Professional Documents
Culture Documents
MOHAMMAD FAIZAN
A Dissertation
Presented to
In Partial Fulfillment
Doctor of Philosophy
Mohammad Faizan
December, 2007
EXPERIMENTAL STUDY AND MODELING OF METAL DISSOLUTION AND
Mohammad Faizan
Dissertation
Approved: Accepted:
______________________________ ______________________________
Advisor Department Chair
Dr. Guo-Xiang Wang Dr. Celal Batur
______________________________ ______________________________
Committee Member Dean of the College
Dr. T. S. Srivatsan Dr. George K. Haritos
______________________________ ______________________________
Committee Member Dean of the Graduate School
Dr. Alex Povitsky Dr. George R. Newkome
______________________________ ______________________________
Committee Member Date
Dr. Ernian Pan
______________________________
Committee Member
Dr. Thein Kyu
ii
ABSTRACT
packaging industry for joining electronic components. The industry is aiming for the use of
environment friendly lead-free solders. All the lead-free solders are high tin-containing
alloys. During the soldering process, an intense interaction of metallization on PCB and tin
from the solder occurs at the joint interface. Intermetallic compound (IMC) is formed at the
interface and subsequently PCB bond-metal (substrate) is dissolved into the molten solder.
The formation of the IMC is desirable for the joint to occur but excessive metal dissolution
and uncontrolled growth of the IMC is detrimental to quality of joint. In the present study the
terms bond-metal and substrate will be used interchangeably and the term ‘substrate’ refers to
the top layer of the PCB which comes in contact with the molten solder during soldering
reaction. During the wet phase of soldering process, the IMC exhibits scalloped
morphology. The growth of scalloped IMC during the solder/substrate interaction entails
Understanding of the physics and the kinetics involved in the formation of IMC phase
and metal dissolution is important in achieving the desired control of the process.
This study presents an in-depth analysis of the physics involved in the formation
of the scalloped intermetallic phase. To substantiate the physics of the process and to
verify the kinetics involved, a mathematical model is setup. The actual physics involved
iii
in the dissolution of metal layer and growth of IMC during soldering process is translated
into mathematical model based on the fundamental equations and known material
properties and thermodynamic constraints. The model was applied to study the
properties of material. The model served as a valuable tool to closely understand the
physics involved in the process of diffusion of various species across the joint, phase
change and kinetics effects during soldering process. Extensive parametric study was
conducted to explore the relative effect of various parameters like interface reaction
kinetics and physical properties such as the diffusivity of solder and IMC. The effect of
these parameters on the dissolution behavior of metallization layer and the growth of
The results from the model were also compared with experimental data for IMC
processing conditions were also considered in the model. A constant value of IMC
diffusivity was able to accurately predict the IMC growth for the cases involving planar
IMC layer, however, the single value of IMC diffusivity did not work very well for
scalloped IMC layer. Due to the complicated evolution of the scalloped IMC growth in
wet phase of soldering, it was suggested that the diffusivity in the IMC phase changes
with time and a varying value of IMC diffusivity be used in the model. The varying IMC
diffusivity gave satisfactory predictions for the IMC thickness and metallization layer lost
during reflow process. The model was also applied to study the dissolution behavior of
the micro-size copper particles in lead-free composite solders during reflow process.
iv
DEDICATION
Dedicated to my
Parents
and
Family
v
ACKNOWLEDGEMENTS
All praise to Allah (swt) who taught the mankind what they did not know. My
prayers for my deceased parents who gave countless sacrifice and did every effort in
reach to nurture me and provided the highest moral values. My sincere regards to my
advisor in the present work, Dr. Guo-Xiang Wang who always helped and supported me
in the time when I needed him. I am also thankful to the committee members for their
valuable suggestions in bringing out the present work. I feel obliged to express my
gratitude towards my brother, Dr. Shakir Husain who launched my academic career while
I was just ten year old. His continuous support, encouragement and valuable advice on
every occasion of my life helped me in achieving high goals in life. This letter would
remain incomplete if I did not mention the support, sacrifice and affection of my wife and
two lovely daughters, Zahra and Maria. Their presence always gave me strength and
Mohammad Faizan
vi
TABLE OF CONTENTS
Page
LIST OF FIGURES...........................................................................................................xii
CHAPTER
I. INTRODUCTION............................................................................................................1
vii
2.5.2 Need to Conserve Metallization Layer.............................................22
viii
3.6 Size Distribution of IMC Scallops...................................................................55
ix
6.2.3 Effect of IMC Diffusivity.................................................................97
7.3 Model Predictions for IMC Thickness and Copper Dissolution for
Scalloped IMC Growth..................................................................................110
x
9.2 Application of Numerical Model...................................................................149
BIBLIOGRAPHY............................................................................................................152
APPENDICES………………………………………………………………………….158
xi
LIST OF FIGURES
Figure Page
1.1 Micrographs showing the section view and top view of the IMC...……..……......4
2.6 Schematic representation of the wetting angle and force balance at the
IMC/solder interface……………………………………………………………..29
4.4 Optical micrograph showing the IMC layer and its trace used for the
measurement of average IMC thickness…………………………………………64
xii
5.1 Schematic and micrographs showing various stages of formation and
growth of IMC during soldering…………………………………........................74
5.2 Schematic showing various regions and interfaces inside a typical solder joint...75
6.1 Comparison of analytical and numerical solutions to show the effect of finite
domain on interface position and interface velocity………………………….….86
6.5 Plots showing effect of interface kinetics on solute distribution in the solder…..92
6.9 Plots showing effect of solder diffusivity on solute distribution in the solder…..96
xiii
6.15 Plots showing copper thickness consumption with time at various values
of IMC diffusivity................................................................................................103
6.16 Plots showing growth of IMC thickness with time for various values
of IMC diffusivity................................................................................................104
7.4 Plots showing comparison of numerical prediction with experimental data for
growth of intermetallic phase in the Pd-Sn binary couple...................................110
7.8 Plot showing variation of IMC diffusivity with time used in present
calculations..........................................................................................................114
7.9 Plots showing comparison of numerical prediction with experimental data for
average thickness of scalloped IMC....................................................................115
7.10 Plots showing comparison of numerical prediction with experimental data for
consumed thickness of copper.............................................................................115
7.11 Plots showing comparison of numerical predictions with experimental data for
consumed thickness of copper after adding 1.5 µm to numerically predicted
values...................................................................................................................118
7.12 Comparison of numerical results with experimental data for average IMC
thickness and consumed copper thickness...........................................................119
xiv
8.2 Plots showing effect of inter-particle distance on particle dissolution in
Sn-Ag-0.9Cu solder alloy………………………………………........................125
8.5 Plots showing effect of initial particle size on IMC thickness during
dissolution of copper particles in Sn-3.5Ag-0.9Cu solder alloy………..............132
8.6 Plots showing effect of initial copper content in SAC solder on IMC
thickness during dissolution of copper particles..................................................133
8.7 Typical thermal reflow cycle considered in the present numerical study............134
8.8 Plots showing effect of heating rate on the dissolution of copper particle
and IMC growth in Sn-Ag-0.9Cu solder alloy.....................................................136
8.9 Plots showing effect of cooling rates on the dissolution of copper particle
and IMC growth in Sn-Ag-0.9Cu solder alloy.....................................................138
8.10 Plots showing effect of holding time interval on the dissolution of copper
particle and IMC growth in Sn-Ag-0.9Cu solder alloy........................................139
xv
CHAPTER I
INTRODUCTION
Soldering is one of the oldest techniques of joining two pieces of metal together.
Historically, the method dates back more than two thousand years. In a most simple
soldering process, the gap between the metals to be joined is filled with an alloy, usually
a mixture of two or more pure metals, which has lower melting temperature than the
members to be joined. Heat is applied so that the filler material melts around the
members to be joined and, upon solidification, forms a permanent joint between them.
The filler material, called the solder, plays a vital role in determining the quality
of the resulting joint. In the past, the soldering industry usually employed lead-based
solders for most of the applications. Traditionally, the lead-based solders are tin-lead
alloys of eutectic composition. These solders have been in use in the soldering industry
for several decades and proved to be an excellent soldering material to satisfy most of the
industrial needs. Due to the fact that lead-based solders have a very high content of toxic
lead (37.0 %wt), they are considered harmful for the environment and must, therefore, be
replaced with eco-friendly materials. Due to the environmental and health concerns,
alternate lead-free solders have been intensively researched in the last two decades.
In the current industrial world, soldering has become indispensable for the
interconnection and packaging of virtually all electronic devices and circuits. The fast
pressed the researchers to investigate and develop the novel solder materials to meet the
Almost all the potential lead-free solder materials are high Sn-containing alloys,
including the most promising Sn-Ag-Cu (SAC) alloy of eutectic composition. During the
soldering process, the metallization layer of the printed circuit board (PCB) or the
component pins react with Sn in the solder to form an intermetallic compound (IMC)
layer at the metallization/solder interface. Although the formation of the IMC layer is
desirable for good wetting and necessary bonding, an excessively thick layer is
detrimental to the health and reliability of solder joints. The problem arises due to the
brittle nature of the IMC which makes it prone to mechanical failure even at low loads. A
thicker IMC layer would also increase the heterogeneity in the physical properties of
material across the joint. Therefore, the growth rate of the IMC layer and subsequent
thickness of the resulting IMC is vital for achieving a better control of the process. The
information can also help in selecting the right solder for a particular application. A
proper mathematical modeling of the process is critical to the understanding and analysis
of various factors affecting the dissolution of substrate and growth of IMC during
soldering process. Researchers have addressed the problem of the IMC thickness
experimental observations. But these models do not reflect the actual physics involved in
the process. Also, these empirical relations need extensive experimentation and are valid
2
only for the particular material for which they are conducted. Most often these empirical
more complicated case of mass diffusion process. Any solder in question has a very
limited solubility of the substrate material which acts as a solute in this case. The early
stage of the process when the solder is in the molten state is most critical to the resulting
joint. Researchers have made many assumptions and approximations to deal with the
the dissolution behavior of the substrate. Many things still need to be addressed and a
As soon as the molten solder comes in contact with the substrate, an intense
chemical reaction takes place at the substrate/solder interface. The main reactants of the
interface reaction are the substrate material and tin in the solder. The product of the
soldering reaction is the IMC phase formed at the interface. Almost all the commercial
substrate-solder combinations result in the IMC formation at the interface. During wet
phase of the soldering process, the IMC grows in irregular morphology. Figure 1.1a
shows the end view of these undulated humps after the substrate/solder interface is cut
perpendicular to the joint interface, while Fig 1.1b shows the top view of the
morphologically irregular IMC phase when the solder is etched away from the interface.
3
Figure 1.1: Micrographs showing the (a) the section view and (b) top view of the IMC [6]
Once the IMC is formed, it acts as a barrier for further contact of the elements of
the substrate and solder. However, the reacting species continue to diffuse through the
narrow channels formed between the IMC scallops. Initially these channels are deep
enough to provide a direct contact of the solder with the substrate. Tin from the solder,
availability of tin results in the formation of the IMC at the bottom of these scallops. The
atoms of the substrate elements also continue to travel to the front of the IMC through
these open channels and contribute to the formation of intermetallic phase. The
dissociation of the intermetallic phase into the constituent elements also takes place
simultaneously at the IMC/solder interface and the substrate atoms (solute) diffuse into
the bulk solder. The process of IMC formation and diffusion of substrate element in the
bulk matrix is fast at the beginning but slows down as the solder starts to saturate with
solute. The mechanism continues until the given volume of solder in contact with the
Micrographs in Fig 1.1 and the above explanation reveal that the process of metal
layer dissolution and the IMC growth during soldering is a complicated process. The
4
problem involves two interfaces moving simultaneously. The substrate/IMC interface
will always move towards the substrate side, while the IMC/solder interface will move to
the substrate side as long the dissolution of IMC takes place. Once the dissolution of IMC
stops, the IMC/solder interface will grow into the solder. The scallops of the IMC grow
larger in size but fewer in number with time, resulting in a net decrease in the channels
area for the element diffusion. The scalloped morphology of the IMC is dictated by the
minimization of surface energy and the Gibb’s free energy concept and these scallops can
The intent of this study is to translate the actual physics involved in the
dissolution of metal layer and growth of IMC during soldering process into mathematical
model based on the fundamental equations and known material properties. In the present
materials and thermodynamic constraints will also be incorporated in the mode. The
differential equations will be integrated using numerical techniques. The model will be
used to analyze the relative effect of various kinetic parameters and physical properties of
material on the dissolution behavior of substrate and the evolution of IMC with time
during soldering process. The results from the model will also be compared with
experimental data. Although the IMC grows in irregular morphology during wet phase of
soldering process, and thus the one-dimensional model will have some limitations for
representing the irregular shape of IMC layer, however, the model will accurately
5
translate the basic physics involved in the process of the diffusion of various species
across the joint, phase change and kinetics effects during soldering process. Extensive
parametric study will be conducted to analyze the effects of process parameters and
conditions on the dissolution and growth behavior of IMC and substrate. Variations in
industrial processing conditions for obtaining these joints including the temperature
change with time and the solder volume will also be simulated by setting the appropriate
geometry variables in the model and the effect of these processing conditions will be
analyzed.
6
CHAPTER II
LITERATURE REVIEW
Soldering is one of the oldest techniques for joining two similar or dissimilar
metals. Historically, the practice dates back to several thousand years before Christ. The
basic process of soldering is simple, which involves the melting of a soft filler metal
around the components to be joined and allowing it to cool. A permanent joint is obtained
upon solidification. The soft filler metal, called the solder, is usually an alloy of two or
more pure metals. Traditional solders are essentially an alloy of tin and lead with the
eutectic composition of Sn-63.0 wt% and Pb 37.0 wt%. The lead-based solders served
excellently for almost all the soldering applications and still continue to be a vital
soldering material in many areas. However, the potential health hazards and toxicity of
lead have become more widely understood and appreciated in recent years. The global
movement to eradicate lead from soldering industry has encouraged the researchers to
explore alternate lead-free solders. The electronic packaging industry, being the largest
consumer of solder material has pressed for eradication of lead from traditional solders.
industry from the use of lead and towards ‘non-toxic’ products. Various alternatives have
replaced the traditional use of lead in casting alloys for toys, as well as solders for certain
7
plumbing applications. Once considered a joke, lead-free ammunition (bullets and shot)
exposure to hazardous materials is high. Added to this, there are now a series of
initiatives worldwide that outline targets for electronic equipment re-use and recycling. In
such initiatives, the use of hazardous materials such as lead is often limited in order to
industries has been passed by the European Commission in the waste from electronic and
outlining targets for electronic equipment re-use and recycling. This legislation also
limits the use of hazardous materials to improve the ease of recycling. For Europe, July 1
2006 has been set as the deadline for the transition. It will impact not just on solder alloys
but component finishes and temperature ratings and board finishes issues. Lead in solders
for automotive applications have a temporary exemption from the lead ban.
The Japanese Ministry of Trade (MITI) passed a recycling law for electrical
appliances with effect from April 2001. This suggests but does not include lead phase-out
timetable. OEM's are removing lead from electronics mainly due to market pressures.
Although there is no federal legislation yet in the US, there are a number of State
waste recycling Bill. In addition, the EPA has proposed a crack-down on lead emissions
from plants that may impact the soldering industry. (Source: www.lead-free.org)
8
Another reason that presses for the search of alternate soldering material is the
A bonding wire is typically of the order of 0.001 inch (25 µm) in diameter [16]. In
a solder interconnection between two copper pads, it is very important to ensure that the
(VLSI) Si chip is 0.5 µm or less. Assuming that the spacing between two wires is also the
same as diameter, the pitch becomes about 1.0 µm. Therefore, on a 1 cm2 area of chip,
one can have approximately 104 wires. To provide the electrical leads to all these wires
on the chip, several thousands of input/output (I/O) pads are needed on the chip. The only
feasible way to provide such a high density of I/O pads is to use area array of tiny solder
there will be 10, 000 solder balls on a 1 cm2 chip surface. A typical cross-section of a
Plastic Ball Grid Array (PBGA) is presented in Fig 2.1. The available copper is around
Figure 2.1: Typical setup for Plastic Ball Grid Array (PBGA)
(Courtesy: National Semiconductors Inc., USA)
9
For these miniature parts, properties like strength, reliability, thermal and
mechanical fatigue of solder joints become critical for the overall integrity of the device.
Traditional lead-base solders are sometimes not able to meet the demands of the modern
day industry. Therefore, there is a quest to improve the lead-free solders also to improve
Solder joints are not merely the electric interconnections between the components
but they also serve as the mechanical support and heat dissipater for the device. Choice of
solder is a very important factor and is critical both to the soldering process and
The quality of solder joints depends strongly on the combination of filler and
component materials, including the processing conditions. It is for this reason that a
that occurs in making solder joint is so vital for the development of a reliable joint.
explanation for the soldering technology, but have rarely provided reliable data for use in
the design of joining processes. In reality the soldering process is extremely complex,
because it brings into play a large number of variables, some of which may not be easy to
recognize and understand. Among the relevant factors are the condition of the solid
surfaces (i.e. nature of other coatings, oxides, sulfides, surface roughness etc.), the
temperature gradient that developed during the joining operation, the metallurgical
10
reaction involving the filler and parent material, and the chemical reactions with fluxes if
The traditional tin-lead solders have a long history in soldering process. This
solder and the alloys developed with it provide many benefits such as ease of handling,
low melting temperature, good workability, ductility and excellent wetting on copper and
its alloys [9]. Lead bearing solders, and especially the eutectic or near-eutectic Sn-Pb
alloys, have been used extensively in the assembly of modern electronic circuits. [9].
exist, none meets the required standards very well. The essentially desired material
equipment (involving fluxes) are optimized for Sn-Pb solder alloys over the past 30
years. The development of proper assembly processes for lead-free solders is also needed
[9].
A number of alloy systems are available to serve as solders for industrial use.
Each alloy is unique with regard to its properties and must be chosen to meet the
and shear strength, and rate of consumption of the base metal in solder and/or substrate
[16].
tin-lead solders and the legislations that limit the usage of lead bearing solders, have
11
provided the much desired impetus for intensive research efforts for developing alternate
lead-free solders [6,9]. The first step in finding suitable alloy candidates is, therefore, to
search for some nontoxic, low melting temperature alloys which can replace the lead in
Almost all the potential candidate materials, including the most promising Sn-Ag
alloy of eutectic composition, are high-tin alloys [6]. The candidate alloy components
involve Sn as the base element, Ag, Bi, Cu, and Zn as the major alloying elements, and In
and Sb as some other minor additions. Among these elements In is known to be the
precious metal with limited world production. Given its scarceness, In cannot become a
α-particles. When these particles pass though a Si device, they generate electrons and
holes. Before these charge carriers recombine, they may affect the charges stored in the
capacitors (memory units) of the system and lead to what is called a “soft error failure”
[30]. To prevent this problem, refined Pb or Pb-free solders are preferred for critical
applications.
Conventional tin-lead solders are also prone to inferior fatigue properties during
thermal excursions and/or cycling. A harder alloy with high melting point and enhanced
mechanical properties is needed to obtain durable solder joints [4]. The Sn-(3-3.9)%Ag
alloy offers improved mechanical properties and good interfacial bonding [9].
12
2.2.2 Process Conditions of Electronic Assembly
The current processing equipment and conditions for electronic assembly are
optimized for Sn-Pb solders. Any new conditions for lead-free alloys must ensure both
productivity and reliability at least equivalent to present level of Sn-Pb solders. One of
the most sensitive parameter for the quality of solder joints is soldering temperature. The
melting temperature of Sn-Pb eutectic alloy is 183oC, and the typical soldering
temperatures are 230oC and 250oC for reflow and wave soldering respectively. The
temperature margin between the melting point of the solder and processing temperature is
around 50oC for reflow soldering. In contrast, melting temperature for typical lead-free
solders are higher than Sn-Pb eutectic solders by about 30oC, which makes the process
interconnection requires good wetting. Good wetting results in well formed solder fillets.
The liquid solder is sometimes constrained from spreading away by surrounding the
surface around the confines of the joint with a non-wetting material (a solder mask). The
ability of liquid to wet a solid surface is measured in terms of the contact angle. The
contact angle is the angle that the liquid solder makes with the solid base metal surface.
