 

. 7.. i). .

+.

i.

i A
1....

u. .4 ..!.:.
it." >

21.91:. .

vi

“a 3..."? .5 5.6...
Badman? cm

{mt-s.

.

:is .
.11. 4!.1 :-

11‘551: :

. . . .5.
A A...» A1» up» . .
musmn in. man.
'7: , Xi .wt.
ﬁﬂnuﬁﬁ.‘ sit...h
, r 1 3. J
P...
h: :9.”-
9.1.3....

..
anti} :1.

.. .35” n: as

.33.... in.

I 3 7 1-1 ‘2: t .1

by}! n.

2...... t...

.. .35 .12.:
. 5-.

.
‘2.)‘1: 5.1.9.1.
.1, u“ I. if):
xvtmﬂsﬁqatzuiii n.
5. 3.... Hut—H31. ax)...”
3... I. n

. i..D\I.stur;
x... {.1 A a
.212!

.. ha

.32.”... 5...: i . a
inn Arﬂmrﬁ: 35.14.”!
H.515

. .
31,3...)‘133‘

: 1... .1.

. .. :31
wwzf>aﬁﬁ
A gnaw-.33...“
m3»,

:u. l :31
01.1.33»..- 153332:-
.0..!E.I.A..I2£.l;.:la
5 3.5 i: 3 S
33.5}...i1.

2....

livoIIs .3).
.47..$.rito..§h1t... ! 2
. 25:5)! 1 . >1.
. . n1. \- 1.1.3113

Izl‘asaz
3.. :5}! a:
.1: as?

 

‘25..er 3.11.,
7...}. I . .

. J! LIBRARIES
AN STATE UNIVERSITY
1 $STIIIJIACIWSING, MICH 48824-1048
r'\ . .
t (2
WM! .,

This is to certify that the
dissertation entitled

INVESTIGATION OF THERMOMECHANICAL
FATIGUE (TMF) BEHAVIOR IN TIN-SILVER
BASED SOLDER JOINTS

presented by

JONG-GI LEE

has been accepted towards fulﬁllment
of the requirements for the

PhD. degree in Materials Science

 

 

' 7
\8< - iv ﬁbrczwawqam .

 

Major Professor’s Signature

Man/ch 30,2005

Date

 

MSU is an Afﬁnnative Action/Equal Opportunity Institution

 

- '- -—- 1-—

 

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/3 e/CincmateDueJndd-ms

 

INVEST
BE

INVESTIGATION OF THERMOMECHANICAL FATIGUE (TMF)
BEHAVIOR IN TIN-SILVER BASED SOLDER JOINTS

By

J ONG-GI LEE

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

2005

[MYSTIC
BEH

ABSTRACT

INVESTIGATION OF THERMOMECHANICAL FATIGUE (TMF)
BEHAVIOR IN TIN-SILVER BASED SOLDER JOINTS

By

J ONG-GI LEE

Currently, there are two major driving forces for considering alternative materials
to lead based products, speciﬁcally interconnections, in electronics applications,
including the impending legislation or regulations which may tax, restrict, or eliminate
the use of lead, and the trend toward advanced interconnection technology, which may
challenge the limits of present soldering technology. The reliability of solder joints is a
concern because fracture failures in solder joints accounts for 70% of failures in
electronic components. Lead- free solders are being investigated as replacements for lead
solders currently used in electronics.

Solder joints experience thermomechanical fatigue, i.e. interaction of fatigue and
creep, during thermal cycling due to temperature ﬂuctuation in service conditions.
Incorporation of allying addition has been pursued to improve the mechanical and
particularly thermomechanical behavior of solders, and their service temperature
capability. Several service parameters, such as temperature extremes encountered, dwell
times at these temperatures, and the ramp-rates representing the rate at which the
temperature changes, were carried out to investigate the roles of these parameters.

Sn-Ag based solder joints that experienced a long dwell time at the high

temperature extreme exhibited less surface damage accumulation and loss of residual

shear strength
temperature 6
amounts of C
during NF v
Faster
surface dam;
signiﬁcant dc
surface dam
faster heating
A cra.
sliding. The:
during mt.
better TMF
extreme due

retard Sn‘sn

shear strength than that of solder joints that experienced a long dwell time at the low
temperature extreme. In particular, quaternary solder joints that containing small
amounts of Cu and Ni in Sn-Ag solder were signiﬁcantly improved the TMF behaviors
during TMF with a long dwell time at the high temperature extreme.

Faster ramp rate during heating segment of TMF cycles caused highly localized
surface damage in eutectic Sn-3.5Ag and Sn—4Ag-O.5Cu solder joints that resulted
signiﬁcant deterioration in residual shear strength in these solder joints. No signiﬁcant
surface damage and drop in residual strength were found in quaternary solder joints with
faster heating rate during TMF.

A crack initiated and propagated along Sn-Sn grain boundaries by grain boundary
sliding. There was signiﬁcant residual stress within the solder joint causing more damage
during TMF. Addition of small amounts of Cu and Ni into Sn—Ag solder resulted in
better TMF performance under TMF with a long dwell time at the high temperature
extreme due to presence in ﬁne, hard, and sub-micron sized IMC precipitates that can

retard Sn-Sn grain boundary sliding.

lwould like to
members. Dr. 1
like to thank ti
my work. I M

and my son for

ACKNOWLEDGEMENTS
I would like to thank my main advisor, Dr. K.N. Subramanian, and the other committee
members, Dr. J. P. Lucas, Dr. T.R. Bieler, and Dr. P. Duxbury, for their support. I would
like to thank the National Science Foundation (NSF) for funding to enable me to pursue
my work. I would also like to give a special thanks to my family and especially my wife

and my son for their love and support.

iv

LlST OF TAJ
LIST OF FI(

CHAPTER 1
INTRODL'C’

{TUNER}.
OWRWEW
BEHAVIOR
2.1. ROL':
PAC]

Id
'4
~c

7::
O

r.) r.) '

OJ
‘1.)

pa ”1'1 it) If)
’ I

I.)

.4“
53'
m

tJf—‘

My

0.).

.N .N .N
+* + e
.w .w r-

TABLE OF CONTENTS

LIST OF TABLES ................................................................................ viii
LIST OF FIGURES ................................................................................ ix
CHAPTER 1
INTRODUCTION .................................................................................. 1
CHAPTER 2
OVERVIEW OF THERMOMECHANICAL FATIGUE (TMF)
BEHAVIOR IN EUTECTIC TIN-SILVER BASED SOLDER JOINTS ................ 5
2.1. ROLES OF SOLDER JOINT IN ELECTRONIC
PACKAGING AND ITS STRUCTURES .............................................. 5
2.2. PROBLEMS INVOLVED IN HIGH TIN SOLDERS ................................ 9
22-1. Anisotropy of ﬁ—Tin ................................................................ 9
22-2. Allotropic Phase Transformation of Tin ...................................... 12
2.3. FATIGUE ................................................................................. 12
23—1. Isothermal Mechanical Fatigue on Sn—Ag Based Solders ..................... 12
2.4. THERMOMECHANICAL FATIGUE ................................................ 19
24-]. Types of Thermal Cycles ......................................................... 20
2.4-2. Deformation Characteristic of Solder Joints During TMF ................... 23
2.4-3. Comparison of Residual Mechanical Behavior
aﬁer TMF between Lead-bearing and Lead-free Solder ..................... 30
2.5. ALLOYING ADDITION IN TIN-SILVER BASED SOLDERS ............... 31
25-1. Eutectic Sn—Ag system ............................................................ 31
25-2. Sn-Ag-Cu system .................................................................. 34
25-3. Sn—Ag-Cu-Bi system ............................................................... 36
25-4. Sn—Ag-Cu-In system ............................................................... 37
25-5. Sn-Ag-Cu-Ni system .............................................................. 37
2.6. ALLOYING APPROACHES TO ENHANCE
MECHANICAL PROPERTIES OF TIN-SILVER BASED SOLDERS ......... 38
CHAPTER 3
EXPERIMENTAL PROCEDURE ............................................................ 40
3.1. MATERIALS USED ..................................................................... 40
3.2. THE PREPARATION OF SOLDER JOINTS ....................................... 40
3.3. THERMOCMECHANICAL FATIGUE (TMF) SET-UP ........................... 43
3.4. SURFACE DAMAGE OBSERVATION AND MECHANICAL TESTS. . .. . ...43
3.5. ETCHING PROCEDURE ............................................................... 43
CHAPTER 4

RESIDUAL-MECHANICAL BEHAVIOR OF THERMOMECHANICALLY
FATIGUED TIN-SILVER BASED SOLDER JOINTS UNDER TMF WITH

LONGER
4.1. [N
4.2. E)
4.3. RE?

4.3-
4.3-.
4.3-.
4.3.~
4.4. cc

(TIMER
EFFECT C
BEHAVIC
5.1. I.\'
5.2. E.\
5.3. RE

5.3-

LONGER DWELL TIMES AT LOW TEMPERATURE EXTREME ................. 46
4.1. INT ODUCTION .......................................................................... 46
4.2. EXPERIMENTAL PROCEDURE ..................................................... 48
4.3. RESULTS AND DISCUSSION ......................................................... 52

43-]. Effect of TMF Cycles on Simple Shear Behavior of Solder Joints ......... 52
43-2. Effect of TMF cycles on the Creep Resistance of Solder Joints ............ 55
43-3. Fracture of Solder Joints .......................................................... 59
43-4. Possible Model of Observed Residual Behavior ............................... 63
4.4. CONCLUSION ........................................................................... 63

CHAPTER 5

EFFECT OF DWELL TIMES ON THERMOMECHANICAL FATIGUE

BEHAVIOR OF TIN-SILVER BASED SOLDER JOINTS ....................... 66
5.1. INTRODUCTION ........................................................................ 67
5.2. EXPERIMENTAL PROCEDURE ..................................................... 68
5.3. RESULTS AND DISCUSSION ........................................................ 71

53-]. Surface Damage Accumulation in Sn-Ag Based Solder Joints
that Experienced Longer Dwell Time at High Temperature

(+150°C) extreme During TMF ................................................. 71
53-2. Comparison of Damage Accumulation in Specimens with
Long Dwell Times at Low and High Temperature Extremes ............... 74

53-3. Comparison of the Effect of Dwell Times at Different
Temperature Extremes on Residual Simple Shear Behavior
of Solder Joints Aﬁer TMF ..................................................... 82
5.4. SUMMARY ............................................................................... 86

CHPTER 6
EFFECTS OF TMF HEATING RATES ON DAMAGE
ACCUMULATION AND RESULTANT MECHANICAL

BEHAVIOR OF TIN-SILVER BASED SOLDER JOINTS ............................. 87
6.1. INTRODUCTION ........................................................................ 88
6.2. EXPERIMENTAL PROCEDURE ..................................................... 90
6.3. RESULTS AND DISCUSSION ........................................................ 93

63-]. Surface Damage Accumulation in Sn—Ag Based Solder Joints
that Experienced Fast Heating Rate During TMF ............................ 93

63-2. Comparison of the Effect of Different Heating Rates Imposed
During TMF on Residual Shear Behavior of Sn-Ag Based

Solder Joints ........................................................................ 97
63-3. Possible Reasons for Better TMF Behavior in Quaternary
Solder Joints than That of Binary and Ternary Solder Joints .............. 104

6.3-4. Possible Reasons for Worse TMF Behavior in Ternary
Solder Joints Than That of Binary And Quaternary Solder Joints. . . . . ...105
6.4. SUMMARY .............................................................................. 1 10

vi

CHI’TER7

MICROSTRI

BEHAVIOR'
7.1. INTI?
7.2. BAC
7.3. EXPE
7.4. REST
7.5. SUI

CHAPTER 8
SUDIARY.

APPENDEX.

REHERENC]

CI-IPTER 7
MICROSTRUCTURAL FEATURES CONTRIBUTING TO ENHANCED

BEHAVIOR OF TIN-SILVER BASED SOLDER JOINTS ........................... 112
7.1. INTRODUCTION ...................................................................... 1 13
7.2. BACKGROUND ........................................................................ 113
7.3. EXPERIMENTAL PROCEDURE ................................................... 116
7.4. RESULTS AND DISCUSSION ...................................................... 117
7.5. SUMMARY .............................................................................. 127

CHAPTER 8

SUMMARY ....................................................................................... 128

APPENDEX ....................................................................................... 133

REFERENCES .................................................................................... 1 52

vii

7311613.

Tab1e 4.1. 1

T311113 7.1.

 

LIST OF TABLES
Table 2.1. Physical properties of B-tin and a-tin [23,24] ..................................... 11
Table 2.2. CTE values of Cu, Cu3Sn, and Cu68n5 [39] ....................................... 26
Table 2.3. Requirements for suitable alternative lead-free alloys [22] ..................... 32

Table 4.1. Comparison of residual strength of eutectic Sn-Ag solder
joints in various specimen conﬁgurations [45,46,75] ............................ 53

Table 7.1. Residual shear strength of Sn-Ag base solder joints
underwent TMF with different dwell times at

low and high temperature extremes after maximum
1000 TMF cycles [99] ............................................................... 113

Table 7.2. Comparison of strain for the onset of tertiary creep
between eutectic Sn-Ag solder joints and Sn-2Ag-1Cu-1Ni
crept at 22°C and 85°C under stresses ranging from 5.5 to 24 MPa [55].....113

Table IX.1. Details of TMF conditions with two
different dwell times at each temperature
extreme to evaluate material-related parameter ‘r ’.
Note: AT is 165°C (~15°C to 150°C) for all TMF studies ..................... 134

Table D(.2. Details of TMF conditions with two different
heating rates to evaluate material-related parameter ‘p ’.
Note: AT is 165°C (-15°C to 150°C) for all TMF studies ..................... 134

Table IX.3. Details of joint geometry and different imposed

temperature difference used to evaluate material-related
parameter ‘m ’ from published literature [45,46] ................................ 145

viii

Figure 2.1.

‘ 1
Figure 2.-.

Figure 2.7. S.

\\

Hem 2.3. E;

ar
US

Figure 2.9, SC;
la}
(b)
(C)
es;
(far
(d)
of.

Fi‘é’ire 2.10. Scl

T.\-

F‘ig‘dfe 2]]. '.

Figure 2.1.

Figure 2.2.

Figure 2.3.

Figure 2.4.

Figure 2.5.

Figure 2.6.

Figure 2.7.

Figure 2.8.

Figure 2.9.

LIST OF FIGURES

(a) Schematic of typical through hole devices (THDs) ........................... 6
(b) Schematic of surface mount devices ............................................ 7
(c) Schematic of surface mount (BGA) devices [22] ........................... 8
Schematic illustration of the body-centered tetragonal structure [23]. . . . . ....10
Schematic illustration of the elongation of grain 1

due to thermal ratcheting [25] ...................................................... 10
Schematic illustration of strain-controlled cycling [29] ......................... 14

Schematic illustration of stable stress-strain hysteresis loop
under strain-controlled cycling [29] ................................................ 14

Schematic illustration of elastic, plastic, and total strain
versus life curve [30] .......................................................... 16

Schematic drawing of hysteresis loop showing load drop
with increasing number of cycles [32] ............................................. 18

Example of isothermal fatigue test conducting Pb-free
and Pb-bean'ng solder joints at 35°C and 150°C
using a 50 % load drop failure criterion [31] .................................... 18

Schematic illustrations of types of thermal cycling.

(a) thermal shock with no ramp rates consideration,

(b) thermal shock with no dwell times at each temperature extremes,

(0) actual thermal cycling that solder joints can experienced during TMF,
especially automotive under-the hood situation

(ramp rates and dwell times will vary with service), and

(d) power cycling which can be effected by turning on and off

of components in ambient environment [36] ............................... 22

Figure 2.10. Schematic illustration of classic model for

TMF of a soldered interconnection on a PCB [3 8] ............................. 24

Figure 2.11. Various forms of thermal strains due to thermal ﬂuctuations [3 8] ........... 25

Figure 2.12. Schematic illustration of CTE mismatches involve

in solder joints made on Cu substrate.

Note: values are CTE mismatches between

(a) Cu substrate and IMC(Cu68n5) layer,

(b) IMC layer and Sn grains in contact with Sn grains adjacent to IMC,
(c) Sn grains in contact with IMC, and

ix

Fig-are 2.1

Figure 2.1

17:23:23.1.

p

~.~ a
15.1.? .12.

In

mm

Figure 2.13.

Figure 2.14.

Figure 2.15.

Figure 2.16.

Figure 2.17.

Figure 2.18.

Figure 3.1.

Figure 3.2.

Figure 3.3.

Figure 3.4.
Figure 3.4.
Figure 4.1.

Figure 4.2.

Figure 4.3.
Figure 4.4.

((1) Sn gains in bulk ................................................................. 26
Schematic drawing of possible failure modes under TMF[40] ............... 28

Graphical illustration of the effect of temperature
on the strength of gain boundary and gain lattice materials [41] ........... 29

Schematic illustration of
a possible surface damage in a gain due to TMF .............................. 29

Comparison of creep rupture times of
eutectic Sn—Ag, Sn-Bi, Sn-In, and Sn-Pb solders [50] ......................... 33

Creep rupture-resistance test of Sn-Ag, Sn—Ag-0.5Cu, and Sn-Pb.
(a) room temperature, (b) 100°C [54] ............................................ 35

Schematic illustration of gain boundary sliding

due to shear forces applies at high temperature.

(a) gain boundary sliding between two adjacent

gains with no obstacles, and

(b) gains undergo lattice deformation due to

presence of an intermetallic particle on the

gain boundary during gain boundary sliding [63] ........................... 39

Schematic drawing of the dimensions of the Cu Substrate ..................... 41

Schematic drawings of
(a) the solder joint ﬁxture assembly and

(b) a fabricated single shear-lap joint .......................................... 41
Temperature versus time proﬁle for manufacturing of

Sn—Ag based solder joints ........................................................... 42
Schematic drawing of TMF set-up ................................................. 42
Schematic drawing of solder specimen preparation for etching. . . . . . . . . ....45
Schematic of setup of dead weight loading miniature creep ﬂame ............ 50
Illustration of analyzing creep strain using

excimer laser-ablation mapping technique ....................................... 50
Schematic drawing of high temperature creep setup ............................ 51

Effect of thermal cycles on the residual-simple
shear strength of four Sn—Ag based solder joints ................................. 53

Figure 4.8. .“

u it
1111:
Inc:
F3...
‘S'il'ES a
.‘)_ SE

Figure 4.5. Normalized residual-simple shear strength as

function of thermal cycles for four Sn-Ag based solder joints ................. 54
Figure 4.6. Normalized residual fracture toughness as function of
thermal cycles for four Sn-Ag based solder joints .............................. 54

Figure 4.7. Comparison of steady-state creep rates of the four solder alloys
subjected to 0, 250, 500, and 1000 TMF cycles as a function stress.

a Eutectic Sn—A ................................................................... 56
( ) g

(b) Sn-4Ag-0.5Cu .................................................................... 57
(c) Sn-2Ag-1Cu-1Ni ................................................................. 58

[Note: In Figure 7 (b), the lower creep resistance of 500 TMF
cycled joints (as compared to those with 250 TMF cycled ones)
is attributed to the presence of large porosity in them]

Figure 4.8. Normalized steady-state creep-strain rate versus normalized
stress for TMF’ed and non-TMF’ed solder joints used
in this study along with Darveaux’s data for aged
eutectic Sn-3.5Ag solder joints [80].
Abbreviations were used to identify each solder joint
(i.e., SA is Sn—3.5Ag, SAC is Sn-4Ag-0.5Cu, SACNl is
Sn-2Ag-1Cu-1Ni, and SACNO.5 is Sn-2.5Ag-0.5Cu—0.5Ni).
Number appearing before the abbreviation indicate
the number of thermal cycles.
(a) at room temperature .............................................................. 60
(b) at 85°C ............................................................................. 61

Figure 4.9. Representative fracture surface of the crept solder joints.
(3) ductile fracture features smeared due to applied pure shear,
(b) ductile dimple ﬁ'acture .......................................................... 62

Figure 4.10. Schematic of possible crack behavior during TMF cycles.
(a) initially have minimum # of cracks at Sn gain boundaries,
(b) # of cracks increase in the initial stage of TMF, and
(c) TMF accident can connect these cracks to cause failure .................. 64

Figure 5.1. Temperature-time proﬁles of
(a) long dwell time at low temperature extreme (prior study)
and (b) long dwell time at high temperature extreme (current study). . . . . ....70

Figure 5.2. SEM images of surface damage accumulation of eutectic Sn-3.5Ag
with long dwell time at high temperature extreme showing increasing
intensity of shear banding near intermetallic interface with
increasing number of TMF cycles ................................................ 72

Figure 5.3. SEM images of surface damage accumulation of Sn-4Ag-0.5 Cu
with long dwell time at high temperature extreme showing clear

xi

Figure 5.4.

‘

(17,"

20:2

Qgﬁg‘

Figure 5.4.

Figure 5.5.

Figure 5.6.

Figure 5.7.

Figure 5.8.

Figure 5.9.

gain boundary sliding in the solder area with increasing
number of TMF cycles (arrows indicate cracks) ................................ 73

SEM images of thermomechanically fatigued Sn-2.5Ag-0.5 Cu-0.5 Ni
with long dwell time at high temperature extreme showing no sigiiﬁcant
surface damage accumulation even after 1000 TMF cycles .................. 75

SEM images of thermomechanically fatigued Sn-2Ag-l Cu-l Ni
with long dwell time at high temperature extreme showing no sigtifrcant
surface damage accmnulation even after 1000 TMF cycles .................. 76

SEM microgaphs showing gain boundary sliding after 1000 TMF cycles.
(a) Sn-Sn gain boundary sliding of eutectic Sn-3.5Ag and

(b) Sn-Sn gain boundary cracks form near intermetallic

interface for Sn-4Ag-0.5Cu ......................................................... 77

Comparison of surface microstructures of 1000 TMF cycled

eutectic Sn-3.5Ag solder joints. (a) long dwell time at low temperature
extreme [adopted from ref. 44] and (b) long dwell time at high

temperature extreme ................................................................. 79

Comparison of surface microstructures of 1000 TMF cycled

Sn-4Ag-0.5Cu solder joints. (a) long dwell time at low temperature

extreme [adopted from ref. 44] and (b) long dwell time at high

temperature extreme ................................................................. 80

Comparison of surface microstructures of 1000 TMF cycled
Sn—2.5Ag—0.5Cu-0.5Ni solder joints. (a) long dwell time

at low temperature extreme [adopted from ref. 44] and

(b) long dwell time at high temperature extreme ................................. 81

Figure 5.10. Comparison of surface microstructures of 1000 TMF cycled

Sn-2Ag—lCu-1Ni solder joints. (a) long dwell time
at low temperature extreme [adopted from ref. 44] and
(b) long dwell time at high temperature extreme .............................. 83

Figure 5.11. Effect of thermal cycles on the residual-simple shear strength

Figure 6.1.

of four Sn-Ag based solder joints showing minimal drop in

simple shear strength in solder joints experienced long dwell time

at high temperature extreme. (a) long dwell time at

low temperature extreme [adopted from ref. 82]and

(b) long dwell time at high temperature extreme .............................. 85

Schematic of various service parameters that can

affect thermomechanical fatigue (TMF) damage
on electronic solder joints .......................................................... 89

xii

Figure 6-2

Figure 6.3.

Figure 6.4.

Time 6.6.

r; “I
* tyre 6.7,

Flilh‘re 6.3,

‘é‘dre 6.9_
F2536 6.19

flan,
:‘.e 6.1]

Figure 6.2.

Figure 6.3.

Figure 6.4.

Figure 6.5.

Figure 6.6.

Figure 6.7.

Figure 6.8.

Figure 6.9.

Temperature-time proﬁles of TMF.

(a) overall temperature-time proﬁle showing ﬁxed

dwell times at each temperature extremes and cooling rate,

(b) a slow heating rate with ﬁxed dwell times and cooling rate, and

(c) a fast heating rate with ﬁxed dwell times and cooling rate .............

Surface damage accumulation on eutectic Sn-Ag solder
joint undergoing TMF with faster heating rate.
(a) 0 TMF cycle, (b) 250 TMF cycles,

(c) 500 TMF cycles, and (d) 1000 TMF cycles ...............................

Surface damage accumulation in eutectic
Sn-Ag solder joints after 1000 TMF cycles.
(a) slower heating rate imposed during TMF, and

(b) faster heating rate imposed during TMF ..................................

Surface damage accumulation in Sn-4Ag-0.5Cu solder joint
undergoing TMF with faster heating rate.
(a) 0 TMF cycle, (b) 250 TMF cycles,

(c) 500 TMF cycles, and (d) 1000 TMF cycles ...............................

Surface damage accumulation in Sn-4Ag-0.5Cu solder joints after

1000 TMF cycles. (a) slower heating rate, and (b) faster heating rate. .

Surface damage accumulation in Sn-2.5Ag-0.5Cu—0.5Ni
solder joint undergoing TMF with faster heating rate.
(a) 0 TMF cycle, (b) 250 TMF cycles,

(c) 500 TMF cycles, and (d) 1000 TMF cycles ...............................

Surface damage accumulation in Sn—2Ag-1Cu-1Ni solder
joint undergoing TMF with faster heating rate.
(a) 0 TMF cycle, (b) 250 TMF cycles,

(c) 500 TMF cycles, and (d) 1000 TMF cycles ..............................

Surface damage accumulation in Sn-2.5Ag-0.5Cu-O.5Ni
solder joints after 1000 TMF cycles.

....92

...94

...95

...96

.98

...99

...99

(a) slower heating rate, and (b) faster heating rate ........................... 100

Figure 6.10. Surface damage accumulation in Sn—2Ag-1Cu-1Ni

solder joints after 1000 TMF cycles.

