Sun Yangyang 200612 PHD

STUDY ON THE NANOCOMPOSITE
UNDERFILL FOR FLIP-CHIP APPLICATION
A Thesis
Presented to
The Academic Faculty
by
Yangyang Sun
In Partial Fulfillment
of the Requirements for the Degree
Doctor of Philosophy in the
School of Chemistry and Biochemistry
Georgia Institute of Technology

December, 2006
STUDY ON THE NANOCOMPOSITE UNDERFILL FOR FLIP-CHIP

APPLICATION
Approved by:
Dr. C. P. Wong, Advisor
School of Materials Science and
Engineering
Dr. Rigoberto Hernandez

Dr. Karl Jacob

School of Polymer, Textile and Fiber
Engineering
Dr. Boris Mizaikoff

Dr. Z. John Zhang

Date Approved: November 8, 2006
II
ACKNOWLEDGEMENTS
I would like to express my sincere gratitude to my advisor, Dr. C. P. Wong for his
guidance, inspiration, and financial support throughout the course of this research. I
would like to extend my gratitude to Dr. Rigoberto Hernandez, Dr. Karl Jacob, Dr. Boris
Mizaikoff and Dr. Z. John Zhang for serving on my Ph. D committee as well as providing
invaluable instructions and recommendations.
I would like to thank the faculty and staff members in the National Science
Foundation Microsystems Packaging Research Center, the School of Chemistry and
Biochemistry, and the School of Materials Science and Engineering. They are Professor
Rao R. Tummala, Professor Jianmin Qu, Professor Suresh Sitaraman, Professor, Z. L.
Wang, Professor David Collard, Dr. Cam Tyson, Dr. Mira Josowicz, Dr. Leyla Conrad,
Mr. Dean Sutter, Ms. Yolande Berta, Ms. Vicki Speights, Ms. Mechelle Kitchings, Mr.
James Cagle, and Mr. Tim Banks.
My special thanks go to my fellow co-workers in Dr. Wongs group, for all the
discussions and helps I received from Dr. Lianhua Fan, Dr. Kyoung-sik Moon, Dr.
Shijian Luo, Dr. Haiying Li, Dr. Zhuqing Zhang, Dr. Jianwen Xu, Dr. Fei Xiao, Dr. Hai
Dong, Dr. Brian Englert, Mr. Suresh Pothukuchi, Ms. Lara Martin, Ms. Yi Li, Ms.
Jiongxin Lu, Mr. Lingbo Zhu, Mr. Hongjin Jiang, Mr. Yonghao Xiu, Mr. Brian Bertram,
Ms. Jessica Burger, Ms. Gusuel Yun. I would like to thank the undergraduate students
and high school intern students who worked with me during the PhD study. They are Mr.
Jonathan Peak, Mr. Jerry Grimes, Mr. David Lorang, Ms. Qian Wan, Ms. Elizabeth
Varner.
Special appreciation is extended to Texas Instruments and Indium Corporation of
America for their interests and supports of this work, and also to Hexion Specialty
Chemicals, Hanse Chemie, and Lindau Chemicals for their supply of the materials. This
III
work is funded by National Science Foundation through Packaging Research Center of

