --. . ‘ n2 = 0.968
.38 1- 1 1 1 1 1 1 M1 1 1 L 1 L 1; 1 1 Q 1 J
2.2 2.4 2.6 2.8 3
In (K") x 1000

Figure 5.5 Arrhenius plot for the growth of Cu68n5 IMC layer
in eutectic Sn-Ag and composite solder joints.

 

    
 
   
  

 

 

 

'28 ' I I I I I I I I I IiT I I I T _
—e— Eutectic Sn-Ag _
E‘ = 1.07 eVIatom —
-30 --------------- - B— Eutectlc Sn-Ag —
wI FeSaneSq :
3 - particles ~
to '32 ”‘ """"""" E' = 1.34 eVlatom —
«T ~ : _
E t ~
3 34 L .............................................. g
x _ .
.5 ~ ~
-36 —— ------------------------------------ :- - -- —

.38 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

2.2 2.4 2.6 2.8 3
1/T (K") x 1000

Figure 5.6 Arrhenius plot for the growth of Cuasn IMC
layer in eutectic Sn-Ag and composite solder joints.

154

 

 

5.4 SUMMARY

The formation and growth kinetics of Cu68n5 and Cugsn layers during isothermal
aging up to 1400 hours were investigated in non-composite and composite Sn-Ag solder
joints. The Ag3Sn particles showed significant coarsening while FeSn/FeSnz and N13SI‘I4
particles remained stable during the aging in this study. Voids were present only within
Cu3Sn layer due to Kirkendall effect. The microstructures of composite solder joints
revealed thinner initial IMC layers than non-composite solder joints, thus suggesting an
influence of FeSn/FeSnz and Ni38n4 particles on the formation of IMC layers.

Both non-composite and composite Sn-Ag solder joints exhibited a similar behavior
in the growth of IMC layers. For the Cu68n5 layer, the layer-growth coefficients at
temperatures above 120°C were higher in composite solder joints, while composite solder
joints exhibited lower layer-growth coefficients at temperatures below 120°C.
FeSn/FeSnz and Ni3Sn4 particles were effective in reducing the growth of CuGSns layer at
lower temperatures. In the present study, it was revealed that Cu3Sn IMC layer grew
slower in composite solder joints at all temperatures. The growth of Cu6Sn5 and Cu3Sn
IMC layers appeared to be controlled by two different temperature-dependent diffusion

related mechanisms.

155

 

CHAPTER 6

CREEP PROPERTIES OF EUTECTIC Sn-Ag SOLDER JOINTS
AND Sn-Ag COMPOSITE SOLDER JOINTS CONTAINING
INTERMETALLIC PARTICLES INTRODUCED BY IN-SI TU METHODS

ABSTRACT

Creep behavior of the eutectic Sn—Ag joints and Sn-Ag composite solder joints
containing 20 Vol.% of Cu6Sn5, Ni3Sn4, and FeSn; intermetallic reinforcements
introduced by in-situ methods was investigated. These creep tests were carried out using
single shear lap solder joints at room temperature, 85, and 125°C. The creep resistance
was similar in magnitude for all alloys, and with increasing temperature, the stress
exponents decreased in a manner consistent with power-law breakdown behavior. The
FeSnz intermetallic reinforced composite solder was found to be the most creep resistant
alloy at room temperature. Creep failure was observed to occur within the solder matrix
in all these solder joints. Although the smearing of the features present in the fracture
surface made a detailed analysis of the processes involved difficult, there were

indications of grain boundary separation, ductile fracture, and interfacial separation.

156

 

 

 

6.1 INTRODUCTION

Global economic and environmental pressures warrant that lead be removed from
electronic solders [143]. Another concern in the electronics industry is that performance
limitations of Sn-Pb solders are being exhausted due to requirements of surface mount
technology. In many electronic systems, the majority of reliability losses were identified
with mechanical failures of the solder joints rather than device malfunctions [3]. Several
Pb-free solders have been considered as alternatives to Sn-Pb solders. Among potential
Pb-free solders, eutectic Sn-Ag solder (Sn-3.5Ag in wt%) is amongst the leading Pb-free
candidates [11,14,21,15]. Also for electronic systems used in harsh conditions such as
automotive under-the-hood and aerospace applications, a composite solder approach has
been proposed to improve the performance of solders [25,92,26].

In thermomechanical fatigue at high homologous temperatures, both thermal and
atherrnal processes are operative [11]. At high frequencies or low temperatures, failure is
dominated by atherrnal processes. However, at low strain-rates or high temperatures,
creep deformation dominates. Creep has been identified as a major deformation mode of
the solder joint. Grain boundary sliding, an important deformation mechanism in creep,
was typically observed in therrnomechanically fatigued solder joints [13].

Creep deformation of the Sn-3.5Ag solder has been investigated in many studies
[92,45,12,46,30,53,48]. However, there is no clear agreement about the responsible
creep mechanism. A brief summary on the creep property of the Sn-3.5Ag solder is
given next to elucidate the current understanding.

Mathew et al. [45] reported that the Sn-3.5Ag bulk solder of 20pm grain size

exhibited the stress exponent of 5 and 60.7 kJ/mol activation energy. They compared this

157

 

data with the creep data of pure Sn of 150um grain size with a stress exponent of 7.6 and
an activation energy of 60.3 kJ/mol and cited lattice diffusion-controlled dislocation
climb as the controlling mechanism. Despite the difference in stress exponents, creep
deformation of Sn-3.5Ag solder was therefore attributed to the same mechanism as that
of pure Sn. Yang et al. [12] and Liang er al. [46] regarded Sn-3.5Ag solder as a particle-
hardened alloy. Dislocation climb over the Ag38n particles was considered as the rate-
controlling mechanism.

Darveaux et al. [30] performed creep tests using actual ball grid arrays of Sn—3.5Ag
solder joints. From the measured stress exponent of 5.5, they concluded that dislocation
climb was the controlling mechanism. However, the observed activation energy of 38.5
kJ/mol was considerably lower than the expected value of lattice (106 kJ/mol) or
dislocation pipe (63kJ/mol) diffusion in pure Sn. They surmised that the activation
energy was stress dependent since the data was in the power law break down regime,
producing a lower activation energy than usual.

Igoshev et al. [53] suggested other mechanisms such as grain boundary sliding
and/or dislocation movement to account for observed stress exponents higher than 5.
Other microstructural studies on crept specimens appear to support this suggestion
[53,48] Frear [48] reported that the creep strain was accommodated through the Sn
matrix of 1 pm grain size, at Sn-Sn grain boundaries. Igoshev et al. [53] observed that
microcracks were accumulated at grain boundaries due to grain boundary sliding, leading
to an intergranular fracture. The accumulation of grain boundary defects started at an

earlier stage of creep and lasted for at least 70 ~ 80 % of the lifetime.

158

 

 

 

In our previous study [92], Sn-3.5Ag solder joints reinforced with Cu58n5 particles
showed two or three order of magnitude higher creep resistance in as-fabricated
condition, compared to the eutectic Sn-Ag solder joint. But with aging, the difference in
creep resistance was less dramatic. In this study, the creep deformations of unaged Sn-
3.5Ag solder and its in—situ composite solders reinforced with Cu6Sn5, Ni3Sn4, or FeSnz
are characterized by constant-load creep testing of single shear lap solder joints. The
effects of creep test methodology and reinforcement particles on measured creep
properties are discussed. The significance of the particle type in creep resistance is

examined. Fracture processes during the creep of the solder joints are also investigated.

6.2 EXPERIMENTAL PROCEDURES

Sn—3.5Ag solder stock was received in the form of cast ingots. Sn-Ag composite
solders were prepared by incorporating 20 Vol.% Cu58n5, Ni3Sn4, or FeSnz intermetallic
particles into the Sn-3.5Ag solder by in-situ methods. The composite solders were made
by intrinsically producing the intermetallic reinforcements within the solder, rather than

by mechanically adding intermetallic particles to the solder. The Sn-3.5Ag and

composite solders were rolled to produce a sheet of about 100 um thick and cut into

2 2

preforms of about 1.0 mm area. Single shear lap joints with an area of about 1 mm
were fabricated with preforms of the solders by soldering together two Cu substrates of
semi—dogbone shape. Further experimental details about fabrication of the solder joint
were described in Chapter 4 [141]. The solder joints were prepared for microstructural

examination by standard metallographic techniques, finishing with 0.05 um SiOz

colloidal suspension.

159

 

 

Constant load creep tests were performed by hanging a dead-load on one side of the
single shear lap solder joint mounted on an aluminum beam. Tests were conducted in air
at room temperature, 85°C, and 125°C, with the applied stress varying from 3 to 25 MPa.
Figure 6.1 shows how a test specimen was set up to do elevated temperature creep
experiments. Such testing was carried out using a small heating pad affixed to the side of
the Cu substrate, which was electrically connected to a variable transformer. The
temperature was measured by a thermocouple soldered closely to the joined region of the
Cu substrates.