(a) slower heating rate, and (b) faster heating rate ............................ 100

Figure 6.11. Residual mechanical behaviors of Sn-Ag based solder joints

undergoing 1000 TMF cycles. (a) shear strength versus number
of TMF cycles with slower heating rate,
(b) normalized shear strength versus number

xiii

ﬁgm6ll

Egm6ll

Egm6li

ﬁgméii

of TMF cycles with slower heating rate, .................................... 102
(c) shear strength versus number of TMF cycles

with faster heating rate, and (d) normalized shear strength

versus number of TMF cycles with faster heating rate ...................... 103

Figure 6.12. Etched microstructures of (a) eutectic Sn-Ag and
(b) Sn-2Ag-1Cu-1Ni. (Cu-Ni-Sn) IMC precipitations
that may key gain boundary movements can be seen in (b) ............... 106

Figure 6.13. Backscatter Electron (BE) images of as-fabricated
Sn-4Ag-0.5Cu showing small amounts of Cu6Sn5 IMC particles.
in as-fabricated condition.
(a) low magriﬁcation, and (b) high magriﬁcation ........................... 106

Figure 6.14. Series of images of Sn-4Ag-0.5Cu solder joint that
experienced 1000 TMF cycles. (a) Scanning Electron (SE) image,
(b) Backscatter Electron (BE) image,
(c) Energy Dispersive Spectroscopy (EDS) mapping showing
Sn in these particles,
(d) EDS mapping showing Cu in these particles, and
(e) EDS mapping showing Ag in these particles .............................. 107

Figure 6.15. SEM images of Sn-4Ag-0.5Cu solder joints after 1000 TMF cycles.
(a) Backscatter Electron (BE) image showing (Cu-Sn) IMC
particles at cracked area near IMC layer,
(b) Backscatter Electron (BE) image showing (Cu-Sn) IMC
particles at cracked area with the solder region ............................... 108

Figure 7.1. Schematic drawing of solder specimens used in the present study ........... 118

Figure 7.2. SEM images of as-fabricated specimens without etching.
(a) Eutectic Sn-3.5Ag, (b) Sn-4Ag-0.5Cu,
(c) Sn-2.5Ag-0.5Cu-O.5Ni, and (d) Sn-2Ag-1Cu-1Ni ........................ 118

Figure 7.3. BE images of etched eutectic Sn-Ag.
(a) low magniﬁcation microgaph of solder
specimen showing Cu substrate and IMC layer,
(b) high magtiﬁcation microgaph of solder specimen
showing Sn-Sn gain boundaries and Ag3Sn precipitates
(typical eutectic microstructure of Sn-Ag solder) .............................. 120

Figure 7.4. EDS of second phase IMC precipitates presented in
(a) Sn-4Ag-0.5Cu, and (b) Sn-2Ag—1Cu-1Ni ................................... 121

Figure 7.5. BE images of etched Sn-4Ag-O.5Cu.
(a) low magniﬁcation microgaph of solder specimen

xiv

figure 7.7.
Figure 1X1

Figure L\’.3.

F 13% DH.

'9
fr

1
5;

Fr)».
‘é‘re r F
I4\.7. EI
Pa

showing Cu substrate and IMC layer,

(b) high magiiﬁcation microgaph of solder specimen

showing Sn-Sn gain boundaries, Ag3Sn precipitates and

Cu6$n5 binary IMC precipitates located at Sn-Sn gain boundaries ......... 122

Figure 7.6. BE images of etched Sn—2.5Ag-0.5Cu-0.5Ni.
(a) low magriﬁcation microgaph of solder specimen
showing Cu substrate and IMC layer,
(b) high magiiﬁcation microgaph of solder specimen
showing Sn-Sn gain boundaries, Ag3Sn precipitates,
and Cu-Ni-Sn ternary IMC precipitates located
at Sn-Sn gain boundaries ......................................................... 124

Figure 7.7. BE images of etched Sn-2Ag-1Cu-1Ni.
(a) solder specimen showing Cu substrate and IMC layer,
(b) solder specimen showing Sn-Sn gain boundaries,
Ag3Sn precipitates, and Cu-Ni-Sn ternary IMC precipitates
located at Sn-Sn gain boundaries .................................................. 125

Figure IX. 1. A schematic of residual properties of solder joints under
TMF conditions classiﬁed into three stages.
Solid line represents region of useful life of solder joint.
The dot line represents the range in which catastrophic
failure takes place and is not addressed in this study ......................... 139

Figure D(.2. Material-related parameter ‘r’ in [Am/AIL]r for
Sn-Ag based single shear-lap solder joints .................................... 141

Figure D(.3. Material-related parameter ‘p’ in [(dT/dt)h]”’ for
Sn-Ag based single shear-lap solder joints .................................... 142

Figure D(.4. Comparison of material-related parameter ‘p ' in [(dT/dt)h]‘p
for eutectic Sn—Ag solder joints evaluated from both actual
TMF test and isothermal shear test ............................................. 144

Figure IX.5. Material-related parameter ‘p’ in [(dT/dt)h]"’ and
‘q ’ in [ (dT /dt)c] *1 for eutectic Sn-Ag solder joints
from isothermal shear test ......................................................... 145

Figure IX.6. Effect of solder alloy on the material-related parameter
‘m ’ in [2177“. Note: Joint geometry was the

same for all these conditions ..................................................... 147

Figure IX.7. Effects of solder joint geometry on material-related
parameter ‘m ‘ in [A7]” .......................................................... 148

XV

1.1. 310‘“
L63
intercom“:
past sex-era
“variability '
det‘elOPmcr
environmen
The
Elmronic I
hazardous n
Japanese El
disposed in
the L'nired S
lead-free star
’Jie world. .\
is to eliminat
:6].
A3013".
capability sir

i

for a »

K

CHAPTER 1

INTRODUCTION

1.1. Motivation

Lead-tin (Pb-Sn) alloys have been the most important solders for the
interconnection and packaging of modern electronic components and devices over the
past several decades due to their low cost, good manufacturability, as well as good
wettability on common substrate such as Cu and Ni used in electronics [1-5]. However,
developments of alternative lead-free solders are required due to increasing
environmental and health concerns of the toxicity of lead.

The effort made in the European Union’s Directive for Waste Electrical and
Electronic Equipment (WEEE) will be in effect from January 2006. The disposal of
hazardous materials from electrical and electronics products will be banned [6]. The
Japanese Environmental Agency has proposed that lead-containing scrap must be
disposed in sealed landﬁlls to prevent lead leaching. Consequently electronics sold by
the United States and other worldwide electronics producers to these countries must meet
lead-free standards to maintain trade. The United States is the largest consumer of lead in
the world. Nevertheless, the trend being set by some of the major countries in the world
is to eliminate the use of lead in consumer products for health and environmental reasons
[6]-

Another concern is that Sn-Pb solders are reaching the limit of performance
capability since the adoption of surface mount technology (SMT). Increasing demands

for geater packaging density and high performance are being realized with the advent of

 

sulfa“e

package

I:

been exit“

tin-51'1“?r
demandln.
smite €01
rear-hem
ph}‘SlC31 p1"
and meea
based lead-1
to improVC‘ F
Rene

in electronic
the printed c
reliability ber
connector and

tire mach

The m‘

c

31133

W
dm on the joi

So" . .
luﬁjomt faili

L: .
‘16 Combinatior
“5171mm of rh

surface mount technology, resulting in an increase in the numbers of solder joints per
package and reduction of joint dimensions [3,10].

In order to satisfy the requirement of replacing lead-bearing solder, there have
been extensive studies of lead-free solders. Among the various lead free solders, eutectic
tin-silver (Sn-Ag) is an attractive candidate alloy to meet the requirement for the
demanding high temperature service environment such as automotive under-the-hood
service conditions. However, the melting temperature of this alloy is higher than that for
lead-bearing solder. For depressing the melting temperature and also improving its
physical properties, some additional elements such as copper, nickel, indium, antimony,
and rare-earth elements are added to Sn-Ag solderas [1 1-16]. Thus, candidate Sn-Ag
based lead-free solder will be a ternary or quaternary alloy containing additional elements
to improve properties of eutectic Sn—Ag solder.

Renewed interest in solder has been driven by reliability concerns of solder joints
in electronic packaging schemes. In SMT the solder joint alone joins the chip carrier to
the printed circuit board (PCB). A major problem of SMT is that the solder joint
reliability becomes increasingly critical, as the joint must act both as an electrical
connector and mechanical bond. The failure of this single joint could render a device, or
an entire machine, in0perable.

The major failure of solder joint arises from the imposition of the mechanical
strain on the joints in service. It was shown that low-cycle fatigue is the primary cause of
solder joint failure in electronic assemblies [17]. This low cycle fatigue is produced by
the combination of ﬂuctuations in temperature acting on the materials of differing

coefﬁcient of thermal expansion (CTE) mismatched between substrate, component, and

solder US
in solder
ﬂuctuatio
Th
smemar
Sn-Ag bis
Creep resis
microjoints
solder joint
solder in se
the processi
dissolution .

the high he

solder used in the electronic package. This CTE mismatch that causes low-fatigue cycle
in solder joints is produced from environmental temperature changes and/or temperature
ﬂuctuations from the joule heating of the device during service [17].

Thermomechanical fatigue (TMF) due to temperature ﬂuctuations during service
is the major low-cycle fatigue mode that the solder joint can result in failure. Therefore,
Sn—Ag based lead-free solder should have mechanical properties such as fatigue strength,
creep resistance, rigidity, and so on, which are sufﬁcient to ensure the reliability of
rrricrojoints in a modern electronic device. However, the deformation behavior of the
solder joint becomes more complicated due to the complexity of microstructure of the
solder in service. The solder is used in an as-solidiﬁed microstructure that depends upon
the processing history [18]. The microstructure of the solder is also altered due to the
dissolution of substrate materials into the solder during soldering [19, 20]. Because of
the high homologous temperature in service for Sn based solders, the as-solidiﬁed
microstructure of the solder changes continuously during service [21]. As a result, the
mechanical properties of the solder joint during service are different from those of the as-
made joint. The microstructural evolution of the solder makes the deformation and
reliability analysis of the solder joint very difﬁcult. There are several factors that can
contribute to TMF during service, such as ramp rates (related to strain rate imposed),
dwell times at each temperature extreme (creep/stress relaxation), static aging (especially
dwell times at the high temperature extreme), reversed repeated shear (due to change in
temperatures with thermal cycle), microstructural features (effect of second phase IMC
precipitates), anisotropy of tin (due to high population of tin in tin-based solders), etc.

Among these factors, it is not clear which is the most dominant factor that can affect

 

fatigue of

undersrand

1.2. Objec

The
behaviors 3
based solder
“1111 Small 2
Parameters. s
and the ramp
0111 10 lm‘egt
Finally, 3 Sir.

TMF will be 1

 

fatigue of solder joints due to TMF. Thus, it is important to establish fundamental

understanding of behavior of solder joints that underwent actual TMF.

1.2. Objective

The objective of this study is to investigate the thermomechanical fatigue (TMF)
behaviors and the mechanisms that are responsible for the deformation of the Sn-Ag
based solder joints. Using the eutectic Sn-Ag solder as a model system, the Sn—Ag solder
with small amounts of Cu and, (Cu + Ni) alloy solders will be used. Several service
parameters, such as temperature extremes encountered, dwell times at these temperatures,
and the ramp-rates representing the rate at which the temperature changes, will be carried
out to investigate the role of individual service parameter in Sn-Ag based solder joints.
Finally, a simple model to predict the service capability of Sn—Ag based solders under

TMF will be investigated.

 

0"ERVIl

rt R0LE
srRt

The p
in the electrc
or drop of “
chips should
package. but
consisting of
be sent to se
This is the n'
electrical to
components.
memblies. 1
Electronic Cor

153d in electr

board (PCB;

CHAPTER 2
OVERVIEW OF THERMOMECHANICAL FATIGUE (TMF) BEHAVIOR IN
EUTECTIC TIN-SILVER BASED SOLDER JOINTS
2.1. ROLES OF SOLDER JOINT IN ELECTRONIC PACKAGING AND ITS
STRUCTURES

The purposes of electronic packaging are mainly to protect semiconductor devices
in the electronic devices. Since silicon chips are very delicate, even a tiny speck of dust
or drop of water can hinder their functions. Thus, ensruing the functionality of silicon
chips should be established. Silicon chips are not only needed to be protected by the
package, but also, they need to exchange signals with the outside. Attaching metal "legs"
consisting of lead frame (soldered balls in the case of BGA), therefore allows signals to
be sent to semiconductor devices from outside, and the results of processing accessed.
This is the major function of solder in electronic packaging. Solder joints not only are
electrical connections, but also, are mechanical connection between substrate and
components. In order to have a good understanding of their functions in electronic
assemblies, it is important to overview the different types of mounting techniques of
electronic components. Figure 2.1 illustrates the types of mounting techniques currently
used in electronic industries. The basic element may be considered as the printed circuit
board (PCB) that is mainly made by a reinforced polymer material. As can be seen in
Figure 2.1(a), the PCB contains holes and components have stiff lead-frame. Using an
auto-insertion machine, the holes in a PCB are ﬁlled with lead-frames while ﬂux is
applied. Wave soldering is performed to attach PCB and lead-frames by exposing a wave

of molten solder.

 

Througl
types have bee

lighter weight.

 

Through hole mount types have been used for many years, but surface mount
types have been developed due to demands of high packaging density, smaller height,

lighter weight, and better performance.

Through-hole Mount Type

DIP (Dual In-line Package):
the most basic package type
used for many years. Leads
extend directly down from the
longer edges of the package.

 

 

 

 

 

 

 

 

 

 

IC package
Solder ] Leadframe
Cu pad
\ /
I: PCB j
/' ﬂ

 

 

 

Schematic of typical through-hole devices (THDs)

Figure 2.1(a). Schematic of typical through hole devices (TI-IDs)
(cont'd)

 

/'r.

Fi

Surface Mount Type

SOP (Small Outline Package):
This is the main type of surface
mounting and is very widely used.

 

QFP (Quad Flat Package): This
is the evolved version of the SOP
package. This is most commonly
used in surface mount type
packaging recently.

IC package

Solder /

PCB

 

Leadframe

 

 

 

 

 

 

 

 

Schematic of a Leaded Chip Carrier Package (LDCC)

1C package

 

Solder

Schematic of a Leadless Chip Carrier package (LLCC)

 

 

 

 

 

 

 

 

 

Figure 2.1(b). Schematic of surface mount devices (SMDs).
(cont'd)

a!

 

 

Surface Mount Type (BGA)

BGA (Ball Grid Array): The
BGA package can be made with a

\./ large number of pin counts

Solder Balls 1c package

    

 

 

 

 

Schematic of a BGA package

Figure 2.1(c). Schematic of surface mount (BGA) devices [22].

Unlike the through-hole mounting, the components are directly mounted on PCB surface.
Figure 2.1(b) illustrates typical surface mount types. Soldering is usually performed
using solder paste (a mixture of solder powder, ﬂux and binder agent) that is applied via a
screen printer. Reﬂow soldering is used to melt and re-solidify for forming the solder
joint.

The rapid development of silicon device technology in recent years demands
sigriﬁcant increases in service performance. Thus, a larger numbers of interconnections
are needed to satisfy such requirements. Ball gid array (BGA) packages (illustrated in
Figure 2.1(c)) have been developed to suit this purpose. They have excellent electrical

performance and are suitable for high frequency and high pin count.

ii-

 

2.2 PR0
Sc

(Sn) and 5

tin (Sn).

2.2.-l. Ant
Tin
tetragonal 1
material b;
directions (
exhibits an
di’ﬁ‘ctions, ;
bfmieﬁ'n latg
mfhettng", 1
btm'em 21de
11131 is [Dana
Vertical (lire c1
(ﬁremen. Th
tries to Expan
aﬁjacem$535111):—
nature

Of tin,

Fat-chem 2 [1 it
s ~‘J'

2.2 PROBLEMS INVOLVED IN HIGH TIN SOLDERS

Solders, especially, lead-free solders mainly consist of more than 90 wt% of tin
(Sn) and some other elements. Thus, it is important to understand physical properties of
tin (Sn).
2.2-1. Anisotropy of B—Tin

Tin exists as the white tin (B-tin) at room temperature and it has the body-centered
tetragonal (BCT) crystal structure as is shown in Figure 2.2. Tin is a highly anisotropic
material because it possesses different physical properties along different lattice
directions (see Table 2.1). Barry et al. [23] suggested that “It will be noted that tin
exhibits anisotropy of thermal expansion properties in different crystallogaphic
directions, and this may lead to steps appearing in a free surface at the boundaries
between large gains in samples subjected to severe thermal cycling”. The term “thermal
ratcheting”, is deﬁned as the accumulation of plastic strain due to different orientations
between adjacent gains during thermal cycling. As can be seen in Figure 2.3, gain 1,
that is located in the center of gain 2, is oriented so that plastic ﬂow occurs easily in the
vertical direction while gain 2 is oriented so that plastic ﬂow is easy in the horizontal
direction. Thus, gain 1 tries to expand in vertical direction during heating and gain 2
tries to expand in horizontal direction. Such different tendency to expansion in two
adjacent gains will result in up-hill of gain 1 during thermal cycling. Due to anisotropic
nature of tin, the surface becomes textured and holes appear resulting from thermal

ratcheting [25].

 

Figure.

Grair

 

F1Elite

fs)

5
.3-

(I)

 

 

 

 

Figure 2.2. Schematic illustration of the body-centered tetragonal structure [23]

Grain boundaries

 

 

 

CTE C E

 

 

 

 

 

 

 

Grain 2 Grain 1 Grain 2 Grain 2 Grain 1

 

 

 

 

 

Grain 2

Before Thermal Cycling After Thermal Cycling

Figure 2.3. Schematic illustration of the elongation of gain 1 due to thermal ratcheting

[25].

 

Table 2.1. Physical properties of B-tin and a-tin [23,24].

 

Parameter

Value

 

Crystal structure of B-tin

BCT (A5 lattice type)

 

Lattice constants B-tin

a=b = 58.32 x10"2 nm, e = 31.82 x10'2 nm

 

Density of B-tin

7.3 gn/cm3

 

Young’s Modulus of B-tin

a axis = 85 GPa , c axis = 54 GPa

 

CTE of B-tin

a axis = 15.4 x 10'°/°C , c axis = 30.5 x 109°C

 

Resistivity of B-tin

12.6 pin-cm (direction unknown)

 

Tensile Strength of B-tin

2600 psi, ~ 18MPa

 

a to B transformation temperature

132"

 

Crystal structure of a-tin

Diamond cubic (A4 lattice)

 

Lattice constants a-tin

a = 64.9 x 10an

 

 

 

 

Density of a-tin 5.75 gn/cm3
Volume change (%), [3 to or 21 %
Resistivity of a-tin 300 pﬂocm

 

 

11

 

 

experience

complete lc
Only a few
137.38]. It:
the Sn~O.SC:
Sn-Ag based
that there is
Phl‘Sieal pro]

faiironmem ;

23.1 [$01119

to ’he malaria

dCVQiOp in [0 cr

“iii beIOH' a

2.2-2. Allotropic Phase Transformation of Tin

The allotropic transformation in tin is not a well understood phenomenon yet. As
can be noted from Table 2.1, tin can change its crystal structure ﬁ'om BCT crystal
structure to diamond cubic crystal structure below 132°C. This will result in a
substantial volume change (21 %) as the transformation occurs and causes the metal to
experience internal stresses which are sufﬁcient to cause fracturing and ultimately the
complete loss of integrity [26]. This allotropic transformation in tin is termed “tin-pest”.
Only a few studies have attempted to understand tin-pest in high tin-based solders
[27,28]. Tin-pest has never been observed in Sn-37Pb alloys, but has been reported for
the Sn-0.5Cu solder alloy aged at -18°C for 46 months [26]. Formation of tin pest in the
Sn-Ag based solders has not been established but long term aging tests at ~18 °C indicate
that there is such a possibility [27]. Tin-pest may cause sigriﬁcant deterioration of
physical properties in high tin-based solder joint. Storage at very low temperature

environment may cause serious problems in the electronic devices due to tin-pest.

2.3 FATIGUE
2.3-1. Isothermal Mechanical Fatigue on Sn-Ag Based Solders

Stresses resulting from cyclic loading can lead to microscopic physical damages
to the materials. These damages can accumulate with continued cycling until they
develop into cracks or other damage that lead to failure of the materials even at stresses
well below a given material’s ultimate strength. Cyclic loading will often ultimately
fracture the material. Fracture due to this cyclic loading is called fatigue. Thus, fatigue

resistance is the ability to endure the cyclic loading in the materials [28].

12

The 1
between c0

amplitude.

 

. reaches a...
time the str:
ﬁxed-freque
continued u1
constant anc
loading is he
and these (‘3
state. a metal

“we 2.

(I)

. 51
be Seen in t1
Constant and
fr

FIOm [h

Fme TC

83.13110“ Can t

"D

“(5°C modulus

The most common fatigue test involves completely reversed (R = -1) cycling
between constant strain limits as given in Figure 2.4 (where R is 0.3/ 0...“). A strain
amplitude, a. = Aa’Z, is selected, and an axial test specimen is loaded until the strain
reaches 5.1,. = - a, and the test is continued, with direction of loading being reversed each
time the strain reaches + a. or - 3.. Fixed strain rate between these limits, or sometimes a
ﬁxed-frequency sinusoidal variation of strain is used. Such cyclic strain tests are
continued until fatigue failure starts. 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 reversible, 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.
Figure 2.5. shows the complete stable stress-strain hysteresis loop with one cycle. As can
be seen in this ﬁgure, from point 1 to 2, the slope of the stress-strain path is at ﬁrst
constant and close to the elastic modulus (E =A0'08). Then the path gadually deviates
ﬁom linearity as plastic strain occurs, for example, path from 2 to 3.

From this ﬁgure, the total strain range can be (As) written as,

where a is plastic strain

P
From relation between Aaand a. (e. = Ae/Z), Aaand 0., and 15,. = A5}, above

. . a' . . . .
e uatron can be re-wrrtten as, a =—"—+£ , where, 3 1s a strain am lrtude, E rs the
q a E pa a P

elastic modulus a, is a stress amplitude, and a pa is a plastic strain amplitude.

13

 

n

éhdte 2'

5. SC

 

 

 

Figure 2.4. Schematic illustration of strain-controlled cycling [29].

 

 

 

 

 

 

 

 

 

Figure 2.5. Schematic illustration of stable stress-strain hysteresis loop under strain-
controlled cycling [29].

14

 

Str:
stresses. .1
tersus life
estimates ir

The

so. equation

11 is also use
35 can be s
benk’een 1 ;
bem‘een poi

lines,

- O’
SIHCE’ f

\ I

‘5 ‘ .
(”x71 11'] ab0\

Strain-based approach to fatigue is more appropriate for case of cyclic thermal
stresses. A plot of strain amplitude versus cycles to failure is called S-N curve (strain
versus life curve). This curve is used in the strain-based approach for making life
estimates in a manner analogous to the use of the S-N curve in the stress-based approach.

The strain amplitude can be divided into elastic and plastic parts.

a. = 8.. + 6p.
. . 0' . . . .
so, equation Will be a, = —E‘i + spa , as explarned 1n prevrous equations.

It is also useful to plot 6,. and a). separately as functions of number of cycles to failure Nf
as can be seen in Figure 2.6. The slope of the elastic part can be plotted from point
between 1 and 2 with each cycle and the plot for plastic can be obtained ﬁom slop
between point 2 and 3 as can be seen in Figure 2.5. Equations can then be ﬁtted to these

lines.

0' 0" . c
8.. =f=—E—’—(2N,)”, 3,. =e,(2N,>

a' .
Since, EL and sf are the intercept constants at Nf =0.5, requiring use of the quantity

(2Nf) in above equation. So strain amplitude can be rewritten as,
a" .
a, =?f(2Nf)° +£f(2Nf)°

Note: 0",, b, a}, and c are considered to be material properties. This equation the so

called Cofﬁn-Manson relationship represents the total strain amplitude (a) in Figure 2.6.
The equation or modiﬁed version of this equation has been used to evaluate life of

solders or solder joints under isothermal strain-controlled cycling.

15

 

F1

Elite 7 6

(“I1

Strain Amplitude (log scale)

Total :80: sea +880

 

 

 

Elastic =5“

 

Plastic =81”

 

 

.5

Nr

Cycles to Failure (log scale)

Figure 2.6. Schematic illustration of elastic, plastic, and total strain vs, life curve [30]

16

 

T111

“1181:? A} 1
are materia
111111 the de\

failure [31].

where JP is
asillustrated
ISOIht
SHE and S
isolhermal f;
frFluency of
“3118116 thin
and Smopb
increasing pla

The most frequently used Cofﬁn-Mason equation is given as
N ffAe = C

where Nf is number of cycles to failure, A5 is total or plastic strain range, and or and C
are materials constants. The load drop in the hysteresis loop is believed to be correlated
with the development and gowth of fatigue cracks and is used as a measure of the fatigue

failure [31]. The load drop is characterized by the load drop parameter (I),
q. = 1- (AL,
AP"!

where AP is the load range at any number of cycle and AP," is the maximum load range
as illustrated in Figure 2.7.

Isothermal fatigue tests were conducted by Solomon et. al. [31], using eutectic
Sn-Ag and Sn-40 Pb single shear lap solder joints under plastic stain control. The two
isothermal fatigue tests conducted in this study were at 35°C and 150°C with cycling
frequency of 0.3 Hz without hold time at each strain extreme. 50% load drop was used as
a fatigue failure criterion. Figure 2.8 shows the comparison of fatigue lives of Sn—3.5Ag
and Sn-40Pb solder joints. The numbers of cycles to failure were decreased with
increasing plastic strain range, following a Cofﬁn-Mason low cycle fatigue relationship.
The number of cycles to failure of Sn-3.5Ag solder joints exhibited longer than that of
Sn-40Pb solder joints at both temperatures over most plastic strain ranges.

Cofﬁn-Manson equation assumes that the product of the number of cycles to
failure and the plastic strain is a constant. This model addresses only plastic stain.

However, solder joints do not deform homogenously due to complex geometry.

17

 

Load (P)

(If
>1
(1)

FT \
Isidre ')
‘.

 

1)

——_—-— -— —— knew;-

 

 

Load (P)

[1|

1 4 10 20 30
N (Cycles)

 

 

 

 

Figure 2.7. Schematic drawing of hysteresis loop showing load drop with increasing
number of cycles [32].

 

10].... T T TTTTTT] T T TIIITTI T T TTHIII I I 111111] T I 1TH”! j—T ITI'T

111111

723
T T TTIITII ﬁr I TIITI
I
I

1 1 LIIIJli

  

I 111111

I TTTTTIT]

 

Sn-3.5Ag, 35°C
102g ----- Sn-40Pb, 35oc
--------- Sn-3.5Ag, 150°C 3
- — $11-4on, 150°C

1 3 1 1 1111111 1 11111111 1 11111111 1 11111111 L1_1111111 1 1111111

10I 10° 101 102 103 104 105
Cycles to Failure

Plastic Strain Range
‘2‘

ITTTJ

 

 

 

Figure 2.8. Example of isothermal fatigue test conducting Pb-free and Pb-bearing solder
joints at 35°C and 150°C using a 50 % load drop failure criterion [31]

18

 

This issue v
in the 39“”
out thiiI inh‘
joint geome
assumption
aWham-0“ C
model as C0:
the actual 50
predictions.
the effects or

for materials :

2.4. THER\

Therm
due to change
expansion (C'
comrmn fatig
relatively low

conditions. F

This issue was criticized by Winterbottom [3]. He reported that the deformation process
in the actual solder joint was highly inhomogeneous after thermal cycling. He pointed
out that inhomogeneous and highly localized deformation was found depending on the
joint geometry under actual thermal cycling. This indicated that the uniform strain
assumption throughout the simpliﬁed solder joints or bulk solders employed in
application of the Cofﬁn-Manson type relation was not appropriate. Thus, 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 will be needed for more accurate life
predictions. In addition to the above issues, Cofﬁn-Manson model does not account for
the effects of hold time and temperature extreme which are very important considerations

for materials stressed above 0.5 of their melting point (Tm).

