Académique Documents
Professionnel Documents
Culture Documents
By
JIA LIU
DOCTOR OF PHILOSOPHY
MAY 2013
The members of the Committee appointed to examine the dissertation of JIA LIU
__________________________________________
Indranath Dutta, Ph.D., Chair
__________________________________________
David P. Field, Ph.D.
__________________________________________
Kerry W. Hipps, Ph.D.
ii
ACKNOWLEDGEMENTS
I sincerely appreciate my advisor, Dr. Indranath Dutta, for his guidance and support
throughout my PhD study and research during the past four and half years. His high expectations
Thanks are due to Dr. Leonel Arana (Intel Co.) for the suggestion on using Au as an
interfacial layer in Cu-In system, to Dr. Rishi Raj (University of Colorado) for suggesting the use
of atomic layer deposited Al2O3 at the Cu/In interface, and to both Dr. Ravi Mahajan and Dr.
Mukul Renavikar (Intel Co.) for suggesting combination of Au and Al2O3 at Cu/In interface. I
would also like to thank Dr. Theva Thevuthasan for providing me the access to all the advanced
I would like to appreciate Dr. David Field, Dr. Kerry Hipps, Dr. Uttara Sahaym, and Dr.
Praveen Kumar for their intellectual contribution to my research progress. ,I would also like to
thank Dr. Daniel Perea, Dr. Dan Schreiber, Dr. Zihua Zhu, Dr. Tamas Vagar and Dr. Bruce Arey
I would also like to appreciate Miles Pepper, Kurt Hutchinson, Gary Held, Henry Ruff,
and Joshah Jennings for their time and technical expertise contributed, my colleagues Zhe,
Kathryn, and Babak for their help and support, and Frederick Brinlee (Research Education
Undergraduates, REU) during summer intern for the contribution on the optimization of thermal
measurement apparatus.
Last but not the least, I would like to thank my family for their emotional support and
encouragement regardless to the location and time. My parents, Wei and Ying, and my
grandfather, Jianchen, have always shown unconditional love even if I could not spend too much
iii
time with them. I would also like to thank my close friends, Yiwen and Hui, for the support of
helping me overcome a period of tough time in my life so that I can focus on my research.
This research was supported by a grant from INTEL Corporation through the Strategic
Research Segment (SRS) program. Partial support from NSF-CMMI-0709506 and NSF-DMR-
0939392 is also acknowledged. The author would also like to acknowledge the financial support
from EMSL as well as instrument time under the user proposal #44754.
iv
LIQUID PHASE SINTERED COMPOSITE SOLDERS FOR NEXT GENERATION
Abstract
It is undeniable that electronics are becoming increasingly powerful and that there is
continual effort towards miniaturization of these devices and thus increasing heat generation
requires a new paradigm in thermal interface materials (TIM) design. This work was aimed at
optimizing the processing parameters and characterizing the performance of Cu-In composite
solders produced by liquid phase sintering (LPS). These composites comprise a high-melting
phase (HMP) such as Cu embedded in a matrix of a low-melting phase (LMP) such as In. Copper
contributes to high thermal and electrical conductivity of composites, whereas the soft In matrix
helps maintain high shear compliance. This combination of high thermal/electrical conductivities
and high shear compliance makes these solders suitable for a range of next-generation thermal
After considering a range of compositions, a solder with 60 volume percent In was found
to possess the requisite combination of high compliance and high conductivity. During the study,
interfacial engineering was introduced to slow down the reaction between Cu and In, and hence
further improve the performance of composite solders. A dual interfacial layer consisting of
Al2O3 and Au was used to mitigate the reaction between Cu and In. A 1 nm Al2O3 layer was used
v
as a diffusion barrier to prohibit the inter-diffusion between Cu and In, while a 20 nm Au layer
was coated on top of the ceramic Al2O3 for wetting enhancement. The dual layer increased the
thermal conductivity of the solder by a factor of ~2 while reducing the yield strength to make the
solder more compliant. The effects of particle size, shape and volume fraction were also studied,
and a simple model was utilized to explain the trends in the mechanical and the thermal
properties.
The optimized Cu-In composite solders were further used to study the performance of
solder joints. Mechanical properties under shear and joint thermal resistance were measured as a
function of joint thickness, thermal excursion history, and different inter-layers between solder
and Cu. Minimizing interfacial contact-resistance is desired, especially when the joint thickness
becomes sub-millimeter, and hence role of inter-layer on the contact resistance was studied.
vi
TABLE OF CONTENTS
Page
vii
4.4.2. CHARACTERIZATION OF THERMAL PROPERTY ........................................ 30
viii
5.3.5. EFFECT OF Au INTERFACIAL LAYER THICKNESS ON THE
ix
LIST OF TABLES
x
LIST OF FIGURES
Chapter 1:
Figure 1-1 Schematic of idealized LPS solder microstructure with a uniform distribution of
Chapter 2:
Figure 2-2 A schematic structure representing the bonding condition of TIM to two components
......................................................................................................................................................... 5
Figure 2-3 Binary Cu-In phase diagram proposed by Bolcavage et al.[35] ................................ 10
Figure 2-4 Binary Au-In phase diagram by Okamoto et al. [44]................................................. 12
Figure 2-5 The ALD mechanism for the deposition of Al2O3 film from TMA and H2O. ........... 16
Chapter 4:
Figure 4-1 Microstructure of the as-received (a) 40μm spherical Cu particles, (b) 40μm irregular
Figure 4-2 Schematic structure of hardened steel die for green pellet processing ...................... 25
Figure 4-3 Schematic structure for multi-layer coating on Cu particles, imbedded in In. .......... 27
Figure 4-4 (a). Lift out of the Cu/In layered sample. (b). TEM grid for welding the specimen.
(c). Sample shape after 30kV cleaning. d. top view of sample shape after final thinning............ 29
Figure 4-5 Schematic of the test apparatus and specimen dimensions ........................................ 31
Figure 4-6 Rendering of experimental facility developed in the present study ........................... 32
Figure 4-7 Schematic of instrumentation and control of experimental facility ........................... 34
xi
Figure 4-8 Structure of the impression test fixture equipped in DMA ........................................ 35
Figure 4-9 Structure of the shear test set-up and the image of test set-up with loaded sample ... 36
Chapter 5:
Figure 5-1 Schematic of LPS solder, containing a high melting phase (HMP) of grain size D and
current I is utilized as a unit cell for the model [3]. ...................................................................... 39
Figure 5-2 Prediction of dependence of Ksolder on f, R thc and D for LPS Cu-In solder. .............. 41
Figure 5-3 A parametric study showing the effect of the particle size of Cu on the thermal
conductivity of the solder, for various thermal contact resistances .............................................. 43
Figure 5-4 SEM micrographs showing the cross-section of (a) non-spherical Cu-50 vol.%In and
(b) spherical Cu-50 vol.%In composite solders. The solders were sintered at 160oC for 45s. The
dark areas represent Cu and the light areas represent In. Arrows point to some of the pores. ..... 45
Figure 5-5 A few representative stress-strain plots showing the effect of particle shape on the
Figure 5-6 Variation in the yield strength and the electrical resistivity of the Cu-In based solders
with In volume fraction. The shaded region in the plot shows a sweet spot where both optimum
Figure 5-7 Variation in the yield strength of Cu-In composite solders with matrix volume
Figure 5-8 The effect of sintering conditions on the extent of densification during LPS for both
(a) non-spherical particles, and (b) spherical particles. The lines in the plots are interpolate fits, to
xii
Figure 5-9 The effect of (a) volume fraction of In and (b) sintering temperature on the sintered
relative density of the Cu-In solders. The lines in the plots are interpolate fits, to guide the eyes
....................................................................................................................................................... 56
Figure 5-10 XRD spectrum of the etched Cu-In and Cu(Au)-In layered samples showing the
relevant IMC phases formed during sintering at 165oC for 45s. The upper spectrum corresponds
to Cu-In sample whereas the lower one corresponds to Cu(Au)-In. Peaks corresponding to IMC
Figure 5-11 (a) Effect of particle size of Cu on the yield strength, and the thermal conductivity
of composites with 60vol% of In. All the samples were sintered at 160 oC for 15s and (b)
variation in the yield strength of Cu-60vol.%In composite solders with particle size of
Figure 5-12 (a)-(d) Microstructure of direct immersion Au coated Cu substrate with various Au
thicknesses, such as (a) 5 nm, (b) 10 nm, (c) 15 nm, and (d) 20 nm. SE image (e) and topography
(f) showing some defects in the Au coating when Au thickness is 20 nm. .................................. 62
Figure 5-13 Microstructure of dual layer coated Cu substrate. The stacking sequence is 20nm
Figure 5-14 Cross-sectional micrographs of the interfaces, showing continuous layers of IMCs
at (a) Cu/In interface, and (b) Cu/200nm Au/In interface. In was then etched away revealing the
IMC morphology at the Cu/In interface: (c) bare Cu substrate and (d) 200nm Au coated Cu
Figure 5-15 Microstructures of as-sintered (a) Cu-In, (b) Cu(Au)-In, and (c) Cu(Al2O3)-In
solders, sintered at 160 oC for 15 s. The black regions indicate pores. ........................................ 67
xiii
Figure 5-16 Glancing incident x-ray diffractograms of (a) Cu/In layered samples, (b) Cu/Au/In
layered samples, and (c) Cu/Al2O3/Au/In layered samples, which were aged at 125oC for 24h and
96h................................................................................................................................................. 70
Figure 5-17 Sequential sputtering using SIMS analysis showing the sequence of IMCs formed at
Figure 5-18 Effect of aging on the yield strength of composites with different interfacial layers
Figure 5-19 Microstructures of the solders with different aging conditions. (a) As-sintered Cu-
In, (b) Cu-In solder after an aging for 96 hrs, (c) As-sintered Cu(200nm Au)-In, (d) Cu(200nm
Au)-In solder after an aging for 96 hrs, (e) As-sintered Cu(1.1nm Al2O3+20nmAu)-In, and (f)
Figure 5-20 Effect of aging on (a) the electrical resistivity and (b) the thermal conductivity of
composites with different interfacial layers between Cu reinforcement and In matrix. ............... 78
Figure 5-21 (a) A Cu(Au)-60vol%In LPS solder joint between two Cu rods, and (b) SE image
showing microstructure of the interface between Cu substrate and LPS solder. A continuous
layer of IMC, as shown by arrows at the prominent places, formed at the interface proving good
chemical bonding between the solder and the Cu substrate. ........................................................ 82
Figure 5-22 Effects of bondline thickness on the thermal resistance of different solder joints and
Figure 5-23 Effect of aging on joint thermal resistance of solders bonded with (a) 50nm Au
Figure 5-24 Shear stress-strain curve of the composite solder joints and In joints ..................... 87
Figure 5-25 Flow stress variation with bondline thickness under shear ...................................... 87
xiv
Figure 5-26 The fractographes of the fractured joint with various bondline thickness under shear
loading........................................................................................................................................... 88
Figure 5-27 Bondline thickness effect on (a) maximum shear stress and (b) maximum
elongation...................................................................................................................................... 89
xv
I. INTRODUCTION
microelectronic devices, the requirements for heat dissipation via thermal interface materials
(TIMs) and current density flow through chip-to-package interconnects (ICs) have drastically
increased. Hence, the packaging of advanced next generation microelectronic devices demands
development of solder alloys with (1) low melting temperature, (2) low flow stresses in shear
with moderate compressive stiffness, and (3) high electrical and thermal conductivities. To
simultaneously attain the above characteristics in a solder, Dutta and co-workers [1-3] proposed
a novel composite architecture, as shown in Figure 1-1. This composite consists of a highly
conductive high melting phase (HMP, e.g. Cu, Sn, etc) uniformly distributed in a highly
complaint and soft low melting phase (LMP, e.g. In, etc) matrix. This architecture proffers high
thermal and electrical conductivities, which are controlled by the high conductivity HMP, along
with low flow stresses in shear, which is controlled by the soft and compliant LMP. The
considerable volume fraction of the hard HMP imparts good compressive strength to the solder.
On the other hand, the shear compliance of this composite solder joints at low strain rate can be
quite high because of localization of strain in the continuous LMP network around the HMP
particles. This high shear compliance is critical for accommodating differential deformation
between the chip and the heat-spreader (for TIM applications), or chip and substrate (for IC
applications). Also, the reflow and the rework temperatures are controlled by the melting
temperature of the LMP, which can be very low as compared to the existing Sn-Ag based lead-
free solders, making these composites suitable for low temperature soldering, especially in
1
HMP LMP
Figure 1-1 Schematic of idealized LPS solder microstructure with a uniform distribution of HMP and
LMP
Recently, liquid phase sintering (LPS) was used to process materials for microelectronic
packaging applications, such as solders and metal-polymer composites which contain a very low
melting phase, such as In and polymers. [1-8]. Dutta and co-workers [1-4] introduced LPS to
produce Sn-In composites with short sintering times (1-2 minutes) at a temperature of 160-162
°C (just above the melting point of indium). By optimizing the In content, a highly compliant Sn-
In solder, with flow stresses close to that of pure In, were produced. The electrical and thermal
conductivity of the LPS solder was found to be approximately half that of pure In. Based on a
semi-empirical model [9], it was demonstrated that for a high conductivity HMP, LPS solders
with very high thermal conductivity may be obtained so long as the HMP particle size is larger
than a critical size (the Kapitza diameter). Hence, a composite system with Cu as the HMP and
Solid Cu can react with liquid In to form several intermetallic compounds (IMCs) at the
Cu/In interface. The most common IMCs, which form in the temperature range of 160 oC to 175
2
o
C, are Cu16In9 and Cu11In9 [10-12]. Since the Gibbs free energy of the formation of Cu16In9 (-
7.58 kJ/mol [11]) is slightly more negative than that of Cu11In9 (-6.79 kJ/mol [11]), Cu16In9 IMC
may preferentially form at the Cu-rich Cu-In interface in the beginning of the sintering process.
However, once a thin continuous layer of Cu-In IMC is formed, the growth of the subsequent
IMC layer is controlled by the diffusion of Cu through the IMC layer [10]. Since the activation
energies for the growth of Cu16In9 and Cu11In9 layers are 59.5 kJ/mol and 16.9 kJ/mol,
respectively [10], the growth rate of Cu11In9 in the later stages of sintering is much faster as
compared to Cu16In9. The formed interfacial IMC layer continues to grow during thermal aging.