Georgia Tech.
Finally, I would like to thank my parents, my brother, and my friend Zhimin Song
for their continuous support and encouragement. Without them, this dissertation would
not be possible.
IV
TABLE OF CONTENTS
ACKNOWLEDGEMENTS
III
TABLE OF CONTENTS
LIST OF TABLES
IX
LIST OF FIGURES
XI
SUMMARY
CHAPTER 1.
XVII
INTRODUCTION
1.1.
ELECTRONIC PACKAGING AND FLIP-CHIP TECHNOLOGY
1.1.1.
Packaging technology development
1.1.2.
Flip-chip technology
1.2.
UNDERFILL MATERIALS AND NANO SIZE FILLER
1.2.1.
Overview of underfill materials
1.2.2.
Underfill classifications
1.2.3.
Composition of epoxy underfill
1.2.4.
Filler in the underfill
1.3.
PARTICLE DISPERSION
1.3.1.
Energy state of particle in the medium
1.3.2.
Attractive force
1.3.3.
Repulsive force
1.3.4.
Filler stabilization in underfill
1.4.
IMPACT OF NANOPARTICLES ON THE RHEOLOGY
1.4.1.
Definition of viscosity
1.4.2.
Einstein Equation for dilute suspension
1.4.3.
Kreigher-Dougherty Equation for concentrated suspension
1.4.4.
Particle size effect to viscosity
1.5.
RESEARCH OBJECTIVES
CHAPTER 2.
NANOSILICA SYNTHESIS AND MODIFICATION
2.1.
SILICA SYNTHESIS
2.1.1.
Pyrogenic silica
2.1.2.
Sol-gel method
2.1.3.
Size control of nanosilica by Stber method
2.2.
SURFACE MODIFICATION OF SILICA BY SILANE
2.2.1.
Contact angle and surface wetting
2.2.2.
Silane coupling agent
2.3.
EXPERIMENT
2.3.1.
Material
2.3.2.
Surface tension measurement after treatment
1
1
3
5
5
7
10
14
18
18
19
20
22
25
25
27
27
29
31
34
34
34
37
40
44
44
47
49
49
50
2.3.3.
Surface modification of nanosilica
2.3.4.
Particle characterization
2.3.5.
Underfill composite preparation and characterization
2.4.
RESULTS AND DISCUSSION
2.4.1.
Surface tension measurement of silicon dioxide after treatment
2.4.2.
Optimal experimental conditions for nanosilica modification
2.4.3.
Characterizations of treated nanosilica
2.4.4.
Viscosity of nanocomposite no-flow underfill
51
51
52
53
53
57
64
71
CHAPTER 3. MATERIAL PROPERTIES CHARACTERIZATION OF THE

NANOCOMPOSITE UNDERFILL AFTER CURING
73
3.1.
EXPERIMENTS
3.1.1.
Materials
3.1.2.
Underfill composite preparation
3.1.3.
Underfill composite characterization
3.2.
RESULTS AND DISCUSSIONS
3.2.1.
Anhydride epoxy polymerization mechanism
3.2.2.
Curing Behaviors and Tg of composite underfills
3.2.3.
Rheological and optical behavior of composite underfills
3.2.4.
Thermal mechanical properties
3.2.5.
Moisture absorption and density measurement
3.2.6.
Morphology
3.2.7.
Wetting test
3.3.
GLASS TRANSITION AND RELAXATION BEHAVIORS OF NANOCOMPOSITES
3.3.1.
Experiments
3.3.2.
Characterization
3.3.3.
Results and discussion
CHAPTER 4. INFLUENCE OF INTERPHASE AND MOISTURE ON THE
DIELECTRIC SPECTROSCOPY OF EPOXY/SILICA COMPOSITES
4.1.
DIELECTRIC PROPERTIES OF COMPOSITE MATERIALS
4.1.1.
Theory and background
4.1.2.
Existing dielectric study for composite material
4.1.3.
Dielectric properties measurement
4.2.
4.2.1.
Dielectric properties
4.2.2.
TTS shifting of dielectric loss curve
4.2.3.
Dielectric loss in composites
4.2.4.
Moisture influence for dielectric properties
CHAPTER 5.
DISPERSION
73
73
74
75
77
77
79
81
83
86
91
92
94
95
96
97
108
108
108
112
113
114
114
116
119
122
THE HARDENER EFFECTS TO COLLOIDAL SILICA

127
5.1.
EXPERIMENT
5.1.1.
Materials
128
128
VI
5.1.2.
Dynamic rheology
5.1.3.
Dielectric constant of liquid sample
5.2.
5.2.1.
Rheology measurement
5.2.2.
Van der Waals interaction
129
131
131
131
137
CHAPTER 6. PHOTO-POLYMERIZATION OF EPOXY NANOCOMPOSITE FOR