2.4. THERMOMECHANICAL FATIGUE

Thermomechanical fatigue usually refers to the cyclic displacement that occurs
due to change in temperature when two materials with dissimilar coefﬁcient of thermal
expansion (CTE) are joined. High-cycle fatigue and low-cycle fatigue are the two
common fatigue modes. High-cycle fatigue occurs when the specimen is subjected to
relatively low loads and the elastic strains are much geater than the plastic strains in such
conditions. Fatigue failure typically occurs after about 104 cycles and it is a stress-
dorrrinated event. Vibrational loads in the microelectronic packages generally cause
high-cycle fatigue failures and this is a stress-based fatigue.

On the other hand, low-cycle fatigue occurs at higher loads where the plastic

component of the strain is comparable to or even geater than the elastic strain.

19

 

Temperatt
failures ir
dominated
joints is te
changes in
mismatch.
fatigue pi]:
and propag
melting p0
can range 1
mechanism
thermally ;
vthich resu]
solderJ'Oint;
2.4-l. T-‘I
In t1

PTDper testi

Temperature cycling and power cycling are the primary causes of the low-cycle fatigue
failures in microelectronic packages. Low-cycle fatigue is characterized as a strain-
dominated event rather than a stress-dominated event [33]. Low-cycle fatigue in solder
joints is termed thernomechanical fatigue (TMF) as the solder joint is experiencing both
changes in temperature as well as a change in the mechanical strain imposed by the CTE
mismatch. Fatigue in solder joints can be considered as a high temperature low cycle
fatigue phenomenon [34]. At low temperatures, fatigue causes damage by the initiation
and propagation of microcracks. But even room temperature is high compared to the
melting point of solder, and in general electronic components see thermal cycling which
can range from 0.5 Tm to 0.8 Tm. As a consequence, creep is an operative deformation
mechanism whenever the temperature is above 0.5Tm. Creep strain is a result of
thermally activated dislocation motion and/or the movement of vacancies and atoms,
which results in the elongation of the material. Thus, the thermal fatigue damage on the
solder joints is stored by a combination of creep and fatigue mechanism [35].
2.4-l. Types of Thermal Cycles

In the automotive and computer industries, there has been a challenge to desigr
proper testing methods of thermal cycling test, which would correspond to service
condition. Failures due to either thermal cycle tests or environmental service conditions
are characterized by thermomechanical fatigue behaviors. Several fatigue models have
been proposed, but none of these models take into account all of the parameters of the
test or service environment. In thermal cycling, there are many factors that can inﬂuence
the failure of solder joints due to TMF, such as the temperature range, ramp rate, dwell

time, and stepped heating and cooling [36].

20

 

T1
mmkﬁi
componer
relatively
be carried
effects of
each of th.
dwell time
more life ti
“Memes
Stock is th;
.10th can ex
2.9121» and (-

ActL
0“ serViCe c
MLIere the at
reach as h1g1
be Vfitted by

I:Ol‘c

dam

a:

ETri'mrlment

There are three means by which thermal cycles may be imposed: (1) thermal
shock, (ii) temperature cycling, and (iii) power cycling. Thermal shock involves cycling
components between two thermal baths at high and low temperature extremes. This is
relatively easier than actual therrnechanical fatigue cycling since the accelerated test can
be carried out in a short period of time. However, thermal shock considers neither the
effects of ramp rates for heating and cooling segnent nor the effects of dwell times at
each of the temperature extreme. Actual solder joints experience these ramp rates and
dwell times during thermal cycling. Kubik et al. [37] reported that the solder exhibited
more life time when it experienced thermal shock than that of the life time of a solder that
experienced thermal cycling with slower ramp rates. The other disadvantage of thermal
shock is that the result of this test might not represent the actual situation that the solder
joint can experience. Schematics of two types of thermal shock are illustrated in Figure
2.9(a) and (b).

Actual TMF cycle must consider appropriate dwell time and ramp rate depending
on service conditions. Figure 2.9(c) represents an automotive under-the-hood situation
where the automotives experience very low temperature (maximum 410°C) and also can
reach as high as 160°C due to heating from the engine after igrition. The ramp rates can
be varied by situations, such as surrounding temperatures or times of service, or both.

For computer industries, power cycling is the major factor that can cause TMF
damage since the power on and off causes thermal cycling rather than the surrounding

environment as can be seen Figure 2.9(d).

21

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

+ +
2 2
:1 a
. . /\ /\
a: is .
:1. :1. /
i: i/ \/ \/
{-4 — ~ L— e- H
"" Time — Time
(a) (b)
+ +
Q) Q)
‘5 S ._
g [‘1 / ‘53 J m
Q) i Q)
1:. :1.
e / e
E L4 1-4
— Time — Time
(C) ((0

Figure 2.9. Schematic illustrations of types of thermal cycling. (3) thermal shock with
no ramp rates consideration, (b) thermal shock with no dwell times at each temperature
extremes, (c) actual thermal cycling that solder joints can experienced during TMF,
especially automotive under-the hood situation (ramp rates and dwell times will vary with
service), and ((1) power cycling which can be effected by turning on and off of
components in ambient enviromnent [36].

22

 

2.4.-2. De
Th1
of thermal
Thermal ﬂ
strains 11 hi
softest mat
cyclic darn
example oi
difference
Sigﬂiﬁcant :
1110”: comp
deeIope-d 1:
adjacent Sn
313d Substratt
separation [3
“ring temp.
Emmi-try, 1.1
SOide-r 101mg
Table 21 311C
rite...
tempera

‘
dis
1; x -

c,

Ural.

Cle
and ‘

2.4.-2. Deformation Characteristics of Solder Joints During TMF

The source of TMF of solder joints is the sigriﬁcant difference in the coefﬁcient
of thermal expansion (CTE), or, that exists between the material components of a joint.
Thermal ﬂuctuations due to change in temperature during a service environment generate
strains which lead to crack initiation, crack propagation, and ﬁnal failure. Solder is the
softest material in the electronic components, and solder accommodates most of the
cyclic damage (strain absorbing material) dming thermal ﬂuctuations. The classic
example of TMF of an interconnection on a PCB is given in Figure 2.10. Due to
difference in the CTE between the material components, solder joints experience
sigriﬁcant straining. However, this example is a global situation. The actual case is
more complex than the global situation. During thermal ﬂuctuations, thermal strains
developed in the material components may be categorized into (i) anisotropy effects in
adjacent Sn gains in the solder area, (ii) solder material and IMC layer, (iii) IMC layer
and substrate, and (iv) a geometrical mismatch determined by the component size and the
separation between individual solder joints. Various forms of these thermal strains
during temperature changes are illustrated in Figure 2.11 [38]. In case of a simpler joint
geometry, i.e. solder joints made with Cu substrates, the possible CTE mismatches in
solder joints are given in Figure 2.12. CTE values of Sn, Cu and Cu6Sn5 are given in
Table 2.1 and 2.2.

Thermal strains caused by TMF can lead to sigriﬁcant surface damages. The
temperature-time proﬁles in service introduce dwell times at each temperature extreme in

the cycle and establish low-cycle creep-fatigue conditions. There are three distinct

23

T.:

mm
During
cooling

 

 

T.°

During Substrate

 
 
 
   

 

 

cooling
LL :1
To:
Solder <— Solder
Tmax:
During
heating

 

Figure 2.10. Schematic illustration of classic model for TMF of a soldered
interconnection on a PCB [38].

24

 

Cont

Sillgle

F1§Ure ;

0‘1

 

 

 

 

 

 

 

0‘2
0L4
0‘3
810C AT01l : 82 cc AT012 : 83 cc AT013 etc

1% . “” . ‘ ._ = i

i T” J .3 ., 1‘ a3 T‘ ‘r m:- d
Contact , - -._ e ab ~

y 0C AT (ora- 01b)

Single joint

OLb
ys oc AT (aa— 011))/h

Figure 2.11. Various forms of thermal strains due to thermal ﬂuctuations [38].

25

 

 

 

 

Fisure 2.1;

Cu SUb:
IMOCuc~..Sr

C.

 

Table 2.2. CTE values of Cu, Cu38n, and Cu6$n5 [39].

CTE value (x1

17.1

19.0103

l6.3i0.03

 

Grains in contact with IMC

_, .1. .. A01
.1 . v- ﬁﬁﬂtﬁﬁEwiv

"5W9 ' ‘ﬁﬂwﬂ ‘ d ashi as —1 '10-6/0C
iﬂgaea‘i‘mﬁﬁgﬁgﬂ

AWiNﬁﬁe-iﬁiﬁﬁ‘ﬁﬂﬁﬁ (b)~0 —2' was

(a) ~1 ' 10'6/°C

    
      
 
 

.,-' M

  
  

 

   
    
 

 

Cu

Figure 2.12. Schematic illustration of CTE mismatches involved in solder joints made on
Cu substrate. Note: values are CTE mismatches between (a) Cu substrate and
IMC(Cu.sSn5) layer, (b) IMC layer and Sn gains in contact with IMC, (c) Sn gains in
contact with Sn gains adjacent to IMC, and (d) Sn gains in bulk

regimes of the failure modes in materials under cyclic loading. Such failure modes are

fatigue dominated, fatigue-creep dominated, and creep dominated. Fatigue dominated

26

 

 

failure ari:
material. V
dominated
lutergraula
2.13031). L
cavitations
observed 11
interact wit
Endurance 3
Sold
kalIlg (l’C
thermal CFC
beCOmes We
ti’mlDel‘alure
”15003121111
the absolute
Iempi’l‘ature
Sliding. andt
As H.
dEvelOp bOtfj

crack-m2 may

failure arises due to the gowth of surface cracks, i.e. fatigue striations, through the
material, with no evidence of any creep damage as illustrated in Figure 2.13(a). Creep
dominated failure is mainly due to dwells at various locations in the thermal cycle.
Intergaular crack resulting ﬁom creep cavitations is major failure mode (see Figure
2.13(b)). During creep-fatigue interactions, surface fatigue damage developed with creep
cavitations damage within the material. The fatigue crack gowth rate is the same as that
observed in a fatigue dominated fracture initially. However, the fatigue crack may
interact with the creep damage, resulting in accelerated crack gowth, a reduction in
endurance and a fatigue crack interaction failure [40].

Solder joints exhibit two simultaneous failure modes which are intergaular
cracking (related to creep crack) and transganular cracking (fatigue crack) resulting from
thermal cycling. At high temperature above 0.5Tm, the strength of gain boundary
becomes weaker than the lattice. As can be noted in Figure 2.14, there may be a certain
temperature (equicohesive temperature) to change the deformation mode in
polycrystalline materials. The equicohesive temperature may be approximately 50% of
the absolute melting temperature of the materials. Thus, above the equicohesive
temperature (depicted as TC in Figure 2.14), the deformation is due to gain boundary
sliding, and below Tc, the deformation is due to dislocation glide on lattice plane [41].

As mentioned earlier, solders will experience creep-fatigue interaction and
develop both intergaular cracking and transganular cracking. The transganular

cracking may be due to deformation in a gain resulting from dislocation glide on lattice

27

 

 

7TH
cra

Int:
crat

lnte
crac
tran
crak

’ ’ ,3
Transganu ar I , .

 

 

 

craking e/ W ‘
*‘a
(a) Fatigue dominated
Interganular I
cracking
(b) Creep dominated
Interganular
cracking and ,—
transganular
craking

 

 

(c) Creep-fatigue interaction

2.13. Schematic drawing of possible failure modes under TMF[40]

28

 

 

Figure 2.14

(1:? '
(if
H
(D
I‘ J
H

.Uv
(f)

Grain boundary

  

“a? / """""
i

Lattice

 

 

 

 

T

C

Temperature

Figure 2.14. Graphical illustration of the effect of temperature on the strength of gain
boundary and gain lattice materials [41].

Grain Boundary

/

 

Strain Direction

/

é

r. . . .
Shear banding Fatigue Smatrons

 

 

 

 

Figure 2.15. Schematic illustration of a possible surface damage in a gain due to TMF.

29

 

 

plane. T111
[41].

Gra
when a 5011
[42]. The s
is increasec
grain bounc

grain bound

ALL Con
bear

Stud
fIOm T31}: 1
Such 35 111161
dECOhesion‘
CXpecred '1
TMF is eSSe:

be?!) 1.6“" ST

espen-EnCed ;

 

plane. This can cause shear bandings and fatigue striations as illustrated in Figure 2.15
[41].

Grain boundary sliding can cause intergaular cracking during thermal cycling
when a solder joint is being strained during temperature excursions in product application
[42]. The solder undergoes plastic deformation due to stresses introduced as temperature
is increased. Grain boundary sliding occurs with initiated cracks propagating along the
gain boundaries resulting increase in surface (crack) area and energy [41]. Keying this

gain boundary motion may be a best way to prevent crack propagation during TMF.

2.4-2. Comparison of Residual Mechanical Behavior after TMF between Lead-
bearing and Lead-free Solder

Studies have clearly documented a sigriﬁcant amount of surface damage resulting
from TMF in Sn-Ag based solder joints [43,44]. Based upon the evolution of damage
such as intense shear bending, solder joint surface upheaval, gain boundary sliding and
decohesion, degadation of residual strength of the solder joints due to TMF should be
expected. Thus, an investigation of residual-mechanical properties of solder joints in
TMF is essential to evaluate their reliability in microelectronic applications. There have
been few studies dealing with residual-strength of lead-free solder joints that have
experienced actual TMF [45,46].

Kariya et a1. [45] reported that eutectic Sn-3.5Ag and Sn—Ag-0.5Cu solders
showed better mechanical properties than eutectic Sn-37Pb after 1600 TMF cycles with
temperature extremes between —30°C and 130°C in quad ﬂat pack leads.

Poon et. al. [46] investigated residual shear strength of Sn-Ag and Sn-Pb solders

and reported that Sn-Ag solder showed better thermal shock resistance than that of Sn-Pb

30

 

 

solder “‘11
minutes 0
reports are

solders are

2.5. ALLC

Mar
replacement
alloys are 11
design 10 re
categorized
alloys are 5

£10.13. Cutect

 

solder when solder joints were cycled 1000 times between —25°C and 125°C with 15
rrrinutes of dwell at each temperature extreme. However, only a limited number of
reports are available, and low-cycle fatigue characteristics and mechanisms of lead-free

solders are not yet fully understood.

2.5. ALLOYING ADDITION IN TIN-SILVER BASED SOLDERS

Many investigators have attempted in the past decades to choose the best possible
replacement candidates for lead-bearing solder. The requirements for suitable lead-free
alloys are listed in Table 2.3 [47]. Various numbers of lead free solder alloys have been
design to replace lead-bearing solders. They are all high tin solders. They can be
categorized into binary, ternary and complex multi component systems. Binary solder
alloys are Sn-Cu, Sn-Ag, Sn-Bi, Sn-Zn, and Sn-Sb, etc. Among these binary solder
alloys, eutectic Sn-Ag solder has shown better mechanical properties than other Sn-based
lead-free solder alloys under creep rupture test in the temperature range of 25-100°C as
shown in Figure 2. 16 [48].
2.5.-1. Eutectic Sn-Ag system
Studies have shown that eutectic Sn-Ag solder exhibited thermal fatigue reliability
equivalent to, or even better than, eutectic Sn-Pb solder [45-46]. Electrical conductivity,
wettability, thermal conductivity, and CTE in eutectic Sn-Ag solder are noted to be
comparable to conventional Sn-Pb solders. The melting temperature of eutectic Sn-Ag
solder is about 40°C higher than conventional Sn-Pb solders. This can be an advantage to
use eutectic Sn-Ag solder in high temperature applications. Also, Sn-Ag system has a

limited solubility of Ag in Sn, making it more resistant to microstuctural coarsening, and

31

 

 

Relatively

l Relatively

No Potenti
> In > 511 <

Capable of
lax
. Ailoys must

Capable of

Solder voids.

i
W
Ompatrble 1

COTTOSTOIT-I’ES

\
[Compaltble \-
1

lead substrate

| 11 ,

1 1183313 [156d 111

13 -
i L“ mehlng It‘r
l

W

W

1
i

/

Ade

. U3 ‘

1 q 16 °tr911:
‘ . il0""

"5 mUSt be

7

¢

/

Table 2.3. Requirements for suitable alternative lead-free alloys [22].

 

Requirements

 

Relatively Low-cost

 

Relatively non-hazardous

 

No Potential Environmental Problems; Ranking in decreasing order of safety is: Bi > Zn

>In>Sn<Cu>Sb>Ag>Pb.

 

Capable of wetting common lead-free component and board ﬁnishes

 

Alloys must function with existing ﬂux technologies and desigrations

 

Capable of forming a reliable solder joint, free of oxide inclusions and with minimal

solder voids.

 

Compatible with relatively low temperature processing.

 

Corrosion—resistant and not prone to electrolytic corrosion potential

 

Compatible with copper substrates, immersion gold-over-nickel, and a variety of non-

lead substrates, as well as leaded substrates

 

Alloys must be available in bar, wire, performs, spheres and paste forms

 

Metals used in alloying must be available in sufﬁcient quantities

 

Low melting temperatures (< 240°C)

 

Good electrical conductivity and Good thermal conductivity.

 

Easy to repair

 

Adequate strength properties.

 

 

Alloys must be able to be easily recycled.

 

32

 

 

 

‘ ]

I

0
at 2.16. (

l 0
1

2:5 v.24 9:35.! ametm.

Fig

 

 

    
 
  

 

 

 

 

 

 

— Sn-Pb (25°C)
105 _ A '— Sn-In (25°C)
33 A ’ ‘ Sn-Bi (25°C)
5 104 SK \ A Sn-Ag (25°C)
0 \ o
5 .\ D Sn-Pb (100 C)
22 1000 — \ ‘ '3 Sn-Pb (60°C)
+2 \ \ O Sn-Bi (60°C)
52 10° ’ \ A Sn-Ag (100°C)
§ \
3: 10 — ~ 4
U1 \
1 _ \ _
0.1 1
1 10 100
Stress (MPa)

Figure 2.16. Comparison of creep rupture times of eutectic Sn—Ag, Sn-Bi, Sn-In, and Sn-
Pb solders [50].

33

 

 

m, A835“
mechanical
Ag solder.
elements to
to r13121a“? 1°;
25:1 5°";
Near
The 1163“?”I
Ass“ and C
(11°02 0.5 to 2
during ““0“
in solderlomti
by adding Sm:
As some“ Sf
pmpmies ”m
Su-Ag‘Cu‘ anC
amperature. 51
Ag. HOWeven
better creep~TU1
found by 0110 <
1011115. He rC’pOr
Solder joints un

1% ~
perature to 85

the AggSn IMC precipitates are uniformly distributed to cause improvement in
mechanical properties [50]. In order to improve the mechanical properties of eutectic Sn-
Ag solder, sigriﬁcant number of studies have been conducted by adding alloying
elements to Sn-Ag system. Currently, Sn-Ag-Cu alloys are believed to be the best choice
to replace lead-bearing solders.
2.5.-2. Sn-Ag—Cu system

Near eutectic Sn—Ag-0.5Cu alloying is the leading candidate for lead-free solders.
The near-eutectic ternary, alloys yield three phases upon solidiﬁcation which are Sn,
Ag38n and Cu68n5 [51]. The melting point of Sn-Ag-Cu alloys (Ag wt%: 3.0 to 4.0, Cu
wt%: 0.5 to 1.0) is typically observed as 217°C. Thickness of IMC layer that formed
during reﬂow can be reduced with adding Cu. Because of the brittle nature of IMC layer
in solder joints, it is believed to be an advantage to reduce the layer thickness of this layer
by adding small amounts of Cu to eutectic Sn-Ag solder [52-53]. Similar to eutectic Sn-
Ag solder, Sn-Ag—Cu solder alloys have shown equivalent to, or better mechanical
properties than Sn-Pb. Grusd [54] have investigated creep rupture resistance of Sn-Ag,
Sn-Ag-Cu, and Sn-Pb solders at both room temperature and at 100°C. Under ambient
temperature, Sn—Ag—Cu solder exhibited better creep rupture resistance than eutectic Sn-
Ag. However, at high temperature (100°C) creep-rupture test, Sn-Ag solder showed
better creep-rupture than eutectic Sn-Ag-Cu (see Figure 2. 17). Similar behavior was
found by Guo et. al [55] in eutectic Sn-Ag and Sn-4Ag-0.5Cu single shear lap solder
joints. He reported that Sn-4Ag—0.5Cu solder joints performed better than eutectic Sn—Ag
solder joints under creep at room temperature. However, upon increasing the test

temperature to 85°C, eutectic Sn—Ag and Sn-4Ag-0.5Cu solder joints switched positions.

34

~“~.L— '

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

+Sn-Pb
I. O
30 - \ \ \ \ " Sn-Ag
\ ‘ \ + Sn-Ag—Cu
g 25 — —
8 20 — a
b
m
E 15 ~ ~
'3'.
n.
< 10 — l
5 e _
0o 1 i 110 1110 1000 10‘
Time to Failure (Hours)
(a)
35 i l 1
+Sn-Pb
30 .— '_ SIP-Ag
+ Sn-Ag-Cu
g 25 — —
1k
U1 \
E 15 " .\ ‘ ‘ ~\ ‘
u—u ‘ \ \ ‘\ ‘
a VA I
< 10 — \ ‘N a
5 _ \ \ ..
2 ~ !\
‘ ‘
0 1 110 1110 1000 104
Time to Failure (Hours)

(13)
Figure 2.17. Creep rupture-resistance test of Sn-Ag, Sn-Ag-0.5Cu, and Sn-Pb. (a) room

temperature, (b) 100°C [54].
35

 

 

Suhl:
compared 11
conducted u:
uith 50 mini
0.701) exhit
*125'3C. Ho
joints for the
such results
substitutions
adding bisrm
PTOperties tin
indicates that
Pmpenjes of 1
253, Sn-Ag
The 1r.
Minced melti
ﬁle 3* Solubili
in bener meci
adds.

10
1:

Bi C311.
30:} ’
be»

“1°19? a me

1 .
soar.
Sn‘Aet

Similar:

 

Suhling et. al. [56] reported TMF reliability of Sn-Ag-Cu-X solder joints
compared to 67Sn-37Pb solder joint using leadless chip carriers. TMF tests were
conducted using two different temperature ranges (40°C to 125°C, and -40°C to 150°C)
with 50 minutes hold times at each temperature extreme. The SAC alloy (95.58n-3.8Ag—
0.7Cu) exhibited similar reliability to standard 63Sn—37Pb after TMF test from -40°C to
+125°C. However, 63Sn-37Pb joints dramatically outperformed the lead-free SAC alloy
joints for the more extreme TMF test between -40°C and +150°C. He pointed out that
such results further emphasize the need to be cautious when proposing lead-free solder
substitutions for SMT conﬁgurations in harsh environments. He also pointed out that
adding bismuth (Bi) or indium (In) into SAC alloy improved the residual mechanical
properties under TMF test with imposing temperature range from -40°C to +150°C. It
indicates that a certain alloy addition can play sigriﬁcant roles to improve mechanical
properties of SAC alloy.

2.5-3. Sn-Ag-Cu-Bi system

The most important advantage to adding bismuth (Bi) in Sn-Ag-Cu system is a
reduced melting temperature that is desired for manufacturing purposes [57]. However,
the Bi solubility is limited to about 1% in Sn-Ag-Cu alloys. Initially, adding Bi resulted
in better mechanical properties than other Sn-Ag based lead-free solders. However,
adding Bi causes drastic degadation on fatigue life after larger number of cycles under
isothermal fatigue cycling or TMF cycling. This is likely due to the fact that B1 is rather
brittle for a metal [45, 57-5 8].

2.5-4. Sn-Ag-Cu-In system

Similarly with adding Bi into the Sn-Ag-Cu (SAC) alloy, the addition of indium

36

_3A

-- \vak-c. -

 

 

an) to the 5‘
indium into i
eifects of In
exhibited gr e.
the melting 16
However. 1nd
using indiumi
2.5-5. Sn-Ag
Study 0
of Sn-l .2Ag-l.
Ag-0.5Cu allt
microstructure
together with
5173111 hardenn
qualmaty sol.

fOllnh alloy ele

2.6. ALI-‘0‘],
0F TIN-s

Hardeni
through the mat
the dislocation
ﬁrm the i

nterac

PDT
‘42:; ,
1038118

115' uh

(In) to the SAC alloy assists in a reducing melting temperature. The ranges of adding
indium into SAC alloy are from 4 wt % to 8 wt %. Hwang [57] et. al. have reported
effects of In addition on fatigue life of SAC alloy. Sn-4.lAg-0.5Cu-8.01n solder joints
exhibited geater strength and fatigue life than eutectic Sn-Pb solder joints (SMT), and
the melting temperature was about 200°C which was about 20°C lower than Sn-Ag solder.
However, Indium is an expensive material. Thus, processing cost may be a problem for
using indium containing solder alloys.
2.5-5. Sn—Ag—Cu-Ni system

Study carried out by Kariya et. al. [59] indicated that low cycle fatigue resistance
of Sn-l.2Ag-0.5Cu-0.05Ni for ﬂip chip interconnections was equivalent to that of Sn-3
Ag-0.5Cu alloy. He also pointed out that this quaternary solder alloy had ﬁne
microstructure and Ag3Sn intermetallic compound which made a network structure
together with ﬁne (Cu,Ni)6Sn5 compound. This microstructure resulted in high cyclic
strain hardening exponents, which lead to good low cycle fatigue endurance of the
quaternary solder alloy. However, only a few studies have investigated the effects of

fourth alloy elements to SAC alloy.

2.6. ALLOYING APPROACHES TO ENHANCE MECHANICAL PROPERTIES
OF TIN-SILVER BASED SOLDERS

Hardening of a material can be achieved by pinning the motion of dislocation
through the material. In the structure, an obstacle is immobile, which exerts a force on
the dislocation. Forces on dislocations arise from externally applied stresses and also
from the interaction of the stress ﬁeld which surrounds every dislocation with any in

homogeneity which produces a distortion in the crystal structure [60]. In order to achieve

37

 

 

high stren,
barriers sh
in most st
strengthene
introduced
solution or '
and system-
can prevent
Sit-Ag based
motion betvv
hard second.
adjacent 5,1,5
been 518mm“;
Sn‘Ag based
generation an

based SOldETj(

high strength in a metal, all dislocations should be eliminated or as many as powerful
barriers should be created for dislocation motion. The latter approach is the one chosen
in most strengthening mechanisms. It is well lcnown that metals and alloys can be
strengthened by the presence of ﬁne second-phase particles [61]. Such particles may be
introduced to the solder either by precipitation from supersaturated solid or liquid
solution or by external addition of foreigr dispersoids [62]. The presence of ﬁne, stable,
and system-compatible second—phase precipitates located at the Sn—Sn gain boundaries
can prevent gain boundary sliding, which is one of major deformation mechanisms in the
Sn-Ag based solder alloys under TMF [63]. As illustrated in Figure 2.18, gain boundary
motion between adjacent Sn—Sn gains can be retarded by introducing ﬁne, stable, and
hard second-phase precipitates as barriers to prevent gain boundary motion between
adjacent Sn-Sn gains [63]. As mentioned in earlier sections (section 2.5), there have
been sigriﬁcant attempts to improve the mechanical and, particularly, TMF behavior of
Sn-Ag based solder alloys by adding alloying elements. However, more studies and data
generation are needed to understand the roles of alloying elements in improving the Sn-

based solder joint reliability.