During thermal aging at high temperature (>100 oC), CuIn may also form along with Cu16In9 and
Cu11In9 [10]. However, the growth rate of Cu11In9 during isothermal aging is also much faster
than that of other Cu-In IMCs [12] and hence the thickness of the interfacial IMC layer between
The reaction between Cu and In can affect the mechanical properties and thermal
properties of Cu-In composite. However, the reaction can be retarded by: (i) minimizing the total
time available for reaction by minimizing the time spent at the sintering temperature, as well as
the time spent above 125°C following sintering (e.g. via rapid cooling from the sintering
temperature), and (ii) utilizing a thin, deliberately produced reaction layer to serve as an
3
II. BACKGROUND
thermal interface material (TIM) is the component of a package, which attaches a heat spreader
to the silicon device (chip). Therefore, the essential requirements for TIMs are: (i) high thermal
conductivity to effectively dissipate heat produced by the chips, and (ii) high shear compliance to
sustain large shear displacement due to the coefficient of thermal expansion (CTE) mismatch
between Cu and Si. Unfortunately, high thermal conductivity of TIM alone is not sufficient to
guarantee the optimal system performance, as shown in Figure 2-2, which shows the situation
when TIM is bonded to two surfaces with finite bondline thickness (BLT), and at the interface
they do not completely fill up the voids. This situation represents most of the TIM bonding
condition in reality. Therefore, the thermal performance of TIMs can be written as:
BLT
RTIM = + Rc1 + Rc 2
kTIM (2-1)
where Rc represents the contact resistance between TIM and each bounding surface, and RTIM is
the sum of thermal resistances contributed by the solder and two bounding surfaces. Both Rc and
RTIM have the unit of cm2-K/W, which is the thermal resistance normalized to a unit area on one
square cm. Based on Equation (2-1), the thermal performance of a TIM will be enhanced by
higher thermal conductivity, lower bondline thickness and lower thermal contact resistance.
Especially for reducing Rc, it is important to eliminate interstitial air gaps from the thermal
interface by conforming to the rough and uneven mating surfaces. In addition to the thermal
performance, there are some other critical criteria for selecting TIM, i.e. ease of use in assembly
4
Cu Heat Spreader
TIM
Si
IC
PCB
Figure 2-2 A schematic structure representing the bonding condition of TIM to two components
thermal paste, and solders. Grease was first to be used as a TIM [13, 14]. It is typically silicone
based matrix loaded with particles to enhance thermal conductivity. Due to the low viscosity, it
can easily fill the air gap between the silicon chips and the heat spreaders, and increase the
contact area between two surfaces; hence, reduce thermal resistances at the interfaces. However,
5
due to the coefficient of thermal expansion (CTE) mismatch between the silicon chip and the
metallic heat spreaders, these thermal greases are often squeezed out when the chips are turned
on and off, in which the grease is forced out of the interface, potentially contaminating the
neighboring interfaces and components [15]. Also, even after the conductivity of grease is
improved by imparting metallic fillers, its thermal conductivity increases up to 7 W/m-K [16],
which limits the efficiency of heat dissipation from the silicon chip.
Besides greases, phase change materials (PCMs) have received considerable attention
during the last decade. PCM have very large latent heat change for solid-to-liquid and liquid-to-
solid transaction at a certain temperature. Hence, they are capable of storing and releasing large
amount of energy. However, most PCMs with high-density storage have unacceptable low
thermal conductivity. Due to the higher viscosity as compared to grease, they generally give
Thermal pastes are predominantly based on polymers, particularly silicone, filled with
high density thermal enhancing particles, such as Ag, Al2O3, and Al. As compared to grease and
PCM, one of the biggest advantages is that there is no pump-out or migration concerns for pastes.
In current packages, metallic solders are often used as TIMs. These solders have
resistances smaller than 0.05 cm2-K/W [17], which is much smaller as compared to that of
traditional greases, PCM and pastes, whose thermal resistances are typically ~1 cm2-K/W [17] or
the latest greases, whose thermal resistances are typically 0.15 cm2-K/W [17]. Indium and
indium based solders have been widely used as TIMs recently. They have the advantages of: (i)
high thermal conductivity (up to 60 W/m-K) [18] making them suitable for the high power
devices with heat dissipation in excess of 1000 W [19], (ii) high shear compliance to
6
accommodate the CTE mismatch between the chips and the heat spreaders without applying
significant stresses on either components, and (iii) low reflow and rework temperature due to the
low melting temperature of In. However, the thermal conductivity of all the indium-based
solders is limited to 60 W/m-K [18] and due to the high compliance of In, the attachment of heat
Recently, Dutta and co-workers introduced LPS to produce Sn-In, Cu-Bi, Al-In, and Cu-
In composite solders [1-4]. By tailoring the In content, the thermal conductivity and shear
compliance can be adjusted to meet the different requirements of TIM in various working
environment.
Nanotechnology has also been introduced in the latest research on TIMs. Since CNTs
have very high thermal conductivity, various CNT-based composite has been proposed and
evaluated [20-24]. However, due to CNT’s small size, the resistance at the interfaces between
CNT and matrix can significantly affect the apparent thermal performance of the composite. Xu
and Fisher [25, 26] grew CNTs directly on the back of Si and then used the heat spreader to press
against the as-grown CNTs. They also used PCM to fill up the gaps between CNTs in order to
reduce the thermal resistance [27]. Later, Hu et al. [27] measured the thermal resistance of these
CNT-PCM composite TIM, which varied from 0.14 to 0.17 cm2-K/W at different testing
temperatures and pressures. These values look very promising. However, the reliability and
performance degradation of these TIMs have not been fully studied yet.
As shown in Figure 2-1, the solder interconnects (ICs) are used to connect the silicon
device, substrate and the PCB to mainly provide the electrical interconnection and mechanical
stability. The most commonly used ICs are Sn-Ag-Cu (SAC) based lead-free alloys [28]. The
minimum reflow temperature of SAC solders is ~235 oC, and their typical electrical resistivity is
7
about 160 nΩ-m, which is twice as much as that of In. Also, these solders are much stiffer than
usual TIMs [28]. Hence, they can accommodate limited amount of differential displacement
produced in the silicon device, substrate and the PCB due to any CTE mismatch. This makes
Liquid phase sintering (LPS) is widely used for densification of hard-to-process and
refractory materials, such as carbides, tungsten, and various ceramics to produce structural
[29-34]. LPS is a low temperature process to form high performance, multiple-phase composites
from powders. During LPS, the liquid phase provides sufficient external force in the form of
capillary force to assist the densification. In a typical LPS process, there are three stages to reach
full densification: firstly, particle rearrangement occurs rapidly due to the capillary force of the
which mass transfer between HMP particles occurs via the liquid phase to reduce the total
surface energy, resulting in rounder HMP particles; and thirdly, grain growth occurs during
which a solid skeleton structure is formed and the densification is very slow [8]. The last stage is
often referred to as solid state sintering [8]. The first stage of the LPS is very rapid [8] and can
produce almost fully dense pellets if the powder preparation and compaction process is
optimized [4]. Hence, short-time sintering is often used during transient LPS, where the liquid is
prone to either react with the solid phase [4-6, 32] or may diffuse into solid phase to form solid
solution [7]. Recently, LPS is being widely used to process materials for microelectronic
8
packaging applications, such as solders, metal-polymer composites, etc., which contain a very
Indium-based alloys are potential candidates for Pb-free solders in electronic packaging,
which attract significant attention in industry due to their environmental friendliness. Copper, is
used as the common substrate in the electronic packaging, which is in direct contact with the In-
based solder. It is therefore important to obtain in-depth knowledge of the reaction between Cu
and In, which may further affect the performance of solder joint and their reliabilities.
The Cu-In phase diagram, as shown in Figure 2-3, has been under study since 1930s. In
1989, Subramanian et al. [35] reviewed the phase relations based on the early work [36-43],
which agrees very well with their latest results [35] for In content <35 at.%. However, in the In
content range of 35-40 at.%, especially in the range composed of η and η’, Jain el al. [43] found
at least five different phases, which were not existing in the study by Hume-Rothery [37], Owen
[38], and Bolcavage [35]. In 1994, Okamoto et al. [44] proved that there are only two stable
phases in the range of 35-38 In at.%, which are η and η’. The phase η is denoted by Cu0.64In0.36
In 1934, Weibke et al. [36] first reported the phase ψ, which was determined later by
Rajasekharan et al. [45] to be Cu11In9. In the light of Bolcavage [35], this phase is stable at room
temperature. Another phase, CuIn2, was also found to form at room temperature by Keppner [46]
and Chen [47]. Bolcavage et al. [35] studied bulk samples annealed at 250 oC for up to 45 days
9
by DSC
C and did nnot find a transition ccorrespondinng to CuInn2. Thereforre, CuIn2 m
might be a
T basic tthermodynaamic and kinetics propperties of Cuu/In reactioon have beeen studied
The
2 2 m
G = ∑ xi Gi0 + RT ∑ xi ln ( xi ) + xCu xIn ∑ L(Cu,
j)
,In ( xCu − xIn )
j
i =1 i =1 j =1
(2-2)
where
10
L(Cu,In
j)
= Aj + B jT + C jT ln (T )
(2-3)
and xi is the mole fraction of the component element I, Gi0 is the molar Gibbs free energy of
( ) j
pure element I, and LCu,In is the interaction parameter. A j , B j and C j are parameters to be
Based on the study by Liu et al. [11], thermodynamic parameters of three stable Cu-In
The kinetic study of Cu-In study has been conducted since 1970s due to its importance in
electronic systems. Barnard [48] investigated the Cu-In diffusion couple produced by the
electroplating of indium on the Cu substrate. A single Cu/In alloy, i.e. Cu11In9, was found at the
interface. The growth kinetics of Cu11In9 follows the parabolic trend with the activation energy
of 24 kJ/mol. Manna [49] studied the solid state reaction between vapor deposited In and single
crystal Cu substrate at the temperature range from 100 oC to 150 oC. The same IMC, i.e. Cu11In9,
was found with the formation activation energy of 26 kJ/mol [49]. In 1995, Vianco et al. [50]
studied the solid state reaction between hot dipped pure In coating and Cu and found an
Other researchers studied the reaction between liquid In and Cu. Kao [51] dipped a Cu
pieces into molten In and heat treated at 200 oC, Cu11In9 was identified. Su et al. [52] found the
same Cu11In9 IMC after heating Cu/In couple at 300 oC. Yu et al. [10] found the formation of
11
CuIn when the Cu/In couple wass heated below 300 oC
both Cuu11In9 and C C, while Cuu16In9 was
found when
w the agiing temperaature is morre than 300 oC. The acttivation eneergy of CuInn, Cu11In9,
and Cu116In9 were 23.5 kJ/mol, 16.9 kJ/mool, and 59.5 kkJ/mol, resppectively.
2.3.22 REA
ACTION A
AT Au/In IN
NTERFACE
ES
attachm
ment. In addiition, Au alsso plays an important
i roole as the oxxidation prootective layeer in under
bump m
metallurgy (U
UBM) and ttop surface m
metallurgy (TSM). Figuure 2-4 show
ws the phasse diagram
of the A
Au-In binaryy system [444].
12
To assess the thermodynamic properties of liquid in Au-In system, Equation (2-2) and (2-
3) were used. After optimization of all the parameters in both equations, the thermodynamic
equations for Au-In alloys at the temperature range of 0-200 oC are expressed as [53]:
(2-7)
Kinetics study of Au/In interfacial reaction was focused on solid-state reactions. Powell
and Braun [54] showed four IMCs, AuIn2, AuIn, Au7In3, and Au4In at the Au/In interface below
the melting point of In. Simic et al [55] pointed out that the Au-In reaction can take place quickly
at room temperature, and the type of compounds relied much on the ratio of Au/In thickness ratio
and diffusion time, whereas AuIn2 was always the first phase formed. Bjøntegaard [56] showed
that only AuIn2 was formed when In atoms were diffused through preformed AuIn2 layer and
react with Au atoms. The reaction kinetics followed parabolic law which indicates that the
reaction is diffusion-controlled process. Liu et al. [57] studied reaction between Au and liquid In,
which shows a continuous AuIn2 layer formed at the interface. The growth of AuIn2, above the
melting point of In, was also proved to be diffusion-controlled process with the activation energy
of 39.42 kJ/mol.
13
In general, chemical reactions at the solid-liquid interface may occur when the surface
energy can be substantially reduced after reaction [58]. The amount of decrease in the surface
energy depends on the free energy per unit area of the chemical reaction [58]. A reaction aided
wetting can cause spreading of the liquid on the solid. However, the amount of the reaction
product formed during liquid phase sintering depends on the diffusivity of the reacting species
through the intermediate phase formed between the liquid and the solid. The kinetics of the
reaction at a solid-liquid interface depends on both the dissolution of solid into the liquid and the
reaction rate [58]. If dissolution is the rate-controlling step, then a sequence of equilibrium
phases, as shown on the corresponding phase diagram, will form at the interface [58]. However,
if the reaction is the rate-controlling process, then the phase with the lowest free energy of
formation will form first [58]. Once formed, this phase will then hinder the inter-diffusion and
homogenization of the microstructure due to the low diffusivity of the reacting species through
it.
Atomic layer deposition (ALD) is a surface-controlled thin film growth technique for
which the formation of the film is atomically controlled. The growth of the film is achieved by
introducing one element of the compound at a time reacts with the growing surface. Because this
is a self-limiting reaction, i.e. the surface will accept one atom at a time, it results in an inherent
homogeneity of the thin film. This technique, known as atomic layer epitaxy (ALE), was first
developed in 1974 by Tuomo Suntola in Finland. It was first used to grow ZnS, a common
material in electroluminescent devices. During the process, the Zn atom from the ZnCl2 vapor
14
will form an ion bond to the O atom on the surface of the substrate (glass or oxide thin film
surface). The ZnCl2 surface is then exposed to the other reactant H2S. The Zn-S bond ensures the
monoatomic S surface coverage. Then the process is repeated. The morphology of the ZnS film
was studied using x-ray diffraction and TEM. It was confirmed that with the deposition
temperature of 300oC and film thickness of 30nm, the film has an extremely well oriented cubic
phase[59]. These films were all grown on amorphous substrates. In about 2000, the name ALE
was replaced by atomic layer deposition (ALD) since most of the films grown were not epitaxial
to their underlying substrates. Many amorphous films were grown using ALD, which were more
ALD of Al2O3 has been developed as a model system because the surface reactions are
very efficient and self-limiting. ALD of Al2O3 is usually performed using trimethylaluminum
(TMA) and H2O. The overall reaction at the substrate surface is described as:
where the starred species are the surface species. Because the reaction happened as
ABAB…sequence, for each half-reaction, e.g. Equation (2-12b), the reactant H2O has limited
number of reactive sites, i.e. AlCH3, the reaction will continue until all the sites have been
reacted and no further. This is the reason why the ALD Al2O3 can be coated uniformly and under
controlled rate. In addition, it was found that –OH termination on the substrate was required for
the nucleation of Al2O3. The Al atom in Al(CH3)3 vapor will bond to the O atom at the surface
15
and initiate the reaction. Copper is used as the substrate material for ALD Al2O3. The ALD
mechanism for the deposition of Al2O3 film from TMA and H2O are shown in Figure 2-5.
Al Al
OH OH OH CH4 O O O
Al
CH4
H2O
OH OH OH
Al Al
O O O
Figure 2-5 The ALD mechanism for the deposition of Al2O3 film from TMA and H2O.