Figure 2.2 illustrates the relationship between the contact angle, θ and the degree of
wetting. A contact angle of more than 90 degrees indicates lack of wetting affinity. To
13
ensure good wetting, the molten solder should wet the contact surface with a contact
θ
θ
Figure 2.2: Representation of the degree of wetting in terms of the contact angle, θ
Most lead-free solders seem to exhibit inferior wetting on Cu than Sn-Pb near
eutectic alloy solders. It is well known fact that small amounts of certain impurities in the
solder alloy influence the degree of wetting of the molten solder and the mechanical
studied as a function of reflow time and the reflow temperature range of 200-240oC by
Kim et al. [26]. The wetting angle for a short reflow time (1 min) was observed to be in
the range from 20o-25o. Regarding the temperature dependency of the wetting angle, the
observations revealed that for the reflow time of less than 10 minutes, the wetting angle
was slightly lower at higher temperatures. However, the wetting angle was not
significantly changed for any temperature when the reflow time was extended beyond 10
minutes. During the same experiment, Kim et al. also studied the wetting behavior of
lead-based solder on Ti substrate using Ni/Ti thin films. It was observed that as the solder
14
came in contact with the Ni layer, the limited thickness of Ni was consumed to form
Ni3Sn4 after 13-15 min. For the extended reflow time when the solder came in contact
with the Ti layer, the dewetting of solder started. Once the dewetting started, the bulk part
of the molten solder began to ball up while a small part of it was left at the periphery
which formed a ring surrounding the ball at the center. The driving force for the
dewetting process was attributed to the minimization of surface energy. The key
parameter for the dewetting process is the high interfacial energy between the molten
solder and the Ti film. A similar phenomenon of dewetting by the Sn-Pb eutectic and
ductility, thermal and mechanical fatigue, creep, and shock resistance. Since most of the
around room and service temperature. Due to the reasons that diffusion of elements is
quite active and creep can occur even at room temperature, the reliability of the solder
joints is a critical issue [9]. Cost is another important factor in selecting a solder for
practical electronic applications. In general, taking cost of raw metals into consideration,
most lead-free solders cost about two to three times more than Sn-Pb solders. In contrast,
the cost of Sn-0.7wt% Cu eutectic solder is only about 1.3 times higher than Sn-Pb
solder, which explains why Sn-Cu has been successfully transferred to practical
15
Table 2.1: Lead-free solder candidates with typical processing
temperature and features* [9]
Process Temperature
Candidate (oC)
Solder Benefits Drawbacks
Material Wave Reflow
Soldering Soldering
High soldering
Sn-0.7% Cu
250-260 Cheaper cost temperature
[Ag, Ni, Au]
Lift-off** with Pb
*Figures in [ ] are the third alloying elements and their typical compositions.
**Lift-off is type of failure when upon cooling of the joint, the fillet is peeled from the
Cu land located on the printed circuit board.
16
2.3 Soldering Techniques in Electronic Packaging Industry
industry. These techniques are specific to particular need and application. Variations in
soldering methods arise due to different schemes in applying heat, flux and solder
material to the joining components. However, all the soldering techniques can broadly be
categorized into two basic methods viz. reflow soldering and wave soldering. Currently,
In reflow soldering, a mixture of solder and flux called the solder paste is applied
held at temperature above the melting temperature of solder followed by cooling. During
the process the solder melts and fills the gaps around, forming the joint. An added benefit
of this approach of soldering components to the printed circuit boards and cards is that
board such as surface mount. The operational sequence of a reflow soldering process is:
Solder Components
paste to be joined
17
2.3.2 Wave soldering
In wave soldering, the printed circuit board populated with the components to be
joined is passed across the crest of a molten, standing solder wave. Only the bottom of
the board is exposed to the molten solder. The molten solder serves both as the source of
heat to the board and components, as well as the solder source for the joints. The
2. Applying heat
Printed Circuit
Board
Components
pins
Components
to be joined
Finished Joints
Standing
Solder Wave
Apply Heat
Apply Heat
and Solder
Apply Flux
18
2.4 Anatomy of Solder Joint and Intermetallic Compound Layer
component to be soldered (the substrate) and the molten solder. This reaction at the
does not occur, there will be no bonding between the substrate and the solder and
consequently no joint will be formed. The control of the interface reaction is crucial both
for the process of soldering and the reliability of the resulting solder joint. The reaction
products, also called the intermetallic compounds (IMC), are formed at the interface of
the substrate and the solder. The formation and growth of the intermetallic layer has a
profound effect on the mechanical properties of the joint and its behavior during its
service life [14]. Any non-metallic surface layer, such as an oxide or sulfide on the
substrate or the solder inhibits the soldering reaction at the interface. Flux is added to
remove the oxide layer and prevent it from forming during soldering. It is important to
note that the flux only facilitates in the reaction to take place but it does not take part in
During a soldering process, the substrate dissolves in the molten solder and reacts
with the solder. As a result a layer of the intermetallic compound is formed (Fig 2.5) at
the interface of the solid substrate and the molten solder. Most of the lead-free solders
contain high amount of Sn. During the wet phase of most soldering processes, the Sn in
the solder predominantly takes part in the reaction to form the intermetallic compound.
Solder
IMC
Copper
the stoichiometric binary compounds containing Sn and the substrate material. The most
common intermetallic compounds formed during soldering are from the Cu-Sn, Au-Sn,
Ni-Sn binary systems [16]. Elemental Pb or Bi does not react with the commonly used
base metals like Cu, Au, Ni or Pd. Thus, in Sn-Ag, Sn-Pb or Sn-Bi solders, only Sn will
react with the base material (most often Cu) to form an intermetallic phase [16]. The
occurrence of Cu-Sn phase upon soldering becomes apparent when viewing the Cu-Sn
phase diagram (Figure 2.5), which reveals the presence of two stoichiometric compounds
corresponding to the compositions of Cu3Sn and Cu6Sn5 [16]. The phase near the copper
substrate has been observed to be Cu3Sn and is normally termed as ε-phase, while the one
next to it is observed to be Cu6Sn5 and is termed as η phase [12]. X-ray diffraction was
used to identify these phases by Lee and Duh [13]. Compositions of IMC layer were
further quantitatively measured. The molar ratio of the Cu/Sn was about 6:5, which is
close to stoichiometric composition of Cu6Sn5 in the compound layer on the solder side.
In the intermetallic layer on the copper side, the molar ratio of Cu/Sn was around 3:1,
Sn-Pb solder on Ni substrate, Kim et al. [26] identified the IMC layer the Ni3Sn4
compound. In the soldering reaction of molten pure Sn and Pd substrate, PdSn4 was
conserve the metallization layer and keep the growth of IMC at minimum.
for good wetting and bonding, an excessively thick layer is detrimental due to its intrinsic
brittleness which makes it prone to mechanical failures even at low loads [12]. The brittle
nature of an intermetallic phase is in part a result of the ordered crystal structure of the
compound phase [16]. Also, an excessively thick intermetallic compound layer results in a
joint having non-uniform physical and electrical properties, like coefficient of thermal
expansion and elasticity [9]. The solderability of pre-tinned circuit boards may be degraded
by the intermetallic compound penetrating the pre-tinned surface and deteriorating its
Uncontrolled growth of IMC poses the weakness in the solder joint and causes
micro-strains due to the mismatch of the thermal expansion coefficient between the solder
and IMC [7]. As the micro-strains exceed a critical value, micro-cracks occur at the
interface of solder/IMC and might propagate, which would result in failure of the joint and
hence influence the reliability of solder joints [7]. Further, the film could continue to grow
during service if the joint attains high temperatures due to internal heat generated by the
chip or heat dissipated from external environment [12]. The growth of IMC in the solid
state is thought to depend, to some extent, on the initial film that formed during reflow
process [12]. It is argued that the degradation of the solder joint is related to the thickness
of IMC and to the kinetics of IMC growth during soldering [7]. In order to prevent the
21
excessive growth of thick intermetallic layers, a barrier layer is often deposited on the
copper base metal contacting the molten solder alloy. Nickel is most often used as a barrier
layer between solder and copper. Ni reacts with the Sn to produce Ni-Sn intermetallic
phases, which grow much more slowly than Cu-Sn intermetallic phases at high
temperatures [16].
During a soldering process, the substrate or base material dissolves into molten
solder. Dissolution of base metals (Cu, Ni, Au etc.) in lead-free solders is observed to be
much higher than in lead-based solders. The increased thickness of IMC during the
reaction involving lead-free solders is due their high content of Sn [8]. In addition,
reaction between the high tin-containing solder and copper is intense during soldering,
which favors quick depletion of copper. Rapid depletion of copper is conducive for
dewetting and failure of the joint. As the thickness of copper is limited and the rework
and/or repair of a solder joint requires a layer of unreacted copper, the loss of copper
during soldering process must be kept under control [6]. Therefore, the dissolution of
The dissolution of base metal during soldering reaction depends upon the relative
amounts of solder and substrate material. When there is a small amount of solder
available compared to the substrate, the solder saturates quickly with solvent, after which
dissolution decreases and fast intermetallic growth can occur. On the other hand, if there
22
resulting IMC thickness is low [8]. Complete dewetting of the scallop shape IMC on Ti
and Cr substrate was observed (contact angle 180o) after the base metal was completely
consumed by the solder [31]. Due to dewetting the scallops took the shape of spheroids.
The phenomenon is known as the “spalling” of the IMC, which sometimes results in the
failure of solder joint. It was further reported that the IMC spheroids were lifted up from
between the solder and the substrate. In solid-state aging planar layer growth is typical
[1,6,20,21]. While the temperature difference between these two reactions is less than
morphology of IMC during the wet phase of soldering reaction. In these experiments,
pure Sn and Sn-Pd solders were reflowed over Pd substrate. A thick planar IMC,
identified as PdSn4, was observed at the Pd/Sn interface. It was reported that this is
probably the fastest growing intermetallic compound in the binary diffusion couple.
However, most of the basic soldering operations are performed at the temperature
higher than the melting point temperature of solders. In most of the base metal-solder
interaction, when the solder is in molten state, the morphology of the IMC has scalloped
shape. The scalloped shape morphology has generally been observed in Cu/Sn, Ni/Sn and
23
Zn/Sn systems [1,2,4]. Once the IMC is formed, the rate of substrate dissolution greatly
decreases, because it is then governed by the rate at which atoms of the base material can
diffuse through the solder intermetallic compound [15]. The IMC formed at the interface
compound is widely studied and addressed in literature. The actual difficulty in dealing
with the subject arises due to the complex nature of the solder-substrate interaction
during soldering process. The soldering reaction, particularly during the wet phase of
soldering process is very intense. The reaction product of such soldering reaction is the
IMC phase, which forms at the substrate/solder interface. The process of intermetallic
The sequence of operations and more importantly the flow of the substrate and/or
solder material in the wet phase of soldering has always been a topic of debate. The
complex physics behind the scallop formation is not fully understood yet. The subject of
discussion is the flow of the material. Is it the molten Sn flowing towards substrate and
forming the IMC at the substrate/IMC interface or the substrate material flows towards
As soon the molten solder wets the substrate, the soldering reaction starts rapidly.
The product of the reaction is the IMC phase formed at the substrate/solder interface. The
intermetallic phase layer attains a scalloped shape due to minimization of surface energy.
With time scallops grow larger but fewer, indicating that the coarsening process takes
24
place [6,34]. The mechanism of coarsening is not fully understood yet. To describe this
process Kim and Tu [6] suggested a two-flux non-conservative Ostwald ripening model,
which is based on the assumption that the rapid growth of the Cu6Sn5 compounds on the
solder side is actually a result of a ripening process. Under this assumption, the scallops
bigger than a certain size grow while the smaller ones shrink. The scallops increase in
size at the expense of the smaller scallops. No solid proof for the mechanism was
provided.
The theory of Ostwald ripening seems to be valid in the condition when the solder
is saturated with solute. However, the theory does not seem to satisfy the extremely fast
growth and dissolution of IMC scallops prior to the saturation stage. During the early
stage of soldering reaction the formation and dissolution process of IMC is very fast.
Also, the diffusion of solute in the molten solder is quite rapid prior to the solder
saturation. It is true that the scallops are formed in various sizes and follow some kind of
size distribution, yet the smaller scallops will dissociate into the constituent elements and
the substrate material thus detached will diffuse in the unsaturated solder matrix.
However, the ripening process dictates a kinetics which involves detaching of the
elements from small scallops and attaching them to the bigger ones. In the early stage of
a diffusion dominated process this process of slow interface kinetics will not be possible.
Schaefer et al. [11] considered a problem of soldering with a saturated solder and
introduced a model for the growth kinetics of IMC layer, which assumes that the grain-
25
2.6.2 Diffusion of Elements across Solder Joint
that Cu diffuses through the channels between the Cu6Sn5 scallops to reach the molten
solder. Due to high diffusivity of Cu in the molten solder, the Cu atoms diffuse rapidly to
the growth front of the IMC scallops, where they can react with Sn in the solder to add to
the IMC phase. Diffusion of substrate and solder elements through a continuous layer of
Cu6Sn5 was studied by Mei et al. [47]. They reported the intrinsic diffusivities of Cu6Sn5
and Cu3Sn compounds in a Cu-Sn diffusion couple. Based on the diffusivity values of
these compounds, it can be inferred that the diffusion of substrate material or solder
through the IMC phase is rather a slow process. Also, the quantitative results of the
substrate dissolution reflect that the substrate thickness reduction is much higher and
bound to follow a faster mechanism. A faster diffusion path must be arising through the
channels formed in between the scallops. These channels are sometime referred as the
short circuit paths to bring the solder directly in contact with the substrate.
Correct description of the IMC grain coarsening is not possible without a clear
picture of the mechanism of scallop formation and growth. Hayashi et al. [50] suggested
that the Cu6Sn5 compound has a scalloped morphology probably due to the fact that the
compound dissolves faster along the grain boundary. However, the authors did not
present a solid proof of that and did not attempt to study the dissolution rate
quantitatively. The reason seems to be valid as the grain boundary is nothing but the
26
crystal defects. It has been reported in literature that the diffusion in grain boundary is
Quantitative assessment of various material fluxes and their direction of flow are
critical to the understanding of the process and setting a mathematical model. The early
stage the process of soldering involves simultaneous growth and dissolution of IMC as
well as a continuous dissolution of the base metal. The solute from the base metal
continues to diffuse in the solder matrix as long the solder is not saturated. Once the
solder is saturated with solute, the base metal still continues to dissolve, but at a very
slow rate. The IMC phase formed at this stage starts to deposit in the open channels,
making it difficult for the molten solder to come in direct contact with the base metal.
However, the IMC still maintains its scalloped morphology at the IMC/solder front.
morphology of IMC during wet phase of soldering process. One of the reasons is based
on gain in maximum free energy. All chemical reactions at constant pressure and constant
temperature occur with a negative Gibbs free energy change. The interfacial reactions
forming IMC should also be governed by free energy change. However, in order to have
the largest negative free energy change in a short period of time, it is the rate of free
energy change that becomes more important. Therefore, the high reaction rate during wet
phase of soldering process can be a result of obtaining a large free energy change. The
reaction path that gives the largest rate of free energy change will be followed in the
course, which in turn depends on the morphology and microstructure of the product
phase. In this respect the scallop type morphology has a higher growth rate and therefore,
is favored during the initial stage of the soldering reaction [30]. Since the atomic
27
diffusivity in a molten solder is much higher than solid solder, a high rate of reaction is
achievable when the morphology of the product IMC phase has scalloped shape. The
scalloped shape morphology provides higher kinetics for the reaction to take place. Yet,
in the solid state of solder, when the high atomic diffusivity is no longer available, the
scalloped shape is no longer stable and thus the IMC transforms to layered-type
morphology.
Another explanation for the layered or scalloped morphology of the IMC can be
based on the minimization of surface energy. Scalloped-type grains of IMC have been
found to be stable in contact with the molten solder, while the layer-type grains are stable
in contact with the solid solder. The stability is due to minimization of surface and grain
boundary energy [5]. Since scallops have a large area than a flat surface, the scallop-type
morphology is unfavorable when the IMC is in contact with the solid solder. However, in
the wetting reaction, the rapid gain in compound formation energy may compensate the
Transition of a planar IMC layer to scalloped shape was studied by Ma et al. [5].
In these experiments, a planar IMC was grown by solid state annealing of Cu/Sn-Pb
solder joint. Upon subsequent reflow cycles in the molten state of the solder, the IMC
partially dissolved in the solder and the planar morphology changed to scalloped shape.
The transformation of the planar to scallop shape of IMC was attributed to thermal
grooving and the minimization of interfacial energy at the IMC/solder interface. The
planar morphology in contact with solid solder and scalloped morphology of IMC in
28
contact with the liquid solder was justified by Ma et al [5] by relating the interfacial
Figure 2.6: Schematic representation of the wetting angle, φ and the force balance at the
IMC/solder interface [5]
The difference in the dihedral angle, φ arises due to the difference in interfacial
energy of IMC in contact with solid and liquid solder (γil < γis). According to the available
data for Sn-Pb solder, γil = 90 mJ/m2 and γis = 200 mJ/m2. If γgb = 164 mJ/m2 is
considered same for both solid and liquid phase solders, the deep grooves in case of
liquid solder can be justified on the basis of difference in the interfacial energies.
Effect of heating and cooling rates on the IMC morphology around micron size Ni
particles in contact with eutectic Sn-Ag solder was studied by Lee et al. [32]. It was
29
observed that regardless the cooling rate, the morphology of IMC around Ni particles was
scalloped (what they called as “sunflower”) shape. It was deduced that the cooling
segment of the thermal reflow cycle does not significantly affect the morphological
development of the IMC around the Ni particles in the eutectic Sn-Ag solder. The IMC
morphology change was influenced particularly by the heating rate of the samples.
Depending on the heating rate, two distinctly different IMC morphologies developed
around the Ni particles in Sn-Ag solder. Higher heating rates resulted in scalloped
morphology of IMC while the slowest rate caused blocky and faceted type morphology.
that decrease in number with time [5,6,20]. In ripening process, due to the Gibbs-
Thomson effect, the smaller intermetallic grains dissolve into the liquid solder, feeding to
further growth of their neighboring larger grains. Therefore, the dissolution of the
The dissolution kinetics may be the limiting factor for the ripening process if the
dissolution kinetic rate (the rate of decomposition of intermetallic compound into its
component atoms) is smaller than (or comparable with) the solute diffusion rate in the
liquid solder. This may happen during high temperature soldering when the diffusivity of
solute is relatively high [5]. It is important to point out here that Tu’s study addressed the
stage when the molten solder is saturated with the base metal. However, prior to the
saturation of solder, a different kinetics might determine the morphology and growth rate
of IMC. It has also been reported that the dissolution of the intermetallic compound
reduces the average thickness of the IMC layer during reflow [1,22]
30
2.7 Factors Governing the IMC Thickness and Substrate Dissolution during Soldering
The quantity of intermetallic present within the solder joint depends on a numbers
during soldering and the length of time during which the substrate was exposed to the
molten solder (dwell time) are most important. Based on these factors, the quantity of
The amount of the substrate material dissolved essentially depends on the volume
of solder, the contact area of substrate and solder, the operating temperature and the
solubility limit of the material in solder at that temperature. In most of the copper
substrates in contact with the tin-based solders, the copper dissolution is proportional to
The most influential factor determining the end IMC thickness is the nature of
base material. All of the common base materials form tin intermetallic compounds. The
most reactive base metal is probably palladium, while gold comes next. In a Pd-Sn binary
couple, PdSn4 phase was reported to be the fastest growing intermetallic compound [27].
Gold dissolution in tin is very high. It has been reported that at 250oC tin can
readily but quite less than gold (around 6%wt in pure tin). However, the solubility of base
Copper is the most widely used base material. The rate of dissolution of copper in
tin-based solders under normal soldering condition is appreciable, but not to the extent
31
that special precaution need to be taken as for gold and silver. The solubility of copper in
pure tin at 250oC is around 1.5wt%. The rate of dissolution of nickel and palladium are
slower than that of copper. Of the common base materials the slowest to be attacked by
tin are those based on iron alloys such as the well known “Alloy 42” (42Ni-58Fe) which
The temperature of the system strongly influences the solubility of the base
materials in the molten alloy. For example, for copper and silver the solubility at 250oC
in pure tin is approximately twice that at their eutectic temperatures (227oC and 221oC
respectively). It has been found experimentally, that the IMC thickness is profoundly
affected by the soldering temperature and dwell time [14,28,29] and increases by
increasing the two parameters. Another important variable is dwell time. The maximum
soldering temperature reached and the dwell time vary widely between different
soldering methods. There is a remarkable difference in dwell time for the wave and
reflow soldering process. In wave soldering, a contact time of approximately two seconds
is typical while in reflow process, this same may be an order magnitude higher.