38

—‘A9,‘

 

Grain
betWCt
gr ﬂirts

 

Flam-18 .

01051an and

Pa;

-—.—-‘A .‘_ Llc‘

Grain boundary sliding
between two adjacent

gains (no obstacles) Particles increase the

resistance to sliding

 

L

.--
.9... p' ‘O 'nun.

 

 

 

 

Grain A M... Grain A s."

“s‘ ’ I o: ‘v. ’ ‘C 5'

K. W ’3 “'1 a a ‘f’

.\‘ I “a. r“ ”l W— 9..‘

{v T ‘ i f ‘C ‘ :'

Grain B Grain B
SM 1 '

Q
~~
s _.

‘5

~
I.’. .....'§ 0'
. ‘-
." ~.'--.-'.~."’.

(a) (b)

o Intermetallic Particle

t Shear Stress
a Area of Lattice Deformation

O
.hn-ul ‘.

Figure 2.18. Schematic illustration of gain boundary sliding due to shear forces applies
at high temperature. (a) gain boundary sliding between two adjacent gains with no
obstacles, and (b) gains undergo lattice deformation due to presence of an intermetallic
particle on the gain boundary during gain boundary sliding [63].

39

 

 

3.1. MATEI

The SI
(iii) Sn-2.5.~\;
Piﬂﬁianeas.
alloys “as pu
3.1 were mac}

[0 STIUUIate Ihg

31- THE PR.

Copper
10 Soldering 10
aluminum ﬁXtt
tech Copper dc
SDread of 501m
all)“ 100 lift.
Sandniched bet

The entire 3le

CHAPTER 3
EXPERIMENTAL PROCEDURE

3.1. MATERIALS USED

The solder materials used in this study are (i) eutectic Sn/Ag (ii) Sn—4Ag-0.5Cu,
(iii) Sn-2.5Ag-0.5Cu-0.5Ni, and (iv) Sn-2Ag-1Cu-1Ni in paste forms. These solder
pastes are as-received. The substrate material used to make solder joints using solder
alloys was pure copper. Cu dogbone halves which had the dimensions shown in Figure
3.1 were machined using EDM, to obtain a reduced width of 1 mm from a copper sheet,

to simulate the actual solder joint geometry in the electronic applications.

3.2. THE PREPARATION OF SOLDER JOINTS

Copper dog-bone halves were pre—cleaned with a 50 % aqueous HNO3 just prior
to soldering to remove the thin tarnish layers. Copper dog-bone halves were placed in an
aluminum ﬁxture desigred to hold the Cu dog-bones in their proper alignment. Between
each copper dog-bone, about 600 um thickness of glass plates are placed to prevent the
spread of solder beyond the 1 mm2 soldered area at the end of each half and produce
about 100 um thickness of solder joints. Proper amounts of solder paste were
sandwiched between the two Cu dog-bone halves. The assembly is shown in Figure 3.2.
The entire aluminum ﬁxture was placed on a preheated (about 500°C) hot plate and
allowed to reach a temperature of 250°C for the solder before being removed to the an
aluminum plate to cool. The Cu6Sn5 intermetallic n-phase remained in the solid state and
only the eutectic Sn-Ag was reﬂowed and solidiﬁed during soldering process. Ten solder
joints can be manufactured in the aluminum ﬁxture at a time following the

heatingcooling history shown in Figure 3.3.

40

 

36.5 mm

 

 

 

 

 

 

Figure 3.1. Schematic drawing of the dimensions of the Cu Substrate

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Cu I .
/ Giana Plates
/ Af 7
2
2L
/1
/f
r
r
/I
HeatedAluminum Fixture
(a)
/ k
f
l V— /
Solder jo'ml: N
~100 um thickness

«4 mm2 join'mg area
(b)

Figure 3.2. Schematic drawings of (a) the solder joint ﬁxture assembly and (b) a
fabricated single shear-lap joint

41

\ Rh.

 

 

Figure 3.3,

Free.

l

 

300 f F ﬁ‘l

‘5‘“ /\ 1
221°C Melting point
200 ~

Temperature (°C)
§ 5

 

g

 

 

 

o 1 1 1 1 1
0 50 100 150 200 250 300

Time (sec)

Figure 3.3. Temperature versus time proﬁle for manufacturing of Sn-Ag based solder
joints.
Dig ital
Tem pe ratu re
Controller

     

     

rill

 
 
  

   
  
 

 

 

 

my Timer
Solder
Joints
.. A -3 ,
<<®>
-— u Plate

     

Heater

 

Figure 3.4. Schematic drawing of TMF set-up.

42

 

 

One 5
from 400 to :
suspensions-

nu'crostmcmr

3.3. THER‘
Figure
controllable r“.
as a heating St
as the cooling
solder joint 5;“
control the ten
POWET to the ‘
heating rate ca
was controlled
Theme

Solder joints vv:

each chapter th

3.4. SL’RFACI

Surface I

fr.
ill-[1'
Sn~Ag ba se

iliCTOScOp

5" (SE

One side of the single shear lap joint was polished using series of grinding papers
from 400 to 2000 grit and then ﬁnal polishing was done by using 0.05 pm SiOz colloidal
suspensions. The specimen was washed with methanol prior to observing the

microstructure in scanning electron microscope.

3.3. THERMOMECHANICAL FATIGUE (TMF) SET-UP

Figure 3.4 shows TMF set-up. A thermomechanical fatigue set-up with
controllable ramp-up and dwell times was desigred. This set-up consisted of a hot-plate
as a heating source inside a common household freezer where temperature is -15°C, used
as the cooling source. A copper plate placed on the hot plate was used as a stage for the
solder joint specimens. Thermocouples were placed on the metal stage to measure and
control the temperature. The heating element was turned on and off by an external timer.
Power to the heating element was provided through a variable transformer so that the
heating rate can be adjusted and controlled. The dwell times at the temperature extremes
was controlled by the external timer, which turned the heater on or off.

Thermal cycling between —15°C and +150°C (AT=165°C) was imposed on the
solder joints with this set-up. Detailed temperature-time proﬁles of TMF will be given in

each chapter that investigates each service parameter during TMF.

3.4. SURFACE DAMAGE OBSERVATION AND MECHANICAL TESTS
Surface damage accumulation at the exact same-targeted area was observed in the
four Sn-Ag based solder joints after 0, 250, 500, and 1000 cycles by Scanning Electronic

Microscopy (SEM). Some of the solder joints were removed after 250 and 500 cycle

43

 

 

intervals to a
screw-driVEIl
room temper
microscope I

easured loa
stress values ‘
fracture toug

obtained for s

35. ECHINt

T0 1m
mese four Sn
1.5mm With 11
Due to di 111 cu
“as “58d to (
location Of Die
the cut Surface

Solder" -

intervals to assess the residual mechanical behavior. The shear test was carried out on a
screw-driven tensile testing machine using a ﬁxed crosshead speed at 0.5 mm/min at
room temperature. The failed solder joints were examined either in the SEM or optical
microscope to determine the true soldered area after excluding the pore area. The
measured load values from the screw-driven tensile testing machine were converted to
stress values by using the true solder joint area. The residual shear strength and residual-
ﬁ'acture toughness (measured by the area underneath the stress-strain curve) were

obtained for solder joint specimens that have experience selected TMF cycles.

3.5. ECHING PROCEURE

To investigate the role of microstructural features, specimens were made with
these four Sn-Ag based solder pastes by reﬂow on copper substrates (20m x 2cm x
1.5cm) with the same heating proﬁle used to fabricate the single shear-lap solder joints.
Due to difﬁculty to etch the small area of solder joint, the following alternative method
was used to observe larger etched solder area to reveal Sn-Sn gain boundaries and
location of precipitates. The solder buttons formed on Cu substrate were sectioned and
the cut surfaces of these specimens were metallurgically polished as same procedure as
solder joint preparations. These specimens were then etched with 5 vol % HNO3 and 95
vol % methanol for 1 second and washed with water. Cross sections of polished and
etched side of specimens were examined using SEM. The solder joints used for creep
and TMF studies had a joint thickness about 100 um. As a consequence, the specimens
cross-sectioned and etched were observed in regions within 50 11m from Cu/solder

interface as illustrated in Figure 3.5.

 

 

SECTION

300 pm
500 [Jul~

Figure 3

 

 

H‘A"

 

 

 

 

     
 

   
 

Solder reﬂow on Section
the Cu Substrate 813901111811

SECTION "‘ j, . M

  
     
 

 

. solder area

’*\ ,I‘

. v
. ‘-. -, }\'“" ’1 n _/
’ ' ’ . v - '- .1 r r
‘- I ./ '_ g.
. " .J‘" i e . ’
a ' . ' ' . _ \‘

v
f ‘

 

Polish and etch
side of the specime

 

 

 

300 rung:

500 Imb— To represent details in a
solder joint of 100 pm
thick

 

 

 

Figure 3.4. Schematic drawing of solder specimen preparation for etching.

45

 

 

RESIDI
FATIG
LO
ABSTRAC
Ther
expansion (
properties 0
TMF on the
Sada-and j
bemeen ~15
for the cooli:
low extreme
TMF Cycles
Alloying Ele

Tlﬂi

LCad.

 

CHAPTER 4
RESIDUAL-MECHANICAL BEHAVIOR OF THERMOMECHANICALLY
FATIGUED TIN-SILVER BASED SOLDER JOINTS UNDER TMF WITH
LONGER DWELL TIMES AT LOW TEMPERATURE EXTREME
ABSTRACT
Thermomechanical fatigue (TMF) due to the mismatch in coefﬁcients of thermal
expansion (CTE) between solder and substrate gadually degades the mechanical
properties of electronic solder joints during service. This study investigated the role of
TMF on the residual-mechanical behavior of solder joints made with eutectic Sn-Ag
solder and Sn—Ag solder with Cu and/or Ni additions. TMF tests were carried out
between —15°C and +150°C with a ramp rate of 25°C/min for heating segnent and 7°C
for the cooling segnent. The hold times were 20 rrrin at high extreme and 300 min at the
low extreme. Residual shear strength was found to drop sigriﬁcantly during the ﬁrst 250
TMF cycles, although it did remain relatively constant between 250 and 1000 cycles.
Alloying elements were found to affect the residual creep strength of solder joints after

TMF.

4.1. INTRODUCTION

Lead-tin solders have been used extensively as joining materials for the
interconnection and packaging in modern electronic components and devices over the
past several decades. However, alternatives for these solders are being sought due to
environmental and economic issues [64-67]. One of the attractive lead-free solder
candidates is tin-silver based solder because the reliability and mechanical properties are

comparable to lead-tin solders. In high temperature applications such as in automotive

46

 

 

under-the-l
due to its
industrial a
elements to
physical me
of adding a
[68-70].
Sold
mismatches
to such ther.
Senice. As
been studies
data is Heede
Studi.
fr0m TMF if
such as int En
dECOhestH ar
10 TMF 51101)]
Solder joints
appliCarjODS.
5011131» Joints 1
eutectic Sn~3_5

534:1?qu 3n~3 "P

under-the-hood, aerospace, military applications, eutectic tin—silver solder is appropriate
due to it’s higher melting point. However, this solder may not be appropriate in some
industrial applications due to cost concerns. It was found that adding third or fourth alloy
elements to tin-silver solder can depress the melting temperature and/or improve some
physical/mechanical properties. There have been several studies to investigate the effects
of adding alloying elements in enhancing the performance of eutectic tin-silver solder
[68-70].

Solder joints in service conditions experience internal thermal stresses due to CTE
mismatches between the component leads, solder, and the substrate. TMF will occur due
to such thermal stresses that develop during temperature excursions encountered during
service. As a result, mechanical deformation will occur in the solder joints. There have
been studies of TMF on lead-tin solder during last decade [71-74]. However, additional
data is needed on the TMF on lead—ﬂee solders.

Studies have clearly documented a sigriﬁcant amount of surface damage resulting
from TMF in Sn-Ag based solder joints [44,65]. Based upon the evolution of damages
such as intense shear bending, solder joint surface upheaval, gain boundary sliding and
decohesion and surface cracking, degadation of residual strength of the solder joints due
to TMF should be expected. Thus, an investigation of residual-mechanical properties of
solder joints in TMF is essential to evaluate their reliability in microelectronic
applications. There have been few studies dealing with residual-strength of lead-ﬁee
solder joints that have experienced TMF [4,45,46]. Kariya et. al. [45] reported that
eutectic Sn-3.5Ag and Sn-Ag-0.5Cu solders showed better mechanical properties than

eutectic Sn-37Pb after 1600 TMF cycles with temperature extremes between —30°C and

47

 

 

130°C in q
Ag and St
resistance 1
with 15 mir
understandi
lead-free so
In th
between ~15
shear and cr

mEChanical b

4.2. EXPE R
Solder
0.5Cu~0.5Nj,
These

C 't T .

pTOCEdUre USC‘d

“ -
3° metallur‘iic

150C and

*150

COOIIHQ S

egmen-

Elite-m
» j e m
e d,
Sm...-
.‘thmm
3 Ex .
- 13c

130°C in quad ﬂat pack leads. Poon et. al. [46] investigated residual shear strength of Sn-
Ag and Sn—Pb solders and reported that Sn-Ag solder showed better thermal shock
resistance than that of Sn-Pb solder when cycled 1000 times between —25°C and 125°C
with 15 minutes of dwell at each temperature extreme. However, there is no guidance for
understanding the role of alloying element addition to improve the TMF resistance of the
lead-free solder joints in the published literature [4446,75].

In this study, four different tin-silver based solder alloys were subjected to TMF
between —15°C and 150°C. The residual-mechanical behaviors were examined by simple
shear and creep tests to evaluate the effects of alloying element addition on the residual

mechanical behavior of TMF’ed tin-silver based solder joints.

4.2. EXPERIMENTAL PROCEDURE

Solder joints made with (i) eutectic Sn-3.5Ag, (ii) Sn-4Ag-0.5Cu, (iii) Sn-2.5Ag-
0.5Cu-0.5Ni, and (iv) Sn-ZAg-lCu-lNi solder pastes were used in this study.

These studies were designed to represent service environments experienced by the
solder joints used in automotive/microelectronic industrial applications. The detailed
conﬁgurations of the dog-bone shaped single shear-lap solder joint and the soldering
procedure used have been reported in other publications [76,77]. The side of solder joints
was metallurgically polished prior to TMF test. The TMF test was carried out between —
15°C and +150°C with a ramp rate of 25°C/min for heating segnent and 7°C/min for the
cooling segnent. The hold times were 20 min at high extreme and 300 min at the low
extreme. The details of the TMF set up used in this study can be found elsewhere [44].

Specimens experienced 250, 500, and 1000 TMF cycles and the microstructure and

48

 

 

mechanical
this study \\
Klee

room tempe
(1) Simple s
The :

using a ﬁXCC
shear strengt

stress-strain

selected T.\l F
(31 Creep lest
Creep

1°3de On a 1
Figure 4-1- A1
5110th in Flgu
elevated 13mm
and the temper
101111 as Shown
duration of the 1
After Si;
011

“Tied USin

7

ﬂ

E

(I

 

mechanical properties were evaluated. The maximum number TMF cycles employed in
this study was 1000.

Mechanical evaluation consisted of (1) simple shear test, and (2) creep tests at
room temperature and at 85°C.
(1) Simple shear test

The simple shear test was carried out on a screw-driven tensile testing machine
using a ﬁxed crosshead speed at 0.5 mm/min at room temperature. The residual—simple
shear strength and residual-fracture toughness (measured by the area underneath the
stress-strain curve) were obtained for solder joint specimens that have experience
selected TMF cycles.
(2) Creep test

Creep tests were conducted at room temperature and at 85°C using dead weight
loading on a miniature creep testing frame ﬁxed on an optical microscope, as shown in
Figure 4.1. An excimer laser ablation mapping technique was used to determine strain as
shown in Figure 4.2. The detail of this technique can be found elsewhere [78,79]. For
elevated temperature tests, a heating pad was afﬁxed to the Cu strip as a heating source
and the temperature was monitored using a thermocouple kept in contact with the solder
joint as shown in Figure 4.3. The temperature was maintained to within i 2°C for the
duration of the creep test.

After simple shear and creep tests, the fracture surfaces of each joint were

observed using a scanning electron microscope.

49

 

 

Figure

 

Flm

adre 42

 

 

 

  
    

Cu Substrates

¢/\

Solder

 

 

   

 

 

 

 

 

 

         
     

Digital
Camera

Optical

Microscope

 

 

  

Top View

Side View
Figure 4.1. Schematic of setup of dead weight loading miniature creep ﬂame.

 

Strain

 

. .9.
159 ' 1‘
“BAD?”

Time (sec)

Figure 4.2. Illustration of analyzing creep strain using excimer laser-ablation mapping
technique.

50

 

 

Com
vari
heatin

Cu

/Substrat
Heating ,

 

 

 

 

 

Pa d\ </ /der
\,/9=-——_;_
:A/
Connect to
vaflac for Thermocouple
heating source to measure
temperature

Figure 4.3. Schematic drawing of high temperature creep setup.

51

 

 

4.3. RES!
43-]. [Ht
lnte
decohesion
strength of :
evaluate the
absolute and
are presenter
strength as 2
However. no
350 and 10W
simple Shear
€16merits couL
drop in the re
Wes [4540}
SMg~0.5Cu 5
after abom 100
4.1. A Similar!
The are;
““15 Studied.

condithn. For

 

 

4.3. RESULTS AND DISCUSSION
4.3-1. Effect of TMF Cycles on Simple Shear Behavior of Solder Joints

Intense shear banding, solder joint surface upheaval, grain boundary sliding and
decohesion and surface cracking, of solder joints resulting from TMF can reduce the
strength of solder joints. The residual simple shear strength can be used to quantitatively
evaluate the effect of TMF damage on solder joints. The effect of thermal cycles on the
absolute and normalized residual simple shear strength of four Sn-Ag based solder joints
are presented in Figures 4.4 and 4.5. There is an initial drop in residual-simple shear
strength as a consequence of 250 TMF cycles in all four Sn-Ag based solder joints.
However, no signiﬁcant decrease of the residual-simple shear strength was noted between
250 and 1000 TMF cycles. Although Sn-4Ag-O.5Cu solder joint exhibited the highest
simple shear strength in the as-joined condition, no apparent inﬂuence of alloying
elements could be realized in the residual strength after 250 and 1000 TMF cycles. The
drop in the residual-simple shear strength was approximately 30~60 % after 250 TMF
cycles [45-46]. According to these studies the residual strength of eutectic Sn-Ag and
Sn—Ag—O.5Cu solder joints were not signiﬁcantly affected by thermally induced stress
after about 1000 TMF cycles [45,46]. Findings of these studies are summarized in Table
4.1. A similar trend was reported in other studies that used different joint conﬁgurations.

The area under the stress-strain curve before fracture was measured for all solder
joints studied, and normalized by corresponding value for the joint in the as-joined

condition. For example, the fracture toughness of as-joined solder joint was set as 1 and

52

 

 

Table 4.1

l Investigator
l

 

r , _
kanya eta}

l Poon et. al.

Ahat et. al.

'Incitlnul k‘:...--l,. L‘l- _ _

Hﬁm44.Ef

 

Table 4.1. Comparison of residual strength of eutectic Sn-Ag solder joints in various
specimen conﬁgurations [45-46,75]

 

 

 

 

 

 

 

 

 

 

 

 

 

S imen Ran e Max No of Initial about Compared
Investigators can? tion ATg c .cles. after 200 between 200 and
gura y cycles Max cycles
. Quad Flat 0 ~ 0 ~ 0
Kariya et. a1. Pack Lead 160 C 1200 cycles 97 A; 78 /o
Poon et. al. PCB 150°C 1000 cycles ~87 % ~67 %
Ahat et. al. PCB 190°C 2000 cycles ~73 % ~ 41 %
50 [l r 1 1
q + Sn-3.5Ag
\ El— Sn-4Ag-O.5Cu
‘\ '9- Sn-2.5Ag-0.5Cu-O.5Ni
4° \ "A" Sn—2Ag-1Cu-1Ni '
\

N
O

Residual-Simple Shear Strength (MPa)
3

 

 

 

 

 

l

 

 

0

250

500

750

Number of Cycles

1000

Figure 4.4. Effect of thermal cycles on the residual-simple shear strength of four Sn-Ag
based solder joints

53

 

 

 

Figure 4.5. ,\

ﬂ

 

P
on

 

P
a

 

.9
A

 

—.— Sn-3.5Ag
El- Sn-4Ag-O.5Cu ‘
O- Sn-2.5Ag-O.5Cu-O.5Ni

' 15‘ 'Sn-2Ag-1Cu-1Ni

l I l

0 250 500 750 1000
Number of Cycles

Figure 4.5. Normalized residual-simple shear strength as function of thermal cycles for
four Sn-Ag based solder joints.

9
N
r

 

 

 

Normalized Residual-Simple Shear Strength

 

 

G

 

ﬂ

 

P
co

P
at

 

p
a
.,..
l

 

+ Sn-3.5Ag
E}- Sn-4Ag-O.5Cu ‘
9- Sn-2.5Ag-O.5Cu-0.5Ni ‘

‘ ‘A‘ ' Sn-2Ag-1Cu-1Ni

N ormagized Residual Fracture Toughness
i9

 

 

 

 

 

i l I I l l l 4 L A l 1 l 1 J

0 250 500 750 1000
Number of Cycles

Figure 4.6. Normalized residual fracture toughness as ﬁmction of thermal cycles for four
Sn-Ag based solder joints.

Y Y T

 

54

 

 

fracture tou
fracture tou;
in Figure 4.(
found to be

toughness is
signiﬁcant d.
a finding ind

ofTMF C}'Clr

43-2. Effect

fracture toughness of solder joints after TMF was indexed. Such normalized residual
fracture toughness was plotted as a function of the number of TMF cycles and presented

in Figure 4.6. The trend of a decrease in the residual fracture toughness due to TMF was
found to be similar to that for residual-simple shear strength. The initial drop in fracture
toughness is found to occur between 0 and 250 cycles. Similarly, there is no additional
signiﬁcant drop in residual fracture toughness between 250 and 1000 TMF cycles. Such
a ﬁnding indicates that the ductility of the solder joints deteriorates during the early parts

of TMF cycles.

4.3-2. Effect of TMF cycles on the Creep Resistance of Solder Joints

Alloy additions have been found to alter of the creep behavior of as-fabricated Sn-
Ag based solder joints [55]. Secondary creep strain rates of the chosen three solder alloy
joints subjected to TMF cycles are presented in Figure 4.7(a)-(c). In these ﬁgures, closed
symbols represent specimens crept at room temperature while open symbols represent
specimens crept at 85°C. Dotted lines were used to identify trends. For comparison the
corresponding values for the as-fabricated joints adapted from our previous studies are
also given in Figure 4.7(a)—(c) [55].

As can be observed from Figure 4.7(a), TMF of eutectic Sn-3.5Ag solder joint
deteriorates its room temperature creep resistance by one order of magnitude. Similar
behavior was also noted during creep at 85°C. The residual-creep resistance of solder
joints after 1000 cycles was not signiﬁcantly different from that of 250 cycles. There was

no signiﬁcant difference in the creep resistance of eutectic Sn-Ag solder joints as a

55

 

3:6 :25

 

  

 

 

LE Eur M. lmllv a:
0.3 .mﬁaam 3&0 82 a. .0. e. .

.2 aﬁaam 3&0 82 4 .. w
0.3 .3095 3&0 8m 0 ... . a
. .. .9 .. S m...

em wﬁaam 326 com o . .. .. s. u
0.8 $2.5 393 can D . ... o m
.E .m<m.m.=m 39$ o8 I OI 3.4.. m.
0.3 .w<m.m.=m 3&8 o .. 4 )

aw . I h e l “-6.“ s

.3 «25303 0. . mu
.- 00 .r

.. . {x

..t

 

 

 

 

 

(a) Sn-3.5Ag (cont’d)

56

 

cm

0.3 585.378 398 82

2 32.9378 3&0 82
0.3 .aon.o.w<ram 396 con

2 .53..»«ram 3&0 8w
0.3 .53.??8 3&0 own

2 .aon.o.w<r=m 398 am
can” sonoastem 396 o

.2 .aom.o.m<rcm 398 c

 

 

00-80044

 

 

 

322V 39%

gems wm

v

Q
s
Q
s
s
4:44

 

 

"’2 “’3 "'2 ”a
(1.098)91921 urn-us

0.5Cu (cont’d)

-4 A g-

(b) Sn

57

 

 

O 0 cycled Sn-ZAg-lCu- lNi, RT

D 250 cycled Sn-2Ag-1Cu-1Ni, 85°C
0 500 cycled Sn-2Ag-1Cu- lNi, RT

0 500 cycled Sn-2Ag-lCu-1Ni, 85°C
A 1000 cycled Sn—2Ag-lCu-1Ni, RT
A 1000 cycled Sn-2Ag-1Cu-1Ni, 85°C

.0 O o cycled Sn-ZAg- lCu-lNi, 85°C
I 250 cycled Sn-2Ag-lCu-1Ni, RT

 

 

 

 

 

"" "‘ '3: “i: ”a: '3:
g g ... .. .. ...
( -098) area "tears

I

(c) Sn-2Ag— 1 Cu— 1 Ni

Figure 4.7. Comparison of steady-state creep rates of the four solder alloys subjected to 0,

250, 500, and 1000 TMF cycles as a function stress. [Note: In Figure 7 (b), the lower
creep resistance of 500 TMF cycled joints (as compared to those with 250 TMF cycled
ones) is attributed to the presence of large porosity in them]

58

consequence of TMF between 250, 500, and 1000 cycles. This trend agrees well with
simple shear test results.

In contrast to eutectic Sn-3.5Ag, ternary Sn-4Ag-O.5Cu and quaternary Sn-2Ag-
lCu-lNi showed only a slight decrease in creep resistance as a consequence of 1000
cycles of TMF. As can be observed from Figure 4.7(b) and (c), the residual creep
resistance of these two alloys at room temperature and at 85°C after TMF is almost the
same as that of specimens that were not subjected to TMF cycling.

Figure 4.8(a) and (b) show normalized strain-rate versus normalized stress for
solder joints crept at room temperature and at 85°C, respectively. Data for aged eutectic
Sn-3.5Ag solder joints adopted from Darveaux’s studies are presented in these plots for
comparison [80]. Closed symbols in Figure 4.8(a) and (b) correspond to solder joints that
were not TMF’ed. As-fabricated solder joints used in the current study showed slightly
better creep resistance than the aged solder joints used in studies of Darveaux and Banerji
[80]. Although the creep behavior of solder joints after TMF did not apparently disobey
the power law relation, it generally exhibited slightly less creep resistance than

Darveaux’s data. Similar behavior was also found for solder joints crept at 85°C.

4.3-3. Fracture of Solder Joints

The representative fracture surface of a crept solder joint is shown in Figure 4.9(a). Most
specimens after fracture showed ductile smeared fracture surface due to failure by shear
through the solder matrix. The highly ductile nature of the fracture can be noted from the

spherical dimples present in Sn phase as shown in Figure 4.9(b). Thus, the

59

S

P
H

N ormallzed Strain Rate (8 /Dv)

 

1000 ;

y—a
5
ler1

0.001

 

ff‘rﬁ'rViﬁ']

Q=38.5 KJ/mol

 

 

 

n
G=2.8x10“-10000T(psi) 9* O
0

CC?