ALD is also a promising technique to deposit uniform and conformal coating on particles
with diameter from a few hundred microns to one-tenth micron [63, 65]. The deposition on Cu
particles was achieved by using fluidized bed reactor technology. A set of particles will sit in a
fluidized bed. While ALD reactants are passing through the particles area, the agitation of the
particles enhances the contact between precursors and particle surfaces. In the meantime, it can
also prevent the individual particles from being “glued together” in the particle bed by the ALD
coating.
16
The growth rate of Al2O3 at 450K has been shown to be ~1.1Å per AB cycle [61]. TEM
study of the as-deposited Al2O3 at 450K is amorphous and was deposited equally well on the
basal planes and edges of the particles [63]. The film was conformal and free of pinholes. It was
also confirmed that the ALD Al2O3 film followed the stoichiometry when the thickness is about
10 nm [66]. No information was mentioned about the stoichiometry when the film thickness of
diffusion behavior of Cu atoms. The pre-factor of Cu diffusion is 5.8x10-4 cm2/s and the
activation energy is 0.9eV [67]. By inserting these values into the Arrhenius equation:
⎛ E ⎞
D = D0 exp ⎜ − a ⎟ (2-13)
⎝ k BT ⎠
where D0 is the diffusion pre-factor, Ea is the activation energy, kB is the Boltzmann constant and
T is the temperature. Using the temperature of 125 oC (or 398K), the diffusivity of Cu in
amorphous Al2O3 is 2.3x10-15 cm2/s, while the interdiffusivity of Cu and In is 2.1x10-8 cm2/s [12].
Assuming the Cu and amorphous Al2O3 is kept at 125 oC for 24h, the distance which Cu diffuses
through Al2O3 is 0.014 μm. This distance is two order magnitudes slower than the distance due
to interdiffusion of Cu and In, i.e. 1.75 μm. It is then a promising evidence that ALD Al2O3 can
A first estimation for the thermal conductivity of a composite is the rule of mixture. If the
composite contains two different identical phases, named as A and B, which have the thermal
17
conductivities of KA and KB, and volume fractions of VA and VB, respectively, the thermal
1 V A VB
= + (2-14)
K composite K A KB
In this equation, only the thermal conductivity and volume fraction of each component
are expected to have an effect on the overall thermal conductivity of the composite. However,
many researchers demonstrated that the size and shape of the reinforcement also play an
important role in determining the thermal conductivity of the composite [9, 68-71]. Every et al.
found that with the constant volume fraction of particles in the composite, the thermal
particles, but lowered after adding sub-micron size particles of diamond. This effect was
attributed to the interfacial thermal resistance between the reinforcement and the matrix.
In 1936, Kurti first discovered a temperature discontinuity at the interface between liquid
helium and copper surface [72], which provided the experimental evidence for the thermal
resistance at the interface. In 1941, Kapitza carried out the first measurements of thermal
resistance at the interface [73], which was named as thermal boundary resistance in later work.
Based on this definition, the thermal boundary resistance, RBd, is expressed as:
ΔT
RBd =
Qinterface (2-153)
where Qinterface represents the amount of heat flowing through the measured interface and ∆T is
the temperature discontinuity across the interface. The existence of thermal boundary resistance
is due to the difference in electronic and vibrational properties in different materials, when an
18
energy carrier (electron or phonon) tries to traverse the interface, it will scatter at the interface.
The probability of the transmission after scattering determines the thermal boundary resistance.
In 1992, Every et al. [9] built a phenomenological model to draw a link between simple
experiments with planar geometries of interfaces and experiments with particulate composite
where interfaces are distributed randomly but uniformly based on the microstructure. In the
model, the boundary resistance was put forwarded into the average thermal conductivity of the
particles. In this geometry the thermal conductivity of the composite becomes particle size
dependent. The thermal interfacial resistance is then expressed as an effective Kapitza radius,
α k , which is:
αk = RBd Km (2-16)
2α k
α = (2-17)
d
where d is the diameter of the dispersed particles. The thermal conductivity of the composite is
then analyzed based on the α value. Thus, in a particulate composite, when α = 1 , i.e. the actual
particle radius is equal to the Kapitza radius, the effective thermal conductivity of the particles is
equal to that of the matrix. Thus, the reinforcement particles have no effect on the thermal
conductivity of the composite. If α > 1 , the actual particle radius is smaller than the Kapitza
radius, the thermal interfacial resistance becomes a dominant factor in determining the thermal
conductivity of the composite. If α < 1 , i.e. the composite has large particle size, the thermal
conductivity of the bulk particle material can enhance the thermal conductivity of the composite.
To calculate the thermal conductivity of the composite, effective media theory was
developed. Hasselman and Johnson [69] and Benvensite [74] modified Maxwell’s theory [75]
19
and considered the thermal conductivity of a composite with low volume fraction reinforcement.
K c [ K d (1 + 2α ) + 2 K m ] + 2 f [ K d (1 − α ) − K m ]
= (2-18)
K m [ K d (1 + 2α ) + 2 K m ] − f [ K d (1 − α ) − K m ]
where K c , Kd and K m are the thermal conductivity of composite, particles, and matrix,
When the volume fraction of spherical particles is high, Bruggeman theory provided a
good estimation of thermal conductivity of the composite using the following equation:
(1+ 2α )/(1−α ) 3/(1−α )
⎧K ⎫ ⎧ K c − K d (1 − α ) ⎫
(1 − f ) = ⎨ m ⎬
3
⎨ ⎬ (2-19)
⎩ Kc ⎭ ⎩ K m − K d (1 − α ) ⎭
Metal matrix composite (MMC) can be tailored to meet various requirements for specific
systems. However, based on the interfacial thermal resistance theory, the interface between the
matrix and reinforcement can significantly affect the performance during fabrication and long-
term usage. The common issues associated with increasing interfacial resistance are the chemical
reaction at the interfaces, degradation of reinforcement, and poor wetting of the reinforcements
with matrix.
20
III. OBJECTIVE AND INTELLECTRUAL MERIT
3.1. OBJECTIVE
In this study, a new paradigm in solder TIM and IC design is proposed, where properties
which are typically mutually exclusive (such as, very high thermal conductivity, high shear
compliance, high compressive flow stress and low reflow temperature) will be obtained
simultaneously by proper selection and structuring of the HMP and LMP constituents.
Systematic model-based sintering kinetics and mechanistic studies will be conducted to refine
the LPS process for solders to yield maximum densification. The kinetics of the interfacial
reactions will be studied and optimization of the interface through chemical treatment or
interfacial coating will be conducted to nullify the undesirable effects of the interfacial reactions
on the properties of the solders. Unification of models showing the effects of particle sizes and
relative volume fractions of the HMP and LMP on the thermal conductivity [1, 2, 9, 19, 76] and
yield strength [77], and incorporation of the effects of the interfacial IMCs and pores in the
model will conducted to optimize the mechanical and thermal properties of the solders.
The improvement of LPS solders in the performance is not simply dependent on the
properties of HMP and LMP and rule of mixture often lead to erroneous predictions. Structural
factors, such as the particle shape, size and volume fraction of HMP, and the interfacial
interaction between HMP and LMP, can dominate the properties of the LPS solders. This study
will provide insight into the interfacial interactions between HMP and LMP, both during
sintering and isothermal aging, and their effects on the mechanical, electrical and thermal
21
properties of Cu-In based solders and solder joints. The experimental as well as the analytical
study of the effects of interfacial IMCs on the mechanical, thermal and electrical properties of
The introduction of a deliberate interfacial layer at the interface between the reinforcing
particles and the matrix is a novel approach for transient liquid phase sintering. This approach
enhances the sintering by improving wetting and stabilizes the microstructure during sintering
and aging by inhibiting reaction between Cu and In; hence, improves the mechanical, thermal
and electrical properties of the solders. The effects of interfacial layer on the reaction kinetics
and the growth of the IMCs are challenging, given the low volume fraction of the layer as
compared to the reacting species, and will provide important insights into liquid as well as solid
The outcome of this study will be significant as it will provide not only the better
understanding of the effects of the interface on the mechanical, electrical and thermal properties
of a composite but will also propose a revolutionary technology for the production of new
22
IV. EXPERIMENTAL PROCEDURE
4.1. MATERIALS
Cu and In powders were used to fabricate Cu-In composite solders. Copper powders with
both spherical and non-spherical particles of different sizes were used. Table 1 summarizes the
characteristics of raw materials. 1 and 40μm spherical powders were obtained from Accumet
Materials Corporation, USA, while 10μm spherical powder was supplied by Alfa Aesar, USA.
Another batch of Cu powder with non-spherical particles with the particle size less than
40μmwas also supplied by Accumet Materials Corporation, USA. Indium powder with 10μm
spherical particles was supplied by AM&M Advanced Machine and Materials Inc., Canada.
Figure 4-1 shows the microstructure of the as-received Cu particles and In particles. It clearly
shows the 40 µm Cu particles with spherical shape (Figure 4-1a) and irregular shape of Cu
Average
Materials Particle Shape Purity Vendor
Particle Size
23
AM&M Addvanced Maachine and
In pow
wder Sppherical 10μm
m 999.99%
Materiials Inc., Caanada
Figure 4-1 Microstructure of tthe as-received (a) 40μm spherical Cuu particles, (bb) 40μm irreggular Cu
parrticles, and (cc) 10μm spheerical In partticles.
4.2. S
SAMPLE P
PREPARATION
R
Removal o oxide layyer: Beforee processingg, Cu powdder was etchhed using 10
of 1 % HCl
24
P
Powder miixing: An appropriate
a amount of Cu powderr was mixedd with In ppowder by
G
Green pelleet formatioon: After thee Cu and Inn powders w
were thorougghly mixed,, a desired
amount of mixed ppowder weree transferredd in a lubriccated hardeened steel diie (see Figuure 4-2) in
1mm heeight. The vvolume of ggreen pelletss remained constant, annd the densiity was conntrolled by
modifyiing the amouunt of mixed powder addded into thhe steel die.
Figure 44-2 Schemattic structure oof hardened ssteel die for ggreen pellet pprocessing
under different connditions. The first masss, m1, is meaasured whenn the samplle is dry andd lying on
top of thhe balance; while the seecond mass, m2, is meaasured whenn the samplee is immersed in pure
25
water. Based on Archimedes principle, the mass difference, m1-m2, is equal to the weight of the
water displaced by the measured sample. So the density of the sample is calculated as:
m1
ρ LPSS = × ρ H 2O
m1 − m2 (4-1)
It is obvious that the volume measured by this method excludes all the open voids at the
sample surface and minimizes the measurement error when the sample shape is deviated from
Flash liquid phase sintering: The green pellets were wrapped in Al foil after density
measurement , and liquid phase sintered at 160 oC by immersing them in pre-heated silicone-oil
bath for different lengths of time (15 s, 30 s and 45 s), followed by quenching in water.
To study the impact of interfacial layers on the properties of the Cu-In solders, two types
of interfacial layers were selected: metallic Au, and ceramic aluminum oxide. Hereafter, the
composite solders made with Au coated Cu particles are referred to as Cu(Au)-In and those made
The Au layer was deposited using direct immersion gold (DIG) plating technique, which
gives linear growth rate as a function of time to the maximum thickness of 50 nm. To deposit
thicker Au layer, the electroless gold plating bath following DIG were used to grow Au thin film
up to 2 μm.
Al2O3 thin film was grown using atomic layer deposition (ALD) technique. One of the
advantages of this technique is that it can grow thin film on flat surface as well as the surface of
26
fine powder particles. ALD on particles depends on the ability of ALD to produce conformal
For the dual layer condition, Cu was first coated with Al2O3 followed by the Au layer, see
Figure 4-3. For all the coating condition, the interfacial layers were directly deposited on Cu
In
Au
Al2O3
Cu
Figure 4-3 Schematic structure for multi-layer coating on Cu particles, imbedded in In.
Some of the sintered pellets were aged at 125 oC for 24 and 96 h to study the stability of
the microstructure as well as the evolution in the mechanical, electrical and thermal properties of
Cu-In solders. The sintered solder pellets were sectioned, and then mechanically polished with
intermittent etching with 10% HCl solution up to a finish of 0.05 μm. The cross-sections were
27
observed under a scanning electron microscope (SEM), using backscattered and secondary
electrons (BSE and SE, respectively), as well as energy dispersive X-ray (EDX) spectroscopy.
Some of the layered samples both aged and unaged were prepared using FEI Helios
focused ion beam/scanning electron microscopy (FIB/SEM). The milling is accomplished using
the focused Ga+ ion beam. Live image can be performed by the ion beam as well as the electron
beam. In this instrument, the ion beam column and the electron beam column are positioned
under an angle of 52o with respect to each other. For the cross-section characterization, the
specimens are prepared in a similar way as TEM specimen preparation. The specimen was first
prepared using the wedge cut followed by an in situ lift-out (INLO) technique with only minor
difference in the applied parameters. The basic steps for the technique includes the deposition of
a central protective Pt/C strip and staircase thinning at 7 nA at both sides of the region of interest
until a specimen with a thickness of approximately 1.2 µm is obtained. Right after this step the
specimen is tilted 38o with respect to the ion beam and an undercut is made to release the sample
by the Omniprobe. Next, the specimen is welded to the tip of the Omniprobe and lifted out of the
substrate, see Figure 4-4 (a); the stage is then moved to a different holder that contains the final
gird onto which the specimen is welded. After attaching using in situ Pt deposition, the
Omniprobe is separated from the specimen, see Figure 4-4 (b). In the following, the specimen is
thinned to about 1 µm using the 30-kV ion beam and a current of 0.3 nA. The final milling is
performed using 2-kV beam and a current of 0.1 nA in order to remove the surface damage due
to the Ga+ ion beam. The final shape of the specimen is shown in Figure 4-4 (c). During this
process, the stage is also tilted 1.2o positive and negative with respect to the direction of the ion
28
Figure 44-4 (a). Lift oout of the Cuu/In layered sample.
s (b). T
TEM grid forr welding thee specimen. (c).
( Sample
shapee after 30kV cleaning. d. top
t view of ssample shapee after final thhinning.