Creep strain was measured using laser ablation patterns placed on the side of the
solder joint by an excimer laser before creep testing [144]. Optical microscopic images
of laser patterns were taken periodically. The creep strain was determined by calculating
the changes in slope of the laser patterns, which distort from the original position with
time. Details of creep data analysis using this method was described in our previous
study [145].

The fracture surfaces of the solder joints were examined by scanning electron
microscopy after the solder joints failed. The actual stress applied to the solder joint was
determined by dividing the applied load by the actual joined area, obtained after

subtracting the large void area in a solder joint cut by the fracture plane.

160

 

 

       

Thermometer

"M” as
Variable
'II'ansformer

   

Figure 6.1 Schematic drawing of creep testing method.

161

 

6.3 RESULTS AND DISCUSSION
Initial Microstructure

The microstructure of a eutectic Sn-Ag solder joint in Figure 6.2(a) shows the Cu
substrate, continuous Cu6Sn5 interfacial layer, and eutectic Sn-Ag solder matrix. The
eutectic Sn-Ag solder matrix consisted of almost pure Sn grains peppered with Aggsn
particles. Within the solder matrix, a few (31.168115 intermetallic particles were observed
near the Cu substrate, which were formed during reflow of the solder joint [84].

Particulate reinforced Sn-3.5Ag composite solder joints, Figures 6.2 (b) ~ (d),
showed a microstructure similar to that of the eutectic Sn-Ag solder joint along with
incorporated Cu68n5, Ni38n4, and FeSnz particles, respectively. The solder joint
reinforced with FeSnz particles revealed that some FeSn; possessed a dual composition
manifested by the difference in contrast, indicating two different interrnetallics of FeSn
(dark contrast) and FeSn; (bright contrast). There were variations in the particle shape

and size distribution in each composite solder joint.

Creep Properties of Solder Joints
Figures 6.3 (a) ~ ((1) show the variation of secondary creep strain-rate with applied
stress at various temperatures for eutectic Sn-Ag solder joints and its composite solder

joints. A power law relationship of the form,

7:141”.

describes the data over the stress ranges and temperatures used in this study, with some

scatter in the data. 7 is the shear strain-rate, 2' and n are the shear stress and stress

162

 

 

 

000.00 m<m.m-:m 000.0050. 0.2000 05000 g .0200 m<-cm 0000000 00
050.. 000.00 05 00 000.000.0220. .002 003,050 000050.00: 2mm 3V a 3 No 0.590

:3. II.|: 0.0.:£__._0 .5 .013. 0:27.200 0.0

I;

- r..\ ..
ix. éflV . 00%;:

J0??— 0&0. v

.0

w .151. -.
0.5100300. . 0v. 0...:—

 

163

 

..00_00 m<m.m-cm 000.250. £00.00 £50". A3 .5060 m<m.m-cm 000.250. 25:00 Emmi A2
050.. .0080 05 *0 0.3002025... 855 0532.0 0000.00.58 2mm AB .0 A2 Nd 0.09“.

.03. 3. 0:27:00 .5

i :l...

. v. c
. .y. t=m .fiév a

E 2.. c _

avg—utm— St 3, v

v ..c .
., =0 :0, ,
.. M ..

P

F...

 

164

 

 

 

..mm. ..0 .0 .000 0.0.. 0.00 5.2, 9.0.0 0.52 .0060 2.000800 05000 3.
.250. .00.00 0035 0:00.00 .0. .0020 00:000 .0> 0.0.5020 0.0.0 .6005 .0. a .0. 0.0 0.00.“.

A032. 000.5@

 

 

. Jfi. . . . . J . . aogfi
............................................. mi:
X B H H
.2. 0.2 .302... x .. 0.0." 0 h 0.
.2. 0.2 .30.. 0.00 a x. . .m ........ . - h
.2. 0.8 .00.... 0.2: 0 a Q.
.2. 0.0" .002 0.02 0 ..mn ........... (.:.I ............... ..:-0.
.2. 0.2.3.0. 0.00. a u 0.0 u 0 \ . m
UCWN‘ II-lll I I .l ......... r\ ..... ". web " = ........ l nucfi
0.0. I II I
0.2 .I-I. \ I. ..:. c .20 as. 0.
........... ..:-E :.:-0.0.... 0.0.0.000? v-
m n00.00 .3 30.0.00
. _ p 0 . r . . . . 9.5."

 

_ 2.

(PS) 0188 luv-us

.002. 000.50
A: m

.0.

2.:

 

   

d —H —J — A .1 Auemflfi II®III
............. m.....--.....-..0..00l®. 1:

“ DOWN-l1 S
........... ..:-120.-..- ..A: w
0.0": u

................. ..V. O 0.... u
x0 .0.

Q .6 a

......... . I 1: 0.,
ml: ..:

002.0 0003......
b. . . . . . . . nuAVfi

 

165

 

 

 

0.50.. .0200 050000.00 N50.... .0. .0050. .0200 050000.00 «0.00.2 .0.
00020 005000 .0> 0.0.5020 0.0.0 0.0005 .0. a .0. 0.0 0.00.”.

 

 

 

 

.002. 08.00 30V .002. 08.00 A0.
. cfifi.4. . . .U WNfi Hau°fi . Av—H. .. . _ . . . Hauafi
. W .. :9.- W m 0.02:4:
........... .::-.:-:--.Uoma IO: «.0. 0900 I1: 0....
. 4
021' N. 0.5-«III N.
............ V w... m 4x 2: m
m. {x 0.0": m.
............ ..‘-. okvfi m ........ .1‘\‘...‘ c.¢.— m
. MW «‘ \\w.h|: a
§ ........................ 0.0. 0..) ...................... ‘ 0 ........ .... ........ 0.0. 0.»
H (U. xx‘ (
~0m0m>$ v.0. : ....... ‘ .................... 0....
0000.00 M 0.0sz 0000-00
r nuAvM . _...0 r? 5 . nuafi

 

 

 

166

exponent, and A is a material-related constant. The stress exponents tended to decrease
with increasing temperature with exception of the Cu68n5 particulate reinforced solder
joint, which exhibited a stress exponent of 6.5 at 85°C. The stress exponents at a given
temperature were found to be more or less similar, ranging from 5 to 12, which are
similar to values given in other studies [12,46]. The solder joint reinforced with FeSnz
particles showed the highest stress exponent values at room temperature.

Data from our prior work [92] on the same Cu6Sn5 composite solder joints shows
substantially better creep resistance than the present study, as indicated in Figure 6.3(b).
The datum points for the earlier study were obtained by unloading the specimen
periodically to take micrographs of a particular scratch to quantify the shear
displacement. This allowed recovery to occur that was not possible in the present study,
where loading was uninterrupted. This implies that when the loading direction was
reversed, recovery effects apparently reduced the mobile dislocation density that could be
regenerated upon reloading, resulting in a reduction of strain over time by 2-3 orders of
magnitude. When these joints were aged, the creep resistance was degraded due to the
effects of coarsening the Ag3Sn particles.

The scatter in data was related to differences in how damage accumulation
developed during creep deformation. Figure 6.4 shows two representative creep curves
of strain vs. time along with schematic diagrams of damage accumulation in the solder
joint during creep deformation for the Ni3Sn4 particle reinforced solder joint crept at
85°C. The creep curve in Figure 6.4(a) shows an upward curvature for most of its
lifetime, resulting in a lower creep strain before fracture, whereas the creep curve in

Figure 6.4(c) exhibits the conventional downward curvature expected in the materials

167

 

 

0000:0200 0:00c0moEoE._ £000.00. .0. 0:0 .0.

.E_0._ .0060 05 000.00 000050.080 0m0E00 2.0.050 9.00002:
00.0605 000E050 0:0 00200 000.0 02.00.000.00. 03... .0. a .0. 0.0 0.090

.0.

 

 

€00.05;

 

0.: N 0.: m.“ 0.: m ..: m
l — _

_._4|—4-.|1.._._ L

   

Ill

...........................................

..........................................

..........................................

llllIrilllllllIllIIIIllllllll

......... £002 .3 u<m.m-:0.
_

. _ _ — — — _ p P — — ~ — p _ _ h -

 

llll

lll1|lllllllllllllllllllllllll

 

 

m...