WAFER LEVEL APPLICATION
143
6.1.
PHOTO-POLYMERIZATION OF EPOXY
6.2.
EXPERIMENTS
6.2.1.
Materials
6.2.2.
Preparation of nanocomposites
6.2.3.
Characterization
6.3.
PREPARATION OF PHOTO-CURABLE NANOCOMPOSITES
6.3.1.
Filler size of nanosilica
6.3.2.
UV absorption of compositions in the photo-curable nanocomposite
6.4.
REACTION MECHANISM AND KINETICS OF PHOTO-CURABLE NANOCOMPOSITE
6.4.1.
Mechanism of cationic photo-polymerization
6.4.2.
Reaction process measured by real-time FTIR
6.4.3.
Two-steps curing of underfill by cationic photo-polymerization
6.4.4.
Reaction kinetics of underfill by photo-polymerization
6.5.
MATERIAL PROPERTIES CHARACTERIZATION
6.5.1.
Optical properties
6.5.2.
Glass transition temperature
6.5.3.
Thermal degradation behavior
6.5.4.
Thermal expansion
6.5.5.
Thermal mechanical properties of photo-cured nanocomposites
6.5.6.
Nanocomposite morphology
6.5.7.
Surface hardness
6.6.
APPLICATION OF PHOTO-CURABLE EPOXY NANOCOMPOSITE IN WAFER LEVEL
143
145
145
146
146
151
151
152
155
155
157
160
164
170
170
173
174
175
176
179
180
PACKAGING
Novel wafer level packaging process

Advantages of photo-curable nanocomposites
Pattern formation with photo-curable nanocomposite
182
182
185
186
CONCLUSIONS AND SUGGESTED WORK
190
6.6.1.
6.6.2.
6.6.3.
CHAPTER 7.
7.1.
CONCLUSIONS
7.2.
SUGGESTED WORK
7.2.1.
Chemical bond between filler and epoxy matrix
7.2.2.
Molecular level reinforcement in epoxy
7.2.3.
High performance polymer matrix
7.2.4.
Nanocomposite polymeric optical waveguide
APPENDIX A AUTHORS AWARDS, PATENTS, AND PUBLICAITONS
VII
190
195
195
197
199
200
201
REFERENCE
207
VIII
LIST OF TABLES
Table 1-1. Trend in the microelectronic manufacturing
Table 1-2. Coefficient of thermal expansion of major materials in flip-chip packaging 6

Table 1-3. Underfill classification
10
Table 1-4. List of Epoxy resin used in this study
11
Table 1-5. List of curing agents used in this study
12
Table 1-6. Lists of catalysts used in this study
14
Table 1-7. Bulk resistivity of underfill formulation (before curing)
23
Table 2-1. Physical, mechanical, thermal and electrical properties of silica
35
Table 2-2. Ingredients for sol-gel synthesis silica
40
Table 2-3. Chemistry structure of silane coupling agents
50
Table 2-4. Contact angles (degree) of three probe liquids and epoxy on SiO2 surfaces at
different treatment conditions
56
Table 2-5. Critical surface tension of SiO2 surfaces with different silane treatment
56
Table 2-6. DOE of modification condition
58
Table 3-1. Chemicals used in the underfill formulations
74
Table 3-2. Moisture absorption kinetics parameter
89
Table 4-1. Constant parameters of WLF equation for three samples
119
Table 5-1. Matrix composition of underfill with different hardener
129
Table 5-2. Summary of dynamic rheology of different systems
135
Table 5-3. Bulk material properties for various components
138
Table 6-1. Comparison between two photo-polymerization approaches
144
Table 6-2. Observed peaks of epoxy with FTIR
157
Table 6-3. Reaction heat and conversion for the nanocomposite measured by photo-DSC
and thermal-DSC
163
IX
Table 6-4. Kinetics data for the photo-polymerization of nanocomposite underfill
167
Table 6-5. Light absorptivity and components concentration in the nanocomposite