29
4.4.2. CHARACTERIZATION OF THERMAL PROPERTY
Thermal properties of Cu-In composites were measured using two sample geometries:(i)
disk shape without bonding to any substrate, and (ii) composite bonded to two Cu substrates on
both sides. Due to the difference in sample geometry, different parameters and techniques were
The effective thermal conductivity (K) of the composite disk was used to determine the
thermal properties of the Cu/In solder. It is determined as K = κCpρs, where ρs is the density of
the solder, κ is the thermal diffusivity, and Cp is the specific heat. The thermal diffusivity (κ) of
solder pellets was measured using a laser flash apparatus (LFA), the specific heat (Cp) was
measured using a differential scanning calorimeter (DSC), and the density of the solder (ρs)
The thermal resistance of solder joints was measured using a self-designed apparatus
based on ASTM D5470 [79], as shown in Figure 4-5. The measurement is based on steady-state
dT
Q = kA
dx (4-2)
eff
Defining the joint thermal resistance, Rjoint , to be the result of resistivity times the
thickness of the joint, in the unit of cm2-K/W, the equation can be rewritten as
AΔT
jo int = ρ l =
R eff
Q (4-3)
30
Figgure 4-5 Schhematic of thhe test apparaatus and speccimen dimenssions
B
Based on A
ASTM D5470 [79], two modificatioons were appplied:
( A bellow
(i) w was used tto provide a vacuum ennvironment ssurroundingg test sample, in order
to minim mple surfacee. The vacuuum pressuree was mainttained at 3.55x10-2 torr
mize heat looss from sam
the joinnt to calculaate heat fluxx through ssolder. The measured tthermal resiistance is thhe sum of
resistancce of Cu-Inn solder. Thhis yields a more stablle and reprooducible vallue becausee of stable
31
contact)). Also, this modificattion enableed us to exxplore the iinformation of our intterest: the
substrattes.
McClinttock [80]. W
When the heeat flow thrrough the soolder is 1.2 W, the unccertainty forr the joint
4
4.4.2.2 DE
ESIGN OF
F APPARA
ATUS FO
OR JOINT MAL RESISTANCE
T THERM
MEASU
UREMENT
T
32
A detailed scale drawing along with the image of the apparatus are shown in Figure 4-6,
while a schematic illustrating the instrumentation is shown in Figure 4-7. The TIM joint to be
tested is sandwiched between two Cu rods (5) and (7) with diameter of 12.7 mm. These rods
have the length of 12.7 mm and 38.1 mm, respectively, and are made of 101 alloy to enhance the
heat transfer and minimize the self-heating due to the thermal resistance. In order to align the
TIM joint, a 2 mm deep and 6mm diameter groove was made at the end of each Cu bar to guide
the joint to be vertically aligned. The bottom Cu bar (5) is bolted to the 25.4 mm x 25.4 mm
water block, which is cooled by a Neslab model Endocal RTE-5B circulator. The circulator
controller detects the temperature with a glass tube thermometer inserted in a grommet located
on the left side of work area, which can be stabilized to ±0.1 oC. The water block is bolted on top
of the load cell (2) through a 12.7 mm thick insulating block and is mounted below the 12.7-mm-
The upper Cu rod (7) is bolted to steel platen (9) with high temperature rubber band to
provide sufficient vacuum sealing. The part which is extruded from the platen (9) is tightly
wrapped by a mica insulated band heater. Another steel platen (12) is bolted to platen (9) using
three supporting rods and is thermally isolated with a 25.4 mm thick alumina silicate machinable
ceramic. The relative position between platen (9) and (12) are then fixed. This allows the upper
Cu rod to be fixed to the upper platen (12) at a set position relative to the lower meter bar.
The upper platens are fitted with ball bushing, preventing any lateral motion. The linear
motion is controlled by a 2C (self-centering) thread fit (14) and stud (16), which converts rotary
motion into smooth, precise linear motion. Three sleeve bearings (10) are mounted on three
precision-ground steel shafts, constraining and guiding the movement of the parts linearly and
33
A typical teest procedurre consistedd of setting the circulaator water teemperature and input
power to the heaater to dessired valuees and incrrementing the actuatoor until thhe desired
were peerformed uusing a dessktop compputer with LabView ssoftware coommunicatinng to the
34
4.4.33. CHA
ARACTER
RIZATION
N OF MECH
HANICAL PROPERT
TY
M
Mechanical properties of solder pellets were characterized using im
mpression teesting [81]
tests weere conducteed at 25oC aat an equivallent uniaxiaal strain rate1 of 3 x 10-44 s-1.
1
Equiivalent uniaxial sttrain rate is equal to δ / φ , wheree δ is the displaceement rate of the punch and φ is tthe diameter of the punch [81].
35
testing machine
m to measure thhe yield streength and fllow stress oof the Cu-Inn solder joinnts with a
36
V. RESULTS AND DISCUSSION
A model was developed for predicting the thermal and electrical transport properties
based on the microstructure of liquid phase sintered solders. In this model, a certain amount of
heat or electrical flux is assumed to be passing through the composite in one direction. As the
flux passes from one side to the other, it encounters several HMP-LMP interfaces, i.e. Cu-In
interfaces. Each of these interfaces is associated with a contact resistance given by Rc for
electrical transport and Rcth for thermal transport. Based on the measurements of solder’s
electrical and thermal conductivity and its overall joint resistance, we estimate the contact
resistance of these internal Cu/In interfaces which were formed during sintering of composite
solders and bonding of the solder to the Cu substrate. These contact resistance values along with
a composite thermal conductivity model are then used to assess the microstructure dependence of
Figure 5-1 shows a schematic sketch of the LPS solder microstructure along with a
current, I, for electrical conduction or a heat flux, q, for thermal transport through a cylindrical
column of cross section A within the solder. For thermal transport, the measured thermal
resistance of the solder in the cylindrical column of cross section A can be written as:
n
th
Rmeas = RCu
th
+ RInth + ∑ Rith (5-1)
37
th th
where Rmeas is the measured thermal resistance of the entire composite solder column, RCu is the
sum of thermal resistance of all the Cu particles within the column, RInth is the total resistance
due to In within the column and Rith is the thermal resistance at each Cu-In interface and n is the
total number of Cu/In interfaces encountered by heat flux in the column. The number of the
n = 2LvCu / D (5-2)
where vCu is the volume fraction of Cu, L is the length of electron or heat transported through the
column, and D is the average particle size of Cu. The total length of all Cu particles within the
composite solder column is equal to LvCu . Based on the definition of contact resistance, Rith is
equal to Rcth / A , where Rcth is the thermal contact resistance due to the Cu/In interfaces and has
Combining Equations (5-1), (5-2) and (5-3), the thermal contact resistance can be obtained as:
D ⎛ 1 v v ⎞
Rcth = ⎜ − Cu − In ⎟
2vCu ⎝ Kmeas KCu KIn ⎠
(5-4)
where K meas is the measured thermal conductivity of composite solder, KCu is the thermal
conductivity of pure Cu, KIn is the thermal conductivity of pure In, and vCu and vIn are the
38
D
Rc = ( ρ meas − ρ Cu vCu − ρ In vIn )
2vCu (5-5)
where ρmeas is the measured electrical resistivity of composite solder, ρ Cu and ρIn are the
I or q
A
HMP
L LMP
I or q
Figure 5-1 Schematic of LPS solder, containing a high melting phase (HMP) of grain size D and an inter-
particle low melting phase (LMP). A cylindrical column of cross-section A carrying a current I is utilized
as a unit cell for the model [3].
Once Rcth is known, the thermal resistance of a composite solder ( R th ), or a solder joint
th
( Rjoint ), can be calculated. In a joint, in addition to all the interfaces in the solder, there are two
more interfaces between In and Cu. Following the procedure outlined earlier, and assuming that
the contact resistance of Cu/In interfaces in the composite is similar to that of Cu/In interfaces at
the joint bonding interface, the thermal resistance of a composite solder is given as:
39
⎛v v ⎞ Lv
Rth = L ⎜ Cu + In ⎟ + 2Rcth Cu
⎝ KCu KIn ⎠ D
(5-6)
⎛v v ⎞ ⎛ Lv ⎞
th
Rjoint = L ⎜ Cu + In ⎟ + 2Rcth ⎜ Cu + 1⎟
⎝ KCu KIn ⎠ ⎝ D ⎠ (5-7)
The value of thermal contact resistance Rcth can be used in conjunction with the model
developed by Every et al. [9], as shown in Equation (2-18) and (2-19), to assess the impact of
volume fraction and particle size of HMP (Cu) on the thermal conductivity of Cu-In composite
solders.
The LPS solder microstructure consists of a high volume fraction of HMP, i.e., Cu
particles, embedded in a continuous matrix of a relatively low volume fraction of LMP, i.e., In
where α k has the dimension of length. It represents the thickness of matrix material that has the
through-thickness thermal resistance equivalent to the interfacial contact resistance R thc . Using
Rcth = 1.5 ×10−7 m2 -K/W , the Kapitza radius α k can be calculated to be 12 μm. We now define
40
2α k K solder K solder K HMP K Cu
α= k= = r= =
D , K LMP K In , and K LMP K In (5-9)
The following equation is obtained for k by combining Equation (2-19) and (5-9):
1+ 2α 3
⎛ 1 ⎞ 1−α ⎡ k − r (1 − α ) ⎤ 1−α
(1 − f ) = ⎜ ⎟
3
⎢ 1 − r (1 − α ) ⎥ (5-10)
⎝k⎠ ⎣ ⎦
where f is the volume fraction of the HMP (i.e., Cu). Solving Equation (5-10), we obtain Ksolder
as a function of (1-f ). Figure 5-2 shows the thermal conductivity, Ksolder of Cu-In composite as a
function of volume fraction of In and particle size of Cu for two different R thc values.
4
Cu-In R
c,1
th -7 2
= 1.5 x 10 m -K/W; α
K,1
= 12 μm
Model R
th -7 2
= 4.5 x 10 m -K/W; α = 36 μm
c,2 K,1
3 th
D=100μm, R
c,1
In
/K
solder
2
k=K
th
D=40μm, R
c,1
th
D=100μm, R
c,2
1 th D=40μm, R
th
D=20μm, R
th D=10μm, R c,2
c,1
c,1
th th
D=20μm, R D=10μm, R
c,2 c,2
0
0.1 0.2 0.3 0.4 0.5 0.6
Volume Fraction of In, (1-f )
th
Figure 5-2 Prediction of dependence of Ksolder on f, R c and D for LPS Cu-In solder.
From Figure 5-2, it is apparent that the thermal conductivity of the LPS Cu-In solder
decreases with increasing In content when the Cu particle size is larger than the Kapitza diameter
41
(D> 2α k ), and increases with increasing In content when D< 2α k . This is because when D> 2α k ,
the impact of interfacial contact resistance is weak and the conductivity is dominated by that of
the high conductive HMP (Cu). When the high conductive Cu is replaced by the lower
conductive In, the thermal conductivity of composite will decrease. Conversely, when D< 2α k ,
the contribution of the interface resistance becomes dominant. The thermal conductivity of the
composite will be lowered by the particles even if the particles themselves have a higher
conductivity than that of the matrix. In the case that D= 2α k , the thermal conductivity of
composite will be unchanged by the particles since the contribution of interface is balanced by
Thus, the advantage of having a highly thermally conductive HMP substance (Cu) can be
realized only when D> 2α k , where the role of Cu becomes dominant and the interfacial
The role of the HMP particle size in relation to the Kapitza radius can be understood in
the following way. When D= 2α k , the HMP thermal conductivity has little effect on composite
conductivity, and the HMP particles simply offer a boundary resistance R thc = α k / Km . If D <
2α k , the thermal conductivity of the Cu-In composite will be even lower than that of LMP In
matrix. Only when D > 2α k , does the positive contribution of the HMP particles becomes
prominent relative to the negative contribution of the contact resistance introduced by having
42
Figure 5-3 shows the variation in the thermal conductivity of solder with particle sizes of
Cu. When D > 2α k , irrespective of the thermal contact resistance, the thermal conductivity of
the solder increases with the particle size of Cu. Interestingly, when the thermal contact
resistance is high, the particle size has a significant effect on the thermal conductivity of the
composite. If the thermal contact resistance is sufficiently low, the composite prepared with even
much smaller Cu powders (e.g. 1μm) can show very high thermal conductivity. In order to make
thin solder joints using smaller HMP particle size, tailoring of thermal contact resistance by
interfacial engineering is essential so that the thermal conductivity of the solder can be
maintained.
th -9 2
R = 4.1x10 m -K/W
c
th -9 2
2 (calc. from the meas. k ) R = 1.4x10 m -K/W
solder c
th -8 2
R = 1.2x10 m -K/W
/K , W/m-K
1.5
th -8 2
R = 4.1x10 m -K/W
In
c
solder
K
Model
Cu(Au)-In
Spherical Cu particles
0.5
0 10 20 30 40
Particle Size of HMP (Cu), μm
Figure 5-3 A parametric study showing the effect of the particle size of Cu on the thermal conductivity of
the solder, for various thermal contact resistances
43
5.1.4. CONCLUSIONS
Contact resistance (electrical and thermal) values for the HMP/LMP interfaces (i.e., Cu-
estimated from the overall resistance value of composites or joints. Based on the deduced contact
thermal resistance values, a model was utilized to predict solder thermal conductivity values as a
function of the LMP (In) volume fraction, HMP (Cu) particle size and the contact resistance of
the HMP/LMP interface. The results indicated that particularly when the HMP has high thermal
conductivity, LPS solders with very high Ksolder may be obtained so long as the grain size of the
Figure 5-4 shows representative microstructures of the liquid phase sintered Cu-In
composite solders with non-spherical and spherical particles. Both mixing and wetting were
adequate, as evident from the complete penetration of liquid In between all Cu particles. The
sintered density of all composites was ~97% and a few scattered pores were always present.
Figure 5-4 shows much less IMC as compared to that reported in ref. [1, 3], and this is attributed
to the improved LPS processing conditions (primarily the reduced sintering and cooling times).
44
Figurre 5-4 SEM micrographs
m showing thee cross-sectioon of (a) non--spherical Cuu-50 vol.%Inn and (b)
sphericcal Cu-50 vol.%In compoosite solders. The solders were sintered at 160oC foor 45s. The ddark areas
represennt Cu and thee light areas rrepresent In. Arrows poinnt to some off the pores.
F
Figure 5-5 demonstrattes the impaact of Cu-pparticle shappe on the strain responnse of the
composite solder coontaining 500 volume peercent In duuring impresssion testingg. All the strress-strain
plots are for uncoaated Cu partticles. Clearrly, when thhe Cu partiicles are nonn-spherical,, the yield
strengthh of the soldder is muchh greater thhan that for spherical pparticles at all test tem
mperatures.
This is largely attriibutable to the greater propensity for mechannical interloocking of addjacent Cu
particless during exttensive plastic flow of tthe intervenning In, althhough some of it may allso be due
non-sphherical partiicles. Sincee an imporrtant goal of the woork is to pproduce sofft, pliable
composites to be suitable for the proposeed TIM andd IC applicaations. Hereeafter, all soolders that
are repoorted in this study were prepared ussing only sppherical Cu ppowders.