., o

r :1

, I

r:

O b. ' <>

‘5‘ 9

El

0 E

a

b

E]

V

 

Darveaux et al
OSA, RT

2508A, RT
SOOSA, RT
IOOOSA, RT
OSAC, RT
ZSOSAC, RT
SOOSAC, RT
IOOOSAC, RT
OSACN 1, RT
ZSOSACNl, RT
SOOSACNI, RT
IOOOSACN 1, RT
ZSOSACNO.5, RT
SOOSACNO.5, RT
IOOOSACNO.5, RT

 

A I l

 

 

0.001
Normalized Stress (cs/G)

(a) Room Temperature (cont’d)

60

 

1000 g
A E
a“ 10 e
a) E
H
a
a: E
a 0.1 g—
o_ E
a r
h
H r
w E
g 0.001 E
.53 E
_
a E
E E
r... '5 ’
2 10
7 E
10'
0.0001

 

 

 

 

QEVEEBZODODIBDEEOO

Darveaux et al
OSA, 85°C

2508A, 85°C
SOOSA, 85°C
IOOOSA, 85°C
OSAC, 85°C
ZSOSAC, 85°C
SOOSAC, 85°C
IOOOSAC, 85°C
OSACN 1, 85°C
250SACN1, 85°C
SOOSACNl, 85°C
lOOSACN 1, 85°C
ZSOSACNO.5, 85°C
SOOSACNO.S, 85°C
1000SACNO.5, 85°C

 

 

LLJ

 

0.01

Figure 4.8. Normalized steady-state creep-strain rate versus normalized stress for

TMF’ed and non-TMF’ed solder joints used in this study along with Darveaux’s data for

aged eutectic Sn-3.5Ag solder joints [79]. Abbreviations were used to identify each

solder joint (i.e., SA is Sn-3.5Ag, SAC is Sn-4Ag-0.5Cu, SACN 1 is Sn-2Ag-1Cu—1Ni,
and SACNO.5 is Sn-2.5Ag-O.5Cu-O.5Ni). Number appearing before the abbreviation
indicate the number of thermal cycles.

 

250 cycled
Sn-4Ag—0.5 Cu

V‘ 500 eyelet!
lea-2.5Ag-0sc' 0.5Ni

 

(b)

Figure 4.9. Representative fracture surface of the crept solder joints. (a) ductile fracture
features smeared due to applied pure shear, (b) ductile dimple fracture.

62

creep fracture mode was found to be ductile for most solder joints indicating that the

creep ductility of the solder joints was not signiﬁcantly affected by the thermal cycles.

4.3-4. Possible Model of Observed Residual Behavior

The observed residual mechanical behavior could be attributed to an increase in
crack density during early stages of TMF cycling. The residual strength of the solder
joints after TMF appears to be controlled by increasing density of cracks that occurs
during ﬁrst few hundred TMF cycles. However, these cracks might be not connected
together, and cause catastrophic failure. The ultimate TMF failure appears to occur when
these crack interconnect. This behavior is very similar to that of thermal spalling in
ceramics, where crack propagation is the controlling factor more than crack nucleation
[81]. The preliminary proposed model is illustrated in Figure 4.10. Detail TMF damage
accumulation modeling using this concept is currently in progress and will be presented

in future publications.

4.4. CONCLUSION
TMF cycling four Sn-Ag based solder joints between —15°C and 150°C inﬂuences
their residual mechanical properties in the following manner
1. Residual simple shear strength of Sn-Ag based solder joints decrease by 30-60%
as a consequence of 250 TMF cycles. Between 250 cycles and 1000 cycles, no

signiﬁcant drop in strength was noted.

63

‘ml‘

 

(C)

Figure 4.10. Schematic of possible crack behavior during TMF cycles. (3) Initially have
minimum # of cracks at Sn grain boundaries, (b) # of cracks increase in the initial stage
of TMF, and (c) TMF accident can connect these cracks to cause failure.

2. Among the solder alloy joints tested, only the eutectic Sn-3.5Ag solder joint

exhibited a degraded creep resistance due to TMF. TMF degrades the room

temperature and 85°C creep resistance of the solder joints made with this alloy by
one order of magnitude
3. TMF did not signiﬁcantly decrease the steady-state creep rates of Sn-4.0Ag-

0.5Cu solder joint specimens, as compared to that found in the as-joined

condition, at both room temperature and at 850C.

4. Even after 1000 thermal cycles, the creep resistance of the solder joints remained
as good as that found before TMF. In Figure 4.7(b), Sn-4.0Ag-O.5Cu solder
joints after 500 thermal cycles exhibited less creep resistance compared to 250-
cycle and 1000-cycle specimens due to the presence of large-sized porosity in the
joints during fabrication.

5. Although there is some increase in steady-state creep rates observed after TMF
cycles in the solder joints made with quaternary alloy (Sn-2Ag-1Cu-1Ni), the

TMF induced degradation of creep resistance was minimal at both room

temperature and 850C.

65

CHAPTER 5
EFFECT OF DWELL TIMES ON THERMOMECHANICAL FATIGUE
BEHAVIOR OF TIN-SILVER BASED SOLDER JOINTS

ABSTRACT

Thermomechanical fatigue (TMF) due to the mismatch in coefﬁcient of thermal
expansion (CTE) between solder and substrate gradually degrades the mechanical
properties of solder joints during service. Solder joints fabricated with eutectic Sn—Ag,
and Sn-Ag solder with Cu and/or Ni, were subjected to TMF between —15°C and +150°C
with dwell times of 115 min. at high temperature extreme and 20 min. at low temperature
extreme. Characterization of surface damage and residual mechanical strength of these
solder joints were carried out after 0, 250, 500, and 1000 TMF cycles. Results obtained
from this study were compared with those obtained with longer dwell time at lower
temperature extreme. The solder joints that experienced longer dwell times at high
temperature extreme exhibited less surface damage accumulation and less decrease in
simple shear strength as compared to those experienced longer dwell times at low
temperature extreme. Quaternary alloys containing small amounts of Cu and Ni exhibit
better TMF performance than binary and ternary alloys under TMF cycling with longer

dwell times at high temperature extreme.

66

5.1. INTRODUCTION

Eutectic Sn—Ag is an attractive candidate for replacing Pb-bearing electronic
solders to meet the requirements for demanding high temperature service enviromnents
such as automotive under-the-hood applications. Since physical and thermal properties
can be improved with additional elements to eutectic Sn-Ag, new Pb-free solders are
likely to have multi-components [12-16]. Sn-Ag based solder alloys with improve TMF
performance may be a ternary or quaternary alloy with additions of small amounts of
alloying elements such as copper and/or nickel [44,82].

Advances in packaging technologies, driven by the desire for miniaturization and
increased circuit speed result in severe operating conditions that raise reliability concerns
of solder joints. Also, the mismatched thermal expansion characteristics of the materials
joined by the solder, and the temperature excursions normally encountered during
service, constitute a TMF condition for the constrained solder present in the joint. The
reliability of solder interconnections under TMF becomes more critical as new electronic
packaging technologies evolve [12,13].

The design of solder joints in electronic packages requires a good understanding
of the mechanisms that control the deformation under service environment. However, the
dominant mechanism, which contributes to TMF, has not yet been fully identiﬁed.
Studies have reported that the deformation mechanisms in TMF are creep/stress
relaxation, aging, reversed shear, and anisotropy of tin [18,78,83-91]. Recent studies
have brought out the importance of reversed shear and anisotropy of tin on damage

accumulation during TMF [89-91].

67

Evaluations of TMF damage accumulation and resultant residual mechanical
properties of Sn-Ag based solder joints that have experienced longer dwell time at low
temperature extreme during TMF cycling have been documented in our prior publications
[44,82]. However, surface damage evaluation and residual mechanical behavior of Sn-
Ag based solder joints that experience longer dwell times at high temperature extreme
during TMF have not yet been documented. The long dwell time at low temperature
extreme during TMF represents conditions such as, automotive under-the-hood
applications, with service role of commuting to work, or shopping trips, etc. The solder
joints operating under such a mode are used for shorter period of time in service with
longer period of time in non-service mode. The current study is aimed at the evaluation
of the TMF damage accumulation and residual mechanical behavior in lead-free Sn-Ag
single shear-lap solder joints with realistic thickness under temperature excursions with
longer dwell time at high temperature extreme. The situations for longer dwell time at
high temperature extremes during TMF can represent long hauling trucks, taxis, or
airplanes. This study is aimed at comparing the relative roles of longer dwell times at
low or high temperature extremes during TMF on the damage accumulation and resultant

mechanical properties of Sn-Ag based solder joints.

5.2. EXPERIMENTAL PROCEDURE
The current study dealt with solder joints made with (i) Eutectic Sn-3.5Ag, (ii) Sn—
4Ag-O.5Cu, (iii) Sn-2.5Ag-O.5Cu-O.5Ni, and (iv) Sn-2Ag-1Cu-1Ni solder pastes to

facilitate a comparison with the results obtained in earlier studies [44,82].

68

The detailed conﬁgurations of the dog-bone shaped single shear-lap solder joint
and soldering procedure used have been reported elsewhere [79,89]. Cross sections of
solder joints were metallurgically polished prior to subjecting them to TMF test.

A thermomechanical fatigue set-up with controllable ramp and dwell times was
designed. This set-up consisted of an electric hot-plate as a heating source inside a
common household freezer used as the cooling source. A copper plate placed on the hot
plate was used as a stage for the solder joint specimens. Thermocouples were placed on
the metal stage to measure and control the temperature. The heating element was turned
on and off by an external timer. Power to the heating element was provided through a
variable transformer so that the heating rate can be adjusted and controlled. The dwell
times at the temperature extremes was controlled by an external timer, which turned the
heater on or off.

Thermal cycling between —15°C and +150°C was imposed on the solder joints
with this set-up. The ramp-rate for the heating segment of the cycle was 25°C/min and
the dwell time at the high temperature end was about 115 minutes. The ramp-rate for the
cooling segment of the cycle was 7°C/min and the dwell at the low temperature extreme
of the cycle was about 20 minutes. This TMF cycling has the same ramp rates and
temperature extremes used in our earlier studies to facilitate the comparison of the role of
longer dwell times at low and high temperature extremes. The temperature proﬁles
obtained with this set-up for prior and current studies are given in Figure 5.1. In the
current study about eight TMF cycles could be imposed per day, and the solder joints

were subjected to for a maximum of 1000 cycles. Surface damage accumulation at the

69

 

 

Temperature (°C)

  

 

A; l J l 1 I

0 100 200 300 400 500
Time (min)
(a) long dwell time (330 min) at low temperature (-15°C) extreme
200 ""FT"'rrrrr1rrr11Twwr

 

----~‘
' r- Q

id
0|
6
l
i
i
1
r
')
i
i
I
t
D
1

pa
5
a

Temperature (°C)
3

d. A .-
‘o c .,
’
A
'
1 J LL41_1 I 1 1 1 1

 

-50

 

01L 1100‘ 1100‘ 1300' ‘hooi‘L300
Time (min)
(b) long dwell time (115 min) at low temperature (150°C) extreme

Figure 5.1. Temperature-time proﬁles of (a) long dwell time at low temperature extreme
(prior study) and (b) long dwell time at high temperature extreme (current study)

70

61'th

\th

ll

be"

hr
in

re
1.

lflIl

in

exact same-targeted area was observed in all the four Sn-Ag based solder joints after 0,
250, 500, and 1000 cycles by Scanning Electronic Microscopy. Some of the solder joints
were removed after 250 and 500 cycle intervals to assess the residual mechanical
behavior with tensile testing machine at a crosshead speed of 0.5 mm/min at room

temperature.

5.3. RESULTS AND DISCUSSION
5.3-1. Surface Damage Accumulation in Sn-Ag Based Solder Joints that

Experienced Longer Dwell Time at High Temperature (+150°C) extreme

During TMF

Binary eutectic Sn-3.5 Ag solder joints showed progress of surface damage
accumulation with increasing number of TMF cycles as shown in Figure 5.2. The
initially polished surface shown in Figure 5.2(a) experienced shear banding near the
intermetallic interface due to TMF as shown in Figure. 5.2(b). The intensity of this shear
banding became stronger with increasing number of TMF cycles (Figure. 5.2(b). thru
(c)). The surface damage accumulation in eutectic Sn-3.5Ag solder joints was found to
be the worst after 1000 TMF cycles in current study.

Microstructures of the ternary alloy, Sn-4Ag-O.5Cu, after 0, 250, 500, and 1000
TMF cycles are illustrated in Figure 5.3. Visible grain boundary sliding occurred at a
region away from intermetallic interface during ﬁrst few hundred TMF cycles and the
damage accumulation intensiﬁed with increasing number of TMF cycles as can be seen

in Figure. 5.3(a). thru ((1). There was an apparent crack growth within the solder between

Sn-Sn grains due to grain boundary sliding between 500 and 1000 TMF cycles.

71

 

(d) 1000 TMF cycles

Figure 5.2. SEM images of surface damage accumulation of eutectic Sn-3.5Ag with
long dwell time at high temperature showing increasing intensity of shear banding
near intermetallic interface with increasing number of TMF cycles.

72

 

Figure 5.3. SEM images of surface damage accumulation of Sn—4Ag—O.5 Cu with
long dwell time at high temperature showing clear grain boundary sliding in the
solder area with increasing number of TMF cycles (arrows indicate cracks).

73

A comparison of Figure 5.2 and 5.3 suggests that the ternary alloy (Sn-4Ag-O.5Cu)
exhibits less surface damage than the binary eutectic Sn-Ag due to TMF cycle imposed in
the current study.

Both quaternary alloys used in this study showed almost no visible surface
damage accumulation even after 1000 TMF cycles. Figure. 5.4 shows an example of the
microstructures of Sn—2.5Ag-O.5Cu-O.5Ni solder joints after TMF. No signiﬁcant
damage was found between 0 and 1000 TMF cycles. Similar behavior can also be noted
for Sn-ZAg-lCu-lNi as illustrated in Figure. 5.5.

The most predominate surface damages after 1000 TMF cycles in eutectic Sn-
3.5Ag and Sn—4Ag—O.5Cu were grain boundary sliding as shown in Figure 5.6(a) and (b).
The quaternary alloys showed no such damage since there was only minimal or no
surface damage present in these solder joints even after 1000 TMF cycles.

5.3-2. Comparison of Damage Accumulation in Specimens with Long Dwell Times
at Low and High Temperature Extremes

Micrographs comparing surface damage accumulation after 1000 TMF cycles in
the four solder alloys studied with different dwell times at low or high temperature
extremes are presented in Figure 5.7 through 5.10. The amount of visible surface damage
in eutectic Sn3.5Ag joints aﬁer 1000 TMF cycles was similar for specimens experiencing
long dwell time at low [44] and high (present study) temperature extremes as illustrated

in Figure 5.7(a). and (b).

74

 

Figure 5.4. SEM images of thermomechanically fatigued Sn-2.5Ag-0.5 Cu-0.5 Ni with
long dwell time at high temperature showing no signiﬁcant surface damage accumulation
even after 1000 TMF cycles.

75

 

w ». ”re-...wv—o—audos d7“
. - .. \ .
' ‘ r i .' I 1‘ ‘

 

(d) 1000 TMF cycles
Figure 5.5. SEM images of thermomechanically fatigued Sn-2Ag-1 Cu-l Ni with long

dwell time at high temperature showing no signiﬁcant surface damage accumulation even
aﬁer 1000 TMF cycles.

76

 

Figure 5.6. SEM micrographs showing grain boundary sliding after 1000 TMF
cycles. (a) Sn-Sn grain boundary sliding of eutectic Sn—3.5Ag and (b) Sn-Sn grain
boundary cracks form near intermetallic interface for Sn-4Ag-0.5Cu.

77

Although a prior study [44] with longer dwell time at low temperature extreme
indicated that Sn-4Ag-O.5Cu solder joints exhibited more visual deformation as
compared to eutectic Sn-3.5Ag solder joints, the present study with longer dwell time at
high temperature extreme revealed that the Sn-4Ag-O.5Cu exhibits less damage as
compared to the eutectic Sn-3.5Ag (Figure 5.8.). As from Figures 5.7 and 5.8, the
surface damage due to TMF in the ternary alloy occurs within the solder matrix while
that in the binary is mainly conﬁned to regions near the solder/substrate interface. Such a
behavior can probably be attributed to the presence of Cu-Sn intermetallic precipitates in
the ternary alloy along the grain boundaries at regions near the solder/substrate region.
These precipitates may retard grain boundary sliding that can occur during the dwell time
at the high temperature extreme of TMF cycling.

Sn-2.5Ag-O.5Cu-0.5Ni solder joints showed the least visually noticeable surface
deformation after 1000 TMF cycles with longer dwell time at low temperature extreme.
As shown in Figure 5.9(b), no signiﬁcant of damage occurs in this alloy with longer
dwell time at high temperature extreme. Sn—2Ag-lCu-1Ni solder joints exhibited no
signiﬁcant surface damage with long dwell time at high temperature extreme. However,
some surface damage observed in this alloy for long dwell time at low temperature
extreme, as can be seen in Figure 5.10(a) and (b).

In general, the quaternary alloys developed less surface damage accumulation
during TMF tests with long dwell times at low and high temperature extremes. However,
their behavior was much better in situation with longer dwell time at high temperature

extreme.

78

‘2

Eutectic Sn-Ag

1000 thermal cycles Eutectic Sn- 3 5A:

1000 TMF Cycles

f
l
3..
l
1
at
£1:

-"Pra-

 

Figure 5.7. Comparison of surface microstructures of 1000 TMF cycled eutectic Sn-
3.5Ag solder joints. (a) long dwell time at low temperature extreme [adopted from
ref. 44] and (b) long dwell time at high temperature extreme.

79

Sn-4Ag-0.5Cu
TMF 1000 cycles

a“

{ c4811-4A'g90 .tSC-u " 7

g 1 000’?ME W93.“

 

(b)
Figure 5.8. Comparison of surface microstructures of 1000 TMF cycled Sn-4Ag-

0.5Cu solder joints. (a) long dwell time at low temperature extreme [adopted from ref.
44] and (b) long dwell time at high temperature extreme.

8O

- . s11.2. Mob 5( 11— 0 51\l13
12.5 111.0 5C 11 0:51." .. 1000 TMF qyclcs
lOlO TVTF cycles ‘ - 3’

2%--“ .

\‘ "I 2 7..-: c.

it
1“
t
1
:1
I
{1,
t'
1'»

 

Figure 5.9. Comparison of surface microstructures of 1000 TMF cycled eutectic Sn-
2.5Ag-0.5Cu-0.5Ni solder joints. (3) long dwell time at low temperature extreme
[adopted from ref. 44] and (b) long dwell time at high temperature extreme.

81

During the heating part of the TMF cycle, the solder becomes softer due to the
raising temperature. In addition, as the temperature raises the peak shear stress that can
develop in the solder joints also decreases. Stress relaxation studies carried out at 150°C
and room temperature have shown that the stress drop is much higher at 150°C [86].
Based on these observations, it is reasonable to expect the stress that develops due to
CTE mismatch during heating part of the TMF cycle to build up to a very low magnitude.
In addition, residual stress that contributes to creep can be expected to be relaxed more
easily at the high temperature extreme of the TMF cycle. As a consequence, only
minimal damage accumulation results in such a case irrespective of the extent of dwell
time at this temperature.

On the other hand, during the cooling part of the TMF cycle, the solder becomes
less compliant with decreasing temperature. In addition, the solder joint can develop
higher peak shear stresses at lower temperatures. Such a stress is not easily relaxed at the
low temperature extreme of the TMF cycle. Under such conditions large stresses present
at the lower temperature can cause more damage during longer dwell time at this
temperature.

5.3-3. Comparison of the Effect of Dwell Times at Different Temperature Extremes
on Residual Simple Shear Behavior of Solder Joints After TMF

The residual mechanical behavior of four Sn-Ag based solder joints that
experienced longer dwell time at high temperature extremes can be compared with those
ﬁ'om a prior study [82] with longer dwell time at low temperature extreme (Figures 11(a)
and (b).). As can be observed from Figure 11(a), there is an initial drop in residual-

simple shear

82

,lgbomf cycles; IQOOTng/lfy’cyclcs - '

3;
x,

a . . ~ 1,“

a
g .
'1:

l‘

l.

,1
1.-

l .

II.

\

4,‘ /\-I ,'
4x

1

 

Figure 5.10. Comparison of surface microstructures of 1000 TMF cycled eutectic Sn-
2Ag-lCu-1Ni solder joints. (a) long dwell time at low temperature extreme [adopted
from ref. 44] and (b) long dwell time at high temperature extreme.

83

strength as a consequence of 250 TMF cycles in all four Sn-Ag based solder joints.
However, no signiﬁcant decrease of the residual-simple shear strength was generally
noted between 250 and 1000 TMF cycles. The drop in the residual-simple shear strength
was approximately 30~60 % after 250 TMF cycles for solder joints that experienced
longer dwell time at low temperature extreme. However, from Figure 5.11(b), the
decrease in simple residual shear strength is less for solder joints that experienced longer
dwell time at high temperature extreme. Although the behavior exhibited is similar to
that which occurred in Figure 5.11(a), the drop in the residual-simple shear strength is
only about 20 % after initial few hundred TMF cycles. As noted from Figure 5.11(b),
almost no signiﬁcant decrease in residual simple shear strength can be observed in Sn-
2Ag—lCu—1Ni solder joints even after 1000 TMF cycles.

Residual mechanical behavior exhibited by the solder joints after TMF are
consistent with the observations presented in the previous sections. Since the specimens
that experience longer dwell time at the high temperature extreme of TMF exhibit less
damage accumulation than those with longer dwell time at the lower temperature
extreme, they retain more of their mechanical capabilities.

The quaternary alloys exhibit less damage accumulation and loss of mechanical
strength due to TMF, especially in the current study. Alloys containing 1% Ni and 1%Cu
do not show damage accumulation or decrease in strength even after 1000 TMF cycles
with longer dwell times at high temperature extreme. These quaternary alloys have also
been found to exhibit better isothermal creep properties [55]. The roles of Cu and Ni,

especially when both are present as alloying elements, for the improved performance of

84

50 , g 1
1 + Sn-3.5Ag

\ 4L Sn-4Ag-0.5Cu

40 \ -~--O Sn-2.5Ag-0.5Cu-0.5Ni
\ -I-Sn-2Ag-1Cu-1Ni

 

 

 

N
G
1

 

Residual-Simple Shear Strength (MPa)

 

 

l

0 250 500 750 1000
Number of cycles

 

 

 

 

 

-o— Sn-3.5Ag

‘ -v— Sn-4Ag—0.5Cu

we Sn-2.5Ag-0.5Cu-0.5Ni
- I ' Sn-ZAg-lCu-lNi

1— d

N
O
1

 

 

 

Residual-Simple Shear Strength (MPa)
8

 

 

 

o 1 1 u
0 250 500 750 1000
Number of cycles
(b)

Figure 5.11. Effect of thermal cycles on the residual-simple shear strength of four Sn-Ag
based solder joints showing minimal drop in simple shear strength in solder joints
experienced long dwell time at high temperature extreme. (a) long dwell time at low
temperature extreme [adopted from ref. 82] and (b) long dwell time at high temperature
extreme.

85

Sn-Ag based solder is not fully understood at this time, and further investigations are in

progress.

5.4. SUMMARY

The following summary can be made based on the present study,

1. TMF cycling using long dwell time at high temperature extreme produced
deformation and damage very similar to that observed with longer dwell time at

low temperature extreme in eutectic Sn-3.5Ag and Sn-4Ag—O.5Cu solder joints.

2. No signiﬁcant surface damage could be noted in Sn—2.5Ag—0.5Cu—O.5Ni and Sn-
2Ag-1Cu-1Ni solder joints that experienced long dwell time at high temperature

extreme even after 1000 TMF cycles.

3. A comparison of damage accumulation and decrease in simple shear strength of
Sn—Ag based solder joints that experienced longer dwell times at low or high
temperature extremes after TMF indicates solder joints that experienced longer
dwell time at high temperature extreme exhibited less surface damage

accumulation and less decrease in simple shear strength.

4. In general, small Ni addition along with small Cu addition to Sn-Ag based solder

to produce quaternary alloys results in better TMF performance.

86

CHAPTER 6
EFFECTS OF TMF HEATING RATES ON DAMAGE ACCUMULATION AND
RESULTANT MECHANICAL BEHAVIOR OF TIN-SILVER BASED SOLDER
JOINTS
ABSTRACT
Solder joint reliability depends on several service parameters such as temperature
extremes encountered, dwell times at these temperatures, and the ramp-rates representing
the rate at which the temperature changes are imposed. TMF of Sn-Ag based solder alloy
joints of realistic dimensions were carried out with dwell of 115 minutes and 20 minutes
at 150°C and —15°C respectively. Different heating rates were obtained by controlling the
power input during heating part of TMF cycles. Surface damage and residual mechanical

strength of these solder joints were characterized after 0, 250, 500, and 1000 TMF cycles

to evaluate the role of TMF heating rate on the solder joint integrity.

87

6.1. INTRODUCTION

Thermomechanical fatigue (TMF) is a major cause of failure in electronic and
power devices [92-95]. A typical thermal cycle that leads to failure of solder joint
consists of two ramp times of comparatively short duration during which the device heats
up to operational temperature (high temperature extreme) or cools down to ambient
temperature (low temperature extreme) and two longer dwell times at each temperature
extreme which separates these two ramps [27,34]. During service, the device may
experience long dwell at the high temperature extreme or long dwell at low temperature
extreme depending on applications. Also, cooling rate and heating rate can be changed
with external environment, internal heating from power dissipation, or both. In addition
to these service parameters, the temperature regime imposed and difference in
temperature extremes may be important factors that can contribute to TMF behavior of
electronic solder joints, as illustrated in Figure 6.1. However, only a few studies have
been attempted to understand the role of such individual service parameters on the TMF
behavior of solder joints [96-98].

There are certain requirements needed in order to investigate roles of each of
these parameters separately. Such requirements include, (i) the same solder joint
geometry, (ii) varying one parameter while maintaining the other parameters constant,
(iii) use of same solder materials, and (iv) comparison after a ﬁxed number of TMF
cycles.

Effect of dwell times on TMF behavior of Sn-Ag based solder joints have been
reported in an earlier publication [99]. In that study, the geometry of solder joints,

imposed temperature range, heating rate, and cooling rate were held and ﬁxed on Sn-Ag

88

 

 

 

 

Time

Figure 6.1. Schematic of various service parameters that can affect thermomechanical
fatigue (TMF) damage of electronic solder joints.

89

based solder joints while the dwell times at each temperature extreme were the varied
service parameters [99].

The aim of this study is to investigate the effects of ramp rates during heating
segment of TMF cycles and to compare it with results from prior TMF study with a slow
heating rate. Such a comparison includes the resultant residual mechanical behavior and

surface damage accumulation.