During the thermal cycle of the soldering process, another important factor
governing the final thickness and quality of the intermetallic compound is the rate at
which the solder joint is cooled. It has been observed that faster the rate of solidification,
lesser the growth of the intermetallic layer and finer the grain structure of solder in the
joint.
32
2.7.3 Volume of Solder
the dissolution rate decreases. From this stage on the dissolution of the base material
contributes solely for the build-up of the intermetallic compound at the joint interface.
The effect of solder volume on the amount of copper dissolution and the thickness
of IMC was studied by Sharif et al. [46]. The experiments were conducted by reflow of
solder balls of two different volumes made of eutectic Sn-Pb solder on the copper
substrate. The contact area of solder to copper was kept same in all the experiments. The
experiments were conducted at three different temperatures and for various time periods.
Reduction in original copper thickness was calculated as an estimate for the amount of
copper dissolution. It was noted that for the solder balls with bigger volume, resulted in
greater reduction in copper thickness. The phenomenon was observed at all temperatures.
The rate of copper dissolution was also higher for the bigger solder volume. The reason is
simple, as the solder volume or the operating temperature increases, more solute can be
dissolved in the solder before saturation. The only source of solute in this case is the
thickness was observed for smaller solder volumes at all temperatures. The reason for
greater thickness in lesser volume was attributed to the early saturation of the solder with
solute. As the copper continues to dissolve through the open channels even after
saturation, it forms the intermetallic compound and contributes to the increase in IMC
thickness. A similar phenomenon was observed in another study [27,28,47] but the
33
2.7.4 Composition of the Solder Alloy
Perhaps the most obvious factor determining the final quality and thickness of the
intermetallic compound during soldering is the composition of the solder alloy itself.
Many alloys are intentionally pre-doped with the element which forms tin intermetallic,
such as copper and silver. The alloys with pre-doped base material will result in lower
base material dissolution and slightly higher amount of intermetallic compound in the
joint. In their reflow experiments, Alam et al. [47] observed a greater thickness of the
IMC when eutectic Sn-Ag solder was doped with 0.5 wt% copper. The same solder but
without and copper added yield in smaller IMC thickness. The increase in IMC thickness
was justified as due to early saturation of copper-doped solder as opposed to the ordinary
solder.
planar layer, but in most of the wet phase of soldering processes, it takes the shape of
hemispherical scallops. A planar IMC acts as a protective barrier for further diffusion of
base material into the solder and hence results in reduced thickness of IMC. This
34
Figure 2.7: Cu-Sn Phase diagram [16]
35
2.8 Recent Progress in Understanding of IMC Growth Kinetics
The exact kinetics of the scalloped morphology during very early stage of the
soldering process is not reported in the literature. The transformation of IMC phase into
scalloped specially takes when the substrate comes in contact with wet solder and
controls the formation of the IMC Lord et al. [52] conducted some experiments recently.
In these experiments pure copper was dipped in molten pure Sn solder for very short
periods of time. The morphology and growth rate of the IMC scallops were studied. The
size of the first nuclei was observed to be 0.4 to 0.8 microns in diameter depending on the
Based on the experimental observations, Lord et al. [52] also suggested the
kinetics of the IMC scallop formation during wet phase of soldering process. The
The IMC grains are separated by the grain boundaries. The IMC/solder interface will be
dissolving most of the just created IMC phase replacing the crystallites with melt and the
grain boundaries with channels. The grain boundaries will be wetted by the solder and
dissolved faster than the grains themselves, because grain boundaries are defects of
crystalline structure. The fast dissolution along grain boundary leads to creating scalloped
morphology of the layer. In this regard the authors agreed with the conclusion made by
Hayashi et al. [50] that not the equilibrium process of grain grooving but the fast
dissolution of the grain boundaries is the primary cause of the formation of undulations of
the IMC layer in the form of scallops. It was concluded that the IMC formation is a
36
In another study [53], Tu et al. addressed the same topic and shed light on the
kinetics of the scallop formation during the wet phase of the soldering process. The
solder reaction of Cu and molten pure Sn was studied by scanning and transmission
electron microscopy. Size distribution and shape of the scallops were determined
experimentally. In light of the very high resolution TEM images the authors revealed a
detailed morphology of the IMC phases. A definite triple point separating the grain
boundary of the IMC scallops were identified and concluded as independent of the reflow
time. The observed wetting angle (Φ in Fig 2.8) was suggested as an equilibrium feature
at the reaction temperature instead of being formed at quenching. The channels formed in
between the IMC (Cu6Sn5) scallops were identified as the ordinary grain boundary.
Between the scallops and the Cu substrate a polycrystalline Cu3Sn layer (around 300 nm
thick) was observed. It was suggested that the Cu3Sn layer formed at very early stage of
the reaction. However, owing to the high density of the grain boundaries it was not
probably the rate limiting factor for the Cu supply to the reactive solid/liquid interface. A
schematic of the solid copper and liquid solder was suggested as shown in Figure 2.8.
It was further suggested that the small scallops dissolved faster and the left over
was seen as a thin disk attached to the copper substrate. The rapid dissolution of the small
37
2.9 Composite Solders
place a challenge for obtaining reliable and successful component joints. These reasons
have pressed the researchers to investigate and develop the novel solder materials, such
as composite solders to meet the requirements of the modern day solder interconnects.
improve the service temperature capabilities and thermal stability of the solder joints.
One way to obtain composite solders is by embedding the solder matrix with
micro and sometimes nano-size metal particles. These particles, after solidification, act as
reinforcement in the solder matrix and enhance the thermo-mechanical properties of the
joint during service. The terminal size of these particles is important in order to assess the
joint strength and reliability. Dissolution behavior of these tiny particles becomes more
complicated due to their high curvature and associated high interfacial energy at the
IMC/solder interface.
Most of the studies in literature aim at the coarsening and growth of the IMC
layer during service and only limited data is available for the dissolution behavior of
particles during wet phase of soldering process. This is particularly true for lead-free
solders, which is an important issue to eradicate lead from traditional lead-based solders
[1-2]. With the increasing miniaturization of the electronic equipment and the
applicability of these joints in severe environmental conditions (e.g. under the hood in
automobiles), the solder joint reliability has become a challenging issue in recent years. It
has been reported in literature that presence of these particles in solder increases the
strength of the resulting joints [9]. Selection of the proper particle composition for a
38
particular solder is very critical. In addition to the composition of the metal particles, the
size is another important factor to be considered. If the particle size is very small, it is
possible that the entire particle dissolves into the solder during wet soldering reaction. On
the other hand, excessively large particles can lead to heterogeneity in the joint. Due to
the intense soldering reaction at the beginning of the wet phase of soldering process,
particle material quickly depletes during early stage of the soldering. Therefore, the
dissolution behavior of the metal particles in liquid solders has been a critical issue in
electronic packaging industry. Researchers have tried to assess their dissolution in the
liquid solder by studying the dissolution behavior of a planar substrate in molten lead-
free solder [10]. Due to the fact that the micron-level size and spherical shape of these
particles contributes a very high curvature effect at the particle/solder interface, their
dissolution characteristic can not be approximated to that of a planar substrate. The high
curvature associated with the resulting interfacial energy might change the dissolution
of such composite solders have been reported in literature. Certain composite solders did
1990s [56–59]. The composite solders were prepared by mixing Cu6Sn5 (10, 20, 30
39
with the eutectic Sn-37Pb solder paste. The microstructure of these bulk composite solder
specimens showed that Cu-Sn, Ag-Sn, and Ni-Sn intermetallics were developed in the
composite solders around Cu, Ag, and Ni particles respectively. Layer of Cu6Sn5
compound was formed around Cu3Sn particles in the Cu3Sn reinforced composite solder,
while no new intermetallic compound was formed in the Cu6Sn5 particle reinforced
composite solder. The microstructural analysis showed good bonding of the particulate
reinforcements to the solder matrix suggesting that the resulting composite solders might
exhibit improved strength. Pinizzotto et al. [60] studied the above composite solder and
reported that intermetallic formation at the solder/copper interface was formed after
Similar studies were carried out by Wu et al. with aging temperatures of 110oC-
160oC for 0–64 days [62]. Addition of these particles was seen to greatly affect the
intermetallic formation near the Cu substrate. Sn sink theory, i.e., the particles act as Sn
sinks which remove Sn from the solder and decrease the amount of Sn for reaction at the
interface, were proposed for the effects of Cu-containing particles and Ag particles on the
produced by induction melting and inert gas atomization, by Sastry et al. [62]. It was
found that, upon reflow of the solder specimens, the fine spherical dispersoids in rapidly
solidified Sn-Pb-Cu alloys coarsen to > 1 µm platelets, however, the dispersoids in Sn-
Pb-Ni alloys remain spherical and be stable with a size of < 1 µm. The difference in
stability of dispersoids in Cu- and Ni-containing solders was explained on the basis of the
40
solders showed an increase of 25–180% in yield stress and 20–80% in the modulus
Betrabet et al. by adding 2.2 wt.% of Ni3Sn4 intermetallic particles into the Sn-40Pb
solder matrix [63]. Mechanical alloying, a solid state high-energy milling process
developed for superalloy manufacture, provided the means to process such dispersion
strengthened solders. The presence of Ni3Sn4 dispersoids resulted in a smaller grain size
amount of copper particles with a commercial Sn–3.5% Ag solder paste was made by Lin
et al. [64] and effect of the copper powder addition on the microstructure was analyzed
after rapid cooling of the samples. It was concluded that copper was dissolved completely
in the solder when 0.5 wt% copper particles were added. The copper particles were not
completely dissolved in the eutectic solder when their addition was in excess of 0.5 wt.%
and sunflower morphology was found in the eutectic solder with porosity located at and
near the top region of the sample. It was also recommended that addition of copper
powder to the eutectic Sn–3.5% Ag solder should be kept in the range of less than 1.0%
In a previous study Lin et al. [65] presented the microstructure and hardness of
blending nano-sized copper powders (average powder particle size 100 nm) with a
powder of a eutectic solder using water-soluble flux. It was observed that copper powder
41
precipitated as intermetallic compounds that were non-uniformly distributed through the
of the composite solder was attained over the conventional un-reinforced eutectic
counterpart.
Kumar et al. [66]. The composite solder specimens were prepared by mechanical mixing
of SWCNTs with solder powders. The composite solders reinforced with the SWCNTs
also exhibited reduction in melting temperature due to the higher surface free energy and
interfacial instability when compared to unreinforced solders. It was also observed that
the SWCNT addition resulted in finer grain size morphology of the solders and SWCNTs
were embedded deeply into the solder matrix. Tensile properties and micro-hardness of
the composite solders were improved with the addition of SWCNT as compared to the
unreinforced counterpart.
In general, composite solders tend to render improved properties. All the reported
investigations were basically exploratory in nature, and the extent of improvement must
be weighed against environmental and economic factors before widespread adoption can
be realized. However, studies on lead-free Sn-Ag based composite solders have received
42
CHAPTER III
SOLDERING
are vital to the quality of a solder joint. Final thickness of the IMC in the terminal joint
during a soldering has been a critical concern in the soldering industry. The assessment of
base metal lost during the process is another important subject being investigated by the
researchers in the last two decades. The quantity of the base metal lost and the thickness
parameters such as temperature cycle together with the type and quantity of solder
The kinetics of copper dissolution and growth of IMC can be quantified based on
their thickness measurement. The power-law relation, relating the average thickness of
the IMC formed to the dwell time and the diffusion coefficient, is widely cited in
x(T , t ) = Dt n (3.1)
43
where the symbols have the following meanings
t dwell time (time for which the solder remains in contact with substrate at
The diffusion coefficient D is related to the process temperature through the Arrhenius-
−Q
D = Do e RT
(3.2)
where:
Q the apparent activation energy of the solute (material from base metal),
the melting point of the metal through which the solute diffuses [15]. The rate-controlling
step for reaction between two metals is the diffusion of atoms between the reacting
phases [15]. In general, the concentration of dissolving metal in the molten filler
increases in an inverse exponential manner with respect to time. That is, the dissolution
rate is initially very fast, but then slows as the concentration of the solute tends toward its
44
The object of the analysis is to find the proper values of D and n in equation (3.1)
for the temperature range. Multivariable linear regression analysis of experimental data
average IMC thickness at various times is done to estimate the parameters. Experiments
are conducted (for example reflow), where the molten solder is brought in contact with
the substrate for a known time period at a set temperature. The temperature and time is
recorded and the samples are observed under microscope after proper mounting to
by equation (3.1), the logarithmic of which gives the equation of a straight line.
1.1
Pure Sn
Sn-3.5% Ag
1.0
0.9
y = -1.6641x + 3.9097
0.8
0.7
ln D
0.6
0.5
0.4
0.3
y = -1.7584x + 3.7781
0.2
1.7 1.8 1.8 1.9 1.9 2.0 2.0 2.1
1000/T (1/K)
Figure 3.1: Arrhenius plot to estimate the values of activation energy (Q) and apparent
diffusivity (Do)
45
The data is then fitted to a standard straight line equation and by comparing the
parameters of the line with equation (3.3), values of D and n are estimated for each
temperature. Once the value of diffusivity is known for each temperature, an Arrhenius-
type plot is drawn as shown in Fig 3.1. The factor of 1000 with temperature is introduced
to adjust the scaling. The plot is a linear fit for the logarithmic of equation (3.2) as given
in equation (3.4).
By comparing the parameters of the straight line of the Arrhenius plot with
equation (3.4), the apparent activation energy Q and the constant Do are obtained.
Q1
ln( D) = ln( Do ) − (3.4)
R T
These parameters are now valid for the given range of temperature. The apparent
activation energy, Q and the constants Do and n have been estimated for various solders
in contact with different substrates and are cited in literature extensively both for wet
However, for the solid phase of soldering reaction, since the process is mostly
diffusion controlled, the value of n is assumed as 0.5 and the form of the equation is
x = Dt (3.5)
The value of D is estimated by regression analysis of experimental data for the IMC
1996. The study was conducted for the reflow of a lead-based solder alloy (62Sn-36Pb-
2Ag) % by wt. on copper substrate. The analysis was based on the same empirical
relation as given in equation (3.1). They considered the reflow process to be non-
isothermal and divided the temperature-time profile of the wet phase of soldering process
into small isothermal regions. Temperature was averaged for the region and the relevant
technique was adopted to estimate the overall values of these parameters, which was
Chada et al. [2] introduced an enhancement in the model of Schaefer [1] in 1999.
The analysis was done for the reflow of eutectic Sn-Ag solder on a Cu substrate. The
improved method took into account the dissolution and precipitation of the intermetallic
phase during the thermal cycle of the reflow process. To take into consideration the effect
of non-isothermal behavior, like the previous model, they also divided the relevant
temperature-time profile of reflow cycle into practically small isothermal regions. Kinetic
parameters were calculated numerically for each isothermal portion of the cycle. A
simultaneous dissolution of the IMC was considered until the time when solder is
saturated with Cu. Net growth of the IMC in each time step was taken as the growth
minus dissolution prior to solder saturation and growth plus precipitation after the solder
was saturated with Cu. The model was primarily based on the power law relationship,
and required the experimentally determined data to calculate parameters for the equation.
47
The amount of copper dissolved during a numerical time interval was calculated
C s − C o KAt
ln s =
C −C V
(3.6)
t dwell time (time for which the solder remained in contact with Cu)
K a constant
− Qd
K = K oe RT
(3.7)
the IMC is the only source of solute, the thickness of the IMC dissolved during a time
step was calculated using the above equations (3.6) and (3.7) together with a mass
48
balance equation which involved the weight fraction of Cu in the IMC and the density of
the IMC phase. The atomic concentration of Cu in the solder was measured with Energy
The contribution of solute precipitation in increasing the thickness of IMC, ∆x, after the
solder was saturated with Cu, was calculated by making use of the diffusivity of solute in
solder Di, the thickness of the boundary diffusion layer of solute at the IMC/solder
interface, δ and the atomic fraction of solute in IMC and solder (Xs). The final relation
DiI X s − X is−1
∆x ip = ∆t
δ X is − X η
(3.8)
where Di again is given by an Arrhenius relation of the form of equation (3.7). The
numerical results were compared with experimental data. It was found that the present
model under-predicted the thickness of the IMC for all the specimens. The difference in
the results was attributed to the values of experimental data used to calculate the
The diffusivity parameter appearing in equation (3.1) is not the real diffusivity of
the solder, rather it has been cited in literature as the apparent diffusivity of solute. In an
attempt to crystallize the subject of diffusivity, Ghosh [22] analyzed the coarsening
kinetics of Ni3Sn4 scallops during the interfacial reaction between liquid solders and Cu,
49
Ni, Pd and Ni-Pd substrates. Eutectic Sn-Ag, Sn-Bi and Sn-Pb solders were reflowed
over the substrates for various times and at different temperatures. Ghosh suggested the
shape and kinetics of the IMC on Ni substrate similar to that of Cu6Sn5. A similar
analysis as done by the previous researchers [1,2] was done and the kinetic parameters
were estimated. Coarsening kinetics of NI3Sn4 scallops was rigorously investigated in this
study. It was demonstrated that in the radial growth of scallops the grain boundary
diffusion might be playing a dominant role. A relation, as given in equation (3.9) was
presented for connecting various diffusivity values involved in the coarsening process.
κδDgb
Dapp = Dbk +
d (3.9)
where Dapp, Dbk and Dgb are apparent, bulk and grain boundary diffusivity respectively.
Grain boundary width, δ and a geometric parameter, κ are also involved in the process.
With the assumption that the thickening of Ni3Sn4 scallops is controlled by the grain
boundary diffusivity only, equation (3.9) was simplified after dropping the first term on
research of soldering process. They have published various papers focusing on the
50
process [6], they analyzed the reflow process of eutectic Sn-Pb solder on Cu substrate.
With the assumption that the amount of copper lost from the substrate is the sum of the
copper present in the Cu-Sn compound and in the liquid solder, they presented a mass
1 nV
∆h = ρ L + f Cu ρ cVc
ρ Cu A 100 (3.10)
where
n wt % of Cu in liquid solder
ρCu density of Cu
When the molten solder gets saturated with Cu, the consumption rate of Cu depends only
on the change of volume of Cu-Sn compounds, and the rate equation is given by
dh f Cu ρ c dVc
= (3.11)
dt A ρ Cu dt
Kim and Tu demonstrated that the experimental data for Cu consumption closely
followed equation (3.11). But their analysis is good for the case where the amount of
solder is very small and the results will be valid only after the solder is saturated with Cu.
51
3.5 Kinetics of Scalloped Morphology of IMC
In another study [34], Kim and Tu investigated the growth of IMC during the
reflow reaction of Sn-Pb alloy solder on Cu substrate. The focus of the study was to
assess the mechanism involved behind the growth of scallops of Cu6sn5 compound. They
found, like other researchers, that the scallops of the IMC compound grow larger but
fewer with time. The basic mechanisms involved behind the growth were proposed to be
the ripening reaction among the scallops and the interfacial reaction between the scallops
and the substrate. However, they suggested that the ripening is not a constant volume
process since it is also accompanied with the soldering reaction at the Cu/IMC interface.