45
Figure 5-5 A few representative stress-strain plots showing the effect of particle shape on the strain
response of the as-sintered solders containing 50 vol.% In
Figure 5-6 shows the effect of volume fraction of In on the electrical resistivity and the
yield strength of the Cu-In composite solders. As shown, an increase in the volume fraction of In
increased the electrical resistivity of the solders, which can be attributed to the replacement of
high conductivity phase, Cu, with the poor conductivity phase, In. The trend shown in Figure 5-6
is consistent with the semi-empirical model mentioned in Section 5.1 that predicts a monotonous
decrease in the thermal conductivity of a composite with an increase in the lesser conductive
constituent [2, 3]. Since electrons are the major carriers of the electric current as well as the
thermal energy in metallic materials, variation in the thermal conductivity of a metallic material
46
Figure 5-6 Variation in the yield strength and the electrical resistivity of the Cu-In based solders with In
volume fraction. The shaded region in the plot shows a sweet spot where both optimum yield strength and
electrical resistivity can be achieved.
Figure 5-6 also shows that: (i) when the In volume fraction is relatively small, the yield
strength of the Cu-In solders decreases rapidly with the In volume fraction, and (ii) at high
volume fraction of In (≥60%), the yield strength of the solders shows saturation. Since most of
the strain in the Cu-In solders is localized in the In matrix, the yield strength of the solder should
be determined by the yield strength of In. Since Cu particles act as nucleation sites for the In
grains, the Cu particle volume fraction may affect the yield strength of In via Hall-Patch
relationship [82]. However, yield strength data at high In volume fraction, as shown in Figure 5-
6, suggests that the effect of Cu particle volume fraction on the In grain size (or, Hall-Patch
47
increase in the yield strength of the composite can be induced by an increase in the dislocation
density in In, which may occur due to: (i) the coefficient of thermal expansion (CTE) mismatch
between the reinforcement particle (Cu) and the matrix (In) [77, 82, 83], and (ii) the geometrical
mismatch between the elastically deforming reinforcement particles and the plastically
deforming matrix [84, 85]. The dislocation density is highest next to the particles, and decreases
rapidly with the distance from the interface [83]. Hence, the effects of the above two interface-
based dislocation sources on the strengthening of the composites become relevant only when the
interparticle spacing (i.e. In channel width between two Cu particles) is small. In addition to the
strengthening due to dislocations, concurrent recovery mechanisms which tend to soften the
matrix during testing also operate. The observed yield strength of the composite thus depends on
the superposition of these two effects (i.e., hardening due to dislocations and creep). These
The increase in yield strength due to thermally induced dislocations is given by [77]:
BbΔTΔα (1 − Vm )
ΔσCTE = α1G
d Vm
(5-11)
where Vm (=1-Vr) is the volume fraction of the matrix, Vr is the volume fraction of reinforcement
particles, ΔT is the temperature difference between the heat treatment temperature of the
composite and the room temperature, Δα is the mismatch between CTEs of the particles and the
matrix, b is the Burger’s vector of matrix, G is the shear modulus, α 1 is a constant used in the
Orowan equation ( Δσ = α1Gb ρ , where Δσ is the incremental increase in the yield strength,
and ρ is the dislocation density), B is a geometrical factor, which is equal to ~12 for cubic
48
particles, and d is the particle diameter. The incremental strengthening due to strain mismatch
bε (1 − Vm )
Δσgeo = K1G
d (5-12)
where K1 is a dimensionless constant with typical values ranging from 0.2 to 0.4[84], and ε is the
applied normal strain. Using B = 12, ΔT = sintering temperature – room temperature =135K,
Δα = α In − α Cu = 1.5 ×10−5 K-1 and assuming α 1 ~ 1, K1 = 0.4 [86] and ε ~ the plastic strain at
the yield strength = 0.002, the increase in the yield strength due to CTE mismatch, as predicted
by Equation (5-11) will be at least one order of magnitude larger than that due to the geometrical
mismatch, as predicted by Equation (5-12), for the volume fraction of matrix (In) used in this
study. Hence, Equation (5-11) adequately predicts the increment in the yield strength of the Cu-
In a particle reinforced composite with a very soft and compliant matrix (such as In), the
contribution of load transfer to composite strengthening is negligible, and the composite yield
strength σ cy equals the matrix yield strength σ my . Immediately following fabrication (prior to
any stress relaxation), the ‘initial’ composite yield strength σ cy,i , is given by:
σ cy ,i ≈ σ my ,0 + Δσ CTE
(5-13)
particles are nominally in iso-stress (i.e., σ m = σ r , where the subscripts m and r represent the
49
matrix and particle reinforcement, respectively), and the strain is partitioned based on the
where ε c , ε m and ε r are strains in the composite, the matrix and the reinforcement, respectively.
During the time that elapses during testing, the In matrix, for which ambient temperature
corresponds to ~70 percent of the melting temperature, undergoes stress relaxation by creep, and
dε c dε dε r
= Vm m + Vr
dt dt dt (5-15)
where ε r is elastic, and ε m has both elastic and plastic components. At a fixed applied strain
balanced by the rate of decrease of the elastic strain components ε me and ε r (and hence the
⎛ 1 dσ m ⎞ V dσ r
Vm ⎜ + Cσ mn ⎟ = − r
⎝ Em dt ⎠ Er dt
(5-16)
where Em and Er are the Young’s modulus of the matrix and reinforcement, respectively, C is the
creep constant ( = A ⋅ (Gb / kT ) ⋅ e − Q / RT , where A is the Dorn constant, Q is the activation energy
for creep of the In matrix, R is the gas constant, and T is the temperature), and n is the creep
Realizing that at time t = 0, the initial matrix stress is σ m,i , integrating Equation (5-15)
yields:
− 1
⎡ ε⎤ n −1
σm = ⎢σm,i − (n −1) + (n − 1) Vm E 'c C ⎥
⎣ ε ⎦ (5-17)
50
where ε / ε is the ratio of applied strain to strain rate, and represents the elapsed time t, and
1/Ec’ = Vm/Em+Vf/Ef. As the matrix stress relaxes, the rearrangement of dislocations also leads
to a decrease in the matrix yield strength, and therefore, the composite yield strength. Thus,
Equation (5-17) may also be used to describe the instantaneous yield strength, as follows:
− 1
⎡ ε ⎤ n −1
σcy = ⎢σcy,i −(n −1) + (n − 1) Vm E 'c C c ⎥
⎣ ε ⎦ (5-18)
Combining Equations (5-11), (5-13) and (5-18) gives, for the composite flow stress at
− (n −1) − 1
⎡⎛ ⎤ n −1
⎢ BbΔTΔα (1 − Vm ) ⎞ ε
σcy = ⎜ σ my,o + α1G ⎟ + (n − 1) Vm E 'c C c ⎥
⎢⎜ d Vm ⎟⎠ ε ⎥
⎢⎣⎝ ⎥⎦ (5-19)
Figure 5-7 shows a plot of Equation (5-19) for Cu-In composite, for two strain rates. At
the lower strain rate, the creep component is overestimated by Equation (5-19), since the
Vm values, the stress-relaxation predicted is somewhat more accurate. Therefore, the actual yield
behavior is expected to transition from the higher strain rate curve at low Vm to the lower strain
rate curve at high Vm. The result thus provides insight into the observed experimental trend
(Figure 5-6) that the yield strength decreases steadily with increasing Vm until ~60 vol.% In, and
then levels off at higher Vm values because of rapid stress-relaxation. Thus, despite the
limitations associated with the assumption of iso-stress condition (and the associated difficulties
reasonable first approximation of the yield strength, including creep effects. It is noticeable that
51
Equation (5-19) still underestimates the yield strength of the composite. To be more accurate on
prediction, the hydrostatic strengthening mechanism should be taken into consideration [87, 88].
During the impression test, if we take a consideration of a unit cell which contains one spherical
reinforcement particle. The load is applied perpendicular to the top surface. During deformation,
the far field strain accumulates. The sidewall in the upper portion of the cell attempts to expand
at a greater rate than in the lower portion of the cell. However, the sidewall constraint forces the
lateral side to remain straight, thereby generating large tensile hydrostatic stress within the lower
portion of the cell and large compressive hydrostatic stresses within the upper portion of the cell.
It is these tensile hydrostatic stresses which appear to increase the yield strength of the composite
Figure 5-7 Variation in the yield strength of Cu-In composite solders with matrix volume fraction, as
predicted by Equation (5-19), at two strain rates
52
For TIM and IC applications, a solder is required to possess both low electrical/thermal
resistivity and high compliance. As shown in Figure 5-6, these two traits go in opposite
directions with In volume fraction. Hence, optimization with respect to the In volume fraction
should be conducted to achieve acceptable values for both electrical/thermal conductivity and the
yield strength for a particular application. An optimum volume fraction of In for TIM and IC
applications was determined to be 60% for Cu-In solders, shown by the highlighted region in
Figure 5-6. At this composition, Cu-In composite solders showed a similar electrical/thermal
conductivity as of pure In with only ~2.5 times increase in the yield strength as compared to pure
In.
Figures 5-8a, b shows the effect of sintering conditions on the extent of densification
during LPS for both non-spherical and spherical particles. With increase in the green density of
the pellet, the sintered density increases, as expected. However, it is readily apparent from the
plots that the extent of densification during sintering is very limited under most conditions (only
~2-5% densification beyond the green density). This is true irrespective of the type of sintering
treatment used (sintering in oil or N2 environment), despite the fact that microstructural
observation reveals nearly fully dense microstructures, with minimal porosity, as shown in
Figure 5-4. Clearly, some artifact in the microstructure tends to reduce the sintered density,
especially when the overall time of exposure to high temperatures (above ~125 oC) is long (>2
minutes), as in these samples (which were sintered for <2 minutes at 160 oC, but cooled slowly
53
Sintered Relative Density (%)
92 Cu30In
o
162 C, 45sec
Cu50In
Cu30In
88 o
162 C, 60sec
o
162 C, 30sec
84 Cu50In Cu50In
o o
162 C, 45sec 162 C, 80sec
80
Non-spherical Cu
All samples are flash oil sintered
76
76 80 84 88 92
(a)
Green Relative Density (%)
94
Sintered Relative Density (%)
Cu30In
92 o
162 C, 2 min
90
88 Cu30In
o
Cu30In
162 C, 1 min o
86 160 C, 2 min
84
Spherical Cu
82 All samples sintered in furnace
under N atmosphere.
2
80
80 82 84 86 88 90 92 94
(b)
Green Relative Density (%)
Figure 5-8 The effect of sintering conditions on the extent of densification during LPS for both (a) non-
spherical particles, and (b) spherical particles. The lines in the plots are interpolate fits, to guide the eyes.
54
Figures 5-9a, b shows the effects of sintering time and temperature on the sintered
density more clearly. From Figure 5-9a, it is apparent that a shorter sintering time (45 s)
produces a greater density than a longer time (60s). But confoundingly, an increase in In content
also reduces density. Figure 5-9b shows that a higher sintering temperature (beyond ~162 oC)
also reduces density, although the density rises at 170 oC for longer sintering and cooling times.
As noted previously and shown in Figure 5-10 via X-ray diffraction spectrum of Cu/In interface,
a series of interfacial IMCs may form during sintering and subsequent cooling, depending on the
time of exposure of the Cu to molten In during LPS, as well as solid-state reactions occurring
during cooling. The IMCs formed include CuIn, Cu11In9, and Cu16In9 (or η), which have
densities of 7.656 x 103 [89], 8.614 x 103 [90] and 7.676 x 103 [45] kg/m3, respectively. While
the densities of In and Cu which have densities of 7.31 x 103 and 8.96 x 103 kg/m3, respectively.
Thus, the formation of CuIn2 and Cu16In9 produces a decrease in density relative to the unreacted
Cu-In composite, whereas the formation of Cu11In9 causes an increase. As sintering occurs, these
reaction products are formed in sequence, and as a result, the apparent relative density may
decrease even as pores are eliminated due to the sintering. For example, as explained earlier,
Cu16In9 forms preferentially at the beginning of the sintering and Cu11In9 forms predominantly
only after a thin continuous IMC layer is formed around Cu particles. Hence, the sintered density
of the solders decreases with sintering time in the beginning (Figure 5-9a) and increases with
sintering time at longer sintering times (Figure 5-9b). A greater amount of liquid In facilitates
greater intermetallic growth, and therefore a larger apparent reduction in relative density as noted
in Figure 5-9a.
55
91
88
87
t = 60 s
sintering
86
Cu-In
85 Sintered in oil at 162oC
Green density = 85%
84
20 30 40 50 60 70 80
Volume Fraction Indium (%)
(a)
94
ρ =89%, t =2min
Sintered Relative Density (%)
90
88
ρ =85%, t =2min
Green pellet sintering
86
ρ =85%, t =1min
Green pellet sintering
84
82 Cu30In
Sintered in furnace with N
80 2
158 160 162 164 166 168 170 172
o
(b)
Sintering Temperature ( C)
Figure 5-9 The effect of (a) volume fraction of In and (b) sintering temperature on the sintered relative
density of the Cu-In solders. The lines in the plots are interpolate fits, to guide the eyes
56
3000
o
Sintering Cond.: 165 C, 45s
2500 a: CuIn
b: Cu In
16 9
b c: Cu In
2000 11 9
d: AuIn
Intensity
1500 b
1000 d
c
Cu-In a c
500 d
Cu(Au)-In
d
In In Au In Cu Au Cu
0
30 35 40 45 50 55 60
2θ (degree)
Figure 5-10 XRD spectrum of the etched Cu-In and Cu(Au)-In layered samples showing the relevant
IMC phases formed during sintering at 165oC for 45s. The upper spectrum corresponds to Cu-In sample
whereas the lower one corresponds to Cu(Au)-In. Peaks corresponding to IMC phases are marked with
lower case alphabets
Importantly, it is apparent from the above discussion that the as-sintered relative density
reflects not just the pore-content of the composite solder, but also the amount and type of
intermetallic compound formed. Indeed, as revealed by the nearly fully dense microstructure,
densification to 95-97% relative density can occur within a short time (<45 s of LPS at 160 oC),
and this density can be maintained as long as the solder is cooled rapidly.
Figure 5-11(a) shows the effect of particle size of Cu on the yield strength and the
thermal conductivity of Cu(Au)-In solders with 60 vol% In. By decreasing the particle size of Cu
from 40μm to 1μm, the yield strength of the solder increased by ~3 times and the thermal
conductivity decreased by ~20%. Figure 5-6 shows that 60 vol.% In is near the limit below
57
which the interfacial strengthening effect becomes significant. Since the interfacial area increases
and the In channel width (i.e. interparticle spacing) decreases with the particle diameter, the
particles. Hence, Equation (5-11) and (5-19) can be used to explain the dependence of the yield
strength on the size of Cu particles, where it is shown that the yield strength of the composites
decreases with an increase in the Cu particle size (σYS ∝ d-0.5). Figure 5-11(b) shows a plot of
Equation (5-19) for Cu-In composite with various particle sizes of reinforcement (Cu). The
volume fraction of In in composite solders in this plot is fixed to be 60 vol.%. In the range
between 1 and 40 μm, the yield strength decreases monotonically with increase amount of In.