IS

[“81

168

 

 

 

0000000000 00000000080 00000.00. .0. 000 .0.
.050. .0200 0.0 000.00 0000380000 0000.00 «00.0.00 0000002:
0E0.m0_0 0000.008 000 00200 000.0 0>0000000.00. 0.5.. .0. a .0. 0.0 0.090

    
    

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. :0. 300.083. AU.
0.... ._ mam 1:0 2:0 .30 ..:N
.m ’ .. T q q 4 M 4 A q _ u q d m q 1 d _. u q
w ....... ......... a...
.155.. ......... W ....... m 0.0 S
1 . . . . 19
- . . H . J
- m u H n m.
.. ....... ......... ....... 3. u
.1 ........... 0.00 S «02 RE .1. 0...
m H 50sz 000.30 m
, p _ b 0 L _L P 0 r h p _ . _ _ p _ fl
4

169

where dislocation climb is the controlling mechanism. The difference between the two
curves correlates with the different deformation history of the solder joint. The solder
joint that exhibited an upward curvature in creep curve had inhomogeneous strain
distribution across the joint thickness, resulting from preferential strain on one side of the
solder joint as shown in Figure 6.4(b). On the other hand, the solder joint shown in
Figure 6.4(d), exhibiting a downward curvature in creep curve, deformed nearly
uniformly across the joint thickness. The inhomogeneous deformation in Figure 6.4(b) is
thought to be influenced by undetected defects present initially in the solder joint such as
porosity or a non-uniform particle distribution produced during specimen fabrication. In
many cases, inhomogeneously deforming joints also showed higher minimum creep rates
than specimens with more homogeneous deformation. This result points out that a better
microstructural control is required to obtain consistent benefit from particle-reinforced
composite solders.

Figures 6.5 (a)~(c) compare the behavior of each composite solder joint with the
eutectic solder joint, and the data of Darveaux and Banerji [30] using a Dorn model for
creep, which is most valid at lower normalized stresses below 170 ~ 7x104, where

power-law breakdown is not apparent.

-_ Pfl’lnép 3'1. :_Q__£"

kT G d A'G kT G

where D is the operative diffusion process for creep with activation energy Q, G is the
temperature dependent shear modulus, b is the Burgers vector, k is Boltzmann’s constant,

T is temperature in K, d is the grain size, n is the stress exponent, taken to be 5.5 (as used

170

 

 

 

.80. 0.0.0000 w x000>.0o 000 .00. ..0 .0 _000 .0 0.00 0.05 0020 0.50.. .0200 000000.00
009.00 000 0<-0w 0000.00 .0 830000 0.0.0 .6090 00N__00..00 00... .0.m.0 0.00E

Sn. ..0 .0 0.000209 l

.8. 000 .0002: .2026
03. 02 .80.. 08 .2030
08. 000 .000... 02: .2030
.00. 08 .0000 002 .2030
.90 000 .80.. 000. .3026
08 .2030

000 .8050

00.2 .3036

000 600300

000 600300

002 608:5

GOO-IIBEEEEX

U...

.:.: N

 

lllllllll 'Illlllll Illllllll lllllllll lllllllll

PHI” I l

[lllllll l

 

35.3003... 0o+mv0.cu.<.
..oE\_J_m.meO
3.0 .0 x000>00n

30.0003: NN+m:o. .H.<
408:0. o~.nO

wflmoH—O.

.030003: m.+moo..n.<
.0050. oonO.

0000.00

..:. .0 .0.—U 0% 000.0 .0000...—

lllHll l l llllllll l Illlllll I Illlllll l lllllll I l

[“1111 l l

I

111404

llllllll

 

 

2.0.

0—

w—

2

0—

m—

w—

n—

..:

.0.

.V/(MIomxam/mé

171

 

 

..om. 0.0.000m .0 0.000200 .0 0.00 0.05 0020 0.50.. .0200 000000.00 «0w0_z
000 0<-0m 0000.00 .0 530000 0.0.0 >000.0 0050.000: 00 0. 3.0... 0.0001.

Sn. ..0 .0 5.00209 |

000 .5002
000 0:002
002 .2002
000 008.00
000 003:5
000. 008.00

000444

 

 

       
       

 

 

0.: .:.: N
_ 0 0 . 0 0 _ QNIQH
.ml ............................................... h .................... m 0.16H
m 1 m
ml. ........................................... 02‘ ................ [m ”MIQH
1 ....................................... 1.2.:
H .00 . H
I ............. q ...... 30.03.00 030000101. 3....
m Beinmmno m
m . Bu» Xflfiu>hfl m
01.11.3341 ....... ZN GI" 07:“
H 4 20.80.00 _N+mz.ou.< H
.25: oeuo
HI ............................ 4. ............ £5..le 0...:
m 00,080.00 2+08..u.< m
.1 .............. 4 ........................... _ 2.03 8101. 2....
0000.00“
m 4 _ .::—.0 2.000.... m

.0.

.VI(.L>1/0)dxa(o/J.)é

172

 

 

.80. 0.0.0000 w x000>.0o .0 0.00 0.0. 0020 0.50.. .0200 0.000080 N0m0u.
000 0.1.5 0000.00 .0 830000 0.0.0 >000.0 005.0000: 00 0. .0.m.o 0.000...

Sn. ..0 .0 0.000.»..09 III.

000 .8000
0mm .250...
002 .8000
000 6000.00
000 .0030:
000. 608.00

 

  
     

11111111 1

l'lllllll l

llllllIT T
lllLll l l l

Inlnltttall:onbntu1||0IIl|QIICIOIOOIUOIOtIII

I [lllllll 1
1.111111! 1

FT

 

0
[Hill] l_L

ImTr

 

..oE\_.0_m.meO
3.3 .0 0300209

lmnrr 1
I
hLllllll

IIIIIIIIIIIIIIIIIIIIIIIIIII

30.08.00 01000.12.
.38.: ciao

 

000...

[Will l1 T

ill.-. IIIQDIIIIC IIIIIIIIIIIIIIIIIIIIIIIIII

I

l

90030.0 2.080"...
.250. 810.

l

 

0
[1111111 L 111111111

-III'IIOOIIIII‘llllollOJOlIIVIIIIOII lllll

 

[mini 1

~..:-.0 .0000...

 

.:.: N

2.0.

0—1Avfi

2.0.

:-:.

0...:

0...:

0...:

:.0.

173

.VI(L)I/o)dxa({)/.L)é

by Darveaux and Banerji), p is the grain size exponent, and A is a geometrical parameter
representing microstructural features. When terms that are assumed to be constant are
combined into A’, a set of data could be plotted using the second equation with various
values of Q. The value of Q that causes the data at different temperatures to collapse
onto a single line represents a good estimate of the controlling activation energy for
creep. For the eutectic solder joints, a value of 60 kJ/mol caused a better fit of the data
than the value of 38.5 kJ/mol used by Darveaux and Banerji. 60 kJ/mol is about half of
the activation energy for tin self-diffusion [45,30], suggesting that grain boundary or pipe
diffusion may control deformation of the eutectic solder joints. A value of Q=120 kJ/mol
was most effective for the Cu68n5 and Ni3Sn4 composite solder joints, but for the FeSn;
composite solder joints, Q=38.5 kJ/mol was more effective than 60 or 120 kJ/mol. Once
the Q value was determined, the A’ parameter could be found so that the solder joints
could be compared on a uniform scale. These values are indicated in the plots in Figure
6.5 (a)~(c).

Figures 6.5 (a)~(c) show how the composite solder joints typically have the same
properties as the eutectic solder joints at elevated temperatures, but not at lower
temperatures. For the Cu68n5 and Ni3Sn4 composite solder joints, room temperature
deformation appears weaker than the eutectic solder joints at equivalent normalized
stresses, while the FeSn; composite solder joints are effectively stronger. This suggests
that the Dom model may be meaningful at high temperatures, but a different composite
modeling strategy needs to be developed for composite solder joints at lower

temperatures.

174

 

 

The morphology of the reinforcement particles differs among the three composite
solders. Cu63n5 particles tend to be elliptical, Ni3Sn4 particles are angular, and FeSn;
particles tend to be rounded although they exhibit irregular interface morphology. These
variations imply that the mechanisms of dislocation climb over particles would be
different for the three reinforcements due to different interface chemistry, geometry, and
interfacial dislocation structure, and this may account for the difference in stress
exponents. Since solder microstructure is very sensitive to the thermomechanical history,

the variation in creep behavior may be attributed to the variation in microstructure.

Fracture of Solder Joints

A representative fracture surface of the crept solder joint is given in Figure 6.6(a),
illustrating the presence of the porosities in the solder joint. The porosity within the
solder joints ranged from 5 to 30 % of the total joint area. The large-scale porosity was
subtracted in calculation of the actual stress applied on the solder joint since voids do not
carry the load. Figure 6.6(b) shows the typical fracture surface of a solder joint crept to
failure in shear. Most of the fracture surface exhibited failure by shear through the solder
matrix. Thus, the fracture surface shown in Figure 6.6(b) illustrates the smeared Sn
phase in the shearing direction.

In the minor areas of the fracture surface, fracture processes that included tensile as
well as shear components of deformation were found. Eutectic Sn-Ag solder joints
exhibited cracks along the grain boundaries of the Sn phase as shown in Figure 6.7(a),
indicating that the creep strain was accommodated by intergranular separation. Spherical

and elliptical dimples of Sn phase illustrated in Figure 6.7(b) were found in the fracture

175

 

 

 

 

..0020 0.00 :_ 0.300.. 2 0:0 000.50 00.00.00 :0
38.8 0.0.2 5? .50.. .028 9.0000 820.50. 22:8 80.50 3
00.0.. .0200 50.0 0.... .0 000.50 0.300.. 0>=0E000a0m 0.0 0.09“.