The ripening flux was calculated following the Gibbs-Thomson effect, which the Cu
concentration in the molten solder at the surface of the IMC grain as follows
2γΩ
C r = Co e rRT
(3.12 a)
2γΩ
and if << 1 then
rRT
2γΩ
C r = C o 1 + (3.12 b)
rRT
where
52
γ interfacial energy per unit area between Cu6Sn5 compound and the molten
solder
T absolute temperature
Due to curvature difference, the concentration gradient of Cu is setup in the molten solder
between grains with different radii. The concentration difference causes the ripening flux
of Cu atoms going from the small grains to the larger grains, which permits the large
grains to grow and the small ones to shrink and disappear. Using the quasi-steady state
approximation, the flux of Cu atoms due the concentration gradient was given as
dC C − Ci
J = −D = −D (3.13)
dx δ
where
The mean separation distance is difficult to estimate directly. Therefore, it was related to
the mean grain radius with a constant, L through equation (3.14). The parameter, L can
have a value from 0 to 1. It was assumed that all scalloped grains tend to achieve a mean
grain radius, r . Combining equations (12) and (13) and considering that
3
ri max = r and δ = Lr (3.14)
2
53
they presented the final ripening flux as
2γΩDCo 1
JR = (3.15)
3LRT r 2
Another flux, contributing to the growth of the IMC phase, called the interfacial reaction
flux was calculated with the assumption that the total Cu lost from substrate is equal to
the total amount of Cu dissolved through the open channels, and was given as
m
ρ A ∆h =
NA ∫ J IT dt
(3.16)
ρ the density of Cu
∆h consumed thickness of Cu
NA Avogadro’s number
NA
J IT = ρ Aν
m (3.17)
54
where ν=dh/dt is the consumption rate of Cu in the reaction. With the assumption that JIT
is uniformly distributed over all the surfaces of hemispherical IMC grains, the atomic
J IT ρN A Aν (t ) 1
JI = =
2π (r ) N P 2πmN P (t ) (r ) 2
2
(3.18)
By making use of the Gauss’ theorem, the two fluxes in equation (3.15) and (3.18) were
combined together to give a final relation for the growth of IMC scallops with time as
γΩ 2 DCo ρAΩν (t )
r 3 = ∫ + dt
3 N A LRT 4πmN P (t ) (3.19)
The first term on the right hand side of equation (3.19) represents ripening and the second
term is due to the interfacial reaction. The above analysis is good only when the solder is
The assumption of a constant mean radius of the IMC scallop as given in equation
3
ri max = r
2 (3.20)
55
In this analysis [36] of the growth of IMC scallops with time, they considered a bell-
shaped size distribution of the spheroids. The total flux of Cu contributing to the growth
of these scallops was again considered to be the sum of the ripening and interfacial
concluded that the ripening process in the IMC growth is a flux driven ripening (FDR)
process. However, the analysis was limited to the condition when the solder is saturated
with solute. The application of the study was stated to be marginal or nil prior to the
saturation stage.
For the dissolution of the solute in the molten solder, a relation by Dybkov [37] is
dc S
= k * (c s − c )
dt V (3.21)
t time
determined parameter and represents the overall rate of dissolution and does not
56
particularly reflect any reaction kinetics at the dissolving interface. It can easily be
demonstrated that equation (3.21) follows immediately from Fick’s equations on the
within the diffusion boundary layer. This means that the solute distribution in the
diffusion boundary layer is close to linear. However, this assumption is not always true
especially at the beginning of the dissolution process and one should use the Fick’s law
The class of Stefan’s problem was extensively studied by the research group of
Vermolen [38-44]. Most of their research work dealt with the dissolution of various
chemical phases into the multi-component metal alloy matrix. Both analytical solutions
[40,42,43] and numerical techniques [38,39,41,44] were used to solve the problems.
Numerical analysis of the dissolution of a spherical particle in binary alloys was done for
a one-dimensional geometry [38]. The diffusion of solute in the bulk of the matrix was
∂c ( r , t ) ∂ 2 c ( r , t ) 2 ∂c ( r , t )
= D + (3.22)
∂t ∂r
2
r ∂r
where c(r, t) is the solute concentration in the bulk of alloy matrix at any time, r is the
57
The effect of interface reaction was incorporated with the assumption that the numbers of
solute atoms detached from the dissolving sphere are finite and their rate depends on the
concentration difference of solute between the particle and the matrix. A general
n =∞
dN (t )
= 4πR 2 (t ) ∑ K n (t )[c( R (t ), t ) − cα / β ]n (3.23)
dt n =0
Equation (3.23) could be an accurate representation of the interface reaction kinetic, but
the coefficients of the infinite series are unknown, and, therefore, it would complicate the
mathematical approach of the problem. A first order interface reaction at the moving
∂c( r , t )
K [c( R(t ), t ) − cα / β ] = D (3.24)
∂r r = R ( t )
In equations (23 and (24) the symbols have the following meanings
58
Another boundary condition at the moving interface was obtained by the mass
dR (t ) D ρα ∂c(r , t )
= (3.25)
dt ρ β β / α ρα M α ∂r
c − c( R (t ), t )
M M
β α
Parametric study for the moving interface concentration was conducted and their
variation with respect to time and the reducing radius of the sphere was reported.
by Vermolen et al. using a numerical approach [39,40]. The analysis was based on the
Fick’s second law of diffusion and equilibrium conditions at the moving interface were
assumed. A numerical approach was presented to solve the problem. The effect of the
diffusivities of various species in the compound on the dissolution behavior was also
studied. In conclusion of the study, it was shown quantitatively that the dissolution
composition and stoichiometry. The compositional parameters, like the ratio of the cell
size and the initial particle radius, initial matrix composition of the matrix and particle
geometry (spherical, cylindrical or planar) also play important role in determining the
59
CHAPTER IV
DISSOLUTION
composition and process temperature was also investigated experimentally by the author
in another study [54]. In the series of experiments two solders, the pure tin and eutectic
Sn-3.5wt%Ag were selected for investigation. The substrate material used was pure
copper. Two types of experiments were conducted to see the effect of the solder volume
on the kinetics of intermetallic phase formation and substrate dissolution. The reflow and
dipping experiments were conducted to imitate the actual industrial joining process.
In one series of experiments, 99.9% pure copper substrates were light ground and
micro-polished to remove any layer of oxide and improve wettability. The substrate
samples were then rinsed with water and alcohol. Solder paste of two different
compositions pure Sn and Sn-3.5 wt %Ag was prepared separately by mixing the solder
powder in RMA flux. Two solder tablets each around 100 milligrams of same solder
paste were dispensed over the copper substrate. A fine gauge chromel-alumel
thermocouple was embedded in one of these balls, which acted as a dummy sample. The
dummy sample was used to record the temperature-time profile of the solder copper
60
assembly during soldering. A closed furnace was maintained at the desired experiment
temperature. The furnace temperature was monitored with another dummy sample of
Copper samples with two solder balls were put in the furnace and the temperature of the
assembly was monitored in real time on computer with the help of a data acquisition
system. The start of the reflow time was calculated from the time when all solder attained
the melting temperature and was in liquid state. The samples were taken out of the
furnace after time periods ranging from 10 seconds to 10 minuets and immediately
quenched in cold water. These experiments were repeated for the two solder
compositions and various furnace temperatures. The four different temperatures selected
for these experiments were the melting points of the solders (221oC for Sn-3.5 wt %Ag
Real Time
Temp. Recorder
Furnace
Actual
Solder
Sample
Furnace
Temp.
250
Copper
Substrates
61
300
Actual Reflow Time
250 Actual Reflow
Temperature
200
End of Melting
Tem p. (C)
150
Start of Melting
100
Room
50 Temperature
0
0 50 100 150 200 250 300
Time (s)
In other series of experiments, around 400 grams each of Sn-3.5 wt %Ag and pure
Sn solders were taken and melted separately in a stainless steel crucible. The molten
solders were then transferred to glass beakers for the two different experiments. Copper
substrate samples of size 10 mm × 20 mm × 0.4 mm thick were cut from a 99.9 % pure
copper sheet. The sample substrates were mechanically ground and then finish polished
to remove any oxide layer on the surface and to enhance the wettability. The samples
62
Glass containers having molten solders were put in an enclosed furnace
maintained at the desired experiment temperature. The temperature of the solder was
solder bath. Solder bath temperature was maintained in a close range of ±2oC of the
250oC, 275oC and 300oC for Sn-3.5 wt %Ag solder, and 232oC, 250oC, 275oC and 300oC
for pure tin. A thin coating of mildly activated rosin (RMA) flux was applied to the
polished samples. Copper samples were then dipped vertically in the solder bath and
taken out after different time intervals ranging from 5 seconds to 10 minutes.
Thermocouple
Copper Substrate
Temperature
Display
Furnace
250
Molten Solder
63
The samples were cut along the length and perpendicular to the copper/solder
interface using a diamond-cutting wheel. All samples were cleaned with methanol to
remove any dirt and/or grease prior to cold mounting in epoxy at room temperature. The
mounted samples were then wet-ground using 320, 600, 800 and 1200 grit SiC
impregnated emery paper. After grinding, the samples were fine polished using 5 micron,
1 micron and 0.05 micron alumina based lubricant (aluminum oxide suspended in
distilled water). The polished samples were then etched for around 1 to 2 seconds using
The polished and etched samples were examined in optical microscope to analyze
the formation and growth of the intermetallic compound. The images at the solder copper
interface were captured with a digital camera attached to the microscope and stored on a
computer. To calculate the average thickness both in case of dipping and reflow process,
the area of IMC, in each micrograph, was manually traced out using a high accuracy
digital planimeter. A thin layer of Cu3Sn was observed in case of reflow experiments at
higher temperatures and longer dwell time. Since the thickness of Cu3Sn layer was very
(a)
(b)
Figure 4.4: Optical micrograph showing (a) the IMC layer and (b) its trace used for the
measurement of average IMC thickness by a digital planimeter
64
To calculate the total amount of copper dissolved in the solder during
copper/solder interaction, it was assumed that a reduction in the thickness of the copper
substrate is a good measure and directly proportional to total copper dissolution. For the
dipping experiments, the reduced thickness of the copper substrate (Fig 4.5) was
measured at three different locations top, middle and bottom of each sample. Their
average was subtracted from the original thickness of the sample to get the reduction in
total thickness. It was noted that in this experiment, the solder tends to attack the
substrate from both sides. Consequently, the measured value of decrease in thickness is
twice the value in an actual soldering process, and therefore, only half of the measured
value is plotted.
Original
Cu
Dipping
Direction
Cu thickness
after dipping
Figure 4.5: Micrograph showing thickness reduction in the copper substrate when dipped
in molten solder
To estimate the dissolution of copper during reflow process, images were taken at
the junction of the solder cap and the copper substrate as shown in Fig 4.6. The images
were taken at the right and left ends of the cap. The dissolution of copper was measured
65
as the depth to which the copper solder interface moved after. Measurements were taken
at both ends. Their average was taken as the thickness reduction during the reflow
process.
Solder
Thickness Copper
Reduction
Figure 4.6: Micrograph showing copper thickness reduction as a measure of copper lost
during reflow process
Growth of the intermetallic compound layer at the copper/solder interface and the
subsequent dissolution of copper substrate were investigated for the two types of
experiments. In general, the average IMC thickness for both solders increased with
increasing the reflow or dipping time and process temperature. A similar trend was
The IMC growth rate at all temperatures was high during the initial stage of the
process and then gradually slowed down during the terminal stages for the solders under
question. A similar trend was observed for the dissolution of the substrate material. In
fact, the IMC growth and the substrate dissolution are closely interrelated to each other
compound layer acts as the diffusion barrier for copper atoms to travel to the IMC/solder
interface. Therefore, the decrease in the substrate consumption rate is actually a result of
In the present study, the two series of experiments differ from each other mainly
in terms of the amount of solder in contact with the copper substrate. In the reflow
experiments there is a limited amount of solder in contact with the substrate, while in the
dipping experiments, the amount of solder around the substrate is practically infinite. The
two experiments were devised to study the important kinetics of the intermetallic
compound formation and the subsequent dissolution behavior of the substrate. The effect
of solder volume and the substrate dissolution was also studied by Ahmed et al. [46] for
the reflow of Sn-Pb solder on copper and by Yu et al. [49] for the dipping of copper
experiments was much lower than that for the reflow process. The observation was
consistent for both solders and at all temperatures. The relevant phase diagrams involving
copper and pure tin or Sn-Ag reveal that both the solders have a limited solubility of
substrate material (which acts as solute) at a given temperature. It can easily be inferred
that a larger volume of liquid solder (dipping process) can dissolve more copper as
compared to a smaller solder volume (reflow process). Moreover, the solute from the
solute was fast at the beginning of the process and slowed down as the solder stated to
saturate with solute. In case of the dipping process, owing to the large amount of solder,
the copper atoms continued to diffuse in the solder even for longer periods of time. A fast
diffusion of copper from the IMC/solder interface would not allow the IMC to grow
beyond a certain thickness. Consequently, a thin intermetallic layer favored for more
67
copper to diffuse to the IMC/solder interface and eventually into the bulk solder. Due to
high volume of the solder, practically large dissolution of solute is possible. Since in the
dipping process, there is no restriction of the amount of solute to diffuse through the
On the other hand, in case of the reflow process, where the solder volume was
very limited, the intermetallic compound started to form as soon the molten solder came
in contact with the copper substrate. Copper started to diffuse into the solder, but after a
short period of time, the solder started to saturate with solute. The saturation of the solder
slowed down further diffusion of solute in the bulk of solder. A decrease in solute
compound at the interface. As a result, the IMC grew thicker after short period of time.
Eventually, the thicker IMC, acted as a barrier for the diffusion of elements, thereby
reduced the copper dissolution at longer reflow periods. The early stage of process where
a sharp increase in the IMC formation and copper dissolution rate is observed is actually
The experimental study was aimed at finding the effect of solder volume, process
temperature, solder composition on the IMC thickness and dissolution of copper during
compound and dissolved thickness of copper was a strong function of solder temperature
68
and dwell time. Copper dissolution and IMC growth both were favored by increasing
solder temperature. The rate of copper dissolution and IMC growth was observed to be
higher at the initial stage of the process but tend to be constant at the terminal stages. The
solder composition had a mild effect on the terminal IMC thickness and dissolved
thickness of copper where the dissolution of copper was observed more in pure Sn than
Sn-Ag solder. Although the solubility of copper in pure Sn and Sn-Ag solder is almost
same, the difference in IMC thickness for the two solder compositions can be attributed
to the difference in the morphology of the IMC exhibited by the two solders. Both solders
revealed scalloped morphology for IMC, but the scallops were round in pure Sn but more
jagged and needle-like in Sn-Ag. The round scallops in the case of pure Sn created deep
open channels in between them, which allowed the solder to come in direct contact with
the copper substrate and thereby resulted a net increase in copper dissolution.
The experiments were conducted for the reflow and dipping setting depicting the
industrial reflow and wave soldering processes. The volume of solder had significant
effect on the IMC thickness and copper dissolution. Experimental observations revealed
that the end IMC thickness was more in case of reflow while copper dissolution was
much greater in the dipping process. These experimental observations point to the
complicated physics and kinetics involved in the growth of the IMC and dissolution of
the metallization layer during soldering process. This is the intent of the present study to
substantiate the physics with the help of numerical modeling and parametric study using
the model.
69
CHAPTER V
In terms of the actual physics involved, the process of soldering can be viewed as
a more complicated case of mass diffusion process. In most cases, tin in the solder reacts
with the substrate to form the intermetallic phase. When molten solder comes in contact
with substrate metallization layer or the leads of electronic components to be joined, the
stage is known as the wet soldering stage. The early stage of the wet soldering is most
critical to the resulting joint quality. During this stage the substrate-solder reaction is very
intense and most of the substrate layer is consumed and IMC is formed. Researchers have
made persistent effort to analyze and evaluate the growth and evolution of the
substrate. Most of these studies are based on the experimental observations involving a
careful study of the microstructure of the IMC phase. Still many questions need to be
The present study pertains to the post-nucleation stage where the evolution of
IMC phase takes place by a simultaneous process of growth and dissolution. The
nucleation of intermetallic during interfacial reaction with the liquid solder and the
substrate is observed to be extremely rapid and consequently, the entire area of the
70
substrate in contact with the liquid solder will be covered with the intermetallic
In case of the substrate/solder interaction, the development of the IMC and the
dissolution of substrate are the integrated processes of the same overall mechanism of
soldering reaction. Progression of each process affects the other. For an in-depth and
detailed analysis, the dissolution of substrate and formation of IMC should be analyzed
the substrate dissolution and IMC thickness growth, they do not give any insight of the
actual physics involved in the process. A mathematical model which can translate the
actual physics of the process would be a valuable tool to gain an in-depth acquaintance of
the simultaneous process of substrate dissolution and IMC growth. A model based on the
first principal method which can take into account the kinetics of IMC growth and
dissolution can help understand the process more accurately. This is one of the intent of
the present study to develop a mathematical model based on the diffusion equations and
also involving the kinetics and thermodynamics factors. Physical understanding of the
basic physics involved in the process is vital for establishing the mathematical model of
the overall mechanism. In the following paragraphs a detailed analysis of various stages
that the wet soldering reaction and the resulting dissolution of substrate and growth of
IMC is a complicated process and researchers are still trying to investigate the process
71
5.1 Kinetics of Substrate Dissolution and IMC Growth
terms of physics. The IMC covers the entire interface of the joint. The solder and
substrate atoms diffuse though the IMC and react to form the fresh compound at the two
moving interfaces. The solubility of most of the substrate elements is very low at low
temperatures. The substrate and solder diffusivities are also very low in solid state. Under
these conditions the rate controlling parameter is mainly the intrinsic diffusivity of the
intermetallic phase. Below the melting temperature of the solder, the IMC diffusivity is
also low. Consequently, the process of IMC growth in solid state solder is very slow.
Although, at elevated temperatures, when the solder is in liquid state, the intrinsic
diffusivity of the IMC doesn’t change much, the scalloped morphology of the
intermetallic phase entirely changes the course of growth and dissolution process. The
evolution of the scalloped morphology involves a complicated physics and therefore, the
subject has been a topic of debate in the scientific community. During wet phase of
soldering process, the rate of formation and dissolution of intermetallic phase is governed
both by the interface kinetics and the solute diffusivity in bulk solder and intermetallic
phase. The diffusivity of molten solder is several orders of magnitude higher than the
IMC diffusivity. Therefore, interface kinetics is the rate limiting parameter at the
beginning of the process followed by the solder diffusivity as the reaction proceeds.
On the basis of the analysis of our own experimental results [28,29] and recent
developments towards understanding of the process of IMC formation during early stage
of wet soldering process, the following mechanism of IMC growth and substrate
dissolution is suggested.
72
1. The very first traces of the IMC phase appear in the form of individual nucleated
grains separated by visible valleys. These grains grow rapidly and within a very
short period of time form a scalloped layer where the individual grains are
separated by channels. Initially these channels extend all the way to the substrate.
These open channels should not be confused with the grain boundary as
experimental observations have revealed that their width is much greater than the
2. Tin from the solder continues to diffuse through these open channels to the
substrate and further diffuses in the inter-phase boundary between IMC and
constituent elements at the IMC/solder interface and the substrate element (Cu)
4. Due to the soldering reaction at the base of the scallops (substrate/IMC interface)
a fresh layer of substrate is converted into the intermetallic phase. The scallops of
IMC grow bigger and the area of the open channels decreases.
5. The substrate/IMC interface moves towards the substrate side at velocity slightly
faster than the dissolving IMC/solder interface in the same direction. The small
difference in the velocities of the two fronts determines the net growth rate of the
IMC layer.
6. Once the solder is saturated with the solute (metal from the substrate), further
diffusion of solute in the solder is inhibited. However, tin from solder and the
73
substrate element from the metallization layer continue to diffuse through IMC
(a)
Solder
IMC Islands
Cu Substrate
(b)
Solder
Open
IMC Scallops Channels
Cu Substrate
(c)
Solder
Cu Cu Cu Cu Cu
Sn Sn Sn Sn
Cu Substrate
(d)
Solder
IMC
Cu Substrate
Figure 5.1 Schematic and micrographs showing various stages of formation and
growth of IMC during soldering
74
5.2 IMC Growth as Moving Boundary Problem
during soldering process can also be viewed as a moving boundary problem involving
complicated boundary conditions at the moving interfaces. The problem also involves
phase change and reaction kinetics at the moving interfaces. As soon the solder comes in
contact with the substrate, an intense soldering reaction takes place and intermetallic
compound is formed. The two moving interfaces viz. the substrate/IMC and IMC/solder
are established very quickly. The two interfaces initially move at greater rate but slow
down as the process is advanced. The IMC is formed at the substrate/solder interface and
dissolved at the IMC/solder interface. Consequently, the two interfaces initially move
IMC
Substrate/IMC
Interface Solder
IMC/Solder
Interface
Substrate
Figure 5.2 Schematic showing various regions and interfaces inside a typical
solder joint
The process of growth and dissolution continues until the solder is saturated with
the substrate elements (solute in this case). Once the solder becomes saturated with solute,
the dissolution of the IMC at the IMC/solder interface stops and IMC starts to grow into
the solder. As a result the substrate/IMC interface always moves towards the substrate
75
side while the IMC/solder front initially moves towards the substrate side but starts to
move into the solder once the solders is saturated with solute in the vicinity of IMC.