Since Equation (5-19) assumes all the particles in the composite are randomly and uniformly
distributed, no particle will be in contact with each other, the yield strength of composite will be
underestimated when the particles nearly touch. In a fixed volume fraction of composite, the
inter-particle spacing is linearly proportional to the particle size. The smaller the inter-particle
spacing, more possible that particles are touching each other. The impression test method can
also raise the apparent yield strength when the punch indented into the composite, causing the
58
40 130
Therm. Cond.
120
30 YS
115
25
110
20
105
15 100
10 95
0 10 20 30 40 50
(a)
Particle Size of Cu (μm)
6
Model
Cu-60vol.%In LPSS
-4 -1
Strain rate = 10 s
Yield Strength (MPa)
5.5
4.5
4
0 10 20 30 40
(b)
Particle Size (μm)
Figure 5-11 (a) Effect of particle size of Cu on the yield strength, and the thermal conductivity of
composites with 60vol% of In. All the samples were sintered at 160 oC for 15s and (b) variation in the
yield strength of Cu-60vol.%In composite solders with particle size of reinforement (Cu), as predicted by
Equation (5-19)
59
An increase in the thermal conductivity with an increase in the particle size, as shown in
Figure 5-11, is attributed to a decrease in the total interfacial/contact area between Cu and In
with an increase in the particle size. This behavior can be explained using the model in Section
5.1.
Using Equation (5-4), the average value of the thermal contact resistance for the Cu(Au)-
60 vol% In was calculated to be equal to 4±2 x 10-9m2-K/W. Now, this Rcth value can be used in
Equation (5-10) to calculate the impact of the Cu particle size on the thermal conductivity of
Cu(Au)-In solders, which is shown in Figure 5-3. The experimental results shown in Figure 5-11
are consistent with those predicted by Equation 5-10. In addition to sintering procedure followed
in this study, the thermal contact resistance of the powders can be further be reduced by
problem associated with fine Cu powders, is prohibited. This will reduce the fraction of pores
5.2.5 CONCLUSIONS
Short-time liquid phase sintering was employed to produce Cu-In based composite
solders for TIM and IC applications. A combination of low sintering temperature, short sintering
time and fast cooling rate produced solders with the highest sintered density. In general,
Yield strength as well asthe thermal conductivity of the composites decreased with an increase in
the volume fraction of In; the optimum compromise for TIM and IC applications occurred at 60
vol.% of In.
60
5.3 ROLE OF INTERFACIAL LAYERS ON THE PERFORMANCE OF LPSS
Figure 5-12 (a-d) shows microstructures of the direct immersion Au coated Cu surfaces
with various thicknesses of Au, ranging from 5 nm to 20 nm. The images clearly show that when
the Au coating is below 20 nm thick, the surface is non-uniform with a number of surface defects.
With increase in coating layer thickness, the surface becomes more uniform. Figure 5-12 (c-d)
shows the defects on Au coating surface, detected under SE mode and topography mode. These
defects can produce the direct contact between Cu and In during processing, and hence form
IMCs with poor performance. Therefore, in order to ensure a uniform coating on the entire Cu
Figure 5-13 shows the microstructure of 20nm Au and 1.1nm Al2O3 coated Cu substrate.
The image was taken by looking at the top surface of the coated Cu substrate. It clearly shows
that a uniform Au coating can be obtained by depositing 20nm thick Au layer on the 1.1nm
Al2O3 coated Cu surface. Hence, 1.1nm Al2O3 combined with 20nm Au was considered
optimum interfacial coating for both composites and flat plate samples.
61
(a) (b)
(c) (d)
(e) (f)
Figgure 5-12 (a)-(d) Microstrructure of dirrect immersioon Au coatedd Cu substratte with variouus Au
thicknnesses, such as (a) 5 nm, (b) 10 nm, (cc) 15 nm, andd (d) 20 nm. SE image (ee) and topography (f)
shoowing some defects in the Au coatingg when Au thhickness is 200 nm.
62
Figure 55-13 Microsttructure of duual layer coaated Cu substtrate. The stacking sequennce is 20nm Au,
A 1.1nm
Al2O3, andd Cu substratte (top view)
5.3.22. ROL
LE OF Au
A AND A
Al2O3 INTERFACIA
AL LAYER
R ON RE
EACTION
B
BETWEEN
N Cu AND In
sampless were then wrapped inn Al foil andd pressed firrmly, immeersed in an ooil bath at 1165 °C for
examineed under SE
EM to studyy the growtth IMC layeer. Later, Inn layer was also etchedd away by
63
Figures 5-14a, b shows the cross section of the layered samples without and with Au
interfacial layers, respectively. In both types of layered samples, a continuous layer of IMC
formed at the interface between In and Cu. However, a 200 nm of Au layer on Cu drastically
reduced the total thickness of the interfacial IMC layer from ~22 μm to ~8 μm. This clearly
demonstrates the effectiveness of the Au as a diffusion barrier between In and Cu. Figures 5-14c,
d shows the morphology of the IMC layer formed over Cu substrate, without and with Au layer,
respectively. As shown in Figures 5-14c, d, all IMCs were faceted with preferred planes of
growth. However, the IMC grains formed over Au coated Cu substrate were significantly smaller
in size than the IMCs formed over bare Cu substrate. As shown in Figure 5-10, XRD analysis of
the surfaces revealed that the majority of the IMC formed over bare Cu substrate were Cu16In9
and Cu11In9 whereas AuIn2 and Cu16In9 IMCs formed over Au coated Cu substrate. The presence
of the IMCs here is consistent with the equilibrium phase diagrams [11, 44, 53] and the earlier
reports on Cu-In [10, 12] and Au-In [57] systems. Prior results on the growth of above IMCs [10,
57] suggests that the growth rate of Cu11In9 is the fastest whereas Cu16In9 grows at the slowest
rate at 160 oC. This is consistent with Figure 5-14. In summary, the following two statements can
be inferred from these results: (i) once a thin layer of Cu16In9 forms over bare Cu substrate,
Cu11In9 forms rapidly with scalloped morphology, and (ii) in the presence of a thin interfacial
layer of Au, firstly, a very thin layer of AuIn2 forms instantaneously at the Cu substrate2, which
then acts as an effective diffusion barrier between Cu and In and impedes the formation of Cu-In
IMCs.
2
Due to much more negative Gibbs free energy of formation of AuIn2 (-21.317 kJ/mol) at 160oC, this IMC
forms first when liquid In comes in contact with the interfacial Au layer [53].
64
(a) (b)
(c) (d)
Figure 5-14 Cross-sectional micrographs of the interfaces, showing continuous layers of IMCs at (a)
Cu/In interface, and (b) Cu/200nm Au/In interface. In was then etched away revealing the IMC
morphology at the Cu/In interface: (c) bare Cu substrate and (d) 200nm Au coated Cu substrate. All
micrographs correspond to the as-reflowed condition
Figure 5-15 shows the cross-section of composite solders, (a) without interfacial layer, (b)
with 50nm Au interfacial layer, and (c) with 1.1 nm Al2O3 interfacial layer, sintered for 15
65
seconds. Even at this short sintering time, the solder with Au interfacial layer attained a sintered
density of ~95% with much less porosity as compared to the solder without Au interfacial layer
as well as the solder with Al2O3 interfacial layer. This proves that Au interfacial layer also
substantially improved the wetting between Cu and liquid In. This is attributed to the very good
wetting between In and Au [10, 57]. For the Cu(Al2O3)-In composite solders, it was observed
that by increasing the amount of sintering time from 15s to 60s, the relative sintered density of
solder dropped from 85% to 65%. Liquid In was observed to have exuded from the solder pellets
and formed small spheres on the surface. This indicative of inactive wetting between Al2O3 and
liquid In. Hence, in order to study the effect of Al2O3 on solder performance, a Au layer (20nm
66
(a)
(b)
(c)
Figure 5-15 Microstructures of as-sinteredd (a) Cu-In, ((b) Cu(Au)-IIn, and (c) Cuu(Al2O3)-In ssolders,
sintered att 160 oC for 15
1 s. The blacck regions inndicate pores.
67
Another set of layered samples were prepared to study the solid state reaction between Cu
and In with different interfacial layers. A 3 µm In layer was electrolytically deposited on three
types of 6mm x 6mm flat Cu pieces, namely (i) without interfacial layer, (ii) with 50 nm Au
interfacial layer, and (iii) with 1.1 nm Al2O3+20 nm Au dual interfacial layer. These layered
samples were aged at 125 oC up to 96 h. As shown in Figure 5-16, XRD analysis revealed that
the Cu-In IMCs formed over bare Cu substrate and 50nm Au coated Cu substrate were CuIn and
Cu11In9, whereas CuIn2 and Cu16In9 formed over (1.1 nm Al2O3+20nm Au)-coated substrate. It
clearly shows that due to the one nanometer thick Al2O3 coating, the inter-diffusion of Cu and In
was effectively inhibited and a Cu-rich IMC, Cu16In9 formed on Cu side. In addition, due to the
limited amount of Cu diffusing into In, only a metastable In rich IMC (CuIn2) formed on the
Indium side. The stacking sequence of these IMCs was confirmed by depth profile of Cu+
obtained by TOF-SIMS analysis, as shown in Figure 5-17. The Cu+ signal shows three plateau
regions. Based on the relative intensities of these regions, along with the types of IMC identified
from XRD results, it can be concluded that these three plateaus represent CuIn2, Cu16In9 and Cu
substrate.
68
(a)
8000
Cu//In Layered Sample As-deposited
7000 o
Aged at 125 C for 24h
o
6000 In Aged at 125 C for 96h
Intensity (counts)
In In
5000
4000
CuIn
CuIn Cu
3000
2000
Cu In Cu In
11 9 11 9
1000
0
32 34 36 38 40 42 44
Two-theta (degrees)
(b)
840
4
3 10 700
Cu/50nmAu/In Layered Sample
Intensity (counts)
Cu
4
2.5 10
As-deposited 560
o
Aged at 125 C for 24h Cu In
420 Cu In 11 9
In o
Aged at 125 C for 96h 11 9
Intensity (counts)
4
2 10 CuIn
280
In
4
1.5 10 140
40 41 42 43 44
Two-Theta (degree)
4 In
1 10 CuIn Cu
5000
0
32 34 36 38 40 42 44
Two-Theta (degree)
69
(c)
390
8000
Cu/Al O /Au/In Layered Sample
Intensity (counts)
2 3 Cu
7000
As-deposited
o 195 Cu In
6000 Aged at 125 C for 24h Cu In
16 9
In o 16 9
Intensity (counts)
2000 In
In
1000
0
32 34 36 38 40 42 44
Two-Theta (degree)
Figure 5-16 Glancing incident x-ray diffractograms of (a) Cu/In layered samples, (b) Cu/Au/In layered
samples, and (c) Cu/Al2O3/Au/In layered samples, which were aged at 125oC for 24h and 96h.
6
10
Cu/1.1nm Al O /20nm Au/In
2 3
o
5
Aged at 125 C for 24h
10
Intensity (# of counts)
4
10 +
Cu
1000
+
AuIn
2
100 +
AlO
+
In
10
1
0 500 1000 1500 2000 2500 3000
In side Cu side
Sputter Time (s)
Figure 5-17 Sequential sputtering using SIMS analysis showing the sequence of IMCs formed at
Cu/Al2O3/Au/In interface
70
5.3.3. EFFECT OF INTERFACIAL LAYERS ON THE MECHANICAL
PROPERTIES OF LPSS
Figure 5-6 compares the effect of Au interfacial layer on the yield strength of the as-
sintered solders. Both Cu-In and Cu(200nmAu)-In composite solders show similar trends in the
yield strength with the In volume fraction. Figure 5-6 shows that the difference in the yield
strength between as-sintered Cu-In and as-sintered Cu(200nmAu)-In solders was insignificant.
Figure 5-18 also shows that irrespective of the interfacial coating, the yield strength of
composites with 60 vol% In fall in the same range. Since most of the plastic deformation was
concentrated in the soft matrix, whose volume fraction did not change during short-time
sintering, and the thicknesses of the IMC layers in Cu-In, Cu(Au)-In and Cu(Al2O3+Au)-In
solders were much smaller than the Cu particles (Figures 5-19a and c), the effect of Au coating
and Al2O3 coating on the yield strength of the solder was insignificant.
71
50
Cu(X)-60In (40μm Cu powder)
Cu(200nmAu)-60In
30
Cu(50nmAu)-60In
Cu(1nmAl O +20nmAu)-60In
2 3
20
Cu(1nmAl O )-60In
2 3
10
o
Aged at 125 C In
0
0 20 40 60 80 100
Aging Time (h)
Figure 5-18 Effect of aging on the yield strength of composites with different interfacial layers between
Cu reinforcement and In matrix.
72
Figure 55-19 Microsttructures of thhe solders w
with different aging condittions. (a) As--sintered Cu-In, (b) Cu-
In solderr after an aginng for 96 hrss, (c) As-sinteered Cu(200nnm Au)-In, (d)
( Cu(200nm
m Au)-In soldder after an
aging for
f 96 hrs, (ee) As-sinteredd Cu(1.1nm Al
A 2O3+20nm
mAu)-In, and (f) Cu(1.1nm
m Al2O3+20nnmAu)-In
solder after an agingg for 96 hrs
73
Solders, in both TIM and IC applications, are exposed to elevated temperatures during
service that may cause microstructural evolution in Cu-In based solders and deteriorate their
mechanical, electrical and thermal properties. Hence, aging study was conducted to evaluate the
effect of the microstructural evolution on the mechanical properties of the Cu-In, Cu(Au)-In, and
Cu(Al2O3+Au)-In solders. Figure 5-19 shows the effect of isothermal aging on the Cu-In,
formation of a very thin layer of IMC at the Cu/In interface, which becomes apparent only at
high magnification. However, as shown in Figure 5-19(b), the Cu-In IMCs in Cu-In solders
thickens very rapidly with aging resulting in a few very small isolated zones of unreacted In. On
the other hand, AuIn2 IMC forms very rapidly during the LPS of Cu(200nmAu)-In solders
(Figure 5-19c), and remains tenaciously intact during aging allowing only limited diffusion of In
through it. This results in the formation of Cu-In IMC in between the “protective ring” of the
AuIn2 IMC and the Cu particles (Figure 5-19d), leaving large regions of unreacted In in the
Cu(Au)-In solder matrix. Whereas, when the inter-layer contains only 1.1nm Al2O3 and 20nmAu,
Au only acts as wetting enhancer and the ceramic Al2O3 inhibited the IMC formation between
As shown in Figure 5-18, the yield strength of all of the composite solders monotonically
increased with aging. This can be attributed to the growth of the IMC layer at the Cu/In,
Cu(Au)/In, Cu(Al2O3)-In and Cu(Al2O3+Au)-In interfaces, as shown in Figure 5-19. In the initial
stages of aging, the increase in the yield strength was very steep due to a corresponding steep
increase in the IMC thickness (IMC thickness ∝ taging0.5, where taging is the aging time [10, 57]).