 

176

 

0.300.. 0.0000 00:00.00. 00.0E.0 A... 000 00000000.. 0.0.0 0m-0m 0020 0000.050 9.00.0 .0.
00000090 0.300.. 00.2.2.0 0000.00.00: 2mm 3. w .0. No 0.30.“.

 

177

surface of Ni3Sn4 particle reinforced solder joint crept at room temperature, indicating
that this region failed in a highly ductile manner. A particle can be found at the bottom
of some dimples, indicating cavity nucleation at the particle/matrix interface. Similar
dimples were observed in the fracture surface of the Cu68n5 particle reinforced solder
joints crept at room temperature.

Fracture by interfacial separation between Cu68n5 intermetallic interfacial layer and
the solder matrix was found in very few regions. Figure 6.7(c) and ((1) show the faceted
surface of the Cu68n5 intermetallic layer in the fracture surface of the eutectic Sn-Ag
solder joint crept at 85°C and 125°C. The exposure of the Cu68n5 intermetallic interfacial
layer in the fracture surface indicates that the crack grew locally by debonding between
the Cu68n5 intermetallic layer and solder matrix. The fracture surface region in Figure
6.7(c) shows complete debonding, but Figure 6.7(d) shows remaining solder ligaments
indicating partial debonding. It is possible that if the solder joint fracture surface were
not smeared by shear during fracture, the favored fracture processes could be observed.
However, it is not easy to ascertain which is the preferred fracture process based on the
complications indicated. If the overall failure process were a mixture of shear and
tension, these failure modes would probably dominate the fracture surface more than the
smearing shown in Figure 6.6(b). The initiation of a crack may develop by local tensile

effects that are evident in Fig. 7.

6.4 SUMMARY
In this study, creep behavior of the eutectic Sn-Ag and three different types of

composite solder joints were investigated. The stress exponents decreased with

178

M.v— -'

 

 

 

00.000000 .0....00 .0. 0.0.0.... .0200 000 .0>0. .0.00..0.0. 000000 .0 00.000000 20.00.00 .0.
00000020 0.300.. 00.3000 0000.00.20. .200 .0. .0 A0. 0.0 0.00....

1.30.

 

t

179

increasing temperature, ranging from 5 to 12. The FeSn; particle reinforced solder joint
exhibited highest stress exponent values. Scatter in the data was correlated with
differences in damage accumulation process during creep. The solder joints that
deformed at an accelerating rate (rather than a nearly constant secondary rate) underwent
inhomogeneous deformation due to a preferential strain on one side of the solder joint.
This instability of the solder joint during creep in shear appeared to be due to undetected
defects in the solder joint such as porosity or a non-uniform particle distribution. The
solder joints reinforced with different particles exhibited different creep behavior
depending upon the type of incorporated particle. Creep resistance is better if loading is
cyclic rather than monotonic, so that recovery occurs during a stress reversal. All solders
deformed similarly at elevated temperatures consistent with a Dom creep model, but
significant differences at room temperature indicate that appropriate composite modeling
is needed at lower temperatures. Some solder joints contained regions exhibiting mixed
tension and shear fracture processes such as ductile fracture identified with dimples,
cracks along Sn-Sn grain boundaries, and debonding along interface between
intermetallic layer and the solder matrix. However, a clear picture could not be obtained
for most of the solder joint due to the rubbing of the fracture surfaces during the failure

process.

180

 

 

 

CHAPTER 7

MECHANISTIC ISSUES ON CREEP STRENGTH OF SOLDER JOINTS

Creep properties of eutectic Sn-Ag and composite solder joints containing
intermetallic particles have been investigated as detailed in Chapter 6. This study shows
two distinct observations in creep strength of the solder joints. One is that when the
solder joints experience cyclic creep deformation conducted by loading and unloading the
specimen periodically, the solder joints exhibit substantially lower creep strain-rate. The
other is that the composite solder joints exhibit weakening or strengthening compared to
the non-composite eutectic Sn-Ag solder joints at room temperature, depending on the
type of particulate reinforcement. The solder joints reinforced with either Cu58n5 or
Ni3Sn4 particulates showed weakening, while ones reinforced with FeSnz particulates
showed strengthening. In this Chapter, possible mechanistic issues that could account for
the observed features in creep strength of the solder joints are discussed, to suggest future

works to verify hypotheses regarding mechanistic issues that determine creep properties.

7.1 EFFECTS OF CYCLIC CREEP ON THE CREEP STRENGTH
The datum points of the Cu68n5 particle reinforced solder joints in the normalized
plot, Figure 6.5(a), are reproduced in Figure 7.1 to clearly illustrate increased creep
resistance in cyclic creep deformation at room temperature. It is apparent that the solder
joints are more creep resistant by 2 to 3 orders of magnitude in cyclic creep conditions.
During unloading, the specimen is apparently subject to conditions similar to strain

relaxation [146]. In the strain relaxation experiment, the specimen is prestrained at a

181

 

 

 

10.3 l l I T I I I I I I I I I I I I I I I
Composite solder joints
_ reinforced with Cuosns

I IIIII
l l 11111

1 0'4
0 continuous creep

I I IIIlIl
1 11 11111]

 

C cyclic creep O

.5 ..........................................................

l 1 [111111

10'6

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

1 0'7

Strain Rate (8")

I I IIIIII] I IIIIIIII I IIIIIIII

10'8

l llllllll l 1 111111] 1 11111111

I I IIIIIII

 

 

10,9 1 1 1-11L411111111111

10 20
Stress (MPa)

 

Figure 7.1 Comparison of steady state strain-rate of the Cu68n5 composite
Sn-Ag solder joints for continuous and cyclic creep deformation.

182

 

fixed stress, and then rapidly unloaded. The length of the specimen is then measured
with respect to time to record how the specimen shrinks. Deformation may be
conveniently divided into elastic, anelastic, and plastic parts. Upon unloading, elastic
strains are immediately recovered, whereas both anelastic and plastic deformation remain
in materials. However, anelastic strains are recoverable with time, while plastic strains
are permanent [147]. Thus, as result of unloading in the strain relaxation experiment, the
elastic strain is fully relaxed, the plastic strains are static, and only the anelastic
deformation causes continued strain reduction.

Eutectic Sn-37wt%Pb and Sn-2wt%Pb alloys exhibit pronounced anelastic behavior
in strain relaxation experiment [148,149]. These alloys were prepared by rolling at room
temperature and heat-treating to produce specimens of various grain sizes. The similar
anelastic behavior in spite of large difference in composition indicates that Sn is
primarily responsible for the anelasticity. In stress relaxation of the eutectic Sn-Ag and
Cu68n5 reinforced composite solder joints, the % stress drop was similar in both solder
joints. This indicates that Cu68n5 particles do not affect stress relaxation and only solder
matrix is involved in stress relaxation behavior [150]. Since the stress relaxation is
caused by deformation that reduces the internal stress within the material, strain reduction
during strain relaxation may be due to similar mechanisms. The anelastic behavior of Sn
is thus applicable to the eutectic Sn-Ag and composite Sn-Ag solder since the eutectic
solder matrix has almost pure Sn phase containing less than 0.05% Ag. If anelastic
deformation causes significant strain reduction during each unloading cyclic, this could

account for the low creep strain-rate observed in cyclic creep deformation.

183

 

 

Another important consideration is the low melting temperature of solder materials.
For solder materials, the room temperature is above half of melting temperature,
indicating that the solder materials are under hot deformation condition even at ambient
temperatures. It was noted that plastic strains are static in strain relaxation. However,
solders may deform to behave plastically during strain relaxation due to effects of
residual stress and recovery. This implies that unloading in cyclic creep allows recovery
and reduces the mobile dislocation density, as the strain is reduced.

A simulation of cyclic creep using the data from continuous creep was attempted to
examine the considered factors that cause strain reduction. The primary and secondary
creep data for the datum point marked by an arrow in Figure 7.2 was utilized. The
primary and secondary creep can be described phenomenologically by the following

equation [151],
8 = 80 (1 — exp(— 0a)) + té‘ss ,

where e is the creep strain at time t, £0 is the amount of primary creep strain, 6"” is the

secondary stage strain-rate, and a is the fitting parameter. The primary and secondary

creep was best fitted with parameters of 8,, = 0.011, 5}“: 4.45 x 10'6 /sec, and a: 0.008

/sec as shown in Figure 7.2. In the simulation of cyclic creep, an unloading event was
introduced every 100 ksec and a constant 30% strain drop at each unloading were
selected with the assumption that the same creep behavior as that in the continuous creep
occurred with each reloading event due to application of the same stress. This choice of

30% is typical of stress relaxation in an hour [150]. Figure 7.3 shows the cyclic creep

184

 

 

     
   

 

 

   

 

0.2 I r I I I I T T T T l I I r ! r I I I

— Steadyistate strain-rate 5 _
0.15_...‘ 4.45X10.6/Sec ”---.--: ........... i
E - ; = :
2 001 ----”-----“§ """"""""" :"°""""-“§ ------------- _
“" : : : -
m ' —4
0.05— ------------ ............. -
0 l 1 1 i 1 1 1 1 fl

0 10° 1 104 2 104 3 10“ 4 10‘

Time (sec)

Figure 7.2 Strain vs. time curve showing primary and secondary
stages of the datum point marked by an arrow in Figure 7.1.