From the standpoint of the physical metallurgy, one of the important aspect of the
soldering problem is the growth of intermetallic compounds (Cu6Sn5, Cu3Sn, Ni3Sn4) etc.,
between the solder and the substrate. As the wet soldering process advances, the IMC
grows thicker. The physical problem involves the diffusion of various species across the
joint and into the bulk solder, the kinetics at the moving interfaces and thermodynamics
be dominating at various stages of the process. A mathematical model which can entail
the above mentioned factors will be helpful in understanding the basic physics of the
process.
Although during the wet phase of the soldering process, the IMC is formed in
irregular morphology however, the one-dimensional model will still be able to represent
the physics of the IMC growth and substrate dissolution. The solution of the diffusion
equation in one dimension will help to analyze the relative effect of various physical
properties and important parameters on IMC growth and substrate dissolution during
soldering. The problem involves two moving interfaces viz. the substrate/IMC interface
and IMC/solder interface. The interface concentrations C12, C21, C23 and C32 are assumed
to be fixed throughout the computation and can be obtained from the relevant phase
diagram at the process temperature. The substrate, IMC and solder are represented in Fig
76
Substrate IMC Solder
1 2 3
C12
Substrate C IMC Solder
21 C23
C32
Subs/IMC IMC/Solder
Interface Interface
x
x=0 x=x1 x=x2 x=x3
77
5.3.1 Assumptions made in the Mathematical Model
the specie.
• All other sources such as electrical, magnetic field etc are negligible.
Peclet number was calculated for the system and was found to be much
can be written using the Fick’s second law of diffusion. In the one-dimensional diffusion
equation, (equation 5.1), the lame coefficient a, can be adjusted as 0, 1 or 2 to set the
∂C j 1 ∂ a ∂C j
= r Dj (5.1a)
∂t r a ∂r ∂r
78
the lame coefficient, a can have the values 0, 1, or 2 respectively for linear, cylindrical or
For one-dimensional planar coordinate system the above equation can be written as
∂C j ∂ ∂C j
= D j (5.1b)
∂t ∂x ∂x
∂C 2 ∂C
Vi12 ( ρ1C12 − ρ 2 C 21 ) = ρ 2 D2 − ρ1 D1 1 (5.10)
∂x ∂x
∂C 3 ∂C 2
Vi 23 ( ρ 2 C 23 − ρ 3 C32 ) = ρ 3 D3 − ρ 2 D2 (5.11)
∂x ∂x
The outermost boundary of the solder and copper domain can be set as the “no flux”
Refer to the coordinate system in Fig 5.1, it is assumed that the boundaries at the
two ends of the domain are isolated with the surroundings, so the following boundary
It is also assumed that the solute concentration in bulk solder is zero prior to the onset of
dissolution process. Assuming that the substrate is made of pure metal, the following
t time (s)
C12 solute (Cu) conc. on the substrate side of substrate/IMC interface (wt%)
C21 solute (Cu) conc. on the IMC side of substrate/IMC interface (wt%)
C23 solute (Cu) conc. on the IMC side of IMC/solder interface (wt%)
C32 solute (Cu) conc. on the solder side of IMC/solder interface (wt%)
80
The one dimensional, three layers model is a reasonable simulation of the three
domains appearing in the physical problem of substrate dissolution and IMC formation
during soldering. The model is expected to effectively predict the IMC growth and
dissolution and the dissolution of substrate for a given process condition. It has been
widely reported in literature that when the diffusion process proceeds below the melting
point of the solder, a planar IMC is formed at the substrate/solder interface. However, in
most cases involving wet soldering conditions, the morphology of the IMC formed at the
substrate/solder interface takes the form of round scallops. It has been consistently
observed in the present study that this scalloped morphology of the IMC phase is a result
irregular morphology of the IMC phase, it will serve a useful tool to understand the basic
physics and kinetics involved in the substrate dissolution and growth/dissolution of the
IMC phase during soldering. In the preset study attempt will be made to utilize the one-
dimensional model to carry out the important parametric study of various rate governing
parameters involved in the basic governing equations. The one dimensional model will
also be utilized to analyze the dissolution behavior of the micron-size particles in solder
matrix. This study is important in the development of composite solders, where the
81
5.4 Solution of Mathematical Model
The moving boundary problem with mass balance at the moving interface falls in
the category of the classical Stefan problem. The class of Stefan problems has been of
practical importance for a long time. Analytical solution of the one-dimensional diffusion
equation with moving interface is available in scientific literature. Researchers have tried
both analytical and numerical approaches to solve the moving boundary problem. While
the analytical approach is easy to implement but has limited applications, a numerical
successfully used to predict the position of the moving interface. The solution is more
involved in mathematical sense yet can accurately predict the results. The mass
conservation boundary condition at the moving front can easily be implemented in the
solution. The method has been applied to study the dissolution kinetics of metal
through the assumption of a semi-infinite domain and in most case can be applied to one-
dimensional geometry only. In the present problem of soldering the solder domain is of
finite size, and therefore, the analytical solution will not provide the accurate results.
overcome if the diffusion equation is solved numerically. Since the implementation of the
solve the problem, the use of a numerical technique for the entire solution becomes more
attractive to solve the diffusion equation over required domain. The real challenge in the
82
numerical technique is to track the moving interfaces. In the present study the one-
technique. In this method the physically moving interface is fixed in the computational
A simple iteration scheme based on the interface velocity to deal with the
complicated interface conditions was used. The iteration procedure for the non-
1. At a given time step, if the kth iteration for the interface velocity is Vi12
k
, the
∆ε k = Vi12
k
∆t
2. The interface concentration on the solder side is either set as the equilibrium
3. The solute diffusion equation is then solved for the solder domain.
4. If the interface velocity is correct, the interface mass conservation equation (6.2)
k
needed. If the corrected velocity is Vi12 + δVi12
k
, then the following mass
∂C 2 ∂C
k
(Vi12 + δVi12
k
)( ρ1C12 − ρ 2 C 21 ) = ρ 2 D2 − ρ1 D1 1 (6.6)
∂x ∂x
83
where the concentration gradient is evaluated from the values of concentration
(6.6) as follows
∂C 2 ∂C
ρ 2 D2 − ρ1 D1 1
∂x ∂x
δVi12k = − Vi12
k
(6.7)
( ρ1C12 − ρ 2 C 21 )
6. A new interface velocity can then be determined using this interface velocity
correction as follows
k +1
Vi12 = Vi12
k
+ ωVi12
k
(6.8)
velocity, Vi12.
7. Using the new interface velocity, steps (1) through (6) are repeated until a
desired convergence is achieved for this time step. The convergence criterion
δVi12k
k
<φ (6.9)
Vi12
A typical value of φ = 0.0001 gave satisfactory results for the present study.
8. Upon convergence, the solution is saved and steps (1) through (7) are repeated
84
CHAPTER VI
The focus of the present study is to understand the physical process of substrate
dissolution and growth of the intermetallic phase at the substrate/solder interface during
the soldering process. The dissolution of the substrate material and formation and growth
of IMC phase was modeled as mass diffusion problem. In order to understand the
underlying physics of the process experiments were conducted and a mathematical model
based on the basic diffusion equation was developed. The one-dimensional diffusion
equations in planar coordinate system were solved numerically and the relative
The problem of substrate dissolution was solved analytically and outcome was
compared with numerical results. Dissolution of a planar copper substrate into liquid tin
solder was considered as the case for making comparison using the following data
0.8
xi(t)/x1 0.7
0.6
0.5
0.4
0 10 20 30 40
(a) t(s)
1.0
Analytical Solution
Numerical Solution
0.8
0.6
Vi(t)/Vo
0.4
0.2
0.0
0 10 20 30 40
(b) t(s)
Figure 6.1: Comparison of analytical and numerical solutions to show the effect of finite
domain on (a) interface position and (b) interface velocity
solution in the present problem is evident from Fig 6.1. After a short time period of
around 10 sec, the two solutions started to deviate. The deviation in the interface position
86
comes from the fact that, as the dissolution of copper substrate progresses and the solder
volume is limited, the solder started to saturate with the dissolving copper, thereby
slowing down further dissolution of copper. However, since the analytical solution
considers the unbounded solder domain, the solder would never saturate with copper and
the interface position and velocity will not be affected over time due to solder saturation.
The present problem of soldering process involves a finite solder domain and therefore,
the solution to the problem may not be accurate using analytical methods. The finite
solder volume around the substrate in actual problem can be implemented using
numerical techniques.
In the present model, one can identify that there are two rate-controlling
parameters, which play critical role in the overall progression of the dissolution of
substrate material in the solder matrix and growth of intermetallic phase. These two
parameters are the interface reaction kinetics coefficient, µ and the diffusivity of substrate
(copper) element in the solder, D2. Parametric study of these physical parameters will be
helpful to understand their contribution and effectiveness in the process of soldering. The
study will also help in understanding the underlying physics of substrate dissolution and
IMC growth. The diffusion equations were solved for a planar geometry to avoid any
87
6.2.1 Effect of Interface Reaction Kinetics
To study the effect of the interface reaction (µ) on the overall dissolution behavior
of substrate, numerical experiments were conducted for Cu-Sn system. Due to the
prevailing interface reaction coefficient, the copper concentration at the moving interface
(C21) can not be assumed at equilibrium value rather the same will be affected with the
value of interface kinetics coefficient and interface velocity. Numerical results were
obtained for different values of µ and using constant value of copper diffusivity in tin
solder (D2 = 1.0 x 10-9 m2/s). Copper concentration and location of the moving interface
together with the dissolution rate were calculated and plotted against time. To simulate
and compare the results in a consistent and convenient manner, the following
Vi (t ) xo C21 (t )
V* = C* = (6.1a)
DL Ceq
x1 (t ) DL t
x* = t* = 2
(6.1b)
xo xo
where V*, C*, x*, and t* are normalized interface velocity, solute concentration, interface
The effect of interface kinetics coefficient (µ) on the dissolution behavior of the
copper substrate in the liquid tin solder was explored (Figs 6.2 - 6.4). The non-
increase rapidly with time (Fig 6.2) after the copper dissolution started and slowed at
later stage. The rate of increase of interface concentration was observed to be a strong
88
function of the interface reaction rate and increased with increasing value of µ. It was
inferred that for an infinitely high value of µ, the interface would attain the equilibrium
concentration very quickly. It was interesting to note that after the onset of dissolution
process, it takes a finite time for the interface concentration to reach a constant value
highlighting the fact that a non-equilibrium conditions exist at the moving inter especially
copper substrate was studied (Fig 6.3 and 6.4). These plots also show the variation of the
normalized variables for the equilibrium interface condition for comparison purpose. It
was observed that for any value of µ, the copper dissolution rate was very high at the
Equilibrium Concentration
1.0
0.8
0.6
C*
0.4 µ increasing
µ = 1 x 10-7 m/s
0.2 µ = 2 x 10-7 m/s
µ = 5 x 10-7 m/s
µ = 10x 10-7 m/s
0.0
0 50 100 150 200
t*
89
The dissolution rate of copper was slow for low values of µ. The low dissolution
rate of copper was due to the extra time it took for the copper atoms to leave the substrate
due to slow kinetics. As the value of µ was increased, the dissolution rate approached
towards the equilibrium condition very quickly. A higher dissolution rate at the beginning
of the process was attributed to the strong interface reaction between copper substrate and
liquid tin. Due to the intense copper-solder reaction, copper concentration started to
buildup rapidly at the copper/solder interface. Since the copper diffusivity in tin is
limited, the dissolution rate, consequently, had to slow down after short period of time to
maintain the mass balance at the moving interface. These observations revealed that at
the beginning, the copper dissolution rate was controlled by interface kinetics and in the
later stages by diffusion mechanism. These results are consistent with the findings of
1.0
µ = 1 x 10-7 m/s
µ = 2 x 10-7 m/s
µ = 5 x 10-7 m/s
0.8 µ increasing µ = 10x 10-7 m/s
Equil. Int. Cond.
0.6
x*
0.4
0.2
0.0
0 200 400 600 800 1000
t*
Figure 6.3: Plots showing effect of interface kinetics on interface location as function of
time
90
0.005
µ = 1 x 10-7 m/s
µ = 2 x 10-7 m/s
µ = 5 x 10-7 m/s
0.004 µ = 10x 10-7 m/s
Equil. Int. Cond.
0.003
V* µ increasing
0.002
0.001
0.000
0 50 100 150 200
t*
Figure 6.4: Plot showing effect of interface kinetics on interface velocity as function of
time
the solder matrix was analyzed for the Cu-Sn interaction time interval of t* = 100. These
results are plotted in Fig 6.5. It was observed that the rate and amount of copper in the
solder matrix increased with increasing value of µ. It was interesting to note that a very
high value of interface reaction constant closely resembled the equilibrium interface
concentration conditions. The copper concentration near the interface was observed to be
highest and then decreased towards the outer boundary of solder domain. If the copper
and solder are allowed to remain in contact for substantially longer time, the copper
concentration in the entire solder matrix would reach the equilibrium concentration.
91
1.0 µ = 1 x 10-7 m/s
µ = 2 x 10-7 m/s
µ = 5 x 10-7 m/s
µ = 10x 10-7 m/s
0.8
Equil. Int. Cond.
0.6
C*
µ increasing
0.4
0.2
0.0
0.00 0.05 0.10 0.15 0.20
x1/X
Figure 6.5: Plots showing effect of interface kinetics on solute distribution in the solder
Any increase in the value of µ from 5 x 10-7 to 10 x 10-7 m/s did not result any
significant change in the dissolution behavior of the substrate. The little or no change in
the dissolution behavior when increasing µ to higher values was due to the limitations
imposed by long-range diffusion to balance the process and satisfy the mass balance at
the moving interface. The dissolution process is mutually controlled both by interface
reaction and long-range diffusion of copper in the bulk of solder. The two parameters
must balance their respective contributions to control the overall process. Any excessive
change in the value of one parameter will not cause significant effect in the overall
behavior, because the process will also be limited by the other controlling parameter.
92
6.2.2 Effect of Substrate (copper) Diffusivity in Solder
dissolution behavior of copper substrate in liquid tin solder, numerical experiments were
conducted for different values of copper diffusivity in liquid tin, D2 and for a constant
moderate value of interface kinetics coefficient, µ = 4.0 x 10-7 m/s. The dissolution
behavior of a planar copper substrate in the liquid solder was explored and plotted in
Figures 6.6 - 6.8. Similar normalized variables as in the study of interface kinetics study
were introduced as follows for the sake of comparison and plotted against time.
Vi (t )Ceq C21 (t )
V* = − C* = (6.2a)
µ Ceq
x1 (t ) µCeq t
x* = t* = (6.2b)
xo xo
The effect of long-range diffusion of copper in solder was studied with respect to the
changing copper concentration at the Cu-Sn interface (Fig 6.6). As the dissolution of
copper started, copper concentration at the Cu-Sn interface also started to increase rapidly
at the early stage and slowed down at longer time period. The rate of increase of interface
concentration was observed to be a strong function of copper diffusivity (D2) in tin solder
It was inferred that for a significantly low value of D2 the interface attained the
equilibrium concentration very quickly while for higher values of D2 the interface copper
93
1.0
0.8
DSn increasing
0.6
C*
0.4
DSn = 0.1x 10-9 m 2/s
DSn = 1 x 10-9 m2/s
DSn = 3 x 10-9 m2/s
0.2
DSn = 5 x 10-9 m2/s
DSn = 7 x 10-9 m2/s
0.0
0 2 4 6 8
t*
It is worth mentioning here that increasing the solute (copper) diffusivity in solder
beyond a certain value didn’t significantly change the copper concentration at the
interface. The reason for the minor change in interface concentration at higher values of
diffusivity is due to the fact that at this stage the dissolution is bounded by the finite
interface kinetics rate. As the dissolution of copper started, there was a sharp rise in
copper concentration at the interface, thereafter the rate decreased and the concentration
remained almost constant. The plateau in the concentration curve at higher reflow time
was a result of the balance maintained between the diffusivity of copper atoms and the
interface kinetics. The phenomenon remains active prior to the saturation state of the
solder. After the solder would be saturated with solute (copper), further diffusion of
94
1.0
0.8
0.6
x*
DSn increasing
0.4
t*
Figure 6.7: Plots showing effect of solder diffusivity on interface position as function of
time
Effect of copper diffusivity in liquid tin solder on the dissolution rate of the
dissolving copper substrate was studied (Fig 6.7 and 6.8). The dissolution rate was slow
for low values of diffusivity. The low dissolution rate was due to the slow transport rate
of atoms from interface to the interior of solder matrix due to low diffusivity. For any
value of diffusivity, the dissolution rate was very high at the beginning and slowed down
after short period of time, as the process advanced. A higher dissolution rate at the
beginning of the process was due to the absence of any copper concentration at the
interface, as the copper concentration started to build-up rapidly at the interface and the
diffusion length started to increase, the dissolution rate slowed down. Also, since the
solute (copper) diffusivity was finite, the dissolution rate, consequently, had to slow
down after short period of time to maintain the mass balance at the interface. These
95
observations revealed that at the beginning, the dissolution rate was controlled by
2.5
DSn = 0.1x 10-9 m2/s
DSn = 1 x 10-9 m2/s
DSn = 3 x 10-9 m2/s
2.0
DSn = 5 x 10-9 m2/s
DSn = 7 x 10-9 m2/s
1.5
V*
DSn increasing
1.0
0.5
0.0
0 2 4 6 8
t*
Figure 6.8: Plots showing effect of solder diffusivity on interface velocity as function of
time
0.6
DSn increasing
C*
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
x1/X
Figure 6.9: Plots showing effect of solder diffusivity on solute distribution in the solder
96
The concentration distribution of the dissolving copper element in the solder
matrix for increasing values of diffusivity is plotted (Fig 6.9). It was observed that the
rate and amount of copper in the solder matrix increased with increasing value of D2. It
was interesting to note that a very high value of diffusivity did not make any significant
effect in the solute distribution in the solder matrix due to the limited interface kinetics.
One of the important parameters to determine the terminal IMC thickness and the
(D2). The diffusivity of the IMC is a measure of the rate at which the substrate and solder
elements can diffuse through intermetallic phase. Effect of IMC diffusivity on the
dissolution behavior of the copper substrate and growth of the IMC was studied for the
wet phase of Cu-Sn interaction. The interface concentrations at the two interfaces were
assumed to be fixed during the entire calculation and were obtained from the Cu-Sn
IMC/solder interface and copper diffusivity in liquid tin solder (D3 = 1 x 10-9 m2/s) was
used in the calculation. These plots show the location and velocity of the two moving
interfaces with time for various values of IMC diffusivity. Similar normalized variables
as in the study of interface kinetics and solder diffusivity study were derived as follows
97
Vi12 (t )Ceq Vi 23 (t )Ceq
*
Vi12 = Vi*23 = (6.3a)
µ µ
xi12 (t ) xi 23 (t ) µCeq t
xi*12 = xi*23 = t* = (6.3b)
x1 x2 x1
where xi12, xi23 are the interface locations and Vi12 and Vi23 are the velocity of the Cu/IMC
and IMC/solder interface respectively. Symbols with (*) are the derived normalized
The substrate dissolution rate for various IMC diffusivity values is shown in Fig
6.11. At any value of IMC diffusivity, the substrate dissolution rate was high at the
beginning of the process and then slowed down after short period of time. The negative
0.0
-0.1
-0.2
Vi12*
-0.3
98
The IMC diffusivity did not have a significant effect of the dissolution rate of
substrate. The dissolution rate was slightly higher for low values of diffusivity. The
higher dissolution rate at low values of IMC diffusivity is attributed to the fact that at low
IMC diffusivity, the IMC thickness was less. Due to thin IMC layer, the copper atoms
quickly diffuse though the IMC and reach the IMC/solder front and eventually diffused in
the solder matrix. The enhanced copper atoms movement consequently, led to the higher
dissolution rate of copper substrate. On the other hand, at high IMC diffusivity the IMC
grew thicker with time, which caused the reduction in the supply of copper atoms to the
0.05
0.00
-0.05
-0.10
Vi23*
-0.15
-0.20
Dimc = 1x 10-12 m2/s
-0.25 Dimc = 1x 10-13 m2/s
Dimc = 1x 10-14 m2/s
-0.30
0 2 4 6 8 10
t*
It was evident from Fig 6.11 that the IMC diffusivity (Dimc) had an opposite effect
on the IMC/solder interface movement. The IMC/solder interface also moved rapidly at
the very beginning of the process and slowed down at the later stage. With increasing
99
value of IMC diffusivity the velocity of the interface increased. A negative value of
interface velocity again indicates the movement of the IMC/solder front in the negative
direction (dissolution of IMC), while a positive sign of the interface velocity indicates the
It was interesting to note that the substrate/IMC interface velocity (Vi12) was
always negative while the IMC/solder interface velocity (Vi23) was initially negative and
then changed sign at later stage of the process. A negative interface velocity of
substrate/IMC interface indicated that the substrate always dissolved at all stages of the
process. Part of the substrate is consumed in the formation of the IMC and part is
dissolved in the solder bulk. A change in the sign of IMC/solder interface velocity is an
indication of the fact that initially the IMC was dissolving and the front of IMC was
receding but at later stage, the positive interface velocity indicated that the IMC started to
1.0
Dimc = 1x 10-12 m 2/s
Dimc = 1x 10-13 m 2/s
Dimc = 1x 10-14 m 2/s
0.9
xi12*
0.8
0.7
0.6
0 2 4 6 8 10
t*
100
The growth of the IMC front in the solder happens when the solder in the vicinity
of the IMC becomes saturated with the dissolving substrate element (copper) and the
IMC can no longer dissolve in the solder. The same phenomenon is also evident from Fig
6.13 which shows the changing location of the IMC/solder interface with time.