With increase in the aging time, the growth of the IMC layer slows down [10, 57], resulting in a
similar trend in the yield strength of the solders as well. As discussed earlier, thick Au interfacial
74
layer acted as an effective diffusion barrier layer, the IMC growth kinetics in the Cu(Au)-In
solders was much slower than that in Cu-In. Hence, the relative rate of change of the mechanical
property of Cu(Au)-In was slower than that of Cu-In solders. Figure 5-18 also shows that
Cu(Al2O3)-In and Cu(Al2O3+Au)-In had the slowest deterioration rate during aging. It clearly
indicates that Al2O3 or Al2O3+Au is more effective in slowing down the reaction at Cu/In
interface during aging at 125 oC. The dense Al2O3 layer, deposited by ALD technique, can
efficiently slow down the reaction without increasing the size of Cu particles after coating. The
inactive wetting between Al2O3 and liquid In (discussed in section 5.3.2) can be improved by
introducing a thin and uniform layer Au between Al2O3 and In. Therefore, 1 nm Al2O3 and 1 nm
Al2O3+20 nm Au interfacial layer provided the most stable mechanical property of the composite
Figure 5-20 shows the effect of different interfacial layers on electrical and thermal
properties of composite solders. As shown in Figure 5-20, due to the reaction between Cu and In
during the reflow, Cu-In composite solders, with 40 vol. % Cu particles, showed only slight
improvement in electrical resistivity and thermal conductivity as compared to pure In. The
formation of Cu16In9 and Cu11In9 intermetallic compounds (IMCs) was found to limit the
improvement of the electrical and thermal properties of composites. Figure 5-20(a) shows that
the electrical resistivity of Cu(Au)-In, and Cu(Al2O3+Au)-In solders was significantly smaller
than that of Cu-In solders. This can be attributed to the Cu(Au)/In and Cu(Al2O3+Au)-In
interfaces as compared to that of Cu/In interface. The contact resistance depends on the wetting
75
and the conductivity of the interfacial IMC layer. Since (i) wetting between In and Au coated Cu
is significantly better than that between In and Cu, and (ii) the electrical resistivity of AuIn2 and
Cu11In9 are 78 nΩ-m [91] and 170 nΩ-m [92], respectively, the contact resistance of Cu(Au)/In
interface is much smaller than Cu/In interface. However, when there is only 1.1 nm Al2O3 on Cu
particles, the electrical resistivity of composite solders increased dramatically to 3.5 times that of
pure In. This is attributed to the poor wetting between liquid In and ceramic Al2O3 as mentioned
earlier, which inhibits liquid In to penetrate into small regions and then forms porous composite
(see Figure 5-15c). Figure 5-20(b) shows the similar trend for thermal conductivity of various as-
sintered composite solders. The as-sintered Cu(1.1nm Al2O3+20nm Au)-In solders have the
thermal conductivities of 125 W/m-K, which is 1.5 times that of pure In. Cu(1.1nm Al2O3)-In
also shows the lowest thermal conductivity. Both thermal and electrical properties of composites
follow the same trend. This is because in metallic composites, electrons are the main carriers for
both heat and current transport, as mentioned by Wiedemann Franz law [93].
Figure 5-20(b) also shows the effects of interfacial layers on the stability in terms of
electrical resistivity and thermal conductivity of composite solders after isothermally aging at
125oC for 96h. The thermal conductivity of Cu-In composite decreased by 23% after aging at
125oC for 96h when there was no interfacial layer on Cu particles. On the other hand, when the
Cu particles were coated with 50nm Au or 1nmAl2O3, the deterioration of thermal conductivity
of composites were 10% and 3.7%, respectively. 50nm Au layer was quickly consumed by liquid
In during reflowing process and formed AuIn2 layer, which then acted as a diffusion barrier
between Cu and In. Most importantly, this thin layer significantly improved the wetting between
Cu and In due to its affinity to In, which facilitated the sintering to complete in a shorter time and
hence reduced the possible reaction time. 1.1nm Al2O3 layer acted as an efficient diffusion
76
barrier. This is suggested based on the fact that the thermal conductivity of composite was
observed to be very stable during aging. However, when Al2O3 layer was used by itself, the
composite solder showed unacceptably low thermal conductivity. This can be attributed to the
porosity introduced in the composite due to the poor wetting between Al2O3 and liquid In. In the
present work, we utilized Al2O3 layer as a diffusion barrier and Au layer as a wetting enhancer.
This produces a composite the highest thermal conductivity, i.e. 125 W/m-K, and smallest
Figure 5-20(a) also shows the yield strength and the electrical resistivity of commonly
used Pb-free Sn-3.8%Ag-0.7%Cu. It is noteworthy that the yield strength as well as the electrical
and thermal resistivity of Cu(1.1nm Al2O3+20nm Au)-In solders remain lower than that of
commonly used lead-free SAC387 solder, even after severe aging. This certainly justifies the use
of Cu-In based composite solders in TIM and IC applications, especially where a soft solder is
required. This is of particular advantage owing to the low reflow or reworking temperature of
77
300
Cu(X)-60In (40μm Cu powder)
o
Sintered at 160 C for 45s
270
240
SAC387
120
Cu-60In
90 In
Cu(50nmAu)-60In
60 Cu(200nmAu)-60In
Cu(1nmAl O +20nmAu)-60In
o 2 3
Aged at 125 C
30
0 20 40 60 80 100
(a)
Aging Time (h)
o
Sintered at 160 C for 45s
In
80
Cu-In -23%
60
Cu(Al O )-In -3.7%
2 3
o
Aged at 125 C
40
0 20 40 60 80 100
(b)
Aging Time (h)
Figure 5-20 Effect of aging on (a) the electrical resistivity and (b) the thermal conductivity of composites
with different interfacial layers between Cu reinforcement and In matrix.
78
5.3.5. EFFECT OF Au INTERFACIAL LAYER THICKNESS ON THE
Figure 5-18 shows that the impact of the thickness of Au layer on the yield strength of the
solder in the as-reflowed and aged conditions is less pronounced; however, a composite with
thicker interfacial Au layer distinctively show lower electrical resistivity for all aging conditions.
As stated earlier and shown in Figure 5-19, the AuIn2 IMC forms almost instantaneously during
LPS (or reflow), consuming all available Au, and then acts as a tenacious diffusion barrier layer
between Cu and In. Since the difference in the thicknesses of Au layer between the two Cu(Au)-
In composites (150nm) is insignificant in comparison to the size of the Cu particles (40μm) and
the average channel width between Cu particles (~50μm), the effect of Au layer thickness has
insignificant effects on the volume fraction of unreacted In, resulting in similar yield strength
behavior at all thicknesses of Au layer. On the other hand, a thicker layer of Au ensures
continuous conformal coating of Cu powders, which also has higher probability of remaining
unbroken during powder mixing and compaction of mixed powders. Also, a thick interfacial
layer of ductile Au can endure plastic deformation to accommodate the differential thermal
expansion (αCu = 1.65 x 10-5 /K and αAu = 1.42 x 10-5 /K) during the rapid heating of the mixed
powders and hence a fewer micro-cracks will form in the thick Au interfacial layer as compared
to a thin one. Hence, the overall wetting between Cu and In as well as the effectiveness of Au
layer in inhibiting the Cu-In IMC formation improve with thickening of Au layer leading to the
reduction in the contact resistance. Also, an increase in the thickness of highly conductive AuIn2
IMC layer (electrical resistivity = 78nΩ-m [57]) at the expense of In (electrical resistivity =
84nΩ-m [94]) would also result in an overall lower conductivity for the Cu(Au)-In solder with
79
act as a diffusion barrier. However, in order to make thin solder joints using smaller Cu particles,
200 nm thick Au layer may not be acceptable. In order to minimize the amount of Au in the
composite, a very thin layer of Al2O3 was selected to fulfill the role of diffusion barrier. A thin
Au layer (20 nm) was coated on top of Al2O3 to enhance the wetting between Cu particles and
liquid In.
In pure metals, the electron contribution to conduction heat transfer dominates, whereas
in nonconductors and semiconductors, the phonon contribution is dominant. The bulk ceramic
Al2O3 usually has the thermal conductivity of 30 W/m-K [95], however, when the thickness of
Al2O3 layer is close to the mean free path of either electron or phonon, the thermal conductivity
will be varied from that of bulk materials. Usually the mean free path of phonon in bulk Al2O3 is
5.08nm at the temperature of 300K. This mean free path is much larger than the film thickness
we used in our study, which is 1nm. When the mean-free-paths of phonons are comparable to
film thickness, phonons make a negligible contribution to heat transport [96-98]. In this case,
electron would be the main limitation during heat transport. On the other hand, due to the
nanoscale thickness of Al2O3 layer, the electron tunneling through thin Al2O3 layer becomes
dominant. If we assume a rectangular barrier 0.85 eV high and an effective mass m*/m=1, then at
room temperature the critical thickness is 35 Å and below this thickness, tunneling will dominate
almost exclusively [99]. Therefore, 1.1 nm thick layer of Al2O3 will not act a thermal barrier. As
discussed in previous sections, significant decrease in thermal conductivity after coating Al2O3
layer on Cu particles is due to poor wetting of coated Cu particles and increase in porosity of the
composite.
80
5.3.7. CONCLUSIONS
Short-time liquid phase sintering was employed to produce Cu-In based composite
solders for TIM and IC applications. The formation of Cu-In IMC has been found to deteriorate
the mechanical and thermal properties of composite solder after fabrication and long term aging.
However, a 1nm of Al2O3 layer significantly reduces the reaction between Cu and In. By coating
an additional 20nm Au layer on top of Al2O3 layer, the wetting behavior between Cu and In can
be improved to provide a fully dense composite microstructure, which produce a material with
high thermal conductivity of ~125 W/m-K and low yield strength of ~12 MPa. The properties
remain stable after 96h aging at 125 oC, with thermal conductivity decreased by 3% and yield
81
5.4 T
THE PERF
FORMANC
CE OF LPS
SS JIONTS
S
5.4.11 QUA
ALITY OF
F LPSS JOIINT
F
Figure 5-21 (a) shows a LPS soldeer joint form
med betweenn two Cu roods of 6mm
m diameter.
The soldder was maade of In andd 40μm Cu coated withh a 50nm thhick layer of Au. Figurre 5-21 (b)
82
5.4.2 THERMAL RESISTANCE OF LPSS JOINT
Figure 5-22 shows the effect of bondline thickness as well as the effect of interfacial
coating on Cu substrate on the thermal resistance of different solder joints. The thermal
L
R jo int = Rsolder + 2Rcontact = + 2Rcontact
kA
(5-20)
where L is the thickness of solder joint, A is the contact area between solder and substrate, and k
is the thermal conductivity of solder. Rjoint and Rcontact represent the joint resistance and the
interfacial contact resistance between solder and substrate in cm2-K/W, respectively. Thus, the
thermal resistance of a joint would increase linearly with the increase in bondline thickness.
Figure 5-22 shows that the joint thermal resistances with different bondline thicknesses fall on a
straight line. The slope of the straight line represents the inverse of thermal conductivity of
solder. The intersection of the straight line with y axis represents two times the thermal contact
resistance between solder and substrate. The thermal conductivity calculated from the slope of
lines in Figure 5-22, is 88 W/m-K for pure In joint. For joints made with Cu(1.1nm Al2O3+20nm
Au)-60vol% In solder on two different substrates, i.e., 50nm Au coated and 1.1nm Al2O3+20nm
Au coated, the thermal conductivity values are, 127 W/m-K, and 134 W/m-K. All of these
thermal conductivity values are consistent with those observed in the solder materials, see Figure
5-20 (b). The thermal contact resistance for the joints made with (1.1 nmAl2O3+20 nm Au)
interfacial layers on both Cu particles and Cu substrate, obtained from the graph in Figure 5-22
83
0.12
Cu(X)-LPSS-Cu(X) Joints
0.08
Pure In
Cu(50nm Au) sub.
0.06
LPSS on
Cu(50nm Au) sub.
0.04
LPSS on
0.02 Cu(1.1nmAl O +20nmAu) sub.
2 3
Figure 5-22 Effects of bondline thickness on the thermal resistance of different solder joints and the
effect of interfacial coating on Cu substrate
Figure 5-23 shows the effect of aging on the thermal resistance of two different joints.
Both solder joints were made of Cu-60vol%In composite with (1nm Al2O3+20nm Au) interfacial
layers on Cu particles. The substrates used in one of the solder joints was Cu rods coated with
50nm Au, and those used in the second one were Cu rods coated with (1nm Al2O3+20nm Au).
After aging these joints for 21 days at 125oC, the increase in joint thermal resistance of the one
with Al2O3+Au layer on Cu substrates was only 10 % whereas this increase was about 35% for
the joint with only Au coating on the substrates. Since the solders used in both the joint were
same, the difference in this increase in joint thermal resistance in the two joints can be largely
attributed to the microstructure evolution the solder-substrate interfaces. This again indicates that
84
(1 nm Al2O3 + 20 nm Au) interfacial layer is most effective in producing dense composites with
0.07
Cu(X)-LPSS-Cu(X) Joints
Joint Thermal Resistance (cm -K/W)
o
T =~60 C o
Aged at 125 C for 4 days
0.05 test
o
Aged at 125 C for 21 days
0.04 35%
10%
0.03
0.01
0
Cu(50nmAu) Sub. Cu(1.1nm Al2O3+20nm Au) Sub
Figure 5-23 Effect of aging on joint thermal resistance of solders bonded with (a) 50nm Au coated and
(b) 1.1nmAl2O3+20nmAu coated Cu substrate
Figure 5-24 and 5-25 show the stress strain curve of solder joint during shear test. The
pure In joint deformed as an elastic-ideal plastic material, with a maximum flow stress of 1.4
MPa. The maximum flow stress of the joints made with Cu(1nm Al2O3+20nm Au)-60vol%In
solder varied between 2.6 and 3.2 MPa with the BLTs varying from 270 to 682 µm. It is
noteworthy here that although 40 vol% of In has been replaced by the hard Cu particles in the
85
composite solder joints, the shear strength increased about 1.5 MPa. This indicates that during
the shear tests of these composite solder joints, most of the shear strain is localized in the soft In
matrix. A study conducted by Kumar et al. using FEM model also showed that the strain is
mostly localized in the In channels which are far away from the Cu particles [100]. Hence, as
long as In exists in continuous channels, the flow stress in shear would not increase significantly.
In addition, as shown in Figure 5-6 , Cu-In composite was optimized to have 40 vol.% Cu, in
which range the volume fraction has a minimum effect on the yield strength of composite during
impression test . This indicates that the continuous In channel in the microstructure could
effectively localize most of the strain and make the materials behave soft.