 

185

curve obtained by connecting datum points of each strain value after the strain drop,
compared with the curve for uninterrupted monotonic creep. The creep strain in cyclic

creep is then expressed by this equation,

8 = 80 (1 _ exp (— at» + téss _ gunloading ,
where summing is the 30% strain drop during unloading event.

The simulated cyclic creep curve (B in Figure 7.3) expects a lower minimum strain-
rate by 2 orders of magnitude than that in monotonic creep. This indicates that the cyclic
creep approach based on strain relaxation upon unloading could account for the low creep
rate. However, the amount of creep strain over time is significantly larger than that
observed in cyclic creep testing represented by the curve C in Figure 7.3. The strain
values over time must be close to those observed in cyclic creep in order for the strain

relaxation approach to be valid. This indicates that there are other operating processes

that reduce the overall creep strain, since the monotonic strain-rate overpredicts the strain.

Greater frequency of unloading events during cyclic creep would result in a lower
cumulative creep strain. In contrast, the data in the composite solder specimen were
obtained with larger time increments which would cause the model in Figure 7.3 to
predict even larger cumulative strain. Furthermore, the magnitude of stress relaxation
decreased with decreasing prestrain [150], and the creep strains were smaller than the
prestrains used in stress relaxation. Reducing the amount of relaxation in each cycle

would also move the model further from experiment.

186

 

 

 

 

 

0.2 I I I I I I I I I I I I I I I I I I I I I
. A .
r- -n
0.15— r
0 _ .. 1. 3 g g :- 2 2 B ~
g L 4.5 x 10'3 [sec -
'3 0.1 ‘-
h - _
H ’ -
m — _
0.05~— ' C 4
0+ -
i— 1 l l l l L l l l l l l l 1 l L I l l l J -i

 

 

 

0 10° 5 104 l 105 1.5 105 2 105

Time (see)

Figure 7.3 Simulated cyclic creep curve with unloading event every 10 ksec and
30% strain drop at each unloading. Curve A is one from continuous creep, curve
B is the simulated cyclic creep, and curve C is the experimentally observed one.

187

 

In this simulation, it was assumed that the same creep behavior as that in monotonic
creep will occur at each reloading, resulting in the same strain-rate. However, if the
microstructure was not allowing the same dislocation substructure to develop, but instead
changed toward increasing creep resistance due to the cyclic loading/unloading, the
strain-rate during reloading would be reduced. If cyclic strain caused hardening, the
strain-rate in the each subsequent reloading event would be decreased, leading to a lower
strain over time than that expected with the assumption that a steady state substructure
consistently developed.

The secondary stage during creep deformation proceeds by maintaining a certain
balance between hardening and softening (recovery) at a given constant stress, leading to
a constant strain-rate. If the recovery during unloading is small, the dislocations may
accumulate with reloading, resulting in increasing dislocation density. A subgrain
network consistent with climb controlled creep would form with the same stress applied,
the subgrain size would be the same, but if dislocations accumulated in subgrain walls,
then the misorientation would increase with each subsequent reloading. If this process
leads to cumulatively larger misorientations, then grain boundary sliding could develop,
resulting in large strain.

During unloading, if the subgrain network is maintained at some degree because of
severely intertwined dislocations allowing small amount of recovery, it will result in
smaller amount of primary strain at each subsequent reloading and secondary strain—rate
stage will develop earlier. A smaller amount of primary strain at subsequent reloading

was considered to be able to produce another hardening effect. However, it was found

188

that it did not produce any pronounced effect in the simulation of cyclic creep shown in
Figure 7.3.

Tin (Sn) is known to have high elastic anisotropy. The high anisotropic property of
Sn can also cause hardening effect. In a polycrystalline Sn alloy, when there are large
misorientations between grains, the grain boundary must be readjusted to a new shape at
reloading to accommodate strains along the grain boundary caused by the high anisotropy.
The readjustment of grain boundaries can occur by generating more geometrically

necessary dislocations at grain boundaries, also leading to an increase in dislocation

density. Hardening can occur at room temperature since % a 0.6, and recrystallization

or recovery processes may be slow, so dislocations can accumulate. This can also
increase hardening at each reloading, causing smaller strain-rate at each subsequent
reloading.

The processes discussed above can occur at the same time or only a single process
may take place during the cyclic creep deformation. In order to verify the suggested
processes, further systematic experimental study on cyclic creep should be conducted. In
particular, measuring the strain reduction during unloading would provide important data
on degree of anelasticity within the solder joint. Since the anelastic strains of Sn—Pb
alloys measured are larger than can be predicted by existing models [149], much'work on
dislocation dynamics is needed before physically based models that describe Sn can be

developed.

189

7.2 EFFECTS OF PARTICLES ON THE CREEP STRENGTH

The effects of particle reinforcement on the creep behavior were different among 3
composite solder joints, causing weakening or strengthening depending on the type of
particle. Both weakening and strengthening behavior were observed in aluminum alloy
depending on matrix microstructural stability and testing temperature [152]. It is not
certain at present whether the particles incorporated in the eutectic Sn-Ag solder affect
the microstructure of solder matrix.

The morphology of the reinforcement particle differs among the three composite
solders. Cu6Sn5 and Ni3Sn4 particles tend to be smooth, whereas FeSn; particles
exhibited considerable roughness. It is considered at present that the difference in
interface morphology and chemistry among the particles may cause the difference in
creep behavior. These variations imply that the dislocation interaction with particles
would be different, and affect the motion of dislocation passing by the particles. The
effect of shape could be investigated on solders reinforced by Ni-Sn intermetallic
particles, as described by Fu Guo [153], that have different interface structure, but the
same composition. Also important future work is to investigate the interaction between

dislocation and particles with different interface property.

190

CHAPTER 8

SUMMARY

Deformation behavior and damage accumulation during thermomechanical fatigue
were characterized using the actual eutectic Sn-Ag and Sn-4Ag-0.5Cu solder joints. The
eutectic Sn-Ag and its in-situ composite solder joints were investigated on creep
deformation and interfacial IMC layer growth. The important findings from those studies
are summarized.

Microstructural studies of actual Sn-Ag or 95.5Ag/4Ag/0.5Cu solder joints that
underwent 250 and 1000 thermal shock cycles between -40°C and +125°C, with a 20
minute hold time at each extreme were performed. Damages during thermomechanical
fatigue initiated on the free surface and propagated through interior solder joints. The
crack that would cause a catastrophic failure of solder joint occurred very near the
ceramic substrate, initiating at the free surface of the solder joint and progressing
inwards. Significant amount of residual stresses was observed after thermomechanical
fatigue, which can cause further damage to solder joints during storage. Small amounts
of copper did not have any pronounced effect on thermomechanical fatigue of Sn-Ag
solder joints. Pb phase in the solder matrix also did not affect the thermomechanical
fatigue behavior of the Sn-Ag solder joints.

Growth behavior of the Cu68n5 and Cu3Sn interfacial IMC layers in the eutectic Sn-
Pb and Sn-Ag, and their corresponding composite solder joints was characterized during
aging at temperatures ranging 50°C to 180°C up to 5000 hours. The eutectic Sn-Pb and

Sn-Ag solder matrices revealed extensive coarsening and the intermetallic layers also

191

grew significantly. Growth of the interfacial IMC intermetallic layers was affected
several factors such as Pb-rich phase accumulation, the layer thickness itself, the aging
temperature, and Cu68n5 particles. For the eutectic Sn-Pb composite solder joint, the
Cu68n5 particles affected the growth of the intermetallic layers only at the initial stage of
aging. However, the Cu5Sn5 particles did not affect the intermetallic layer growth in the
Sn-Ag composite solder joint during aging at 150°C for long periods of time. The Cu68n5
particles were found to be effective in reducing the layer growth for low temperature
aging, but at high temperatures above 150°C significant coarsening of the Cu68n5
particles was observed and caused the enhanced intermetallic layer growth.