The interface velocity changed sign earlier for higher values of IMC diffusivity
indicating that the IMC started to grow earlier for higher value of Dimc. The early onset of
IMC growth into the solder was related to the fact that at higher diffusivity, the transport
of atoms through the IMC was higher but the diffusion of solute copper atoms into the tin
matrix was limited and reduced with time. Consequently, the concentration of copper
started to build-up at the IMC front and the layer started to grow into the solder.
1.00
Dimc = 1x 10-12 m 2/s
0.90
0.85
xi23*
0.80
0.75
0.70
0.65
0 2 4 6 8 10
t*
The negative slope of the lines showing the changing locations of the two moving
interfaces (Figures 6.12-6.13) indicated the dissolution of the substrate and the IMC.
101
However, it was interesting to note that the slope of the IMC/solder interface location
became positive at later stage of the process, indicating that the IMC started to grow in
As the substrate comes in contact with the molten solder, the two moving fronts
are established very quickly. The difference in the velocity of the two interfaces
determines the net growth rate of the IMC at the substrate/solder interface.
0.10
Dimc = 1x 10-12 m2/s
Dimc = 1x 10-13 m2/s
0.08 Dimc = 1x 10-14 m2/s
0.06
abs(Vi12*-Vi23*)
0.04
0.02
0.00
-0.02
0 2 4 6 8 10
t*
Figure 6.14: Plots showing difference of velocity of the two moving interfaces as
function of time
Effect of the change in IMC diffusivity on the net dissolution of the substrate and
terminal thickness of the IMC was also analyzed in the present study (Figures 6.15-6.16).
Plots in Fig 6.14 indicate that initially the difference in the velocity of the two moving
interfaces was high but reduced rapidly with time. However, the difference started to
102
It was evident from numerical results that an increase in the IMC diffusivity
increased the IMC thickness and the consequential increase in the dissolved copper
thickness. The low IMC thickness at the lower diffusivity value was a result of the slow
diffusion of species through IMC. The net growth of the IMC thickness is governed by
the difference in the rates of the IMC formation and dissolution. At low diffusivity, the
IMC dissolved at almost the same rate as it was formed resulting in a very thin IMC
thickness. This was true particularly during the early stage of the process.
12
Dimc = 5x 10-12 m 2/s
Dimc = 1x 10-12 m 2/s
10 Dimc = 1x 10-13 m 2/s
Cu Thickness Consumed (µ m)
0
0 100 200 300 400
t (s)
Figure 6.15: Plots showing copper thickness consumption with time at various values of
IMC diffusivity
103
12
Dimc = 5x 10-12 m 2/s
Dimc = 1x 10-12 m 2/s
10 Dimc = 1x 10-13 m 2/s
IMC Thickness (µ m)
8
0
0 100 200 300 400
t (s)
Figure 6.16: Plots showing growth of IMC thickness with time for various values of IMC
diffusivity
It was interesting to note that a change in the IMC diffusivity value changed
significantly the IMC thickness but the corresponding change in the copper consumption
thickness was not that great. The small change in the copper consumption thickness can
be attributed to the matter that at high IMC diffusivity, although copper and tin elements
can diffuse through IMC at a much greater rate, the diffusivity of copper in tin is limited.
The addition copper would be consumed in the formation of the IMC at the IMC/Solder
interface consequently, the IMC layer grows thicker. When the solder became saturated
with substrate (copper) element, copper consumption rate decreased, then-after, any
104
CHAPTER VII
DURING SOLDERING
reduction in copper substrate thickness during the reflow process of pure molten Sn was
compared with the present model. The substrate thickness reduction was a result of
dissolution of the material during reflow process. Part of the material dissolved from
substrate is consumed in the formation of IMC and part is diffused in the bulk solder. In
the present comparison, the formation of the IMC is not considered for the sake of
simplicity. Volume of the solder in contact with the substrate together with the area of
the solder cap was transformed to a cylinder for the same contact area, as shown in Fig
7.1. It was assumed that one dimensional model with the space axis perpendicular to the
The height of the transformed cylinder, h was taken as the width of the solder domain.
solder cylinder
solder cap substrate
h
Original Transformed
Figure 7.1: Schematic showing transformation of solder cap to an equivalent cylinder
105
Results from the model were compared with the available experimental data
[28,29] together with the power law equation. The comparison of thickness reduction of
pure copper substrate in contact with pure tin during a reflow process is shown for four
10 14
9
12
Thickness Red. (mic
3 Experimental 4 Experimental
2 Power Law Power Law
2
1 Present Model Present Model
0 0
0 60 120 180 240 300 360 420 480 540 600 0 60 120 180 240 300 360 420 480 540 600
Time (s) Time (s)
(a) (b)
18 20
16 18
Thickness Red. (mic
14 16
14
12
12
10
10
8
8
6
Experimental
6 Experimental
4 4
Power Law Power Law
2 Present Model 2 Present Model
0 0
0 60 120 180 240 300 360 420 480 540 600 0 60 120 180 240 300 360 420 480 540 600
Time (s) Time (s)
(c) (d)
Figure 7.2: Plots showing comparison of numerical results with experimental data for
copper substrate thickness consumption at (a) 232oC, (b) 250oC, (c) 275oC and (d) 300oC
106
The results from the two-layer (Cu and Sn) model were plotted for the
equilibrium condition at the interface. For this model, the IMC formation was not
considered and the problem was viewed as the diffusion of substrate element in the bulk
solder. However, in actual soldering process, most of the substrate element diffusing in
the solder comes from the dissociation of IMC phase, which might change the interface
kinetics. Since the present model deals with an isothermal case at this stage, the effect of
temperature is to only determine the solubility limit and other interface concentrations of
copper in the solder. The copper (solute) solubility in the solder can be obtained from the
relevant phase diagram. The dissolution results are sensitive to the solute solubility limit
and therefore, these limits should be picked very carefully. For the present comparison
the solubility limits of copper in pure Sn at various temperatures are taken from Cu-Sn
The plots in Figure 7.2 also show the results from the fitting of the data with the
power law equation using the parameters calculated for these experiments. The prediction
from power law was good for the overall reflow process at all temperatures. However, the
107
power law is not based on any physics involved in the process. The actual reflow process
is governed both by the interface kinetics and long-range diffusion of solute. The two
mechanisms control the process and are the rate limiting parameters at various stages of
the process. The model based on the diffusion equation takes into account both these
Although, the results from the present simple mathematical model under-
predicted the consumed thickness of copper substrate, however, the present model takes
into account the process physics. The observed difference in the experimental results and
model prediction can be a result of not considering the formation of IMC at the
the formation of IMC phase. This would further increase the substrate consumption
during the process. The formation of the IMC would also change the kinetics of copper
dissolution. Once the IMC is formed, the dissolution of substrate is mainly due to the
mass diffusion through the open channels in between the scallops. These scallops slow
down the diffusion and hence the rate of substrate dissolution. At higher reflow time, it
has been observed that a thin layer of the Cu3Sn compound is formed on the substrate
side of the substrate/IMC interface. The Cu3Sn compound being rich in Cu, would
consume more copper than Cu6Sn5 compound consequently further consuming the copper
substrate.
The present model was applied to simulate the planar intermetallic growth during
the solid state aging of the copper-solder binary couple. Plots in Fig 7.3 show the IMC
108
prediction through the model at various time periods and for different temperatures. The
growth of the intermetallic phase during solid state aging is a diffusion controlled process.
The present model accurately predicted the IMC growth at various temperatures. It can be
inferred from the binary phase diagram that the solute solubility in solder does not change
significantly as long the binary couple is held below the melting point temperature of the
solder. However, it is the intrinsic diffusivity of the intermetallic compound phase which
thickness of the intermetallic phase. The IMC diffusivity values used in the study of solid
12
170 C
125 C
10 Experimental [13]
Intermetallic Thickness (microns)
0
0 2 4 6 8 10 12 14 16 18 20
t, sqrt(hrs)
Figure 7.3: Plots showing comparison of numerically predicted IMC thickness with
experimental data during thermal aging process
109
The present model was also applied to predict the growth of PdSn4 intermetallic
phase for the Pd-Sn binary couple at 250oC [27]. It was reported that PdSn4 is probably
the fastest growing intermetallic phase. The unique feature of the compound is its planar
morphology even during the wet phase of the process. Plots in Fig 7.4 show the evolution
of the IMC phase with time and its comparison with the predicted values from the present
mathematical model. The present model was able to predict the IMC thickness
reasonably well.
60
Present Numerical Model
Experimental Data [27]
50
IMC Thickness (µ m)
40
30
20
10
0
0 10 20 30 40
t (min)
Figure 7.4: Plots showing comparison of numerical prediction with experimental data for
growth of intermetallic phase (PdSn4) in the Pd-Sn binary couple at 250oC
7.3 Model Predictions for IMC Thickness and Copper Dissolution for Scalloped IMC
Growth
The present model was also applied to study the IMC growth during wet phase of
soldering process. When the solder is in molten state, the IMC does not grow as
110
4
Present Numerical Model
Experimental Data [28, 29]
IMC Thickness (µ m) 2
0
0 100 200 300 400
t (s)
Figure 7.5: Plots showing comparison of numerical prediction with experimental data for
growth of scalloped intermetallic phase (Cu6Sn5) at 250oC
10
Present Numerical Model
Experimental Data [28, 29]
Cu Thickness Consumed (µ m)
0
0 100 200 300 400
t (s)
Figure 7.6: Plots showing comparison of numerical prediction with experimental data for
copper consumption thickness at 250oC
111
Plots in Fig 7.5 show the comparison of the results from the present model with
the experimental data of the IMC thickness during reflow soldering. The IMC and solder
diffusivity value used for this numerical experiment were 1 x 10-12 m2/s and 1 x 10-9 m2/s
respectively. The numerical predictions for IMC thickness were low during the early
stage of the process. The difference in the theoretical and experimental data is attributed
to the non-planar development of the IMC. During the wet phase of the copper-solder
interaction, the IMC grows in irregular morphology. The consumed thickness of copper
during reflow experiments at 250oC was also compared with the numerical predictions
from the present model. The consumed thickness of copper substrate was very well
The high disparity in the experimentally observed and numerically predicted IMC
thickness during reflow soldering process is a result of the highly complicated physics
involved in the growth of the IMC and metal dissolution during Substrate/Solder
interaction. During the wet phase of the soldering process, the IMC layer evolves as a
complex kinetics and surface energy at the IMC/solder interface. The irregular
Since the IMC layer is not continuous, the diffusion of substrate and tin elements
across the IMC layer is not straight-forward, rather, they follow a complicated course.
The scalloped structure of the IMC leaves the open channels in between the scallops.
These channels are much wider than the grain boundary and decrease with increasing
112
time as the IMC scallops grow bigger. These channels serve as short circuit paths for the
tin atoms to come in direct contact with the substrate and diffuse in between the IMC and
substrate and form fresh intermetallic compound. During the process, the diffusion rate of
various species is different for each mechanism. The fastest being the direct diffusion of
substrate elements in the bulk of solder while the slowest is the diffusion of copper and
tin elements in the IMC layer. Besides, these mechanisms, the diffusion of substrate and
tin elements in the inter-phase boundary is also very fast while the diffusion of the
elements across the IMC phase is slower than the inter-phase diffusion. As the IMC
scallops grow bigger with time, the area of open-channels decreases and the overall rate
Sn Solder
Cu Cu Cu Cu Cu
Inter-phase
Diffusion
Diffusion Sn Sn Sn Sn
through IMC
Cu
Direct diffusion
in Solder
Figure 7.7: Schematic showing various diffusion mechanisms in IMC layer during Cu-
molten solder interaction
Considering the IMC layer as a region in between the substrate and solder layers
where the diffusing elements cross the region through various mechanisms. The IMC
113
layer has an overall effective diffusivity. The effective diffusivity of the IMC layer is an
in the IMC layer. The effective diffusivity of IMC layer was very high at the beginning
due to greater area of open channels but decreased with time. The effective diffusivity of
the IMC should not be thought as equivalent to the apparent diffusivity used in the Power
law equation [1]. The apparent diffusivity used in the Power law does not pertain to any
specific region (e.g. IMC or solder), rather it is a measure of the overall diffusivity of
substrate elements across the IMC and in the solder. In light of the parametric analysis
from the present model and the experimental observations, effective diffusivity of IMC
was estimated (equation 7.1) and its change with time is shown in Fig 7.8. The IMC
6
m /s)
2
5
-12
IMC Diffusivity ( x10
0
0 100 200 300 400 500 600
t (s)
Figure 7.8: Plot showing example of variation of IMC diffusivity with time for variable
IMC diffusivity used in this study (Do = 6x10-12 m2/s, α = 0.008)
114
6
Experimental [28, 29]
300oC
5
275oC
IMC Thickness (µ m)
4
250oC
221oC
3
0
0 100 200 300 400
t (s)
Figure 7.9: Plots showing comparison of numerical predictions with experimental data
for average thickness of scalloped IMC using variable IMC diffusivity
18
Experimental [28, 29]
16
Cu Thickness Reduction (µ m)
14
300oC
12
275oC
10
8 250oC
6 221oC
0
0 100 200 300 400
t (s)
Figure 7.10: Plots showing comparison of numerical predictions with experimental data
for consumed thickness of copper using variable IMC diffusivity
115
Numerical predictions from the present model for the average thickness of
scalloped IMC and copper consumption during reflow of Sn-Ag solder on copper
substrate were plotted (Figures 7.9-7.10) and compared with the experimental results
[28,29]. Application of varying IMC diffusivity in the model predicted the results very
well for the reflow temperature range of 221oC to 300oC. The corresponding results of
consumption could be a result of the local precipitation and formation of IMC compound
in the bulk of solder. It has been reported that chunks of intermetallic were observed in
the solder bulk [47]. These IMC could be the result of local precipitation in the solder.
The precipitation of intermetallic consumes additional copper and depletes the copper
additional copper comes from the substrate and eventually the net consumed thickness of
copper increases slightly. A similar phenomenon was observed by Alam et al. also during
their experiments of copper dissolution in Sn-Ag and Sn-Ag-Cu solders [47]. The values
of various important parameters used to obtain the plots in Fig 7.9 and 7.10 are listed in
Table 7.2. The equilibrium concentration of copper at various temperatures used in this
Table 7.2: Parameters used in the modeling of IMC growth and copper
dissolution during reflow process of Sn-3.5% Ag solder
temperature but did not change at elevated temperature. The index, α remained almost
constant at all temperatures but Do was seen as a strong function of temperature and
with temperature was observed to follow a linear relationship and can be related as given
time could be due to formation of Cu3Sn compound at the copper/IMC interface. In the
present model the composition of the entire IMC is considered as Cu6Sn5. The Cu3Sn
compound, being rich in copper would consume more copper than Cu6Sn5 compound
resulting in slight increase in copper consumption. Assuming that the equivalent copper
thickness consumed in the formation of IMC in the bulk solder is represented as δ. When
the value of δ ≈ 1.5 µm was added to the consumed thickness of copper, the total copper
thickness consumed is plotted in Fig 7.11. Plots in Fig 7.11 reveal a very good fit with the
experimental data and further that approximately 1.5 µm copper thickness from the
metallization layer was consumed in the formation of Cu6Sn5 compound in the solder
bulk.
117
16
Experimental [28, 29]
300oC
14
Cu Thickness Reduction (µ m)
12 275oC
10
250oC
8
221oC
0
0 100 200 300 400
t (s)
Figure 7.11: Plots showing comparison of numerical predictions with experimental data
for consumed thickness of copper after adding 1.5 µm to numerically predicted values
Numerical results from the present model were also obtained to predict the
average IMC thickness obtained experimentally by Chadha et al. [1] for the Sn-Ag solder.
Plots in Fig 7.12 show the numerical predictions for the average IMC and their
comparison with experimental data [1]. The values of various parameters used in these
numerical experiments were obtained from the established values of these parameters as
given in Table 7.2. These plots reveal that the model was able to predict the average IMC
thickness very well during reflow process. The experimental data for copper consumption
for the study was not provided in the reference [1], however, the corresponding copper
thickness consumption from the present numerical model is also shown in the plots (Fig
7.12 b).
118
7
Experimental [1]
6 292 oC
5
IMC Thickness (µ m) 269oC
4 251oC
226oC
0
0 2 4 6 8
t (min)
14
292oC
12 269oC
µ m)
Cu Thickness Reduction (µ
10 251oC
8 226oC
0
0 2 4 6 8
t (min)
Figure 7.12: Comparison of numerical results with experimental data of Chadha et al. [1]
for (a) average IMC thickness and (b) consumed copper thickness
119
The introduction of the variable diffusivity not only predicted the IMC thickness
and copper thickness consumption well but also substantiated the complicated physics of
the diffusion of substrate and solder elements across the IMC layer. The numerical results
using variable IMC diffusivity also revealed that a single value can not represent the
diffusivity of the IMC especially during early stage of soldering process. It was also
revealed by the application of variable IMC diffusivity that at the beginning or reflow
process grain boundary diffusion and diffusion through open channels is dominated
which losses its influence with followed by the diffusion of elements through IMC layer
120
CHAPTER VIII
The ever increasing miniaturization of electronic devices has pressed the need to
find the alternate solders that can deliver the necessary strength and reliability of the
solder joints. In this view the development of composite solders has become the focus for
many researchers in recent years. One way to obtain the composite solders is by
embedding the solder matrix with micro and some times nano-size metal particles. These
particles, after solidification, act as reinforcement in the solder matrix and enhance the
thermo-mechanical properties of the joint during service. The terminal size of these
particles is important in order to assess the joint strength and reliability. Dissolution
behavior of these tiny particles becomes more complicated due to their high curvature
and associated high interfacial energy at the IMC/solder interface. The present
mathematical model was also utilized to simulate the dissolution behavior of metal
particles in composite solders during reflow process. Thermal effects and related
dissolving micro-size metal particles were also considered. The growth of intermetallic
compound (IMC) and terminal size of embedded particles for various reflow conditions
were analyzed using the model. Dissolution behavior of micro-size Cu particles in lead-
121
free, Sn-Ag-Cu (SAC) alloy solder was studied. The end particle size and the thickness of
the IMC layer around the particles are presented for various initial particle sizes and
reflow conditions. Effects of initial copper content in Sn-Ag-Cu (SAC) solder alloy and
the interface reaction kinetics on the dissolution of copper particles and growth of
The model is based on the basic Ficks law of mass diffusion. The one-
∂C k 1 ∂ 2 ∂C k
= 2 r Dk (8.11)
∂t r ∂r ∂r
where the symbols have the usual meanings as given in chapter 6, the value of k as 1, 2
or 3 represents the value of various physical parameters in the substrate, IMC or solder
domains respectively. The copper (solute) concentration for the present problem in the
particle, IMC and solder is also shown in Fig 1. Referring Figure 1, the mass balance
equation at the particle/IMC interface (I12) and IMC/solder interface (I23) can be written
∂C 2 ∂C (8.2)
Vi12 ( ρ 1C12 − ρ 2 C 21 ) = ρ 2 D2 − ρ1 D1 1
∂r ∂r
∂C ∂C 2 (8.3)
Vi 23 ( ρ 2 C 23 − ρ 3C32 ) = ρ 3 D3 3 − ρ 2 D2
∂r ∂r
122
In addition, at the IMC/solder interface, the following kinetic relationship was also
Vi 23 = µ (C32 − Ceq )
(8.4)
Solder
IMC
Particle
0 r
r1(t)
r2(t)
R
I12 I23
C
C12
C21 C23
1 2 C32 3
r
Figure 8.1: Coordinate system and distribution of copper concentration in particle, IMC
and solder domains
The interface concentrations C12, C21, C23, and Ceq were obtained from the Cu-Sn
phase diagram at the operating temperature. The interface concentration values C12, C21
and C23 are fixed and considered to remain constant throughout the calculation.