6.2. EXPERIMENTAL PROCEDURE

The current study deals with solder joints made with eutectic Sn—3.5Ag, Sn-4Ag-
0.5Cu, Sn-2.5Ag-O.5Cu-O.5Ni, and Sn-2Ag—lCu-1Ni solder pastes to facilitate a
comparison with the results obtained in earlier studies [99]. These solder joints, with
~100um thickness and l x 1 mm2 area, have been used in prior TMF studies [44,82,99].
The detailed conﬁgurations of the dog-bone shaped single shear-lap solder joint geometry
and soldering procedure used have been reported elsewhere [87,88]. The edge of the
solder joints were metallurgically polished prior to TMF testing.

A thermomechanical fatigue set-up with controllable ramp-up and dwell times
was designed. This set-up consisted of a hot-plate as a heating source inside a common
household ﬁeezer with inside temperature of -15°C. A copper plate placed on the hot
plate was used as a stage for holding the solder joint specimens. Thermocouples were
placed on the metal stage to measure and control the temperature. The heating element
was turned on and off by an external timer. Power to the heating element was provided

through a variable transformer so that the heating rate can be adjusted and controlled.

9O

The dwell times at the temperature extreme was controlled by the external timer, which
turned the heater on or off.

Thermal cycling between —15°C and +150°C (A=165°C) was imposed on the
solder joints with this set-up. The dwell time at the high temperature extreme was about
115 minutes and the dwell at the low temperature extreme of the cycle was about 20
minutes. The ramp-rate for the cooling segment of the cycle was ~0.07°C/sec. This TMF
cycling has almost the same temperature extremes and cooling rate used in our earlier
studies to facilitate the comparison of the role of heating rates. The temperature proﬁles
obtained with this set—up for prior and current studies are given in Figure 6.2. In the
current study, the ramp-rate for the heating segment of the cycle was ~l.1°C/sec. This
heating rate is twice as fast as heating rate used in earlier studies (~O.55°C/sec) as can be
noted in Figure 6.2(b). The solder joints were subjected to a maximum of 1000 TMF
cycles.

Surface damage accumulation at the exact same-targeted area was observed in all
the four Sn—Ag based solder joints after 0, 250, 500, and 1000 cycles by Scanning
Electronic Microscopy. Some of the solder joints were removed after 250 and 500 cycle
intervals to assess the residual mechanical behavior in shear with a tensile testing

machine at a crosshead speed of 0.5 mm/min at room temperature.

91

 

 

  

 

Y I I I r T Y Y r7 T T Y I T I I I T I
1 1 t 1 .
l I l 1
l 1 l 1 1
l g l 1 .1
I 1 i 1 q
- l 1 l
150 f.- "g ..:;-;e-;: U ‘ o -1“. f.
A 5' ' l
‘ > 5’ ’s l " i
I . \i n l
er 1. : ll‘ 5
. .
a) 10 '——~--:- j—‘ _______ 1;- ........
LI 9 t
t- | 0 I
= i I ! 1
a l . l i
h —- --L—l—— _____ L ........
Q) I 1 i
l
D. I , .. ‘
.1 l l
E I 1 1
I i
G) H ----1 I 1‘_
H 1‘] 1
I
I 1
i
1
-5. r 1 ill 1 1.1

 

 

 

 

   

 

 

 

 

      

 

 

 

I
I
I I
I l
I l
I l
I
200' 4 . 3.1.. . . 1 r1
. 5 H g 3
E ' i
A 151 . ..g " A ------- ——:
U r i U '
°v L : °V . 1 . :1
a 100 . .. ‘ 2 .. ................. .............. _
3 "20-55°C/.S€C < E ~l.l°C/sec z 3
a . g l a i f l
a so 7...... .. ................... ................ _: g . .................. é ........ . . .: .............. 5
E E Ramp time: 5 min E Ramp time: 2.5 min 4
P o . .............. 1.. .................. '. ...... .. ....... -4 H 0 .. 1-.. “...:
2 nge11 tame: 115 mm Dwell time: 117.5 mm.
-50’. .411 _ .-.1 41...
0 5 10 15 20 500 5 10 15 20

Time (min)

(b)

Time (min)

(C)

Figure 6.2. Temperature-time proﬁles of TMF. (a) overall temperature-time proﬁle
showing ﬁxed dwell times at each temperature extremes and cooling rate, (b) slow
heating rate with fixed dwell times and cooling rate, and (c) fast heating rate with ﬁxed
dwell times and cooling rate.

92

6.3.RESULTS AND DISCUSSION

6.3-1. Surface Damage Accumulation in Sn-Ag Based Solder Joints that
Experienced Fast Heating Rate During TMF

Eutectic Sn-3.5 Ag solder joints showed progress of surface damage accumulation
with increasing number of TMF cycles as shown in Figure 6.3. The initially polished
surface, shown in Figure. 6.3(a)., exhibits shear banding near the intermetallic interface
layer due to TMF as shown in Figure 6.3(b). The intensity of this shear banding became
stronger with increasing number of TMF cycles (Figures 6.3(b). through (d)). This
highly localized shear banding was also found in an earlier study with slow heating rate
imposed during TMF. The SEM images of eutectic Sn-Ag solder joints after a maximum
of 1000 TMF cycles with different heating rates imposed during TMF are presented in
Figure 6.4. More highly localized damage was observed in solder joints that experienced
TMF with faster heating rate than that of solder joints experienced TMF with slower
heating rate as noted in Figures 6.4(a) and (b). Rhee et. al.[100] have reported effects of
shear-strain rates at various temperatures using same geometry of eutectic Sn-3.5Ag
solder joints. They have reported that concentrated deformation was found in localized
region with higher imposed rates of shear-strain while lower rates of simple shear-strain
resulted in diffused deformation characteristics under isothermal shear tests [100].

Microstructures of the ternary alloy, Sn-4Ag-O.5Cu, after 0, 250, 500, and 1000
TMF cycles are illustrated in Figure 6.5. Apparently grain boundary sliding occurred at a
region away from the intermetallic interface during the ﬁrst few hundred TMF cycles and
the damage accumulation intensiﬁed with increasing number of TMF cycles which can
be seen in Figures 6.5(a) through (d). There was an obvious crack growth within solder

between Sn-Sn grains due to grain boundary sliding between 500 and 1000 TMF cycles.

93

 

(d)

Figure 6.3. Surface damage accumulation on eutectic Sn-Ag solder joint undergoing
TMF with faster heating rate. (a) 0 TMF cycle, (b) 250 TMF cycles, (c) 500 TMF cycles,
and (d) 1000 TMF cycles.

94

f 1000 cycles
.1 Eutectic Sir-Ag

l fEutectic Sir-Ag I ‘11
V‘!‘

f

r
,H/

 

More localized damage

(b)

Figure 6.4. Surface damage accumulation in eutectic Sn-Ag solder joints aﬁer 1000 TMF
cycles. (a) slower heating rate imposed during TMF, and (b) faster heating rate imposed
during TMF.

95

250 cycles

1000111171112

T

u”

 

(d)
Figure 6.5. Surface damage accumulation in Sn-4Ag-O.5Cu solder joint undergoing TMF

under faster heating rate. (a) 0 TMF cycle, (b) 250 TMF cycles, (c) 500 TMF cycles, and
(d) 1000 TMF cycles.

96

Micrographs comparing surface damage accumulation after 1000 TMF cycles in Sn-4Ag-
0.5Cu studied with different heating rates imposed during TMF are presented in Figure
6.6. Similar to surface damage that found in eutectic Sn-3.5Ag solder joints, highly
localized surface damage was found in Sn-4Ag-O.5Cu solder joints after TMF with faster
heating rate.

Both quaternary alloys used in this study showed ahnost no visible surface
damage accumulation even after 1000 TMF cycles. Figure 6.7 shows examples of the
microstructures of Sn-2.5Ag—O.SCu—O.SNi solder joints after TMF. No signiﬁcant surface
damage was found to occur between 0 and 1000 TMF cycles. Similar behavior can also
be noted for Sn-2Ag-1Cu—1Ni as illustrated in Figure 6.8.

Irrespective of slow or fast heating rate, the surface damage accumulation was
found to be minimal in both quaternary solder joints studied, as compared to the binary
and ternary alloy joints after 1000 TMF cycles, as can be seen in Figures 6.9 and 6.10.
6.3-2. Comparison of the Effect of Different Heating Rates Imposed During TMF

on Residual Shear Strength of Sn-Ag Based Solder Joints

The residual mechanical behavior of the four Sn—Ag based solder joints that
experienced faster heating rate can be compared with those from a prior study [10] with
slower heating rate (Figure 6.11).

For eutectic Sn-3.5Ag solder joints, the residual shear strength was initially
decreased about 20 % within ﬁrst few hundred TMF cycles with additional gradual drop
between 250 and 500 cycles. No signiﬁcant drop in residual shear strength was noted
between 500 and 1000 TMF cycles with the slow heating rate imposed during TMF,

indicating that the residual shear strength for eutectic Sn-Ag solder joints stabilized after

97

’ 1000 cycles
Shs4rAg—O5Cu

 

More localized damage
(b)

Figure 6.6. Surface damage accumulation in Sn—4Ag-O.5Cu solder joints after 1000 TMF
cycles. (a) slower heating rate, and (b) faster heating rate.

98

20 11111

 

(C)

Figure 6.7. Surface damage accumulation in Sn-2.5Ag-0.5Cu-0.5Ni solder joint
undergoing TMF with faster heating rate. (a) 0 TMF cycle, (b) 250 TMF cycles, (c) 500
TMF cycles, and (d) 1000 TMF cycles.

10.00 cycles-V ‘ _

 

Figure 6.8. Surface damage accumulation in Sn-2Ag-lCu—1Ni solder joint undergoing
TMF with faster heating rate. (a) 0 TMF cycle, (b) 250 TMF cycles, (c) 500 TMF cycles,
and (d) 1000 TMF cycles.

99

. smzs'sagjiois'Cugtr;5111:"

 

Figure 6.9. Surface damage accumulation in Sn-2.5Ag—0.5Cu—0.5Ni solder joints after
1000 TMF cycles. (a) slower heating rate, and (b) faster heating rate.

. 311-211‘g.._1cu.vd1111

.1‘

 

Figure 6.10. Surface damage accumulation in Sn-2Ag-lCu-1Ni solder joints after 1000
TMF cycles. (a) slower heating rate, and (b) faster heating rate.

100

500 TMF cycles. A total of 35 % decreases in residual shear strength were noted for
eutectic Sn-3.5Ag solder joints after a maximum of 1000 TMF cycles (Figure 6.11(b)).
However, with a faster heating rate the residual shear strengths of eutectic Sn-3.5Ag
solder joints'decreased dramatically with increase in number of TMF cycles, as can be
seen in Figure 611(0) and ((1). Drop in residual shear strength 1000 TMF cycles was
about 42 %. Although this is only about 10 % worse after 1000 TMF cycles, the major
difference arise from the fact that the residual shear strength continues to decrease even
after TMF cycles with no stabilization. Continued intensiﬁed highly localized damage
that results from TMF with fast heating rate could be a contributing fact for such
continued deterioration.

As can be observed from Figures 6.11(a) and (b), Sn-4Ag-0.5Cu solder joints
posses better initial shear strength than other Sn-Ag based solder joints studied. The
decrease in shear strength was about 20 % after 250 TMF cycles with a slower heating
rate. About 10 % additional decrease in shear strength was noted between 250 and 1000
TMF cycles under such condition (Figure 6.11(b)). After 1000 TMF cycles, the drop in
shear strength was about 35 %. Under TMF with a faster heating rate, the residual shear
strength of joints made with this alloy dropped linearly with increasing number of TMF
cycles as can be seen in Figures 6.11(c) and (d). This decrease was about 60 % after a
maximum of 1000 TMF cycles with a faster heating with no sign of stabilizing on further
TMF cycling. Such observations indicate that Sn-4Ag-0.5 Cu solder joints are highly

sensitive to change in imposed heating rates during TMF.

101

 

Residual Shear Strength (MPa)

 

 

 

 

 

 

 

 

 

20 ... .................... E ....................... ;. ..................... .1
+Sn-3.5Ag
“- Sn-4Ag—0.5Cu 5
10 - "'O"Sn-2.5Ag-0.5Cu-0.5Ni ..................... 5
- I - Sn-2Ag-lCu-1Ni 5
0 i i i
0 250 500 750 1000
Number of TMF Cycles
(3)

 

 

 

20

Normalized Shear Strength (%)

-O- Sn-3.5Ag

-V— Sn-4Ag-0.5Cu

~--O ~ Sn-2.5Ag-0.SCu-0.5N i

 

 

 

- I - Sn-ZAg-lCu-lNi

 

 

o 1 l l i
0 200 400 600 800 1000
Number of TMF Cycles
(b)

102

 

U
G

 

 

N
G

 

I Sn-3.5Ag

'V— Sn-4Ag—0.5Cu
' - O -‘ Sn-2.5Ag-0.5Cu-0.5Ni
' I - Sn—2Ag-l Cu-lNi

u—s
Q
1

.......

Residual Shear Strength (MPa)

 

 

 

 

 

0 i i i
0 250 500 750 1000
Number of TMF cycles

 

é

 

on
O

8

8

 

 

I Sn-3.5Ag
’- Sn-4Ag-0.5Cu

...... Sn-2.5Ag-0.5Cu-0.5Ni

' I 'Sn-2Ag-1Cu-1Ni

Normalized Shear Strength (%)
N
O

 

 

 

 

i i 1

 

o 1
0 200 400 600 800 1000
Number of TMF Cycles
((0

Figure 6.11. Residual mechanical behaviors of Sn-Ag based solder joints undergoing
1000 TMF cycles. (a) shear strength versus number of TMF cycles with slower heating
rate, (b) normalized shear strength versus number of TMF cycles with slower heating
rate, (0) shear strength versus number of TMF cycles with faster heating rate, and (d)
normalized shear strength versus number of TMF cycles with faster heating rate.

103

Unlike binary and ternary solder joints, quaternary solder joints exhibited minimal
drop in residual shear strength. The residual shear strength was decreased initially with
few hundred TMF cycles. There was no signiﬁcant drop in residual shear strength was
noted between 250 and 1000 TMF cycles in quaternary solder joints, irrespective of
different heating rates imposed. The decrease in residual shear strength was within 20 %
even after 1000 TMF cycles with both slow and fast heating rates. It suggests that
quaternary solder joints were not inﬂuenced signiﬁcantly by change in heating rates

during TMF.

6.3-3. Possible Reasons for Better TMF Behavior in Quaternary Solder Joints than
That of Binary and Ternary Solder Joints

Quaternary alloys containing both Ni and Cu did not show signiﬁcant surface
damage accumulation or a decrease in strength even after 1000 TMF cycles, irrespective
of imposed heating rates. These quaternary alloys have also been found to posses better
high temperature creep properties than that of eutectic Sn-Ag [44]. The second phase
IMC particles present in these quaternary solder joints are (Cu-Ni-Sn) IMC particles and
these IMC particles did not coarsen even under long dwell time exposure at the high
temperature extreme during TMF. Rhee et. al. [100] reported that in eutectic Sn-Ag
solder joints, high temperature deformation is dominated by grain boundary sliding.
Keying the grain boundary movement, to retard their relative movement can improve
mechanical properties of Sn-Ag based solder joints, especially at high temperature.

Figures 6.12(a) and (b) are micrographs of eutectic Sn-Ag and Sn-2Ag-1Cu-1Ni
solders after etching with 5 vol % HNO; and 95 vol % methanol for 1 second to reveal Sn

grain boundaries in the solder area. Eutectic Sn-Ag exhibited Ag3Sn IMC precipitates as

104

can be seen in Figure 6.12(a). For Sn-2Ag-1Cu-1Ni solder joints, the (Cu-Ni-Sn) ternary
IMC precipitates are present at the Sn-Sn grain boundaries, and smaller amounts of
continuous narrow Ag3Sn precipitates that are present within Sn grains can be noted in
Figure 6.12(b). These (Cu-Ni-Sn) IMCs do not coarsen during TMF, probable due to the
very slow diffusion of Ni in Sn [101]. The presence of these ﬁne, stable, hard, sub-
rrricron sized (Cu-Ni-Sn) IMC’s at the grain boundaries can retard the relative grain
boundary movements by keying the grain boundaries resulting in better TMF resistance,
irrespective of heating rate imposed.
6.3-4. Possible Reasons for Worse TMF Behavior in Ternary Solder Joints Than

That of binary And Quaternary Solder Joints

The second phase IMC precipitates present in Sn-4Ag-0.5Cu are Cu6Sn5 that
formed during reﬂow. Initially, only a small number of Cu68n5 IMC particles formed
during reﬂow as seen in Backscatter Electron (BE) image of an as-fabricated Sn-4Ag—
0.5Cu solder joint (Figure 6.13). Cu6Sn5 IMC particles presented in the as-fabricated
joint were generally very small, and they enhance initial mechanical strength of joint
made with that solder. However, a large number of (Cu-Sn) IMC particles were found
after 1000 TMF cycles. These (Cu-Sn) IMC particles were not easily detected by
Scanning Electron (SE) images (Figure 6.14(a)), but BE images allowed to distinguish
these (Cu-Sn) particles, as can be noted in Figure 6.14(b). Energy Dispersive
Spectroscopy (EDS) analysis was carried out to determine the composition of these IMC
particles, as can be seen in Figures 6.14(c) through (e). The size of (Cu-Sn) particles
increased with increasing number of TMF cycles and resulted in signiﬁcant coarsening

during high temperature dwell in TMF. Accumulation of aging time at the

105

 

(b)

Figure 6.12. Etched microstructures of (a) eutectic Sn-Ag and (b) Sn-2Ag—lCu-1Ni.
(Cu-Ni—Sn) IMC precipitations that may key grain boundary movements can be seen in
(b)-

(‘u .. . ....
\L 11911,

/ \

50 11111

\//

(11511. N

/ 2011111

(b)
Figure 6.13. Backscatter Electron (BE) images of as-fabricated Sn-4Ag-O.5Cu showing
small amounts of Cu6Sn5 IMC particles in as-fabricated condition. (a) low magniﬁcation,
and (b) high magniﬁcation.

 

81‘. image

7 11m

 

7 11m
(6)
Figure 6.14. Series of images of Sn-4Ag—0.5Cu solder joint that experienced 1000 TMF
cycles. (a) Scanning Electron (SE) image, (b) Backscatter Electron (BE) image, (c)
Energy Dispersive Spectroscopy (EDS) mapping showing Sn in these particles, (d) EDS

107

mapping showing Cu in these particles, and (e) EDS mapping showing Ag in these
particles.

BF. image

C u-Sn IMC

 

Figure 6.15. SEM images of Sn-4Ag—0.5Cu solder joints after 1000 TMF cycles. (a)
Backscatter Electron (BE) image showing (Cu-Sn) IMC particles at cracked area near
IMC layer, (b) Backscatter Electron (BE) image showing (Cu-Sn) IMC particles at
cracked area with solder region.

108

temperature extreme during TMF cycle imposed was about 2000 hours after 1000 TMF
cycles. This may cause signiﬁcant coarsening of Cu-Sn IMC particles during TMF
Cracks have been observed to form under such condition. Such a crack may propagate
more easily with the growth of (Cu-Sn) IMC particles with increasing number of TMF
cycles because of higher stress at Sn-Sn grain boundaries. These Cu-Sn particles were
mostly found at highly damaged solder areas as can be seen in Figure 6.15. Cracks
developed due to Cu-Sn IMC particles in Sn—4Ag-O.5Cu may result in signiﬁcant
deterioration in residual shear strength. Under a faster heating rate, Sn-4Ag-0.5Cu will
experience a higher strain rate. High temperature strain rate sensitivity of Sn-Ag-Cu
solder has been noted to be higher than those of other solders studied [102]

Creep studies carried by Grusd [54] indicated that Sn-4Ag-0.5Cu solder joints
exhibited better creep-rupture resistance at room temperature than eutectic Sn-Ag solder
joints. However, creep-rupture resistance of Sn-4Ag-0.5Cu solder at 100°C was the
opposite of the room temperature creep behavior. Similarly, Guo et. al. [55] have
reported that the eutectic Sn-Ag solder joints showed better creep resistance in terms of
the maximum strain for onset of tertiary creep than Sn-4Ag—0.5Cu at high temperature
(85°C) while Sn-4Ag-0.5Cu joints exhibited better room temperature creep resistance
than Sn-Ag solder joints. Such ﬁndings are consistent with the observed poor
performance of Sn-Ag-Cu solder under TMF with a fast heating rate.

Suhling et. al. [56] have reported effects of fourth alloying addition to 95.5Sn-
3.8Ag-0.7Cu solder joints on TMF under two different temperature ranges (-40 to 125°C
and -40 to 150°C). Sn-3.8Ag-O.7Cu solder joints exhibited worse TMF resistance than

Sn-37Pb under TMF with a temperature range from -40°C to 150°C. However, the TMF

109

resistance of SAC alloys were signiﬁcantly improved with the addition of a forth element
under TMF with a temperature range from -40°C to 150°C. Such ﬁndings indicate that a
certain alloy addition can play signiﬁcant roles in improving TMF resistance of Sn-Ag-

Cu alloy.

6.4. SUMMARY

TMF with a fast heating rate produced signiﬁcant visible damage in eutectic Sn-
3.5Ag and Sn—4Ag-0.5Cu solder joints. Highly localized surface damage accumulation
was found in eutectic Sn-3.5Ag and Sn-4Ag-0.5Cu solder joints under TMF with faster
heating rate as compared to the same solder joints that under slower heating rate. About
a 42 % of drop in residual shear strength was noted after 1000 TMF cycles with a fast
heating rate in eutectic Sn-Ag solder joints. Residual shear strength of Sn—4Ag—0.5Cu
continues to decrease even after TMF cycles, resulting in a decrease of about 60 % in the
residual strength under faster heating rate. It exhibited only about 35 % drop due to TMF
with slower heating rate.

Quaternary solder joints (Sn-2.5Ag-0.5Cu-0.5Ni and Sn-2Ag-lCu-1Ni) showed
minimal drop in simple shear strength even after a maximum of 1000 TMF cycles for
both slow and fast heating rates. Sn-2.5Ag-0.5Cu-0.5Ni exhibits minimal surface
undulations while Sn-2Ag—lCu-1Ni exhibits almost no surface damage even after 1000
TMF cycles for both fast and slow heating rates. No signiﬁcant interface IMC layer
growth and particle coarsening were observed even after 1000 TMF cycles in these
quaternary alloys.

Solder joints made with ternary Sn-4Ag-0.5Cu exhibited the worst TMF behavior

among Sn-Ag based solder joints. Among all Sn-Ag based solder joints studied, ternary

llO

solder joints that contained small amounts of Cu were highly sensitive to change in

heating rates during TMF.

lll

CHAPTER 7

MICROSTRUCTURAL FEATURES CONTRIBUTING TO ENHANCED
BEHAVIOR OF TIN-SILVER BASED SOLDER JOINTS

ABSTRACT

Role of microstructure of solder joints made with eutectic Sn-Ag solder and Sn-Ag solder
with Cu and/or Ni additions on the creep and thermomechanical fatigue (TMF) properties
was investigated. Quaternary alloys containing small amounts of Cu and Ni exhibited
better creep resistance at higher temperatures, and also better TMF resistance with longer
dwell times at high temperature extreme than eutectic Sn-Ag, and Sn-Ag-Cu ternary alloy
solder joints. Microstructural studies of the quaternary solder alloys revealed the
presence of ternary Cu-Ni-Sn intermetallic compound (IMC) second-phase at Sn grain
boundaries. These precipitates can retard grain boundary sliding that will result during

TMF with longer dwell times at high temperature extreme, and high temperature creep.

112

7.1. INTRODUCTION

Significant research efforts are being expended along various avenues to ﬁnd
suitable substitutes for the lead-bearing electronic solders e.g. [3,67,103,104]. Using
eutectic Sn-Ag solder as a basis, several approaches have attempted to modify its
properties to address in service needs. One such approach involves the development of
ternary and quaternary alloys with additions of Cu, Ni, Bi, Sb, and rare-earth elements etc
[ll-l6]. Thermomechanical Fatigue (TMF) failures that occur as a consequence of
stresses that develop due to coefﬁcient of thermal expansion (CTE ) mismatches between
component leads, the solder, and the substrate during temperature excursions experienced
in service, are the a common cause of failure in electronic circuitry [104]. In order to
understand the basic operating mechanisms and contributing factors to such failures in
Pb-free solders, several focused studies have been conducted e.g. [44,55,82,85,86].
Such studies include the tests on the isothermal creep [55,85]/stress-relaxation [86]
properties and TMF tests with different dwell times [44,82,99]. They have shown that
quaternary solder alloys containing small amounts of Cu and Ni possess better high
temperature creep resistance and better TMF resistance [55,99]. The presence of both
Ni and Cu as alloying elements in Sn-Ag solder seems to be beneﬁcial. The main
objective of this study is to investigate the microstructural features that contribute to the

enhanced performance of these quaternary alloys.

7.2. BACKGROUND
This background information is based on studies conducted with solder joints

made with (i) eutectic Sn-Ag, (ii) Sn-4Ag-O.5Cu, (iii) Sn-2.5Ag-O.5Cu-O.5Ni, and (iv)

113

Sn-2Ag—1Cu-1Ni solder pastes. The detailed conﬁgurations of the dog-bone shaped
single shear—lap solder joint and soldering procedure used in this study have been
reported elsewhere [55,85,86].