The formation and growth kinetics of Cu6$n5 and Cu3Sn layers during isothermal
aging up to 1400 hours were investigated in the eutectic Sn-Ag and in—situ composite Sn-
Ag solder joints. During aging, FeSn/FeSn; and Ni3Sn4 particles remained stable, while
Ag3Sn particles showed significant coarsening. Voids were present only within Cugsn
layer due to Kirkendall effect. FeSn/FeSnz and Ni3Sn4 particles appeared to affect
thickness of IMC layers during reflow, resulting in thinner initial interfacial IMC layer in
composite solder joints. Both non-composite and composite Sn-Ag solder joints
exhibited a similar behavior in the growth of IMC layers. FeSn/FeSnz and Ni38n4
particles were effective in reducing the growth of Cu63n5 layer at lower temperatures.
Cu3Sn IMC layer grew slower in composite solder joints at all temperatures.

Creep study of the eutectic Sn-Ag joints and Sn-Ag composite solder joints was
carried out using single shear lap solder joints at room temperature, 85, and 125°C.
Creep data of solder joints followed power law relationship over the stress ranges and

temperatures employed. The stress exponents decreased with increasing temperature,

192

 

 

ranging from 5 to 12. Different creep behaviors were observed at room temperature
depending upon the type of incorporated particle, while creep strength of the composite
solder joints was similar to the eutectic Sn-Ag solder joints at high temperatures. This
indicates that appropriate composite modeling is needed at lower temperatures. Creep
resistance was found to be better when loading is cyclic rather than monotonic. Some
solder joints exhibited mixed tension and shear fracture processes. The fracture
processes were identified by dimples indicating ductile fracture, cracks along Sn-Sn grain
boundaries, and debonding along interface between intermetallic layer and the solder
matrix. However, the rubbing of the fracture surfaces during the failure process impeded
a clear picture about fracture process of solder joints.

Mechanistic issues that may be responsible to observed cyclic creep behavior and
different creep strength depending on type of reinforcement were discussed. These
include anelastic strain and recovery during unloading, and change in microstructure
upon reloading due to formation of subgrain network and high anisotropy of Sn. In order
to verify the suggested hypotheses regarding mechanistic issues, further systematic

experimental study on creep deformation of the Sn-Ag solder should be conducted.

193

LIST OF REFERENCES

194

10.

ll.

12.

13.

14.

15.

REFERENCES

S. Jin, JOM, 45(7), 3(1993).

W.L. Winterbottom, JOM, 45(7), 20(1993).

P.T. Vianco and DR. Frear, JOM, 45(7), 14(1993).
C. Melton, JOM, 45(7), 33(1993).

M. McCormack and S. Jin, JOM, 45(7), 36(1993).

.l.S. Hwang, Modern Solder Technology for Competitive Electronic Manufacturing,
McGraw-Hill, New York, NY, 21(1996).

J .W. Morris, Jr., Proceedings of the 2nd Pacific Rim International Conference on
Advanced Materials and Processing, eds. K.S. Shin et al., The Korean Institute of
Metals and Materials, Seoul, 715(1995).

WJ. Plumbridge, J. Mater. Sci, Vol. 31, 2501(1996).

J.K. Tien, B.C. Hendrix, and AI Attarwala, IEEE Transactions on Components,
Hybrids, and Manufacturing Technology (CHMT), 12(4), 502(1989).

1F. Hua and J. Glazer, Design & Reliability of Solders and Solder Interconnections,
Proceedings of a Symposium held during the TMS Annual Meeting at Orlando, FL,
Feb. 10-13, 1997, The Minerals, Metals & Materials Society (TMS), Warrendale,
PA, 65(1997).

H. Mavoori, J Chin, S. Vaynman, B. Moran, L. Keer, and M. Fine, J. Electr.
Mater., 26(7), 783(1997).

H. Yang, P. Deane, P. Magill, and KL. Murty, 1996 46th Electronic Components
and Technology Conference ( ECT C ), IEEE, Piscataway NJ, 1136(1996).

S. Choi, K.N. Subramanian, J.P. Lucas, and TR. Bieler, J. Electr. Mater., 29(10),
1249(2000).

S. Chada, A. Herrmann, W. Laub, R. Foumelle, D. Shangguan, and A. Achari,
Soldering & Surface Mount technology, No. 26 July, 9(1997).

W. Yang, LE. Felton, and R.W. Messler, Jr., J. Electr. Mater., 24(10), 1465(1995).

195

16.

17.

18.

I9.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

W. Yang, R.W. Messler, Jr., and LE. Felton, J. Electr. Mater., 23(8), 765(1994).
W.B. Hampshire, Soldering & Surface Mount Technology, No. 14 June, 49(1993).
J. Glazer, Int. Mater. Rev., 40(2), 65(1995).

M. Abtew, G. Selvaduray, Mater. Sci. & Eng, Vol. R27, 95(2000).

M. Ohring, Reliability and Failure of Electronic Materials and Devices, Academic
Press, Chestnut Hill, MA, 439(1998).

M. McCormack and S. Jin, 1994 lEEE/CPMT 16th Int ’1 Electronics Manufacturing
Technology Symposium, Vol. 1, IEEE, Piscataway NJ, 7(1994).

W. Steurer, R. Ferro, and A. Saccone, Physical Metallurgy 4th ed., Vol. 1, eds.
R.W. Cahn and P. Haasen, Elsevier Science B.V., Amsterdam, The Netherlands, 25
& 284(1996).

C.W. Fairhurst and J .3. Cohen, Acta Cryst., Vol. B28, 371(1972).

Metals Handbook 9th ed., Vol. 9, ASM, Metals Park, OH, 455(1985).

A.W. Gibson, S.L. Choi, K.N. Subramanian, T.R. Bieler, Design & Reliability of
Solders and Solder Interconnections, Proceedings of a Symposium held during the
TMS Annual Meeting, The Minerals, Metals & Materials Society (TMS),
Warrendale, PA, 97(1997).

A.W. Gibson, S. Choi, T.R. Bieler, and K.N. Subramanian, IEEE Int’l Symposium
on Electronics & the Environment, IEEE, Piscataway, NJ, 246(1997).

W.D. Callister, Jr., Materials Science and Engineering An Introduction 4th ed.,
John Wiley & Sons, Inc., New York, 220(1997).

R. Arrowood, A. Mukherjee, and W.B. Jones, Solder Mechanics A State of the Art
Assessment, eds. D.R. Frear et al., The Minerals, Metals & Materials Society
(T MS), Warrendale, PA, 108(1991).

R.W. Evans and B. Wilshire, Introduction to Creep, The Institute of Materials,
London, 20(1993).

R. Darveaux and K. Banerji, 1992 42nd Electronic Components and Technology
Conference (ECTC), IEEE, Piscataway NJ, 538(1992).

G.E. Dieter, Mechanical Metallurgy 3rd ed., McGraw-Hill Inc., 445( 1986).

196

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

45.

46.

MA. Meyers and K.K. Chawla, Mechanical Metallurgy principles and
applications, Prentice-Hall Inc., Englewood Cliffs, NJ, 667(1984).

R. Arrowood, A. Mukherjee, and W.B. Jones, Solder Mechanics A State of the Art
Assessment, eds. D.R. Frear et al., The Minerals, Metals & Materials Society
(TMS), Warrendale, PA, 111(1991).

R.W. Evans and B. Wilshire, Introduction to Creep, The Institute of Materials,
London, 45(1993).

G.E. Dieter, Mechanical Metallurgy 3rd ed., McGraw—Hill Inc., 446(1986).

R.W. Evans and B. Wilshire, Introduction to Creep, The Institute of Materials,
London, 46(1993).

R.W. Evans and B. Wilshire, Introduction to Creep, The Institute of Materials,
London, 47(1993).

F.A. Mohamed and T.G. Langdon, Phil. Mag, Vol. 32, 697(1975).

P.L. Hacke, A.F. Sprecher, and H. Conrad, Thermal Stress and Strain in
Microelectronics Packaging, eds. J .H. Lau, Van Nostrand Reinhold, New York,
NY, 472(1993).

T.G. Langdon and FA. Mohamed, Scripta Metall., Vol. 11, 575(1977).

R. Arrowood, A. Mukherjee, and W.B. Jones, Solder Mechanics A State of the Art
Assessment, eds. D.R. Frear et al., The Minerals, Metals & Materials Society
(T MS), Warrendale, PA, 119(1991).

R.W. Evans and B. Wilshire, Introduction to Creep, The Institute of Materials,
London, 59(1993).

M.A. Meyers and K.K. Chawla, Mechanical Metallurgy principles and
applications, Prentice-Hall Inc., Englewood Cliffs, NJ, 672(1984).

R.W. Evans and B. Wilshire, Introduction to Creep, The Institute of Materials,
London, 61(1993).

M.D. Mathew, S. Movva, and K.L. Murty, Key Engineering Materials, Vol. 171-
174, 655(2000).

J. Liang, P.S. Lee, N. Gollhardt, S. Schroeder, and W. Morris, Mat. Res. Soc. Symp.
Proc., Vol. 445, 307(1997).

197

47.

48.

49.

50.

51.

52.

53.

54.