The present model can be used to analyze the dissolution of metal particles by
inserting the relevant physical properties for the system composition of interest. The
model in the present study was tested by considering the dissolution of micro-size copper
123
particles in Sn-Ag-Cu (SAC) solder alloy. It can be identified that besides various rate-
controlling parameters, there are three important conditions which play a critical role in
the overall progression of the dissolution of particles in the solder. These three conditions
are the inter-particle distance, the initial copper content in the SAC solder and the
interface reaction kinetics. In the present study the reflow process was considered
isothermal at 250oC and the copper diffusivity in IMC and molten SAC solder were taken
The inter-particle distance represents the distance between the centers of any two
neighboring spherical particles and will be equal to twice the solder cell radius, (2R) in
Fig 1. For the present analysis it was assumed that the particles in the composite solder
are embedded in such a way that the concentration fields of dissolving particles do not
follows
r1 (t )
r* = (8.5)
ro
where r1(t) is the radius of the dissolving particle at any time instant, t and ro is the initial
radius of the particle. Effect of inter-particle distance or the solder cell size was studied
(Fig 8.2). The plots show the dissolution of copper particles in Sn-Ag-0.9Cu solder alloy
124
for different initial copper particle radii. It was evident from the plots that the dissolution
1.0
0.8
5 µm
r* = r1(t)/ro
0.6
3 4
2
ro=1 µm
0.4
0.2
(a) 0.0
10-3 10-2 10-1 100 101 102 103
t (sec)
1.0
0.8 12 µm
0.6 10 µm
r* = r1(t)/ro
9 µm
ro=1 µm 5 µm
0.4
0.2
(b) 0.0
10-3 10-2 10-1 100 101 102 103
t (sec)
Figure 8.2: Plots showing effect of inter-particle distance on particle dissolution in Sn-
Ag-0.9Cu solder alloy for (a) 2R = 50 µm, (b) 2R = 100 µm
125
Numerical results revealed that smaller particles dissolved in the molten solder
completely in a short time period however, the bigger particles first dissolved very
quickly but slowed down after short period of time. Rapid dissolution of the particles can
be attributed to their small size. The complete dissolution of the smaller particles
indicated that the solder cell around the dissolving particle was not saturated with copper
can be a result of the saturation of the solder cell with copper. Once the solder around the
particle was saturated with dissolving copper, further diffusion was stopped which
slowed the dissolution kinetics. Another important but expected result was revealed that
the bigger solder cell size or greater inter-particle distance was able to dissolve larger
copper particles completely. The result can be justified again to the fact that the greater
volume of solder was able to dissolve a larger amount of copper in it before the saturation
stage.
During the wet phase of soldering the overall dissolution of the particle in the
molten solder is governed by the interface kinetics and bulk diffusivity of the solder. This
is particularly critical when the solute (copper) diffusivity in the solder is high. The
valid, especially at the early stage of the process. In the present model, the interface
kinetics was introduced through equation (8.4). In this equation the kinetics coefficient, µ
is a measure of the interface kinetics. A higher value of µ represents faster kinetics and in
126
the limiting case an infinite value of µ will result in the equilibrium concentration at the
interface. The parametric study related to the dissolution of a planar copper substrate in
the liquid solder in the present study suggested that a value of µ ≈ 4 x 10-7 m/s can be
1.0
µ = 2 E-7m/s
0.8 µ = 4 E-7m/s
µ = 6 E-7m/s
0.6
r* = r1(t)/ro
Equlbm Cond.
0.4
0.2
0.0
(a) 10-3 10-2 10-1 100 101
t (sec)
1.0
µ = 2 E-7m/s
0.8
µ = 4 E-7m/s
µ = 6 E-7m/s
0.4
0.2
0.0
(b) 10-2 10-1 100 101 102
t (sec)
127
Figure 8.3 continued…
1.0
µ = 2 E-7m/s
µ = 4 E-7m/s
0.8
µ = 6 E-7m/s
Equlbm Cond.
0.6
r* = r1(t)/ro
0.4
0.2
0.0
10-2 10-1 100 101 102
(c)
t (sec)
Figure 8.3: Plots showing effect of interface kinetics on particle dissolution in Sn-Ag-
0.9Cu for (a) ro = 1 µm, (b) ro = 3 µm, (c) ro = 5 µm
Kinetics effects at the particle/IMC interface were not considered in this study due
to the low diffusivity of Cu and IMC layer. Under these conditions, the assumption of
In the present study the effect of interface kinetics at the IMC/solder interface was
studied (Fig 8.3) and compared with the assumption of equilibrium concentration. The
solder cell size was taken as R = 25 µm. It was evident from Fig 8.3 that the interface
kinetics had a significant effect on the dissolution of the copper particles. The model with
the equilibrium concentration at the interface (solid lines) showed very fast dissolution of
particle. The introduction of interface kinetics actually retarded the dissolution process
following that slower the interface kinetics, slower the dissolution. It was also evident
from the plots that interface kinetics had more profound effect on the smaller particles
128
than bigger ones during the entire dissolution process. However, for the large particle
(ro=5 µm, Fig 8.3 c), the convergence of these plots at longer reflow time indicated that
the kinetics had an important effect at the beginning but diffusion was the governing
In the growing trend of lead-free solders, the Sn-Ag-Cu (SAC) alloy has now
become the standard solder in the electronic packaging industry. Other additives to the
SAC alloy are also researched to improve the performance of the solder. Besides the
eutectic composition of the Sn-Ag-Cu, these alloys are also researched for the effects of
1.0
0.8
r* = r1(t)/ro
0.6
Cu = 0.0 wt%
0.5%
0.9 %
0.4
0.2
0.0
10-1 100 101
(a)
t (sec)
129
Figure 8.4 continued…
1.0
0.8
Cu = 0.0 wt%
r* = r1(t)/ro
0.6 0.5%
0.9 %
0.4
0.2
0.0
(b) 10-1 100 101
t (sec)
1.0
0.8
Cu = 0.0 wt%
r* = r1(t)/ro
0.6
0.5%
0.9 %
0.4
0.2
0.0
(c) 10-1 100 101 102
t (sec)
Figure 8.4: Plots showing effect of initial copper concentration in Sn-Ag-Cu alloy on
particle dissolution (a) ro=1 µm (b) ro=3 µm (c) ro=5 µm
130
Since a given solder volume can only dissolve a limited amount of copper, any
predoping of the solder with copper will reduce this capacity. With the same notion, the
knowledge of the effect of the initial copper content in the SAC alloy on the dissolution
of copper particles in composite solder becomes an important issue. In this view the
effect of particle dissolution in Sn-Ag-Cu solder with different initial copper content was
studied. Moderate interface kinetics of µ23=4x10-7 m/s was considered at the IMC/solder
interface. It was evident from the plots in Fig 8.4 that although the smaller particles
completely dissolved in the solder, the dissolution was slower in the solders pre-doped
with copper. It was an obvious observation that higher the copper content of the SAC
alloy, slower was the dissolution process of the particle. The change in the dissolution
process with the change in the copper content of the solder was due to the same reason
that an increase in the copper content in the SAC alloy decreased its capacity to dissolve
the copper from the dissolving particles and hence retarded the dissolution process.
The present model can also predict the thickness of the IMC around the dissolving
particle. Growth of IMC around copper particles dissolving in the Sn-Ag-0.9Cu solder
was studied (Fig 8.5) for various initial particle sizes. For these plots equilibrium
concentration was considered at the IMC/solder interface. It was evident from the
numerical results that a very thin IMC layer was formed around the smaller particles
however, for the larger particles the IMC was relatively thicker. At the beginning of the
reflow process, the IMC was thin but starter to grow thicker at the longer dwell time and
131
101
100
IMC Thickness (µ m) 12 µm
10-1
10 µm
10-2
ro=9 µm
10-3
10-4
10-1 100 101 102 103
t (sec)
Figure 8.5: Plots showing effect of initial particle size on IMC thickness during
dissolution of copper particles in Sn-3.5Ag-0.9Cu (wt %) solder alloy
Effect of the initial copper concentration in the SAC alloy on the terminal
thickness of IMC was also analyzed (Fig 8.6). The initial particle radius was taken as ro =
5 µm. It was apparent from the plots that the solders with higher initial copper
concentration yielded a thicker IMC. Solders pre-doped with copper had lower capacity
to dissolve additional copper and therefore, saturated earlier. An early saturation of the
solder retarded the IMC dissolution and thus resulted in a thicker IMC layer during
particle dissolution.
132
100
µm)
10-1
IMC Thickness (µ
Cu = 0.9 %wt
Cu = 0.5 %wt
10-2
Cu = 0.0 %wt
10-3
10-1 100 101 102
t (sec)
Figure 8.6: Plots showing effect of initial copper content in SAC solder on IMC thickness
during dissolution of copper particles
8.6 Effect of Non-isothermal Reflow Cycle on Particle Dissolution and IMC Growth
In the previous sections, the isothermal reflow process was considered to proceed
taken from the phase diagram at the temperature. An underlying assumption in the study
was of instant heating/cooling which means that at the start of reflow process, the
particle-solder system achieved the process temperature instantly and after the desired
reflow time the system was cooled down to room temperature instantaneously. However,
in a real reflow process, it takes a finite time for the system to achieve the process
temperature. This is sometimes called the temperature ramping. The heating rates and
thereby the time it takes the solder to achieve the desired process temperature could be
different depending upon the actual furnace temperature and the volume of the solder etc.
133
The thermal cycle of a typical reflow soldering might involve a heating ramp, a holding
245
Heating Holding Cooling
240
235
T (oC)
230
225
220
215
0 10 20 30 40 50
t (sec)
Figure 8.7: Typical thermal reflow cycle considered in the present numerical study
The present model was used to analyze the effect of the thermal history of the
actual reflow process on the dissolution of metal particles in composite solders. The
model was tested by considering the dissolution of micro-size copper particles in Sn-Ag-
Cu (SAC) solder alloy. It can be identified that besides various other rate-controlling
parameters, the thermal history of the soldering process also plays a critical role in the
overall progression of the dissolution of particles in the solder. A typical thermal cycle of
the process would involve the heating stage, holding stage and the cooling stage. The
actual heating and cooling rates may vary for different industrial processes. The process
was assumed to remain at the peak temperature of 240oC before cooling down.
134
In the present study, the inter-particle distance represented the distance between the
centers of any two neighboring spherical particles and would be equal to twice the solder
cell radius, (2R). For the present analysis it was assumed that the particles in the
composite solder are embedded in such a way that the concentration fields of dissolving
Effect of the heating rate of the composite solder on the dissolution behavior (Fig
8.8 a) and IMC growth (Fig 8.8 b) around the copper particle was studied. To analyze the
effect of heating rate, the cooling rate and holding time was kept same for all three
heating rates. The plots also compare the results with the condition of instant heating. It
was evident from the plots that the dissolution of Cu particles in the molten solder was
rapid for the higher heating rates. The highest heating rate was very close to the instant
heating case. Numerical results also revealed that for the high heating rate, the dissolution
of copper particles was very rapid first and slowed down with later stage of reflow,
however, for slower heating rates the dissolution of particles was steady.
135
1.0
0.9
0.8
r* = r1(t)/ro Medium Slow Heating
Fast
0.7
0.6
Instan
t Heati
ng/Coo
ling
0.5
0.4
(a) 0 20 40 60 80 100 120 140 160
t (sec)
1.2 ng
ooli
C
g/
atin
1.0 e
tH
t an
Ins
IMC Thickness (µ m)
0.8
Slow Heating
0.6
Fast Medium
0.4
0.2
0.0
(b) 0 20 40 60 80 100 120 140 160
t (sec)
Figure 8.8: Plots showing effect of heating rate on the dissolution of copper particle (a)
and IMC growth (b) in Sn-Ag-0.9Cu (wt %) solder alloy
Rapid dissolution of the particles at high heating rate can be attributed to the
greater solubility of copper in solder at elevated temperature. It was observed that the
136
heating rate also had significant effect on the growth of IMC around the copper particles.
It was interesting to note that although the end thickness of the IMC was almost same for
all the heating rates, however, the slower heating rates resulted in higher initial thickness
of the IMC layer. A higher thickness of IMC can also be attributed to the lower value of
various cooling rates after the joining process is completed. The cooling rate has a
the present study, the effect of cooling rate of the solder on the dissolution of micron-size
Cu particle in composite solder was also studied. Plots in Fig 8.9 show the dissolution
behavior of copper particle and the growth of IMC at the Cu/solder interface in eutectic
Sn-Ag-Cu solder. In this study the heating rate and holding time was same for all three
cooling rates. It was evident from the plots that the dissolution rate of copper particles
was practically same for all the three cooling rates but was different from the case of
instant heating/cooling of the solder. The identical dissolution rate of copper particles for
all the cooling rates could be a result of the saturation of the solder with copper prior to
the cooling stage. It was evident from the plots that the IMC gained significant thickness
during the cooling stage. A slower cooling rate yielded even thicker IMC.
137
1.0
0.9
Fa
st
/M
0.8 ed
iu
m
r* = r1(t)/ro /S
lo
w
0.7 Co
ol
in
g
0.6
Instan
0.5 t Heati
ng/Coo
ling
0.4
(a) 0 20 40 60 80 100 120 140 160
t (sec)
1.4
1.2
Slow Cooling g
lin
oo
1.0 g/ C
µm)
ti n
Medium ea
IMC Thickness (µ
tH
0.8 tan
Ins
0.6
Fast
0.4
0.2
0.0
(b) 0 20 40 60 80 100 120 140 160
t (sec)
Figure 8.9: Plots showing effect of cooling rates on the dissolution of copper particle (a)
and IMC growth (b) in Sn-Ag-0.9Cu (wt %) solder alloy
The thickening of the IMC layer during cooling is also observed experimentally
by researchers and is said to be a result of precipitation effect. This is particularly true for
the stage when the solder was saturated with copper before cooling started.
138
8.6.3 Effect of Holding Time
The effect of holding time of the Cu-particle in the composite solder at the peak
1.0
0.9
10
/2
0/
0.8 30
se
c
r* = r1(t)/ro
ho
ld
0.7 in
g
t im
e
0.6
Instant
0.5 Heatin
g/Cooli
ng
0.4
0 20 40 60 80 100 120 140 160
(a)
t (sec)
1.4
1.2
30s
20s
1.0
µm)
10s
IMC Thickness (µ
0.8
0.6 g
o lin
Co
g/
0.4 tin
Hea
t
an
st
0.2 In
0.0
(b) 0 20 40 60 80 100 120 140 160
t (sec)
Figure 8.10: Plots showing effect of holding time interval on the dissolution of copper
particle (a) and IMC growth (b) in Sn-Ag-0.9Cu solder alloy
139
Plots in Fig 8.10 revealed that the holding time interval did not have a significant
effect on the dissolution of copper particles, however, the IMC grew slightly thicker at
longer holding durations. The phenomenon follows a similar explanation as for the
cooling rate on the dissolution behavior and IMC growth. At longer reflow time, after the
solder was saturated with copper any addition diffusion of Cu or Sn through the IMC
actually contributed towards the thickening of the IMC, hence the IMC grew thicker for
In the growing trend of lead-free solders, the Sn-Ag-Cu (SAC) alloy has now
become an industry standard. Other elements as additives to the SAC alloy are also
researched to improve the performance of the solder. Besides the eutectic composition of
the Sn-Ag-Cu, these alloys are also researched for the effects of varying amount of initial
Since a given solder volume can only dissolve a limited amount of copper, any
predoping of the solder with copper will affect this capacity. In this view, the knowledge
of the effect of the initial copper content in the SAC alloy on the dissolution of copper
particles in composite solder becomes an important issue. In this view the effect of
particle dissolution in Sn-Ag-Cu solder with different initial copper content was studied.
A moderate heating and cooling rate were considered and the kinetics effects at the
IMC/solder interface around the copper particle was also considered. It was evident from
the plots in Fig 8.11 that dissolution of copper particles was slower in the solders pre-
140
1.0
Ci = 1.1w
t%
0.8
Ci = 0.9
wt%
0.6
r* = r1(t)/ro
0.4
Ci
0.2 = 0.7
wt
%
0.0
(a) 0 20 40 60 80 100
t (sec)
1.2
1.0
µm)
0.8
IMC Thickness (µ
0.6
Ci = 1.1wt%
0.4
0.2 Ci = 0.9wt%
Ci = 0.7wt%
0.0
(b) 0 20 40 60 80 100
t (sec)
Figure 8.11: Plots showing effect of initial copper concentration in Sn-Ag-Cu alloy on Cu
particle dissolution (a) and IMC growth (b) during non-isothermal reflow cycle
The decline in the dissolution process with the increase in the copper content of
the solder was due to the reason that an increase in the copper content in the SAC alloy
141
decreased its capacity to dissolve the copper from the dissolving particles and hence
retarded the dissolution process. The growth of IMC showed an interesting behavior. The
initial copper concentration of 0.7 % and 0.9 % showed a uniform growth on IMC during
heating and holding stage and the IMC grew thicker during cooling phase of the process.
For the initial copper concentration of 1.1 wt %, the IMC thickness plot showed an
interesting trend. During the initial stage of heating, the IMC grew rapidly for short time
period followed by a decrease in the thickness during the uniform temperature holding
stage. The thickness of IMC started to grow again during the cooling stage of the process.
The possible explanation for the phenomenon during the heating and holding stage could
be that in this case the solder was pre-doped with the copper content close to the
equilibrium concentration at the peak temperature of the process. As the temperature was
increased gradually during the heating stage, the equilibrium concentration also increased
but remained lower than the initial copper content in the solder for most of the heating
stage. A higher copper concentration in the solder than the equilibrium concentration at
the temperature would not allow any dissolution of copper, consequently, any diffusion
of Cu or Sn through the IMC was consumed in the formation of IMC. As the temperature
reached the normal process temperature and the solubility of copper also increased, the
IMC started to dissolve and the net IMC thickness decreased during the constant
temperature holding stage. The IMC again started to grow during the cooling stage of the
142
8.7 Summary of Dissolution Behavior of Micro-Particles in Composite Solders
The numerical model was applied to simulate the dissolution of micro-size copper
particle in Sn-Ag-Cu composite solder during reflow process. The overall rate of
dissolution of copper particles was high at the beginning of the process but slowed down
with time. Smaller particles were observed to dissolve completely in the solder. Moderate
interface kinetics was considered at the IMC/solder interface. It was deduced from the
numerical results that the interface kinetics played a significant role in the dissolution rate
of the particles. Increase in initial copper concentration in Sn-Ag-Cu slowed down the
dissolution of the particles and increased the IMC thickness around the particles as
non-isothermal reflow conditions were considered and the effect of heating rate, holding
time and cooling rate on the dissolution of particles and IMC growth was analyzed. The
dissolution rate of particles was observed to be greater at higher heating rates but the
terminal IMC thickness remained unaffected with the heating rate. The cooling rate effect
on the dissolution of particles was negligible, however, the final IMC thickness was
observed to be higher for slower cooling rates during reflow. Similar behavior as cooling
143
CHAPTER IX
predict the intermetallic phase growth and substrate dissolution during the substrate-
solder interaction. The boundary conditions at the two moving interfaces were derived
from the mass balance equations. An interface kinetic coefficient, governing the rate of
the interface reaction, was applied at the IMC/solder interface. The complete one-
dimensional model consisted of two moving interfaces viz. substrate/IMC interface and
IMC/solder interface. The model was solved numerically and the movement of the
parametric study of the effect of interface reaction kinetic coefficient (µ), long-range
diffusion of substrate elements in the solder (DSn) and the intermetallic diffusivity (Dimc)
present case) concentration at various interfaces and its solubility in solder at the given
temperature were obtained from the relevant phase diagram. The binary system of Cu as
case and the study was conducted for the reflow conditions. However, the model can be
applied to any system by supplying the relevant physical and thermodynamic properties.