Figure 5-25 also shows both the maximum stress of composite solder joints and the
damage rate increases with BLT. Figure 5-26 shows the fractographs of the joints with different
BLTs fractured under shear loading. The elongated dimples on the fracture surface indicate that
the joint underwent ductile fracture under shear. This again confirms that most of the shear strain
has been localized in the In matrix. Also, the fractographs from a joint with 410 µm BLT show
larger proportion of ductile fracture surface as compared to that from the one with 670 µm BLT.
This indicates that thinner joints are more ductile and hence have larger maximum elongation.
86
Figure 55-24 Shear sttress-strain ccurve of the ccomposite sollder joints annd In joints
4
Cu(1n
nm Al-O+20nm m Au)-60In Soolder Joints Beffore damage
Bondeed on Cu(1nm Al-O+20nm Au)
A Substrates
3.5 X Aftter damage occcuring
-3 -1
Shearr Strain Rate = 10 s , RT
Plasticc Strain at: 0.002
3
Shear Stress (MPa)
0.100
X X
2.5 0.200
X
2
X
X
1.5
1
X
0.5
0
LPS
SS,270μm LPSS,419μm
m LPSS,682μm In,429μm μm
In,731μ
Figgure 5-25 Floow stress varriation with bbondline thicckness under shear
87
Figure 55-26 The fraactographes oof the fractureed joint with various bonddline thickneess under sheear loading
F
Figure 5-277 shows thhe effect off BLT on the maximuum shear sstress and m
maximum
elongatiion. As show
wn in Figurre 5-27, wheen the BLT increases frrom 270 µm
m to 419 µm
m, which is
88
decreased by 35%. Whereas when the BLT increases from 419 µm to 682 µm, the maximum
shear stress increased only by about 8% and the maximum elongation decreased by 40%.
5
Cu(Al O +Au)-60In LPSS
2 3
Bonded on Cu(Al O +Au) substrate
4.5 2 3
Maximum Shear Stress (MPa)
-3 -1
Strain rate = 10 s
4
3.5
2.5
2
200 300 400 500 600 700
0.6
0.4
0.2
0
200 300 400 500 600 700
89
5.4.4 CONCLUSIONS
A 100 μm solder joint has been produced by jointing Cu(1 nm Al2O3+20 nm Au)-60
vol.%In with two pieces of Cu substrate coated with 1nm Al2O3+20nm Au. The thermal
resistance can go down to as low as 0.021 cm2-K/W. Due to the protected interfacial layer on
both Cu particles and Cu substrate, the increase in joint thermal resistance was only ~10% after
aging at 125 oC for 21 days. A compliant joint bonded with Cu(1 nm Al2O3+20 nm Au)-
60vol.%In composite was tested after 30% of shear strain and remained bonded. Thinner joint
90
VI. SUMMARY AND FUTURE DIRACTIONS
Short-time liquid phase sintering was employed to produce Cu-In based composite
solders for TIM and IC applications. A combination of low sintering temperature, short sintering
time and fast cooling rate produced solders with the highest sintered density. The formation of
Cu-In IMC has been found to deteriorate the mechanical and thermal properties of composite
solder and its joint after fabrication and long term aging. However, a 1nm of Al2O3 layer
significantly reduces the reaction between Cu and In. By coating an additional 20nm Au layer on
top of Al2O3 layer, the wetting behavior between Cu and In can be improved to provide a fully
densed composite microstructure, which produce a material with high thermal conductivity of
A 100μm solder joint has been produced by jointing Cu(1nm Al2O3+20nm Au)-
60vol.%In with two pieces of Cu substrate coated with 1nm Al2O3+20nm Au. The thermal
resistance can go down to as low as 0.021 cm2-K/W. Due to the protected interfacial layer on
both Cu particles and Cu substrate, the increase in joint thermal resistance was only ~10% after
tested after 30% of shear strain and remained bonded. Thinner joint with BLT of ~200 μm is
Gold layer was found to have limited adhesion on ceramic Al2O3. Therefore the reaction
product, AuIn2, after reflow might spall off and IMC grains can distribute into the soft matrix
and harden the composite materials. One of the solutions is to find another inert metallic
interfacial layer to act as a diffusion barrier. Tantalum has already been widely used in the IC
91
applications to act as a diffusion barrier between Si and Cu. Phase diagram shows that there is
not reaction occur even when the temperature is above the melting point of In. It is then
suggested that Ta can be a good candidate for interfacial layer to act as a diffusion barrier.
92
LIST OF REFERENCES
1. Liu, J., et al. Next Generation Materials for Thermal Interface and High Density Energy
Storage Applications via Liquid Phase Sintering. in Proceedings of the 12th Electronics
2. Kumar, P., et al. Novel Liquid Phase Sintered Solders with Indium as Minority Phase for
on Thermal Issues in Emerging Technologies (ThETA 2). 2008. Cairo, Egypt: IEEE.
3. Dutta, I., et al., Liquid Phase Sintered Solders with Indium as Minority Phase for Next
4. German, R., P. Suri, and S. Park, Review: liquid phase sintering. Journal of Materials
5. Qiao, X. and S.F. Corbin, Development of transient liquid phase sintered (TLPS) Sn–Bi
6. Palmer, M., N. Erdman, and D. McCall, Forming high temperature solder joints through
liquid phase sintering of solder paste. Journal of Electronic Materials, 1999. 28(11): p.
1189-1193.
7. Gallagher, C., G. Matijasevic, and J.F. Maguire. Transient liquid phase sintering
8. Shearer, C., et al., Transient liquid phase sintering composites: Polymer adhesives with
93
9. Every, A.G., et al., The Effect of Particle-Size on the Thermal-Conductivity of Zns
10. Yu, C.L., S.S. Wang, and T.H. Chuang, Intermetallic compounds formed at the interface
between liquid indium and copper substrates. Journal of Electronic Materials, 2002.
31(5): p. 488-493.
11. Liu, H., et al., Thermodynamic assessment of the Cu-In binary system. Journal of Phase
12. Kim, D.G., et al., Reaction diffusion and formation of Cu11In9 and In27Ni10 phases in
13. Cunnington, G.R., Thermal Conductance of Filled Aluminum and Magnesium Joints in a
Vacuum Environment, in ASME Winter Annual Meeting1964: New York. p. ASME Paper
No. 64-WA/HT-40.
14. Getty, R.C. and R.E. Tatro, Spacecraft Thermal Joint Conduction, in Thermophysics
15. Chiu, C.P., et al., Application of Phase-Change Materials in Pentium ® III and Pentium ®
16. Prasher, R., C.P. Chiu, and R. Mahajan. Thermal Interface materials: A Brief Review of
http://www.electronics-cooling.com/2004/02/thermal-interface-materials-a-brief-review-
of-design-characteristics-and-materials/.
94
17. Blazej, D. Thermal Interface materials. 2003; Available from: http://www.electronics-
cooling.com/2003/11/thermal-interface-materials/.
solutions/index.php.
20. Biercuk, M.J., et al., Carbon nanotube composites for thermal management. Applied
21. Hone, J., et al., Thermal properties of carbon nanotubes and nanotube-based materials.
22. Liu, C.H., et al., Thermal conductivity improvement of silicone elastomer with carbon
23. Thostenson, E.T., Z.F. Ren, and T.W. Chou, Advances in the science and technology of
carbon nanotubes and their composites: a review. Composites Science and Technology,
24. Huang, H., et al., Aligned Carbon Nanotube Composite Films for Thermal Management.
25. Xu, J. and T.S. Fisher, Thermal Contact Conductance Enhancement With Carbon
26. Xu, J. and T.S. Fisher, Enhancement of thermal interface materials with carbon nanotube
arrays. International Journal of Heat and Mass Transfer, 2006. 49(9–10): p. 1658-1666.
27. Hu, X., et al. Thermal characterization of aligned carbon nanotubes on silicon. in IEEE
95
28. Frear, D.R., et al., Pb-free solders for flip-chip interconnects. Jom-Journal of the
29. Corker, D.L., et al., Liquid-phase sintering of PZT ceramics. Journal of the European
30. Froschauer, L. and R. Fulrath, Direct observation of liquid-phase sintering in the system
31. Kingery, W.D., E. Niki, and M.D. Narasimhan, Sintering of Oxide and Carbide-Metal
32. Northcutt, W.G., O. Ridge, and W.B. Snyder, Process for fabricating articles of
Y<sub>2</sub>O<sub>3</sub>-
ceramics synthesized by rapid liquid phase sintering. Applied Physics Letters, 2004.
84(10): p. 1731-1733.
35. Bolcavage, A., et al., Phase equilibria of the cu-in system I: Experimental investigation.
36. Weibke, F. and H. Eggers, The Phase Diagram of the System Copper-Indium. Zeitschrift
96
37. Hume-Rothery, W., G.V. Raynor, and H.K. Packer, The Constitution and Structure of
38. Owen, E.A. and D.P. Morris, The Application of X-Ray Methods to the Deterimination of
39. Reynolds, J., W.A. Wiseman, and W. Hume-Rothery, The Equilibrium Diagram of the
System Copper-Indium in the Region 25-35 at.% Indium. Journal of the Institute of
40. Jones, R.O. and E.A. Owen, X-Ray Determination of the Alpha-Phase Boundary of the
Copper-Indium Alloy System. Journal of the Institute of Metals, 1953. 82: p. 445-448.
41. Straumanis, M.E. and L.S. Yu, Lattice parameters, Densities, Expansion Coefficients,
25A: p. 676-682.
42. Jain, V.K.C., M. Ellner, and K. Schubert, On the Phases Occurring Near the
43. Jain, K.C., M. Ellner, and K. Schubert, On the Phases Occurring Near the Composition
44. Okamoto, H., Au-In (Gold-Indium). Journal of Phase Equilibria, 1993. 14(4): p. 532.
97
46. Keppner, W., et al., Studies of Comound Formation at Cu-In, Ag-In, and Au-In Interfaces
201-215.
47. Chen, J.S., E. Kolawa, and M.A. Nicolet, Cu/In Deposited at Room Temperature-
49. Manna, I., et al., Interdiffusion Between in Layer and Bulk Cu or Cu-In Alloy. physica
50. Vianco, P.T., A.C. Kilgo, and R. Grant, Solid-State Intermetallic Compound Layer
Growth between Copper and Hot Dipped Indium Coatings. Journal of Materials Science,
51. KAO, C.R., Microstructures developed in solid-liquid reactions : using Cu-Sn reaction,
Elsevier. 222.
52. Su, L.H., et al., Interfacial reactions in molten Sn/Cu and molten In/Cu couples.
53. Liu, H.S., et al., Thermodynamic reassessment of the Au-In binary system. Calphad-
54. Powell, G.W. and J.D. Braun, Trans. AIME, 1964. 230: p. 694.
55. Simić, V. and Ž. Marinković, Thin film interdiffusion of Au and In at room temperature.
98
56. Bjøntegaard, J., et al., Low temperature interdiffusion in Au/In thin film couples. Thin
57. Liu, Y.M. and T.H. Chuang, Interfacial reactions between liquid indium and Au-
58. German, R.M., Thermodynamic and Kinetic Factors, in Liquid Phase Sintering1985,
59. Oikkonen, M., et al. in Proc. 11th Nordic Semiconductor Meeting. 1984. Espoo, Finland.
60. Dillon, A.C., et al., Surface chemistry of Al2O3 deposition using Al(CH3)3 and H2O in a
61. Ott, A.W., et al., Al3O3 thin film growth on Si(100) using binary reaction sequence
62. Ott, A.W., et al., Atomic layer controlled deposition of Al2O3 films using binary reaction
63. Ferguson, J.D., A.W. Weimer, and S.M. George, Atomic layer deposition of ultrathin and
conformal Al2O3 films on BN particles. Thin Solid Films, 2000. 371(1–2): p. 95-104.
64. Higashi, G.S. and C.G. Fleming, Sequential surface chemical reaction limited growth of
55(19): p. 1963-1965.
65. Wank, J.R., S.M. George, and A.W. Weimer, Coating Fine Nickel Particles with Al2O3
99
67. Kiroubanand Sankarana, et al., Modeling of Copper Diffusion in Amorphous Aluminum
68. Hasselman, D.P.H., K.Y. Donaldson, and A.L. Geiger, Effect of Reinforcement Particle
Matrix Composite. Journal of the American Ceramic Society, 1992. 75(11): p. 3137-
3140.
69. Hasselman, D.P.H. and L.F. Johnson, Effective Thermal Conductivity of Composites with
508-515.
70. Nan, C.-W., et al., Effective thermal conductivity of particulate composites with
71. Yu, S., P. Hing, and X. Hu, Thermal conductivity of polystyrene–aluminum nitride
composite. Composites Part A: Applied Science and Manufacturing, 2002. 33(2): p. 289-
292.
72. Kürti, N., B.V. Rollin, and F. Simon, Preliminary experiments on temperature equilibria
73. Kapitza, P.L., The study of heat transfer in helium II. Journal of Physics (USSR), 1941. 4:
p. 181.
75. Maxwell, J.C., A Treatise on Electricity and Magnetism. 3rd ed. Vol. 1. 1904.
100
76. Liu, J., et al., Liquid phase sintered Cu-In composite solders for thermal interface
7012-7025.
77. Arsenault, R.J. and N. Shi, Dislocation Generation Due to Differences between the
175-187.
78. Young, R.J., et al. High yield and high throughput TEM sample preparation using
focused ion beams. in Proceeding of the 24th International Symposium for Testing and
79. Standard Test Method for Thermal Transmission Properties of Thermally Conductive
81. Pan, D., et al., Miniaturized impression creep testing of ball grid array solder balls
82. Miller, W.S. and F.J. Humphreys, Strengthening Mechanisms in Particulate Metal Matrix
83. K., C.K. and M. M., Intitial dislocation distributions in tungsten fibre-copper composites.
101
85. Ashby, M.F., The Deformation of Plasctically Non-homogeneous Materials.
90. Roy, R., et al., Structural Characterization of the Cuin Intermetallic Phase Produced by
Interfacial Reactions in Cu/in Bimetallic Films. Thin Solid Films, 1993. 229(2): p. 140-
142.
91. Jan, J.P. and W.B. Pearson, Electrical properties of AuAl2, AuGa2 and AuIn2.
92. Nakano, T., et al., Alloying and electrical properties of evaporated Cu-In bilayer thin
93. Franz, R. and G. Wiedemann, Ueber die Wärme-Leitungsfähigkeit der Metalle. Annalen
http://www.makeitfrom.com/material-data/?for=Alumina-Aluminum-Oxide-Al2O3.
102
96. Cahill, D.G., K.E. Goodson, and A. Majumdar, Thermometry and thermal transport in
97. Kim, E.K., et al., Thermal boundary resistance at Ge2Sb2Te5/ZnS : SiO2 interface.
98. Lee, S.M. and D.G. Cahill, Heat transport in thin dielectric films. Journal of Applied
99. Hartman, T.E. and J.S. Chivian, Electron Tunneling Through Thin Aluminum Oxide
100. Kumar, P. and S. Awasthi, Mechanical and Thermal Modeling of In-Cu Composites for
103