55.

56.

57.

58.

59.

60.

61.

62.

63.

CH. Raeder, G.D. Schmeelk, D. Mitlin, T. Barbieri, and W. Yang, 1994
IEEE/CPMT Int ’1 Electronics Manufacturing Technology Symposium, IEEE,
Piscataway NJ, 1(1994).

D.R. Frear, JOM, 48(5), 49(1996).

J .E. Breen and J. Weertman, J. of Metals, Vol. 7, 1230(1955).

F.A. Mohamed, K.L. Murty, J .W. Morris, Jr., Metall. Trans, Vol. 4, 935(1973).

M. Harada and R. Satoh, IEEE Transactions on Components, Hybrids, and
Manufacturing Technology (CHMT), 13(4), 736(1990).

V.I. Igoshev, J .I. Kleiman, J. Electr. Mater., 29(2), 244(2000).

V.I. Igoshev, J.I. Kleiman, D. Shangguan, C. Lock, S. Wong, and M. Wiseman, J.
Electr. Mater., 27(12), 1367(1998).

W.D. Callister, Jr., Materials Science and Engineering An Introduction 4th ed.,
John Wiley & Sons, Inc., New York, 203(1997).

G.E. Dieter, Mechanical Metallurgy 3rd ed., McGraw-Hill Inc., 377(1986).

M.A. Meyers and K.K. Chawla, Mechanical Metallurgy principles and
applications, Prentice-Hall Inc., Englewood Cliffs, NJ, 691(1984).

M.A. Meyers and K.K. Chawla, Mechanical Metallurgy principles and
applications, Prentice-Hall Inc., Englewood Cliffs, NJ, 692(1984).

G.E. Dieter, Mechanical Metallurgy 3rd ed., McGraw-Hill Inc., 378(1986).

M.A. Meyers and K.K. Chawla, Mechanical Metallurgy principles and
applications, Prentice-Hall Inc., Englewood Cliffs, NJ, 689(1984).

G.E. Dieter, Mechanical Metallurgy 3rd ed., McGraw-Hill Inc., 379(1986).

M.A. Meyers and K.K. Chawla, Mechanical Metallurgy principles and
applications, Prentice-Hall Inc., Englewood Cliffs, NJ, 695(1984).

M.A. Meyers and K.K. Chawla, Mechanical Metallurgy principles and
applications, Prentice-Hall Inc., Englewood Cliffs, NJ, 697(1984).

G.E. Dieter, Mechanical Metallurgy 3rd ed., McGraw-Hill Inc., 393(1986).

M. Kitano, T. Shimizu, T. Kumazawa, and Y. Ito, Current Japanese Materials Res.,
Vol. 2, 235(1987).

198

65.

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

77.

H. Mavoori, S. Vaynman, J Chin, B. Moran, L.M. Keer, and ME. Fine, Mat. Res.
Soc. Symp. Proc., Vol. 390, 161(1995).

S. Vaynman, M.E. Fine, and DA. Jeannotte, Solder Mechanics A State of the Art
Assessment, eds. D.R. Frear et al., The Minerals, Metals & Materials Society
(TMS), Warrendale, PA, 155(1991).

J .S. Hwang, Modern Solder Technology for Competitive Electronic Manufacturing,
McGraw-Hill, New York, NY, 513(1996).

S. Vaynman, M.E. Fine, and DA. Jeannotte, Solder Mechanics A State of the Art
Assessment, eds. D.R. Frear et al., The Minerals, Metals & Materials Society
(TMS), Warrendale, PA, 156(1991).

S. Vaynman, M.E. Fine, and DA. Jeannotte, Solder Mechanics A State of the Art
Assessment, eds. D.R. Frear et al., The Minerals, Metals & Materials Society
(T MS), Warrendale, PA, 159(1991).

D.R. Frear, Solder Mechanics A State of the Art Assessment, eds. D.R. Frear et al.,
The Minerals, Metals & Materials Society (T MS), Warrendale, PA, 193(1991).

S. Vaynman, M.E. Fine, and DA. Jeannotte, Solder Mechanics A State of the Art
Assessment, eds. D.R. Frear et al., The Minerals, Metals & Materials Society
(TMS), Warrendale, PA, 195(1991).

W. Englemaier, Circuit World, Vol. 11, 61(1985).

D. Tribula, D. Grivas, D.R. Frear, and J.W. Morris, Jr., Welding Res. Supp. in
Welding Journal, Oct., 68(10), 4048(1989).

D.R. Frear, Solder Mechanics A State of the Art Assessment, eds. D.R. Frear et al.,
The Minerals, Metals & Materials Society (TMS), Warrendale, PA, 201(1991).

H.D. Solomon, J. Electr. Packaging, Vol. 113 June, 102(1991).

Z. Guo, A.F. Sprecher, Jr., H. Conrad, and M. Kim, Materials Developments in
Microelectronic Packaging .° Performance and Reliability, Proceedings of the
Fourth Electronic Materials and Processing Congress, eds. P.J. Singh, ASM,
155(1991).

H. Mavoori, S. Vaynman, J Chin, B. Moran, L.M. Keer, and ME. Fine, Mechanics
and Materials for Electronic Packaging .° Vol. I - Design and Process Issues in

Electronic Packaging, Proceedings of 1994 International Mechanical Engineering
Congress and Exposition, eds. WT Chen et al., ASME, New York, 165(1994).

199

78.

79.

80.

81.

82.

83.

84.

85.

86.

87.

88.

89.

91.

92.

H. Mavoori and J. Chin, 1992 45th Electronic Components and Technology
Conference ( ECT C ), IEEE, Piscataway NJ, 990(1992).

D. Shangguan, Soldering & Surface mount Technology, 11(3), 27(1999).

D. Shangguan and A. Achari, 1994 IEEE/CPMT Int'l Electronics Manufacturing
Technology Symposium, IEEE, Piscataway NJ, 25(1994).

P.G. Harris, K.S. Chaggar, Soldering & Surface Mount Technology, 10(3),
38(1998).

D.R. Frear, P.T. Vianco, Metall. & Mater. Trans A, Vol. 25A July, 1509(1994).
J. London and D.W. Ashall, Brazing and Soldering, no. 11, 49(1986).

H.D. Blair, T.Y. Pan, J.M. Nicholson, R.P. Cooper, S.W. Oh, A.R. Farah, 1996
IEEE/CPMT Int’l Electronic Manufacturing Technology Symposium, IEEE,
Piscataway, NJ, 282( 1996).

G. Lucey, J. Marshall, C. Handwerker, D. Tench, and A. Sunwoo, Proceedings of
the Technical Program, vol. 1, presented at National Electronic Packaging and
Production Conference (NEPCON) WEST 91, Cahner Exhibition Group, Des
Plains, IL, 3(1991).

W.K. Choi and HM. Lee, J. Electr. Mater., 29(10), 1207(2000).

R.J. Field, S.R. Row, and GK. Lucey, Jr., The Metal Science of Joining, eds. M.J.
Cieslak et al., The Minerals, Metals & Materials Society (TMS), Warrendale, PA,
165(1991).

H.P.R. Frederikse, R.J. Field, and A. Feldman, J. Appl. Phys, 72(7), 2879(1991).

R.F. Pinizzotto, Y. Wu, E.G. Jacobs, and LA. Foster, Proceedings of the Technical
Program, vol. 3, presented at National Electronic Packaging and Production
Conference (NEPCON) WEST 92, Cahner Exhibition Group, Des Plains, IL,
1284(1992).

Y. Wu, J.A. Sees, C. Pouraghabagher, L.A. Foster, J.L. Marshall, E.G. Jacobs, and
RF. Pinizzotto, J. Electr. Mater., 22(7), 769(1993).

D.R. Flanders, E.G. Jacobs, R.F. Pinizzotto, J. Electr. Mater., 26(7), 883(1997).
S.L. Choi, A.W. Gibson, J.L. McDougall, T.R. Bieler, and K.N. Subramanian,
Design & Reliability of Solders and Solder Interconnections, Proceedings of a

Symposium held during the TMS Annual Meeting, The Minerals, Metals &
Materials Society (TMS), Warrendale, PA, 241(1994).

200

93.

94.

95.

96.

97.

98.

99.

100.

101.

102.

103.

104.

CC. Guo, S.M.L. Sastry, and KL. Jerina, Microstructural and Mechanical
Properties of Aging Material, Proceedings eds. P.K. et al., The Minerals, Metals &
Materials Society (TMS), Warrendale, PA, 409(1993).

C.G. Guo, S.M.L. Sastry, and KL. Jerina, Microstructural and Mechanical
Properties of Aging Material, Proceedings eds. P.K. et al., The Minerals, Metals &
Materials Society (TMS), Warrendale, PA, 417(1993).

W.J. Tomlinson and N .J . Bryan, J. Mater. Sci, Vol. 21, 103(1986).
C.T. Ho and D.D.L. Chung, J. Mater. Res., 5(6), 1266(1990).