144
9.1 Model Development and Parametric Study
Numerical results revealed that the consideration of the interface kinetics at the
IMC/solder interface was essential to the accurate modeling of the process. At the
beginning of the soldering process the growth of IMC and dissolution of the metal layer
was both governed by the interface kinetics and long-range diffusion of substrate
elements in the solder bulk. The two parameters balance their respective contribution to
achieve the critical mass conservation at the moving interfaces. The solubility limit of the
copper in solder, also known as the equilibrium concentration, is the maximum possible
concentration which can exist at the IMC/solder interface. Application of the interface
reaction kinetics (µ) decreased the interface concentration, thereby slowing down the
Dissolution of copper substrate was analyzed for cases of slow, medium and fast
interface kinetics (µ) and for fast, moderate and slow diffusivity of copper in solder (DSn).
Normalized parameters were derived for the interface velocity, interface concentration
and time. The results were also obtained and compared with the equilibrium conditions at
Effect of the interface kinetics (µ), on the interface concentration and dissolution
copper at the moving interface was observed to increase rapidly with time after the
copper dissolution started but slowed down at later stage. The rate of increase of interface
concentration was observed to be a strong function of the interface reaction kinetics and
increased with increasing value of µ. It was inferred that for an infinitely high value of µ,
the interface would attain the equilibrium concentration very quickly. It was interesting to
145
note that after the onset of dissolution process, it took a finite time for the interface
conditions existed at the moving interface especially during early stages of wet soldering
process. Effect of interface reaction kinetics on the dissolution rate of the dissolving
copper substrate was studied and the results were also compared with the equilibrium
interface conditions. It was observed that for any value of µ, the copper dissolution rate
was very high at the beginning and slowed down as the process advanced. The
dissolution rate of copper was relatively slower for lower values of µ. As the value of µ
was increased, the dissolution rate approached the equilibrium condition very quickly. A
higher dissolution rate at the beginning of the process was attributed to the strong
interface reaction between copper substrate and liquid solder. Due to the intense copper-
interface. Since the copper diffusivity in solder was limited, the dissolution rate,
consequently, had to slow down after short period of time to maintain the mass balance at
the moving interface. These observations revealed that at the beginning, the copper
dissolution rate was controlled by interface kinetics and in the later stages by diffusion
mechanism.
Numerical experiments revealed that the solder diffusivity (DSn) also had the
similar effects on dissolution rate of copper as the interface kinetics (µ). The dissolution
rate of copper was rapid at the beginning of the process but decreased with time. Higher
dissolution rate was observed for the higher values of solder diffusivity. However, the
solder diffusivity had an opposite effect on the interface concentration. The rate of
that similar to the interface kinetics, increasing the solder diffusivity beyond a certain
value didn’t significantly change the dissolution rate or copper concentration at the
interface. The reason for the minor changes at higher values of interface kinetics or solder
diffusivity was due to the fact that during the wet stage of soldering the dissolution
diffusion of copper in the bulk of solder. The two parameters must balance their
respective contributions to maintain the necessary mass balance at the interface. Any
excessive change in the value of one parameter will not cause significant effect in the
overall behavior, because the process will also be limited by the other controlling
parameter.
parameters which affects the terminal IMC thickness and the dissolution of substrate is
the intrinsic diffusivity of the intermetallic compound (Dimc). The diffusivity of the IMC
is a measure of the rate at which the substrate and solder elements can diffuse through
intermetallic phase. Effect of IMC diffusivity on the dissolution behavior of the copper
substrate and growth of the IMC was also studied for the wet phase of Cu-Solder
interface and copper diffusivity in liquid tin solder (DSn = 1 x 10-9 m2/s) was used in the
calculations. Numerical results showed that the dissolution rate of copper for various
IMC diffusivity values was high at the beginning of the process but slowed down after
short period of time. The copper dissolution rate was slightly higher for low values of
IMC diffusivity.
147
Effect of different values of IMC diffusivity on the movement of the two
interfaces viz. the Cu/IMC interface and IMC/solder interface was analyzed. Tracking of
the copper/IMC interface through numerical model revealed that the recorded velocity of
copper/IMC interface always showed a negative sign for any value of IMC diffusivity.
However, the velocity of IMC/solder interface was negative upto a certain limit but then
copper always dissolved at all stages of the process. Part of the substrate was consumed
in the formation of the IMC and part was dissolved in the solder bulk. A change in the
sign of IMC/solder interface velocity was an indication of the fact that initially the IMC
was dissolving and the front of IMC was receding but at later stage, the positive interface
velocity indicated that the IMC started to grow at the IMC/solder interface into the solder
matrix. The growth of the IMC front in the solder happened when the solder in the
vicinity of the IMC became saturated with the dissolving copper elements and the IMC
can no longer dissolve in the solder. Based on the results of the numerical actual reflow
experiments it was inferred that as the substrate comes in contact with the molten solder,
the two moving fronts are established very quickly. The copper/IMC front moves at
slightly higher rate than IMC/solder interface. The difference in the velocity of the two
interfaces determines the net growth rate of the IMC at the copper/solder interface.
Effect of change in IMC diffusivity on the net dissolution of the copper and
terminal thickness of the IMC was also analyzed in the present study. It was evident from
the results that an increase in the IMC diffusivity increased the IMC thickness and the
consequential increase in the dissolved copper layer. The low IMC thickness at the lower
diffusivity value was a result of the slow diffusion of species through IMC. The net
148
growth of the IMC thickness is governed by the difference in the rates of the IMC
formation and dissolution. At low diffusivity, the IMC dissolved at almost the same rate
as it was formed resulting in a very thin IMC thickness. This was true particularly during
The model was applied to estimate the IMC thickness and copper dissolution
during reflow process. Results from the model were compared with the experimental
data. It was observed that the model was able to accurately predict the IMC growth for
the cases where the intermetallic phase grows in planar morphology. For the cases
involving planar morphology of the IMC, a constant value of IMC diffusivity worked
well. However, use of a constant value of IMC diffusivity grossly under-predicted the
thickness of the IMC with scalloped morphology during the wet phase of the Cu-Solder
interaction. Careful observation of the morphology and the support provided with the
parametric study helped conclude that the diffusion of elements across the IMC is not
constant and can not be estimated directly. The overall IMC diffusivity is a combination
morphology and phase change with time. A variable IMC diffusivity value was then
diffusivity. The exponentially varying IMC diffusivity, when used in the model was able
to predict the IMC thickness and the corresponding copper dissolution very well.
The numerical model was also applied to simulate the dissolution of micro-size
copper particle in Sn-Ag-Cu composite solder during reflow process. The overall rate of
149
dissolution of copper particles was high at the beginning of the process but slowed down
with time. Smaller particles were observed to dissolve completely in the solder. Moderate
interface kinetics was considered at the IMC/solder interface. It was deduced from the
numerical results that the interface kinetics played a significant role in the dissolution rate
of the particles. Increase in initial copper concentration in Sn-Ag-Cu slowed down the
dissolution of the particles and increased the IMC thickness around the particles as
non-isothermal reflow conditions were considered and the effect of heating rate, holding
time and cooling rate on the dissolution of particles and IMC growth was analyzed. The
dissolution rate of particles was observed to be greater at higher heating rates but the
terminal IMC thickness remained unaffected with the heating rate. The cooling rate effect
on the dissolution of particles was negligible, however, the final IMC thickness was
observed to be higher for slower cooling rates during reflow. Similar behavior as cooling
The author wishes to extend the present study to further explore the variable IMC
diffusivity and to establish the relationships pertaining to the experimental conditions and
solder composition. Experimental study especially related to the size distribution and
during reflow is also an area of interest to extend the present study. Dissolution behavior
of the leads of electronic components, which are essentially cylindrical shaped, and the
150
prediction of IMC thickness and the dissolved lead element by the model also seems an
interesting part for future work. The author also plans to explore the extension of the
151
BIBLIOGRAPHY
11. Y. C. Chan, C. K. Alex, and J. K. L. Lai (1998) “Growth kinetic studies of Cu-Sn
intermetallic compound and its effect on shear strength of LCCC SMT solder
152
joints,” Mat. Sc. Eng. B, Vol, B55: 5 – 13.
12. K. H. Prakash and T. Sritharan (2001) “Interface Reaction Between Copper and
Molten Tin-Lead Solders,” Acta Mater., Vol. 49: 2481 – 2489.
13. Y. G. Lee and J. G. Duh (1999) “Interfacial morphology and concentration profile
in the unleaded solder/Cu joint assembly,” J. Mater. Sc., Vol. 10: 33 – 43.
15. Giles Humpston, David M. Jacobson (1993) “Principles of Soldering and Brazing,”
ASM International.
17. Armin Rahn (193) “The Basics of Soldering,” John Wiley & Sons Inc.
18. A. J. Sunwoo, J. W. Morris, Jr., and G. K. Lucey, Jr. (1992), Metall. Trans. A:23:
1323.
19. A. J. Sunwoo, H. Hayashigaani, J. W. Morris, Jr., and G. K. Lucey, Jr. (1991), JOM
43 (6): 21.
22. G. Ghosh (2000) “Coarsening kinetics of Ni3Sn4 scallops during interfacial reaction
between liquid eutectic solders and Cu/Ni/Pd metallization” J. of Appl. Phys. 88:
6887-6897.
25. J. London and D. W. Ashall, (1986), Brazing and Soldering 11: 49.
153
26. P. G. Kim, J. W. Jang, T. Y. Lee and K. N. Tu (1999) Interfacial reaction and
wetting behavior in eutectic Sn-Pb solder on Ni/Ti thin films and Ni foils, J. of App.
Phys. 86: 6746-6751.
30. K. N. Tu and K. Zeng, (2001) “Tin-Lead (Sn-Pb) solder reaction in flip chip
technology, Mat. Sc. And Engr. R 34, 1-58.
34. H. K. Kim and K. N. Tu (1996) “Kinetic analysis of the soldering reaction between
eutectic Sn-Pb alloy and Cu accompanied by ripening” Phys. Rev. B, 53:23, 16027
– 16034.
36. A. M. Gusak and K. N. Tu (2002), “Kinetic theory of flux driven ripening”, Phys.
Rev. B, 66:11, 1- 14.
154
38. F. J. Vermolin and S. van der Zwaag (1996), A numerical model for the dissolution
of spherical particles in binary alloys under mixed mode control, Mat. Sc. & Eng.
A, Vol. 220, pp. 140-146.
39. F. J. Vermolin C. Vuik and S. van der Zwaag (1998), The dissolution of a
stioichiometric second phase in ternary alloys: a numerical analysis, Mat. Sc. &
Eng. A, Vol. 246, pp. 93-103.
40. S. P. Chen, M. S. Vossenberg, F. J. Vermolin, J. van der Langkruis and S. van der
Zwaag (1999), dissolution of β particles in an Al-Mg-Si alloy during DSC runs,
Mat. Sc. & Eng. A, Vol. 272, pp. 250-256.
42. F. J. Vermolen and C. Vuik (2000), A mathematical model for the dissolution of
particles in multi-component alloys, J. of Comp. and App. Math., 126: 233-254.
43. F. J. Vermolin, C. Vuik and S. van der Zwaag (2002), A mathematical model for
the dissolution of stoichiometric particles in multi-component alloys, Mat. Sc. &
Eng. A, Vol. 328, pp. 14-25.
44. F. J. Vermolen, C. Vuik and S. van der Zwaag (2005), Cross-diffusion controlled
particle dissolution in metallic alloys, Comput and Visual in Sci, 8: 27-33.
46. Ahmed Sharif, Y. C. Chan, Rashed Adnan Islam (2003), Effect of volume in
interfacial reaction between eutectic Sn-Pb solder and Cu metallization in
microelectronic packaging, Mat. Sci. & Eng. B, 106: 120-125.
47. M. O. Alam and Y. C. Chan (2003), Effect of 0.5 wt% Cu addition in Sn-3.5%Ag
solder on the dissolution rate of Cu metallization, J. of App. Phy., 94:12, 7904-
7909.
48. S. V. Patankar (1980), Numerical Heat Transfer and Fluid Flow, Hemisphere, New
York.
155
50. A. Hayashi, C. R. Kao, and Y. A. Chang (1997), Reactions of solid copper with
pure liquid tin and liquid tin saturated with copper, Scripta Mater. 37:4, 393-398.
52. R. A. Lord and A. Umantsev (2005), Early stages of soldering reactions, J. of App.
Phy. 98: 1-11.
53. J. Görlicin, G. Schmitz and K. N. Tu (2005), On the mechanism of the binary Cu/Sn
solder reaction, App. Phy. Lett. 86, 053106, 1-3.
55. Fu Guo, (2007), Composite lead-free electronic solders, J Mater Sci: Mater Electron
18:129–145
57. J. L. Marshall, J. Calderon, J. Sees, G. Lucey, J.S. Hwang, IEEE Trans. Comp. Hyb.
Manufact. Technol. 14(4), 698 (1991)
59. J.L. Marshall, J. Calderon, (1997) Sold. Surf. Mount Technol. 9(3), 11–14
60. R.F Pinizzotto, Y. Wu, E.G. Jacobs, L.A. Foster, (1992) Microstructural
Development in Composite Solders Caused by Long Term, High Temperature
Annealing, Proceedings of Technical Program, Nepcon West Conference, Anaheim,
CA, Cahners Exhibition Group, Des Plains, IL, 3, p. 1284
61. Y. Wu, J.A. Sees, C. Pouraghabragher, L.A. Foster, J.L. Marshall, E.G. Jacobs, R.F.
Pinizzotto, J. Electron. Mater. 22(7):769 (1993)
62. S.M.L. Sastry, T.C. Peng, R.J. Lederich, K.L. Jerina, C.G. Kuo, (1992)
Microstructures and Mechanical Properties of In-situ Composite Solders,
Proceedings of Technical Program, Nepcon West Conference, Anaheim, CA,
Cahners Exhibition Group, Des Plains, IL, 3, p. 1266
63. H.S. Betrabet, S.M. McGee, J.K. McKinlay, (1991) Scripta Metall. Mater. 25,
2323–2328
156
64. D.C. Lin, T.S. Srivatsan, G.-X. Wang and R. Kovacevic, (2006) Microstructural
development in a rapidly cooled eutectic Sn–3.5% Ag solder reinforced with copper
powder, Powder Technology 166, p 38–46
65. D. C. Lin, G.X. Wang, T.S. Srivatsan, Meslet Al-Hajri, M. Petraroli (2002), The
influence of copper nano-powders on microstructure and hardness of lead-tin
solder, Materials Letters 53, p 333– 338
66. K. Mohan Kumar, V. Kripesh, Lu Shen, Andrew A.O. Tay (2006), Study on the
microstructure and mechanical properties of a novel SWCNT-reinforced solder
alloy for ultra-fine pitch applications, Thin Solid Films 504, p 371 – 378
67. Paul G. Harris, Kaldev S. Chaggar (1998), The role of intermetallic compounds in
lead-free soldering, Soldering and surface mount technology, 38-52
68. M. E. Loomans and M. E. Fine (2000), Metall. Mater. Trans. A 31A, 1155
157
APPENDICES
158
APPENDIX A
transforming the original physical coordinates into computation coordinates (Fig A1).
Assuming the substrate and the solder as η and ξ domain respectively, the following
r r −ε
η= and ξ= (A1)
ε b −ε
where r = ε(t) represents the substrate/solder interface and b is the total thickness of the
substrate and solder. A typical control volume element is explained in Fig A1.
Substrate Solder
Vi
b
r
r=0 ε(t)
η ξ
η=0 η=1 ξ =1
ξ=0
159
∆η w ∆η e
w e
J-1 J J+1
∆η J-1 ∆η J ∆η J+1
To treat the large concentration gradient at the interface, a non-uniform grid can
be used with higher grid density near the interface. The position of the faces of the
ps
J −1
ηs (J ) = J = 1, 2, 3,…M-1 (A2 a)
M −2
pl
J −1
ξ s (J ) = J = 1, 2, 3,…N (A2 b)
N −1
where M – 1 and N are the total number of control volume faces in the solid and liquid,
respectively, and the subscripts s means the faces of the elements i.e. the locations
denoted by e or w in Fig A2. The grid nodes are placed at the center between two
adjacent faces for a total of M – 2 internal nodes in the substrate region and N – 1 in the
solder region. The interface and each of the two boundary elements are considered as
zero-volume elements. There are, therefore, a total of M + N grid nodes. The exponent ps
and pl are arbitrary parameters used to control the variation of element width in the
substrate and solder domain respectively. A typical value of these parameters between 1.5
The central differencing of the convective term in the concentration equation after
universal function derived for these schemes by Patankar [48] can be used to derive the
difference equations for node concentration. The concentration diffusion equation after
∂C S 1 ∂ a ∂C dε ∂
ε2 = a (η DS ) S + ε (ηC S ) − C S (B1a)
∂t η ∂η ∂η dt ∂η
∂C S ∂ ∂C dε ∂C S
η aε 2 = (η a DS ) S + η a +1ε (B1b)
∂t ∂η ∂η dt ∂η
if b is the combined width of the substrate and solder domains, i.e. b = ro + Ro then in the
∂C L ∂ ∂C L dε ∂C L
κ (b − ε ) 2 = κDL + κ (b − ε )(1 − ξ ) (B2)
∂t ∂ξ ∂ξ dt ∂ξ
In equation (B2)
161
κ = [(b − ε )ξ + ε ]a
The finite difference equations for the nodal concentrations are obtained directly by
integrating the transformed differential equations (B1) and (B2) from time t to t + ∆t over
each control volume element. Equations (B1) and (B2) in terms of solute fluxes GS and
∂C S ∂GS ∂FS
η aε 2 + − CS = 0 (B3)
∂t ∂η ∂η
∂C L ∂G L ∂FL
κ (b − ε ) 2 + − CL = 0 (B4)
∂t ∂ξ ∂ξ
where
∂C S
GS = FS C S − DSη a (B5 a)
∂η
∂C L
G L = FL C L − κDL (B5 b)
∂ξ
dε
FS = −η a +1ε (B5 c)
dt
dε
FL = −κ (b − ε )(1 − ξ ) (B5 d)
dt
Following Patankar [48], the difference equation for concentration at node J (Fig 4.3) can
then be derived as
162
where the coefficients in equation (B6) are as follows
E J = De ( A(| Pe |) + − Pe − 0 ) (B7 a)
WJ = Dw ( A(| Pw |) + Pw − 0 ) (B7 b)
PJ = PJo + E J + WJ (B7 c)
S J = PJo C Jn (B7 d)
and
Fe Fw
Pe = Pw = (B8 a)
De Dw
η Ja ε 2
P =
J
o
∆η J for substrate region (B8b)
∆t
κ (b − ε ) 2
PJo = ∆η J for solder region (B8c)
∆t
The definitions for De, Dw, Fe and Fw are given as follows for substrate region
η ea DSe η wa DSw
De = Dw = (B9a)
δη e δη w
dε dε
Fe = −η ea+1ε Fw = −η wa +1ε (B9b)
dt dt
κDLe κDLw
De = Dw = (B9c)
δξ e δξ w
163
dε dε
Fe = −κ (b − ε )(1 − ξ e ) Fw = −κ (b − ε )(1 − ξ w ) (B9d)
dt dt
The function A(|P|) in equations (B7) was derived by Patankar [48] and depends on the
selection of a differencing scheme. For the power-law scheme used in the present study,
164