Although the full details of test regime have been given [44,55,86], some of the
important details are repeated here so as to provide a quick review. TMF cycling
between —15°C and +150°C was imposed on the solder joints with controlled ramp rates
and dwell times. A 25°C/min heating rate and 7°C/min cooling rate were imposed for
two different dwell times to investigate the effect of dwell times. Firstly, TMF cycles
with longer dwell times at the low temperature extreme were carried out, with a dwell for
300 min at the low temperature extreme and a dwell for 20 min at the high temperature
extreme. Secondly, a longer dwell time at the high temperature extreme was imposed.
i.e. 115 min at the high temperature extreme and 20 min at low temperature extreme.
Mechanical tests to measure the residual shear strength of the solder joints aﬁer TMF
were carried out using a simple shear test. A more comprehensive description of the
experimental details of these shear tests is available elsewhere [44,82,99]. Details of the
creep test procedure have been reported in several earlier publications [55,85]. These
tests were carried out 22°C and 85°C with stress range from 5.5 to 24 MPa [55]

Quaternary alloy solder joints that were subjected to the TMF with a longer
dwell time at the high temperature extreme exhibited better TMF resistance compared to
eutectic Sn-Ag, as seen in Table 7.1 [99]. Even after 1000 TMF cycles, almost no
surface damage was noted in joints made with the quaternary alloys. Surface damage
due to TMF in joints made with eutectic Sn—Ag and the ternary alloy tends to increase

signiﬁcantly with an increasing number of TMF cycles. Joints made with Sn-2Ag-1Cu-

114

Table 7.1. Residual shear strength of Sn-Ag base solder joints underwent TMF with
different dwell times at low and high temperature extremes after maximum 1000 TMF

 

 

 

 

 

 

cycles [99].
% drop in residual shear % drop in residual shear

Materials strength with long dwell time strength with long dwell time

at low temperature extreme at high temperature extreme
Eutectic Sn-Ag ~52% ~ 40 %
Sn-4Ag-O.5Cu ~45% ~ 27 %
Sn—2.5Ag-O.5Cu-O.5Ni ~39% ~ 20 %
Sn-2Ag-1Cu-1Ni ~55% ~ 20 %

 

 

 

 

Table 7 .2. Comparison of strain for the onset of tertiary creep between eutectic Sn-Ag
solder joints and Sn-2Ag-1Cu-1Ni crept at 22°C and 85°C under stresses ranging from

 

 

 

5.5 to 24 MPa [55].
Materials Crept at 22°C Crept at 85°C
Eutectic Sn-Ag ~O.3 ~O‘.3
Sn-2Ag—1Cu-1Ni ~0.3 ~0.5

 

 

 

 

 

115

lNi solder exhibited a drop in residual simple shear strength of about 20 % after 1000
TMF cycles, while eutectic Sn—Ag solder exhibited a drop of 40 % after 1000 TMF cycles
with a long dwell time at the high temperature extreme [99]

All these four solder alloy joints exhibited surface damage accumulation due to
the TMF cycles with longer dwell time at the lower temperature extreme. Among these
alloys under such conditions, minimum surface damage was noted in joints made with
quaternary alloys [44]. The residual mechanical strength of all these alloys under such
condition decreased by about 40—55 % after 1000 TMF cycles [99]

All these four solder alloy joints exhibited surface damage accumulation due to
the TMF cycles with longer dwell time at the lower temperature extreme. Among these
alloys under such conditions, minimum surface damage was noted in joints made with
quaternary alloys [44]. The residual mechanical strength of all these alloys under such
condition decreased by about 40-55 % after 1000 TMF cycles [99].

In high temperature (85°C) creep tests, the quaternary alloy (Sn-2Ag-1Cu-1Ni)
showed better creep resistance in terms of the lowest secondary creep rate and the highest
strain for onset of tertiary creep. Strain for the onset of tertiary creep in the quaternary
alloy was about 66 % higher than that for the eutectic Sn-Ag solder joints as indicated in
Table 7.2 [55]. However, the creep resistance of this quaternary solder alloy is not

signiﬁcantly improved during room temperature creep [55].

7.3. EXPERIMENTAL PROCEDURE
To investigate the role of microstructural features contributing to the observed

behaviour, specimens were made with these four Sn-Ag based solder pastes by reﬂow on

116

copper substrates (2cm x 2cm x 1.5cm) with the same heating proﬁle used to fabricate the
single shear-lap solder joints. Although results cited from prior studies are based on
tests carried out on actual solder joints, it was too difficult to etch the small solder area of
the single shear-lap joints. As a consequence, the following alternative method was
used to observe a larger etched solder area to reveal the Sn-Sn grain boundaries and the
location of any precipitates. The solder buttons formed on the Cu substrates were
sectioned and the cut surfaces of these specimens were metallurgically polished. The
details of the specimen conﬁguration and the experimental procedure for polishing can be
found elsewhere [105]. These specimens were then etched with 5 vol % HNO3 and 95
vol % methanol for 1 second and washed with water. The polished and etched face of
cross section specimens then were examined using SEM. The solder joints used for

creep and TMF studies had a joint thickness of about 100 um. As a consequence, the

specimens cross-sectioned and etched were observed in regions within 50 pm from the

Cu/solder interface, as illustrated in Figure 7.1.

7.4. RESULTS AND DISCUSSION

The microstructures of polished and un-etched solder specimens made with these
solder alloys on Cu substrates are shown in Figure 7.2. The microstructures of the
eutectic Sn—Ag and ternary alloy (Sn-4Ag-O.5Cu) are similar, since Sn cells separated by
wide bands of Ag38n precipitates characterize these solders. However, the difference
between these two solders is the presence of small amounts of Cu6$n5. Since there is
less Ag in these alloys, the Ag3Sn bands are narrower and more uniform than in eutectic

Sn—Ag and Sn-4Ag-O.5Cu. Especially, Sn—2Ag-lCu-1Ni has much thinner Ag3Sn bands

117

Solder reﬂow on
the Cu Substrate oecrmen

   
  

SECTION

 

 

 

 

To represent details in a
solder joint of 100 pm
thick

 

 

 

Figure 7.1. Schematic drawing of solder specimens used in the present study.

1(a)

10 pm

 

Figure 7.2. SEM images of as-fabricated specimens without etching. (a) Eutectic Sn-
3.5Ag, (b) Sn-4Ag-O.5Cu, (c) Sn-2.5Ag-O.5Cu-0.5Ni, and (d) Sn—2Ag-1Cu-1Ni.

118

and more Cu-Ni-Sn intermetallic precipitates due to the presence of more Cu and Ni, and
less Ag, since both have the same alloying elements. As seen in Figures 7.2(c) and
7.2(d), multiple Cu-Ni-Sn ternary IMC precipitates are present through out the solder
area.

Backscatter electron (BE) images of etched eutectic Sn-Ag solder specimens
show wide primary Ag3Sn bands and Sn grain cells as shown in Figure 7.3. No second-
phase precipitates, except AggSn, are present at the grain boundaries and this solder
exhibits the worst high temperature creep resistance and TMF resistance.

In order to identify the composition of second phase IMC precipitate in ternary
and quaternary solder alloys, Energy Dispersive Spectroscopy (EDS) analysis was carried
out across these IMC precipitates as seen in Figure 7.4. Figure 7.4(a) shows second
phase IMC precipitates, which consist of mainly Cu and Sn that is identiﬁed as Cu6$n5 in
Cu containing Sn-Ag based solder joints. The second phase IMC precipitates found in
the quaternary alloy solder joints consisted of Cu-Ni-Sn ternary IMC as seen in Figure
7 .4(b). This Cu-Ni-Sn ternary IMC was identiﬁed Chuang et. al, [106] as (Cu, Ni)6Sn5
in Ni containing Sn7Ag based solders.

BE images of ternary Sn-4Ag-O.5Cu are provided in Figure 7.5. As seen in
Figure 7.5(b), Cu68n5 binary IMC precipitates are located at the Sn-Sn grain boundaries.
The TMF resistances of these solder joints have been shown to be slightly better than
those of eutectic Sn-Ag [99]. A possible explanation for this is that the binary Cu6$n5
second-phase precipitates located at Sn-Sn grain boundaries may retard grain boundary
sliding that can occur during the dwell times at the high temperature extreme of TMF

cycling and high temperature creep [55,99]. The creep properties of this ternary alloy

119

 

Figure 7.3. BE images of etched eutectic Sn—Ag. (a) Low magniﬁcation micrograph
of solder specimen showing Cu substrate and IMC layer, (b) High magniﬁcation
micrograph of solder specimen showing Sn-Sn grain boundaries and Ag38n precipitates
(typical eutectic microstructure of Sn-Ag solder).

120

Sn-4Ag-0.5Cu

 

    
       

 

Sn-2Ag-1Cu- lNi

 

(b) _ 

Figure 7.4. EDS of second phase IMC precipitates presented in (a) Sn-4Ag-0.5Cu, and
(b) Sn-2Ag-1Cu-1Ni.

   

121

._ :373'7

Cu Substrate

 

Figure 7.5. BE images of etched Sn—4Ag—0.5Cu. (a) Low magniﬁcation micrograph of
solder specimen showing Cu substrate and IMC layer, (b) High magniﬁcation micrograph
of solder specimen showing Sn-Sn grain boundaries, Ag38n precipitates and Cu68n5
binary IMC precipitates located at Sn-Sn grain boundaries.

122

are comparable with, but slightly better than, those of eutectic Sn-Ag due to the presence
of small amount of Cu68n5 binary IMC precipitates at Sn—Sn grain boundaries preventing
grain boundary sliding. However, the quaternary alloys developed less surface damage
accumulation during TMF tests. Especially, quaternary solder joints subjected to a long
dwell at the high temperature extreme during TMF exhibited no surface damage
accumulation and the smallest decrease in residual mechanical strength [99]. The Cu-
Ni-Sn ternary IMC precipitates present at the Sn—Sn grain boundaries, and smaller
amounts of continuous Ag3Sn precipitates that are narrow present within Sn grains can be
noted in Figures 7.6 and 7.7.

The small white precipitates seen in Figure 7.6(b) and Figure 7.7(b), are Ag38n,
and the dark precipitates are Cu-Ni-Sn ternary IMC. Some of the Ag38n precipitates
that are located at grain boundaries are continuous and very small. The presence of such
precipitates in the eutectic matrix seems to improve the mechanical behaviour of
materials, although not to a signiﬁcant extent.

The presence of ﬁne, stable, and system-compatible second-phase precipitates
located at the grain boundaries can retard coarsening, enhance mechanical fatigue
behavior, and reduce creep rate by decreasing grain-boundary sliding tendency [107].
Precipitates present at a grain boundary represent obstacles that resist sliding between the
Sn-Sn grains that share the boundary. Temperature can also affect the fracture behavior
of Sn-Ag based solder joints. Studies have shown that shear banding is the dominant
mode of deformation in solder joints deformed at room temperature, while grain
boundary sliding is the dominant mode of deformation at high temperature [107]. As a

consequence, during high temperature creep and TMF with a longer dwell times at the

123

Cu-Ni-Sn
precipitates ~‘

 

Figure 7.6. BE images of etched Sn-2.5Ag-O.5Cu-0.5Ni. (a) Low magniﬁcation
micrograph of solder specimen showing Cu substrate and IMC layer, (b) High
magniﬁcation micrograph of solder specimen showing Sn—Sn gain boundaries, Ag38n
precipitates, and Cu-Ni-Sn ternary IMC precipitates located at Sn-Sn gain boundaries.

124

. 1.. “ ..-.-- I Mi .1 7* '
.A t! / if? I. ' ‘
. . ‘ -' L \ ‘3' " ‘ in»:
i r l I” ~
‘ ' .. precipitates
: I; . ~' .

...

 

Figure 7.7. BE images of etched Sn—2Ag-1Cu-1Ni. (a) Solder specimen showing Cu
substrate and IMC layer, (b) Solder specimen showing Sn—Sn gain boundaries, Ag;Sn
precipitates, and Cu-Ni-Sn ternary IMC precipitates located at Sn—Sn gain boundaries.

125

high temperature extreme, the dominant deformation mechanism is gain boundary
sliding, and this can be minimized with the presence of discrete Cu-Ni-Sn ternary IMC
precipitates.

Other studies documented in various alloy systems indicate that such precipitates
may constrain plastic ﬂow during creep or TMF at high temperature because the height of
the climb barrier is larger, hindering dislocation climb and retarding gain boundary
sliding motion. For example, in nickel based super alloys, discrete second-phase
precipitates present at gain boundaries increase the strength and high temperature creep

resistance of these alloys [108].

7.5. SUMMARY
1. Slightly better creep and TMF properties were noted in Sn-4Ag-O.5Cu compared
to eutectic Sn-Ag, due to the presence of small amounts of Cu68n5 binary IMC
precipitates at the Sn-Sn gain boundaries. Ag3Sn may inﬂuence the mechanical
behaviour of Sn-Ag based solder joints to some extent, although it may not be

sigiiﬁcant.

2. Discrete Cu-Ni-Sn IMC precipitates present at Sn-Sn gain boundaries were
observed in Sn-2.5Ag-O.5Cu-O.5Ni and Sn-2Ag-1Cu-1Ni. These quaternary
alloys exhibited a lower secondary creep rate and higher strain for onset of
tertiary creep at high temperature (85°C), and better TMF performance under a
longer dwell times at the high temperature extreme, indicating that the precipitates

present at Sn—Sn gain boundaries can improve the performance of the solder

126

joints, especially at high temperature when the deformation is dominated by gain

boundary sliding.

127

CHAPTER 8

SUMMARY

Deformation behavior and damage accumulation during thermomechanical
fatigue (TMF) under different service conditions were characterized using the single
shear-lap eutectic Sn-3.5Ag, Sn-4Ag-O.5Cu, Sn-2.5Ag-O.5Cu-O.5Ni, and Sn-2Ag-1Cu-
lNi solder joints. The important ﬁndings from those studies are summarized.

Microstructural studies of four Sn-Ag based solder joints that underwent O, 250,
500 and 1000 TMF cycles between —15°C and +150°C, with a 20 minute hold time at the
low temperature extreme and with a 330 minute hold time at the high temperature
extreme were performed. Damage during thermomechanical fatigue initiated on the
surface and intensiﬁed with additional number of TMF cycles. The crack that would
cause a catastrophic failure of solder joint occurred very near the IMC layer in the shear
direction. Residual shear strength of Sn-Ag based solder joints decrease by 30-60% as a
consequence of 250 TMF cycles. Between 250 cycles and 1000 cycles, no sigtiﬁcant
drop in strength was noted due to TMF with a long dwell time at the low temperature
extreme. Small amounts of alloy additions (Cu or Cu and Ni) did not have any
pronounced effect on Sn-Ag solder joints under TMF with a longer dwell times at the low
temperature extreme. Signiﬁcant amounts of surface damage were developed with
increasing number of TMF cycles. Surface damage features due to TMF under a long
dwell time at the low temperature were concentrated shear banding, Sn-Sn gain

boundary sliding, and gain decohesion.

128

 

TMF with a long dwell time at the high temperature extreme, where dwell times
at the high temperature extreme was 115 minutes and dwell times at the low temperature
extreme was 20 minutes, resulted in sigiiﬁcant surface damage in eutectic Sn-3.5Ag and
Sn—4Ag-O.5Cu solder joints. TMF cycling using a long dwell time at the high
temperature extreme produced deformation and damage very similar to that observed
with longer dwell time at low temperature extreme in eutectic Sn—3.5Ag and Sn-4Ag-
0.5Cu solder joints. Overall residual shear strengths of all four Sn-Ag based solder joints
were better than that of solder joints that underwent a longer dwell time at the low
temperature extreme. Less surface damage and loss of residual shear strength were found
in quaternary solder joints that containing small amounts of Cu and Ni.

TMF with a faster heating rate produced more visible damage in eutectic Sn-
3.5Ag and Sn—4Ag-0.5Cu solder joints (with a long dwell time at the high temperature
extreme). The damage accumulation of these solder joints was more localized as
compared to solder joints that were subjected to slower ramp rate during heating segnent
of TMF cycles About 40 % drop in residual strength was noted aﬁer 1000 TMF cycles
with the faster heating rate in eutectic Sn-Ag solder joints. Residual shear strength of Sn-
4Ag—O.5Cu decreases with additional TMF cycles, which results in a drop of about 60 %
in the residual strength due to faster heating rate as compare to solder joints that
experienced slower heating rate during TMF. Residual shear strength in Sn-4Ag—O.5Cu
decreased signiﬁcantly with additional TMF cycles with faster heating rate imposed. Sn-
2.5Ag-O.5Cu-0.5Ni and Sn-2Ag-1Cu-1Ni showed minimal drop in simple shear strength
even after a maximum of 1000 TMF cycles for both slow and fast heating rates imposed.

Quaternary solder joints, Sn-2.5Ag-O.5Cu-O.5Ni exhibits minimal surface undulations

129

while Sn-2Ag-lCu-1Ni exhibits almost no damage even after 1000 TMF cycles for both
fast and slow heating rates. No signiﬁcant interface IMC layer gowth and particle
coarsening were observed even after 1000 TMF cycles in quaternary solder joints. Solder
joints made with ternary Sn-4Ag—O.5Cu exhibited the worst TMF residual strength among
Sn-Ag based solder joints under TMF with faster heating rate. Among Sn-Ag based
solder joints studied, ternary solder joints containing small amounts of Cu were highly
sensitive to change in heating rates during TMF with a long dwell time at the high
temperature.

The microstructure of eutectic Sn-Ag and Sn-4Ag-0.5Cu were similar in as-
fabricated conditions. However, sigiiﬁcant coarsening of (Cu-Sn) IMC particles were
found in Sn—4Ag-O.5Cu solder joints under TMF with a longer dwell time at the high
temperature extreme under faster heating rate imposed due to long accumulation aging
times. Sn-Sn gain boundary sliding was the major deformation features during high
temperature shear test in eutectic Sn-3.5Ag solder joints. It suggests that preventing the
Sn-Sn gain boundary sliding will result in minimum plastic strain during TMF under a
long dwell time at the high temperature extreme. Discrete (Cu-Ni-Sn) IMC precipitates
present at Sn-Sn gain boundaries were observed in Sn-2.5Ag-O.5Cu-O.5Ni and Sn-2Ag-
lCu-lNi. Such precipitates were not observed in the other alloys, so it is hypothesized
that these (Cu-Ni-Sn) precipitates are responsible for the improved TMF performance
especially at high temperature.

Alloying additions that may be responsible to observed TMF behaviors were
discussed. Under different TMF conditions, Sn-Ag based solder joints containing small

amounts of Cu and, Cu and Ni exhibited superior TMF behaviors due to presence of (Cu-

130

Ni-Sn) IMC precipitates in the solder joints. These precipitated keyed the Sn-Sn gain
boundaries and made sliding difﬁcult. In ternary alloys, the shapes of precipitates were
not as effective in keying gain boundaries. In order to verify the suggested hypotheses
regarding microsturctural issues, further systematic experimental study on TMF

deformation of the Sn-Ag based solder joints should be conducted.

131

APPENDEX

132

 

EVALUATION OF MATERIAL-RELATED PARAMETERS AFFECTING THE
SERVICE LIFE OF TIN-BASED SOLDER JOINTS EXPERIENCING
THERMOMECHANICAL FATIGUE

ABSTRACT

A simple parametric empirical approach to predict service life of Sn-based solder joints
can be developed. This paper evaluates the material-related parameters in such an
empirical equation, that affect the service life of Sn-based solder joints under realistic
thermal excursions encountered in service, by incorporating ﬁndings from studies related
to material and service issues. It also examines the roles of these parameters and their
relative importance by analyzing results from studies on Sn-based solder joints of
realistic dimensions under realistic thermomechanical fatigue (TMF) cycling, and

isothermal shear testing.

1. Introduction

Estimation of service life of solder joints is essential to predict the reliability of
electronic devices since the solder joints are the weakest link in electronic devices
[12,13,110,111]. The major cause of failure in electronic devices in service is TMF that
arises due to differences in coefﬁcient of thermal expansion (CTE) between the entities
present in the system [1 1 1-1 14].

There are no simple models available to quickly check the useful service life of a
solder joint under different service conditions. Such service variables can be dwell times
at temperature extremes, heating rate, cooling rate, temperature difference, frequency of
temperature cycles, etc. Each of these affects the useful service life of Sn-based solder

joints. Various material-related parameters have to be taken into account in such an

133

estimation since they will have signiﬁcant inﬂuences on the roles of these service
variables. Prior studies have indicated the importance of such issues in controlling the
useful service life of Sn-based solder joints [115]. Complexities involved in such an
analysis are comparable to those encountered in estimation of tool life in machining
processes. Tool life is affected by many variables and can be estimated using an
empirical equation suggested by Taylor [116]. Similarly a simple parametric empirical
approach can be used to evaluate the useful service life of Sn-based solder joints.

Useful service life prediction is different from predicting total lifetime of solder
joints. Total lifetime of solder joint may correspond to failure of solder joint that can not
carry any mechanical/electrical load. Methods suitable to evaluate the total lifetime
prediction could be Cofﬁn-Manson or Weibull distribution types of approaches. Weibull
distribution, especially will be more appropriate to estimate the total lifetime of solder
joints. In actual practice, one would like to remove the component out of service prior to
accelerated catastrophic events begin to pro gess. As a consequence, evaluation of useful
service life is more appropriate since solder joint is usually taken out of service before
rapid progession of the damage. Such an approach is very similar to that used in tool life
predictions, where one would like to change the tool before it begins to deteriorate
rapidly [116]. Thus, the useful service life prediction might meet the industrial
requirement rather than the total lifetime. Latter will have limited practical implications.

In this paper, actual evaluation of material-related parameters in an empirical
parametric equation has been carried out from assessment of TMF performance of Sn-
based solder joints. Results from realistic TMF tests with different ramp-rates and dwell

times at temperature extremes, and supporting isothermal shear tests were used to

134

evaluate these material-related parameters. Materialcrelated parameters evaluated from
extremely fast supplementary studies, such as isothermal shear testing and stress
relaxation, are compared with those from realistic TMF tests providing a possibility to

evaluate these material-related parameters in a rapid manner.

2. Experimental Details

Sn-Ag based solder pastes were sandwiched between two 0.5 mm thick half dog
bone shaped Cu strips and subsequently reﬂowed to produce single shear-lap solder joints
with ~ 100 pm thickness and 1mm2 solder joint area. The detailed conﬁgurations of the
dog-bone shaped single shear-lap solder joint and soldering procedure used have been
reported elsewhere [78,87,88]. Different TMF test conditions used, such as different
heating rates and dwell times at each temperature extreme, are listed in Tables IX] and
2. Further details can be found ﬁom earlier publications [99,117].

Isothermal shear tests using the same joint geometry were carried out with a ﬁxed
pre-strain of 0.1 imposed with different strain-rates (0.001/sec. and 0.0001/sec.) at 22°C
and 150°C. Peak shear stress and resultant residual shear stress after stress relaxation
were measured and used to evaluate some of the material-related parameters. Further
details of isothermal shear tests have been reported in prior publications [100,118].

3. Results and Discussion

Apparent relationship between service life of solder joint and residual properties
of solder joint should be inversely related to each other. Studies have shown that residual
mechanical strength or electrical conductivity resulting from TMF tends to follow a

logarithmic power-law type behavior, prior to stabilization followed by domination of

135

Table IX. 1. Details of TMF conditions with two different dwell times at each

temperature extreme to evaluate material-related parameter ‘r

15°C to 150°C) for all TMF studies.

’. Note: AT is 165°C (-

 

 

 

 

 

 

 

 

. . . Dwell time at . Dwell time at
Condrtrons Heating rate high T Cooling rate low T
Long dwell ~0.55°C/sec 20 min. ~0.07°C/sec 330 min.
time at low T
Long dwell ... o . ~ 0 .
. e at high T 0.55 C/sec 115 min. 0.07 C/sec 20 mm.

 

Table IX.2. Details of TMF conditions with two different heating rates to evaluate
material-related parameter ‘p ’. Note: AT is 165°C (-15°C to 150°C) for all TMF studies.

 

Dwell time at

Dwell time at

 

 

 

 

 

 

 

conditions Heating rate high T Cooling rate low T
:33 heatmg ~0.55°C/sec 115 min. ~0.07°C/sec 20 min,
:38; heating ~l .1°C/ sec 117.5 min. ~0.07°C/sec 20 nrin.

 

136

 

 

catastrophic events [82,99,117]. Since the catastrophic events progess rapidly,
appropriate useful service life should be estimated within the regime of the stabilized
behavior. As a ﬁrst approximation each contributing service parameter can be assumed
to inﬂuence the useful service life of solder joint in a power-law type relationship in this
service regime. Following empirical equation can be employed to evaluate the useful
service life of the solder joint:
Useful service life at [AT]"" [ﬂ'" [(dT/dt);, ]"’ [(dT/dt)c]*’ [AtH/AtJ [AM]‘

where, AT is imposed temperature difference, f is the frequency of thermal cycles,
(dT/dt);. is heating rate, (dT/dt)c is cooling rate, (Am/AIL) is the ratio of dwell times at
high and low temperature extremes, and (AM) is any other material parameter such as
microstructure changes during TMF. m, n, p, q, r, and s are material-related parameters
that can be determined from experimental results. The constant of proportionality can
account for the inﬂuences of processing parameters and the joint geometry

In this paper, (A M)’ is assumed to remain constant since no sigriﬁcant surface
microstructural changes were noted to result from TMF prior to catastrophic damage
accumulation. (AM)’ will probably play signiﬁcant roles in later stages of TMF when the
relative inﬂuences of these material-related parameters on each other will become very
dominant. Since, this paper deals with useful service life of the solder joint rather than
total life, the role of (AM)’ is not included in the discussion. Solder joints are usually
taken out of service before the damage progesses in such a catastrophic manner, just as
tools that are taken out of service prior to rapid wear that occurs in the ﬁnal stages of tool
life. Appropriate material—related parameter affecting the role of frequency of thermal

cycles, 72, could not be evaluated at this time due to lack of available data.

137

As seen in Figure. IX.1, the lifetime or any other reliability related issues of a
solder joint, could be depicted in three stages (internal damage build-up, surface
manifestation of damage with no signiﬁcant property changes, and progess of
catastrophic failure). Such ﬁndings are very similar to those encountered in the life of a
tool [116]. Tools exhibit rapid initial wear, followed by gadual wear and accelerated
wear during service. Since machining by tools undergoing accelerated wear will be of
poor quality, and the life time of the tools when machining under such conditions is
highly unpredictable, they are taken out of service in the regime of gadual wear after
undergoing certain extent of wear. So tool life prediction is based useful service life in
the gadual wear regime.

Weibull distribution may be a better method to predict total lifetime of solder
joints [119]. Although parametric empirical approach proposed in this study is simple, it
will be more appropriate to predict useful service life of solder joint in a relatively quick
manner. Thus, it provides an estimation of useful service life, rather than total lifetime,
of Sn-based solder joints is the main focus of this paper.

In order to evaluate a given material-related parameter, certain conditions need to
be met: (i) maintain process, and joint geometry parameters ﬁxed, (ii) vary only the
service parameter inﬂuenced by the chosen material-related parameter, and (iii) hold all
other service parameters ﬁxed.

Method to evaluate a given material-related parameter from actual TMF results
consisted of the following steps:

(i) take ratio of percent decrease in residual properties for two different

conditions,

138

 

 

 

 

 

 

 

8 Internal damage build-up
E
Q)
8“
a Surface manifestation of damage
'23 with stabilized residual properties
T52 :‘\
8 : \ Progess of
m : \atastrophic failure
l \
: \
. \
i x
I I
l l
I Useful Service life i E
Total life time ' '
I v.
Number of TMF cycles

Figure D(.1. A schematic of residual properties of solder joints under TMF conditions
classiﬁed into three stages. Solid line represents region of useful life of solder joint. The
dot line represents the range in which catastrophic failure takes place and is not addressed

in this study.

139

(ii) use this ratio to represent ratio of damage accumulation under these _
conditions, and evaluate the approximate material-related parameter using

these two conditions.

Evaluation at ‘r’ in [AtgAtd and ‘ ’ in (”7:11 tram Actual TMF Results

Material-related parameters r in [AtH/AILY and p in [(dT/dt);. ]"’ were evaluated by
using results from actual TMF tests on solder joints made with four different Sn—Ag
based solder alloys.

As mentioned in the earlier section (see Table IX.1), dwell times at each
temperature extreme were the only varied service parameter in this case. Figure D(.2
provides the values of parameter ‘r ’ in [AtH/AtIJ' evaluated four Sn—Ag based solder alloy
joints. The exponent ‘r’ ranged from 0.11 to 0.21 depending on the solder alloy. Solder
joints made with quaternary alloys were more sensitive to dwell time changes than the
binary and ternary solders. Prior studies [99,117] have indicated that quaternary solder
alloy joints containing small amounts of Cu and Ni possess better TMF performance
under long dwell times at high temperature extreme [82].