J .L. Marshall, J. Calderon, J. Sees, G. Lucey, and J.S. Hwang, IEEE Transactions
on Components, Hybrids, and Manufacturing Technology (CHMT), 14(4),
698(1991).

H.S. Betrabet, S.M. McGee, and J.K. McKinlay, Scripta Metall. Mater., vol. 25,
2323(1991).

J .A. Wasynczuk and GK. Lucey, Proceedings of the Technical Program, vol. 3,
presented at National Electronic Packaging and Production Conference (NEPCON)
WEST 92, Cahner Exhibition Group, Des Plains, IL, 1245(1992).

R.B. Clough, R. Patel. J.S. Hwang, and G. Lucey, Proceedings of the Technical
Program, vol. 3, presented at National Electronic Packaging and Production
Conference (NEPCON) WEST 92, Cahner Exhibition Group, Des Plains, IL,
1256(1992).

S.M.L. Sastry, T.C. Peng, R.J. Lederich, K.L. Jerina, and CG. Kuo, Proceedings of
the Technical Program, vol. 3, presented at National Electronic Packaging and
Production Conference (NEPCON) WEST 92, Cahner Exhibition Group, Des
Plains, IL, 1266(1992).

H.S. Betrabet and S. McGee, Proceedings of the Technical Program, vol. 3,
presented at National Electronic Packaging and Production Conference (NEPCON)
WEST 92, Cahner Exhibition Group, Des Plains, IL, 1276(1992).

J .L. Marshall, J. sees, and J. Calderon, Proceedings of the Technical Program, vol.
3, presented at National Electronic Packaging and Production Conference
(NEPCON) WEST 92, Cahner Exhibition Group, Des Plains, IL, 1278(1992).

J .L. Marshall and J. Calderon, Soldering & Surface Mount Technology, no. 26, July,
22(1997).

201

105.

106.

107.

108.

109.

110.

111.

112.

113.

114.

115.

116.

117.

118.

119.

120.

121.

122

J.L. Marshall and J. Calderon, Soldering & Surface Mount Technology, 9(3),
6(1997).

J .L. Marshall and J. Calderon, Soldering & Surface Mount Technology, 9(3),
11(1997).

C.M. Miller, I.E. Anderson, and J .F. Smith, J. Electr. Mater., 23(7), 595(1994).
K.F. Seeling and DC. Lockard, US Patent Number 5 405 577, (1995).

Y. Kariya, Y. Hirata, and M. Otsuka, J. Electr. Mater., 28(11), 1263(1999).

Y.H. Pao, S. Badgley, R. Govila, and E. Jih, Fatigue of Electronic Materials,
gig/[95W 1153, SA. Schroeder and MR. Mitchell, Eds., ASTM, Phil. PA,

R.K. Govila, E. Jih, Y.H. Pao, and C. Lamer, J. Electron. Packaging, 116,
83(1994).

Y.H. Pao, E. Jih, V. Siddapureddy, X. Song, and R. Liu, Sensing, Modeling and
Simulation in Emerging Electronic Packaging, 1996 ASME Int’l Mechanical
Engineering Congress and Exposition, Nov 17-22, 1996 Atlanta, GE, 1(1996).

D.R. Frear, D. Grivas, and J .W. Morris, Jr., JOM, 40, 18(1988).

D.R. Frear, D. Grivas, and J .W. Morris, Jr., J. Electr. Mater., 17, 171(1988).

H. Mavoori and J. Chin, 1995 IEEE 45th ECT C, 990(1995).

W. Engelmaier, Design & Reliability of Solders and Solder Interconnections,
Proceedings of a Symposium held during the TMS Annual Meeting, The Minerals,
Metals & Materials Society (TMS), Warrendale, PA, 9(1997).

W. Englemaier and A.I. Attarwala, IEEE Trans. on CHMT, 12, 284(1989).

H.J. Frost and R.T. Howard, IEEE Trans. on CHMT, 13, 727(1990).

T.Y. Pan, J. Electronic Packaging, 116, 163(1994).

P.L. Hacke, A.R. Sprecher, and H. Conrad, Thermomechanical Fatigue Behavior of
Materials, ASTM STP 1186, H. Sehitoglu, ed. ASTM, Phil. PA, 91(1993).

Q. Guo, E.C. Cutiongco, L.M. Keer, and ME. Fine, J. Electronic Packaging, 114,
145(1992).

Z. Guo and H. Conrad, J. Electronic Packaging, 118, 49(1996).

202

123.

124.

125.

126.

127.

128.

129.

130.

131.

132.

133.

134.

135.

136.

137.

138.

139.

140.

CH. Raeder, R.W. Messler, and LE. Coffin, J. Electr. Mater., 28, 1045(1999).

P.T. Vianco, S.N. Burchett, M.K. Nielsen, J.A. Rejent, and DR. Frear, J. Electr.
Mater., 9, 1290(1999).

R.L. Schalek and LT. Drzal, ASM Advanced Materials & Processes, 152(1),
21(1997).

A. Gickler, C. Willi, and M. Loomans, Surface Mount Technology, 11(11), 44(1997).
K.W. Moon, U.R. Kattner, W.J. Boetinger, and CA. Handwerker, Abstract in Final
Program, The Minerals, Metals & Materials Society (TMS) Fall Meeting, Cincinnati,
OH, 32(1999).

L.G. Earle, J. Inst. Metals, 72, 403(1946).

Z. Mei, P. Callery, D. Fisher, F. Hua, and J. Glazer, ASME Advances in Electronic
Packaging, 2, 1543(1997).

Z. Mei, M Kaufmann, A. Eslambolchi, and P. Johnson, 1998 IEEE 48th ECT C,
952(1998).

Z. Mei, P. Johnson, M. Kaufmann, and A. Eslambolchi, 1999 IEEE 49th ECT C,
125(1995).

S. J adhav, private communication.

A.J. Sunwoo, J.W. Morris, Jr., and GK. Lucey, Jr., Metall. Trans, 23A,
1323(1992).

K.N. Tu, Acta Metall. 21, 347(1973).
D. Grivas, D. Frear, L. Quan, and J .W. Morris, Jr., J. Elect. Mater., 15, 355(1986).
Z. Mei, A.J. Sunwoo, and J .W. Morris, Jr., Metall. Trans, 23A, 857(1992).

R.F. Pinizzotto, Y. Wu, E.G. Jacobs, and LA. Foster, Proc. Tech. Program,
NEPCON West 92, Feb 23-27, Anaheim, CA, 3(1992).

M.E. Warwick and SJ. Muckett, Circuit World, 9, 5(1983).
J .O.G. Parent, D.D.L. Chung, and I.M. Berstein, J. Mater. Sci, 23, 2564(1988).

S.F. Dirnfeld and J .J . Ramon, Weld. J. Res. Suppl., 3738(1990).

203

141

142.

143.

144.

145.

146.

147.

148.

149.

150.

151.

152.

153

. S. Choi, T.R. Bieler, J .P. Lucas, and K.N. Subramanian, J. Elect. Mater., 28(11),
1209(1999).

Binary Alloy Phase Diagrams, vol. 1, ed., T.B. Massalski, (Metals Park, OH :
American Society for Metals), 1774(1986).

B.P. Richards, C.L. Levoguer, C.P. Hunt, K. Nimmo, S. Peters, and P. Cusack,
“Lead-Free Soldering”, UK’s Department of Trade and Industry (DTI) publication,
(1999).

J .P Lucas, F. Guo, J. McDougall, T.R. Bieler, K.N. Subramanian, and J .K. Park, J.
Electr. Mater., 28(11), 1268(1999).

J. McDougall, S. Choi, T.R. Bieler, K.N. Subramanian, J .P. Lucas, Mater. Sci. Eng,
A285, 25(2000).

J .H. Schneibel and PM. Hazzledine, Scripta Metall., Vol. 11, 953(1977).
G.E. Dieter, Mechanical Metallurgy 3rd ed., McGraw-Hill Inc., 434(1986).
C. Homer and B. Baudelet, Scripta Metall., Vol. 11, 185(1977).

J .H. Schneibel and PM. Hazzledine, Acta Metall., Vol. 30, 1223(1982).

S.G. Jadhav, T.R. Bieler, K.N. Subramanian, and JP. Lucas, J. Electr. Mater.,
30(9), 1197(2001).

Dongyi Seo, Ph.D. Dissertation, Michigan State University, 86(1998).

P. E. Krajewski, Deformation, Processing, and Properties of Structural Materials,
Proceedings of a Symposium held during the TMS Annual Meeting, The Minerals,
Metals & Materials Society (TMS), Warrendale, PA, 207(2000).

. F. Guo, Ph.D. Dissertation, Michigan State University, (2002).

204

M IIIIIIIIIIIIIIIIIIII

ill/WM12111111111111Ill/ll”!!!llllllllll/Hl

3 1 93 02305 3584