With two different heating rates during the TMF cycles (see Table VHII.2),
material-related parameter ‘p ’ in [(dT/dthl'p can be evaluated. As can be seen in Figure
V1113, value of parameter ‘p ’ is larger than ‘r’ for all the four Sn—Ag based solder alloys
studied. Such a ﬁnding indicates that varying heating rate could have more inﬂuence on
the useful service life of Sn-Ag based solder joints than changes in dwell times at
temperature extremes. This data also indicates that useful service life of quaternary

alloys are less sensitive to heating rate than eutectic Sn-Ag.

140

 

0.3 r l l

   
    

025 — ...................................................................... Sn—ZAg —
3. -1Cu-1Nr
3
3 0.2 ~
g ; _ Sn2.5Ag;
is; 0,15 —Eutectic ..... ---------------------- #0— SCH-0.51“
E “Ag Sin-4Ag-0.5C:u
3 0.1 i 5
N
>

0.05

0

 

 

Figure D(.2. Material-related parameter ‘r' in [AtH/AtL]r for Sn-Ag based single shear-lap
solder joints.

141

 

    
 
   

2 l 1 I
In 1.5 _ ...................................................................................... _
.33 2 a
“é Sir-4Ag-0.5Cir
a 2 i i
g 1 ~.,Eutectic....§.. ..i...S.‘.l..'..2.'.SA.gi.; .................. _
“5 Sn- Ag ‘ 3-0.5Cu-0.5N1
E I Sn-2Ag
§ 0.5 ..E._e.1.Cu.-.1Ni.2

 

 

0

Figure IX.3. Material-related parameter 'p ’ in [(dT/dt)h]"’ for Sn—Ag based single shear-
lap solder joints.

142

Evaluation at Material-Related Parameters from Sugeorting Studies (ﬁom Isothermal
§h ear Tests)

In isothermal shear tests, peak shear stress can be assumed to represent damage
accumulation for a given strain. Similarly, the resultant residual shear stress aﬁer stress
relaxation can also be assumed to represent damage accumulation resulting from an
imposed pre-strain.

In order to check the effect of heating rates, supplementary data used was the peak
shear stresses at 150°C with two different imposed strain-rates of 0.001/sec. and
0.0001/see, while maintaining the pre-strain (0.1) ﬁxed. Material-related parameter ‘p ’
in [(clT/dt);.]1° was evaluated by using both (i) ﬁom results of isothermal shear tests, and
(ii), residual strength data of specimens that underwent actual TMF tests, as can be seen
in Figure IX.4. Values of ‘p ’ obtained for eutectic Sn-Ag solder joints ﬁom both these
approaches are fairly close. Such a ﬁnding suggests that material-related parameter ‘p ’
could be evaluated from isothermal shear tests with different imposed strain-rates.

However, a scheme to evaluate the effects of ratio of dwell times at high and low
temperature extremes by using isothermal tests is not devised at this stage.

Similarly, effects of cooling rate can be evaluated from results of isothermal shear
tests. Such an evaluation used the peak shear stresses, for two different strain-rates at
lower temperature extreme. Again, the pre-strain (0.1) was ﬁxed for this computation.
As can be seen in Figure D(.5, varying cooling rates has almost the same effect as
varying heating rates on useful service life of eutectic Sn—Ag solder joints.

One of the important ﬁndings from evaluation of parameters from isothermal

shear tests was the ratio representing different strain-rate conditions for ﬁxed pre-strain,

143

 

 

    

 

 

2 I
3:. E
v: E
§ 1.5 rrrrrrrrrrrrrrr ,, ,, , Eﬂectof ............ I
E heating rate
g Effect of from
& heating rate isothermal
.3 1 _ """"" f romaetual """"" """"""" 'S'héaf't'e‘s‘t """"""" —
cu TMF
H
cc
2
9-1
° 0.5 ~ -
Q
E
as
>
0 A

a '1’ (IT -P

( dt h (Eh

Figure IX.4. Comparison of material-related parameter ‘p ' in [(dT/dt);.]'p for eutectic Sn—
Ag solder joints evaluated from both actual TMF test and isothermal shear test.

144

 

I

I

Value of Material Parameters, 'p' & q

0.5 r"

 

1;
dt h

< )"’

E ﬂeet of
1-5 ' heating rate
from
isothermal
1 ............ shear test . ,.

  

Effect of

; cooling rate
""" ' "from
' isothermal

shear test

Figure IX.5. Material-related parameter ‘p ' in [(dT/dt)h]"’ and ‘q ' in [(dT/dOJ" for
eutectic Sn—Ag solder joints from isothermal shear test.

145

 

were about the same irrespective of whether it was calculated using peak shear stress or

the stabilized stress after relaxation.

Wtion of ‘m ’ in (4124' Mublished Work of Other Investigators

Results from two TMF studies, published by other researchers, were used to
evaluate the effect of temperature difference during TMF on Sn-Ag based solder joints
for this part of analysis, since temperature difference was not varied in our TMF studies
[45,46]. In each of these published studies, service variables, such as heating rates,
cooling rates, and dwell times at each temperature extreme were ﬁxed. Details of joint
geometry, material used, temperature differences employed by these investigators are
provided in Table IX.3. The material-related parameter ‘m ’ in (AT)'"' obtained from this
analysis is much geater than other material-related parameters as can be seen in Figures.
R6 and IX.7. It is expected to be large since thermal stress causing the damage will be
directly proportionally to AT. As expected different solder materials used to fabricate
solder joints affect material-related parameter ‘m ’.

Most well-known approach to predict the total lifetime of solder joints under TMF
is Cofﬁn-Manson equation. However, this equation deals with strain range, especially
plastic range within single cycle under isothermal condition, and it might overestimate
the actual lifetime of solder joints [3,119]. This model does not account for the effects of
time and temperature which are important factors for materials stressed at above one half
of the melting point in degees absolute [120]. In order to consider the effects of

temperature and frequency, modiﬁed Cofﬁn-Manson (Norris-Landsberg) equation has

146

 

Table IX.3. Details of joint geometry and different imposed temperature difference used
to evaluate material-related parameter ‘m ’ from published literature [45,46].

 

 

 

 

 

 

 

 

 

 

  
 
      

Joint .
Ref. geometry AT] AT; Materials used
Kari a e t Quad Flat Sn-Ag,
3] [f1 5] Pack Lead 160°C (-30°~130°C) 100°C (0°~100°C) Sn—Ag-0.5Cu,
' (QFP) Sn-Ag-lCu
288268; SMT 150°C (-25°~125°C) 110°C (-25°~85°C) Sn-Ag
10 1 I
s 8 _ ...................................................................................... _
.5
in-
d) .
H . .
° 6 __ ................... 5. ............................. 3, ........................... _
§ QFP % a
3 Eutectic Sn-Agi ;
Q. E g
g 4 L— . - E, .................... L .................. .—
Q ; QFP g QFP
.3 Sn-Ag—0.5Cu Sn—Ag—l Cu
> 3 i
2 ,_ .......... _

  

 

 

Figure IX.6. Effect of solder alloy on the material-related parameter ‘m ’ in [AT] ‘"’.
Note: Joint geometry was the same for all these conditions.

147

 

 

 

Eutectic Sn-Ag

     

"' """"""""" QFP """""""" _
‘ Eutectic Sn-Ag

Value of parameter 'm '

 

 

Figure IX.7. Effects of solder joint geometry on material-related parameter ‘m ' in [AU'

148

been used to predict total lifetime of solder joints [120]. Studies carried out by Shohji et
al.[l 19] found that the material-related parameter ‘m ’ in (AT)"" was about 1.9 after
evaluating TMF lives of Sn-37Pb solder joints under various TMF conditions using
modiﬁed Cofﬁn-Manson equation (Nf = C - F" - (A7)” exp (Q/RTmax), where Nf: thermal
fatigue life, C is a constant, F is frequency, AT is temperature range, Q is activation
energy, R is gas constant, and T max is maximum temperature). This study reported that
the inﬂuence of (A7) on lifetime of solder joints were more sigriﬁcant than other service
variables. However, they also reported that the TMF life of solder joints does not follow
the modiﬁed Cofﬁn-Manson equation under certain conditions for total lifetime
predictions. The reason, why material-related parameter ‘m ’ between results of the
present study and modiﬁed Cofﬁn-Manson equation is different may be attributed to the
difference between useful service life and total lifetime. Further studies are needed to
clarify this issue.

Joint geometry affects the estimated values of material-related parameter ‘m’, as
can be seen in Figure D(.7. Proportionality constant in the proposed parametric equation
that can account for the effect of joint geometry can be evaluated provided results from
identical TMF tests carried out on solder joints with different geometries are compared.
At present such a data is not available. Unluckily, the two published works used to
calculate the parameter ‘m’ could not be used for this purpose since the heating rates,
cooling rates, dwell times and temperature differences imposed are different in these two
studies.

The purpose of the present study is to develop a simple method to predict useful

service life of solder joints under various service conditions. The simple parametric

149

approach proposed in this paper might be one of the possible effective methods to quickly
obtain the useful service life of solder joints, and might suit the needs of industrial

desigrers.

4. Summary

Material-related parameters in a parametric empirical equation to check
reliability/useful service life of solder joint were carried out. By using this approach one
can judge the effectiveness of the various material modiﬁcation approaches, such as
alloying and use of reinforcements, to withstand the effects of various service parameters
to be encountered. Material-related parameters evaluated from extremely fast
supplementary studies, such as isothermal shear testing and stress relaxation, are about
the same as those obtained with time consuming TMF studies carried out with realistic

service conditions.

150

LIST OF REFERENCES

151

10.

ll.

12.

l3.

14.

15.

16.

17.

18.

19.

20.

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).

http://ko.home.ag°lent.com/upload/cmc upload/All/Oresio Ling pbfreetest 082004.
pif

Environmental Protection Agency, National Air Quality and Emission Trend Report.
1989. EPA-450/4-9l-003. Research Triangle Park. NC. 1991

W.J. Plumbridge,Soldering and Surface Mount Technology, 12 (1), 32 (2000)
D. Napp, Sampe Journal, 32 (2), 59 (1996)

CL. Hernandez, P.T. Vianco, and J.A. Rejent, IPC/SMTA electronics Assembly
Expo 1998, S-19-2-1(1998).

C.M.L Wu and ML. Huang, Journal of Electronic Materials, 31 (5), 442 (2002)
DR. Frear, J .W. Jang, J .K. Lin, and C.Zhang, J OM, 53 (6), 28 (2001)

S.K. Kang and T.G. F erence, Journal of Materials Research, 8 (5), 1033 (1993)
H. Takao and H. Hasegawa, Journal of Electronic Materials, 30 (9), 1060 (2001)

CS. Oh, J .H., Shim, 8.]. Lee, and D.N. Lee, Journal of Alloys And Compounds, 238
(1-2), 155 (1996)

2D. Xia, Z.G. Chen, Y.W. Shi, N. Mu, and N. Sun, Journal of Electronic Materials,
31 (6), 564 (2002)

R.N. Wild, Welding Research, 521-8 (1972).

H.J. Frost and RT. Howard, Transactions on Components, Hybrids, and
Manufacturing Technology, 13 (4), 727 (1990).

W.F. Feng, C.Q. Wang, and M. Morinaga, Journal of Electronic Materials, 31 (3),
185(2002)

L.J. Matienzo and RR. Schaffer, Journal of Materials Science, 26 (3), 787 (1991).

152

21

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

S. Russell, A. Grigals, and E. Clemente, Journal of Electronic Materials, 28 (9), 1008
(1999)

http://www.necel.com/pkg/en/pk 02.html.

B.T.K. Barry and C.J. Thwaites, “Tin and Its Alloys and Compounds”, John Wiley
& Sons: New York, 1985; pp. 16

W. Steurer, R. Ferro and A. Saccone, Physical Metallurgy 4th edition, Vol. 1, editors,
R. W. Calm, and P. Haasen, Elsevier Science B.V., Amsterdam, The Netherlands, 25
& 284 (1996).

M.J. Sullivan and SJ. Kilpatick, “Degadation Phenomena”, Puttliz/Handbook of
Lead (Pb) - Free Technology for Microelectronic Assembly, Chapter 22, pp. 926-
927, (2004).

Y. Kariya, C.R. Gagg, and W.J. Plumbridge, “Tin pest in solder alloys”, Soldering
and Surface Mount Technology, 13, pp. 39-40 (2001).

W.J. Plumbridge, “Long term mechanical reliability with lead-free solders”,
Soldering and Surface Mount Technology, 16 (2), pp. 13-20 (2004).

N. E. Dowling, “Mechanical Behavior of Materials ”, Ed by D. Humphrey, Prentice
Hall, Upper Saddle River, New Jersey, pp. 339 (1993)

M.A. Meyers and K.K. Chawla, Mechanical Metallurgy principles and applications,
Prentice-Hall Inc., Englewood Cliffs, NJ, 692(1984).

N. E. Dowling, “Mechanical Behavior of Materials ", Ed by D. Humphrey, Prentice
Hall, Upper Saddle River, New Jersey, pp. 625 (1993).

H.D. Solomon, J. Electr. Packaging, 113 (June), 102(1991).

C. Kanchanomai C, Y. Miyashita, and Y. Mutoh, J. of Electron Mater, 31 (2), 142
(2002).

A. Dasgupta, Handbook of Electronic Packaging, Edited by Michael Pecht, 1991,
p.495.

C.H. Raeder, R.W. Messler Jr. and LP. Cofﬁn. Jr., J. of Electron Mater, 28 (9), 1045
(1999) '

D.A. Miller, R.H. Priest, and E.G. Ellison, High temperature Materials and
Processing, 6 (3&4), 155 (1984).

153

36.

37.

38.

39.

40.

41.

42.

43.

45.

46.

47.

48.

49.

50.

51.

T.Y. Pan, IEEE Transactions on Components, Hybrids and Manufacturing
Technology, 14(4), 824 (1991).

EC. Kubik and T.P.L. Li, “Thermal shock and Temperature Cycling Effects on
Solder Joints of Hermetic Chip Carries Mounted on Cu Thick F ilrn”, Matin Marietta
Aerospace Internal Rep., Microelectronic Center, Orendo, F 1., 32855

W.J. Plumbridge and Y. Kariya, Fatigue Fract Engng Mater Struct, 27, 723 (2004)

DR. Frear, The Mechanics of Solder Alloy Interconnects, Ed by H.S. Morgan and
SN. Burchett, 1993, p. 60-61.

D.A. Miller, R.H. Priest, and E.G. Ellison, High temperature Materials and
Processing, 6 (3&4), 157-158 (1984).

M.J. Sullivan and SJ. Kilpatick, “Degadation Phenomena”, Puttliz/Handbook of
Lead (Pb) - Free Technology for Microelectronic Assembly, Chapter 22, pp. 928,
(2004).

J .D. Lubahn and RP. F elgar, “Plasticity and Creep of Metals”, John Wiley & Sons,
New York, 1961, pp. 217 -220

S. Choi, K.N. Subramanian, J .P. Lucas, and TR. Bieler, Journal of Electronic.
Materials. 29, 1249 (2000).

S. Choi, J .G. Lee, K.N. Subramanian, J.P. Lucas, and TR. Bieler, Journal of
Electronic Materials, 31 (4), 292 (2002).

Y. Kariya, Y. Hirata, and M. Ostuka, J. Electron. Mater, 28 (l 1), 1263 (1999).

N. M. Poon, C.M. Lawarence Wu, Joseph K. L. Lai, and Y. C. Chan, IEEE
Transactions on Advanced Packaging, 23 (4), p. 708 (2000)

http://www.kester.com/en-us/leadfree/requirements.aspx

J. Glazer, J. of Electron. Mater, 23 (8), 693 (1994)

P. Viswanadham, S. O. Dunford, and J. Kivilahti, “Reliability Aspects of Lead-Free
Solders in Ekectrorric Assemblies”, Puttliz/Handbook of Lead (Pb) — Free
Technology for Microelectronic Assembly, Chapter 19, pp. 788, (2004).

C. Mackay and W. von Voss, Mater. Sci. Technol. 1 (3), 240 (1985).

M.E. Loomans and M.E. Fine, Metal]. Mater. Trans. A, Phys. Metall. Mater.Sci. 31A,
1155(2000)

154

52.

53.

54.

55.

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

66.

67.

SH. Huh, K.S. Kim, and K. Suganuma, Mater. Trans. (Japan) 42, 739 (2001).
K.S. Kim, SH. Huh, and K. Suganuma, Mater. Sci. Eng. A 333, 106 (2002).

AZ. Miric and A. Grusd, Soldering, Soldering & Surface Mount Technology, 10(1),
19 (1998)

F. Guo, S. Choi, K.N. Subramanian, T.R. Bieler, J .P. Lucas, A. Achari, and M.
Paruchuri, Materials Science and Engineering A, 351 (1-2), 190 (2003)

J .C. Suhling, H.S. Gale, R. W. Johnson, M. N. Islam, T. Shete, P. Lall, M.J. Bozack,
J. L. Evans, P. Seto, T. Gupta, J. R. Thompson, Soldering and Surface Mount
Technology, 16(2), 77 (2004)

J .S. Hwang, Soldering and Surface Mount Technology, 14 (8), 20 (2000)

C. Kanchanomai, Y. Miyashita, and Y. Mutoh, J. of Electron Mater, 31 (5), 456
(2002)

Y. Kariya, T. Hosoi, T. Kimura, S. Terashima, and M. Tanaka, Materials
Transactions, 45(3), 689 (2004)

M.A. Meyers and K.K. Chawla, Mechanical Metallurgy principles and applications,
Prentice-Hall Inc., Englewood Cliffs, NJ, 402 (1984).

R. M. Brick, A.W. Pense, and R. B. Gordon, “Structure and Properties of
Engineering Materials ”, 4th edition, Ed by M. B. Bever, McGraw-Hill, New York,
NY, pp. 93 (1983).

H. Mavoori and S. Jin, J. of Electron Mater, 27 (11), 1216 (1998).

M.J. Sullivan and SJ. Kilpatick, “Degadation Phenomena”, Puttliz/Handbook of
Lead (Pb) — Free Technology for Microelectronic Assembly, Chapter 22, pp. 922,
(2004)

K.N. Subramanian, T.R. Bieler and J .P. Lucas, J. Electron. Mater. 28 (1 l), 1176
(1999)

S. Choi, K.N. Subramanian, J .P. Lucas, and T.R. Bieler, J. Electron. Mater. 29, 1249
(2000)

K. N. Subramanian, A. Lee, S. Choi, and P. Sonje, J. Electron. Mater. 30(4), 372
(2001)

A.W. Gibson, S. Choi, T.R. Bieler, and K.N. Subramanian, IEEE International
Symposium on Electronics and Environment, p. 246 (1997).

155

68.

69.

70.

71.

73.

74.

75.

76.

77.

78.

79.

80.

81

83

 

68

69

70.

71.

72.

73.

74.

75.

76.

77.

78.

79.

80.

81.

82.

83.

Y. Kariya and M. Otsuka, J. Electron. Mater., 27 (7), 866 (1998)

M. McCormack and S. J in, 1994 IEEE/CPM T 16” International Electronics
Manufacturing Technology Symposium, 1. The Institute and Electrical Electrons
Engineers Piscataway, NJ, p.7 (1994a)

M. McCormack and S. Jin, J. Electron. Mater., 23 (7), 595 (1994b)

A. Dasgupta, P. Sharma, and K. Upadhyayui, International J oru'nal of Damage
Mechanics, 10, 101 (2001)

Y.H. Pao, S. Badgley, R. Govila, and E. Jih, Fatigue of Electronic Materials, Edited
by S.A. Schroeder et. al., Publication of American Society for Testing and Materials,
60-81 (1994).

H.L.J. Pang, K.H. Tan, X.Q. Shi, and Z.P. Wang, Materials Science and
Engineering, A307, 42 (2001).

W.W. Lee, L.T. Nguyen, and GS. Selvaduray, Microelectronic Reliability, 40, 231
(2000)

s. Ahat and M. Sheng, and L. Luo, J. Mater. Res, 16 (10), p. 2914 (2001).

S. Choi, T.R. Bieler, J .P. Lucas, and K.N. Subramanian, J. Electron. Mater. 28 (11),
1209 (1999).

A. W. Gibson, S. Choi, K. N. Subramanian, and T. R. Bieler, Reliability of Solders
and Solder joints, R. K. Mahidhara et. al., eds., TMS Warrendale, PA, p.97 ( 1997).

J .P. Lucas, F. Guo, J. Mcdougall, T. R. Bieler, K. N. Subramanian, and J. K. Park, J.
Electron Mater, 22, 1268 (1999).

F. Guo, J. Lee, J. P. Lucas, K. N. Subramanian, and T. R. Bieler, J. Electron. Mater.
30(9), 1222 (2001).

R. Darveaux and K. Banerji, 1992 42"° Electronic Components and Technology
Conference (ECTC), IEEE, Piscataway, NJ, p. 538 (1992).

D.P.H. Hassehnan, J. Am. Cerarn. Soc., 52 (11), 600 (1969).

J .G. Lee, F. Guo, S. Choi, K.N. Subramanian, T.R. Bieler, and J .P. Lucas, J of
Electronic Materials, 31 (9), 946 (2002).

H. Yang, P. Deane, P. Magill, and K. L. Murty, IEEE Electronic Component and
Technology Conference (ECTC), 1136 (1996).

156

84

85.

86.

87.

88.

89.

90.

91.

92.

93.

94.

95.

96.

97.

98.

99

. H. L. Reynolds, S. H. Kang and J. W. Morris, Jr., J of Electronic Materials, 28(1),
69 (1999).

S. Choi, J. Lee, F. Guo, T. R. Bieler, K. N. Subramanian and J. P. Lucas, JOM, 53
(6), 22 (2001).

S. G. J adhav, K. N. Subramanian, T. R. Bieler and J. P. Lucas, J of Electronic
Materials, 30(9), 1197 (2001).

F. Guo, J. Lee, S. Choi, J .P. Lucas, T.R. Bieler, and K.N. Subramanian, J of
Electronic Materials, 30(9), 1073 (2001)

F. Guo, S. Choi, J. P. Lucas and K. N. Subramanian, Soldering and Surface Mount
Technology, 13 (l), 7 (2001).

J. Howell, A. Telang, J .G. Lee, and K.N. Subramanian, J of Materials Science:
Electronic Materials, 13, 315 (2002).

KC. Chen, A.U. Telang, J .G. Lee, and K.N. Subramanian, J of Electronic Materials
31 (11), 1118 (2002).

J .G. Lee, A.U. Telang, K.N. Subramanian, and T.R. Bieler, J of Electronic Materials
31 (11), 1152 (2002)

W.J. Plumbridge, Soldering & Surface Mount Technology, 11 (3), 8 (1999).

M. F arooq, C. Goldsmith C, R. Jackson, and G. Martin, J. of Electron Mater, 32 (12),
1421 (2004). '

J. Lau, W. Dauksher, J. Smetana, R. Horsley, D. Shangguan, T. Castello, I. Menis,
D. Love, and B. Sullivan, Soldering & Surface Mount Technology, 16 (l), 12 (2004).

G. Di Giacomo and S. Oggioni, Microelectronics Reliability, 42 (9-11), 1541 (2002)

C.J. Zhai, Sidharth, and R. Blish, IEEE Transactions on Device and Materials
Reliability, 3 (4), 207 (2003).

T. Goswarni and H. Hanninen, Materials & Design, 22 (3), 199 (2001).

B2. Hong and LG. Burrell, IEEE Transactions on Components Packaging and
Manufacturing Technology Part A, 20 (3), 280 (1997).

. J .G. Lee and K.N. Subramanian, J. of Electron Mater, 32 (6), 523 (2003).

100.H. Rhee, K.N. Subramanian, A. Lee, and J .G. Lee, Soldering &Surface Mount

Technology, 15 (3), pp. 21 (2003).

157

101.J.G. Lee and K.N. Suramanian, Soldering and Surface Mount technology (in press)
102.H. Rhee, private communication (2004).

103.P.T. Vianco (Proceeding of Surface Mount International Conference, San Jose, CA,
Surface Mount International, Edina, MN 55424, USA), 2, (1993), p. 725

1041.8. Hwang and Z.F. Guo (Proceedings of Surface Mount International Conference,
San Jose, CA), Pub. Surface Mount International, Edina, MN 55424, USA, 1.2,
(1993) p 732.

105.Guo F., Choi 8., Lucas J .P., and Subramanian K.N., J of Electron Mater, 29 (10),
1241(2002)

106.C. M. Chuang, P.C. Shin, and KL. Lin, J of Electron Mater, 33 (1), 1 (2004).
107.F.J. Humphreys and RN. Kalu, Acta Metall, 35, 2815 (1987).

108.W.F. Smith in Structure and Properties of Engineering Alloys, 2"° edition, ED by R.
Gibala, M. Tirrell, and C. Wert, McGraw-Hill, New York, NY, (1993) p.498-509.

109.W.B. Hampshire, Soldering and Surface Mount Technology, 5 (14), 49 (1993).

110.F. Hua and J. Glazer, “Lead-free solders for Electronic Assembly”, in R.K.
Mahidhara et. al. (Eds), Design & Reliability of Solder and Solder Interconnects,
The Minerals, Metals, & Materials Society (TMS), Warrendale, PA, 65 (1997).

111.Z. Xia, Y. Shi, and Z. Chen, Journal of Materials Engineering And Performance, 11
(1) 107 (2002).

112.C. Basaran and R. Chandaroy, Journal of Electronic Packaging, 124 (1), 60 (2002).

113.L.J. Zhang, L. Wang, X.M. Xie, and W. Kempe, IEEE Transactions on Electronics
Packaging Manufacturing, 25 (4), 284 (2002).

114.P.T. Vianco, J .A. Rejent, and AC. Kilgo, J. of Electron Mater, 32 (3), 142 (2003).
115.K.N. Subramanian, Journal of Electronic Materials (in press).

116.Pollack and W. Herman “Tool Life”, Englewood Cliffs, New Jersey: Prentice Hall,
1988.

117.J.G. Lee and K.N. Subramanian, “Effects of TMF Ramp Rates on Damage

Accumulation and Resultant Mechanical Behavior of Sn-Ag Based Solder Joints”,
will be submitted to Journal of Electronic Materials.

158

118.H. Rhee and K.N. Subramanian, Journal of Electronic Materials, 32 (11),
13 10(2003).

119.X. J. Zhao, G.Q. Zhang, J .F .J .M. Caers, and L]. Ernst, Journal of Electronic
Packaging, 125(4), 582 (2003).

120.M.J. Sullivan and SJ. Kilpatrick, “Degation Phenomena”, Puttliz/Handbook of Lead
(Pb) — Free Technology for Microelectronic Assembly, Chapter 22, 967 (2004).

121.1. Shohji, H. Mori, and Y. Orii, Microelectronics Reliability, 44, 269 (2004).

159

 

  

Il]llli]]i]ll]]ljlj]l11]]!le

1