underfill
167
Table 6-6. TGA measured for various filler loadings in the nanocomposites
174
Table 6-7. Materials Constant
179
LIST OF FIGURES
Figure 1-1. Scheme of electronic packaging hierarchy
Figure 1-2. General configuration of wire bonding package
Figure 1-3. General configuration of flip-chip package
Figure 1-4. Conventional capillary flow underfill process
Figure 1-5. No-flow underfill process
Figure 1-6. Curing mechanism of primary amine
13
Figure 1-7. Scheme of underfill flow and possible filler clog between chip and substrate
14
Figure 1-8. Micron size silica entrapped between the solder and the contact pads
16
Figure 1-9. Optical microscope picture of the flip-chip assembly with nanocomposite noflow underfill
16
Figure 1-10. Energy diagram of particle surface with distance
19
Figure 1-11. Electrostatic force in the dispersion system with ionic strength
21
Figure 1-12. Adsorption-dissociation mechanism of ions on the silica surface in the

aqueous medium
21
Figure 1-13. Steric stabilization of particles by adsorbed polymer
21
Figure 1-14. Viscosity definition model
26
Figure 1-15. Viscosity of concentrated suspensions
28
Figure 1-16. Calculated viscosity at low shear rate as a function of particle diameter: (1)
100nm; (2) 200nm; (3) 300nm; (4) high-shear limit
30
Figure 1-17. Viscosity of underfill with silica filler (nanosilica: 100nm; micron silica:
3m, theoretical calculation is based on the Equation 1-8)
30
Figure 2-1. Synthesis of fumed silica
36
Figure 2-2. TEM picture of fumed silica structure[56]
37
Figure 2-3. Reaction process of sol-gel method for silica generation (with basic catalyst)
38
XI
Figure 2-4. Polymerization behavior of silica [60]
39
Figure 2-5. SEM picture of silica synthesized with different ammonia concentration (a)
0.1M, (b) 0.3M, (c) 0.6M, (d) 1.0M (magnification: 20,000)
42
Figure 2-6. Particle size distribution of as-synthesized silica
43
Figure 2-7. Relation between particle size and ammonia concentration
43
Figure 2-8. Relationship between interfacial tension and contact angle
45
Figure 2-9. Wetting phenomenon of silica filler in the underfill
46
Figure 2-10. General structure of silane coupling agents
48
Figure 2-11. Scheme of surface modification for nano-size filler
49
Figure 2-12. Proposed mechanism for the silane reaction onto the glass slides
54
Figure 2-13. Zisman plot to determine the critical surface tension
55
Figure 2-14. Nanosilica dispersion with different pre-treatments
59
Figure 2-15. Nanosilica dispersion with epoxy-silane treatments
59
Figure 2-16. Average size of nanosilica with different treatment conditions
60
Figure 2-17. Reaction mechanism of silane treatment to nanosilica surface
61
Figure 2-18. Nanosilica dispersion with amino-silane treatments and with sonication
62
Figure 2-19. Dispersion of #004 (amino silane treated)
63
Figure 2-20. Dispersion of #007 (epoxy silane treated)
63
Figure 2-21. Dispersion of #007 (enlarged)
64
Figure 2-22. FTIR spectra of nanosilica with different surface modification
65
Figure 2-23. Physical water decreases and silanol groups condense [83]
67
Figure 2-24. Re-absorption of physical water below 400C[83]
67
Figure 2-25. Irreversible elimination of adjacent silanol group [83]
68
Figure 2-26. Weight loss of silica at different drying condition
70
Figure 2-27. Weight loss of nanosilica with different surface modification
70
Figure 2-28. Viscosity of nanocomposite underfills
71
XII
Figure 3-1. Reaction scheme of anhydride/epoxy polymerization with imidazole catalyst

78
Figure 3-2. Curing behaviors of base underfills and composite by DSC
80
Figure 3-3. Glass transition temperatures of composite underfills by DSC
80
Figure 3-4. Viscosity of silica filled composite underfills
81
Figure 3-5. Effect of filler size on the UV-Vis spectra of the composite underfills
82
Figure 3-6. CTE of silica filled composite underfills
83
Figure 3-7. Dynamic moduli of composite underfills with untreated nanosilica
85
Figure 3-8. Comparison of dynamic moduli of composite underfills with different

nanosilica
85
Figure 3-9. Moisture uptake evaluations for underfill with different silica: (a) 24h (b) 48h
(c) 72h (d) 96h
87
Figure 3-10. Kinetics of moisture uptake for the samples
89
Figure 3-11. Density measurement for silica filled composite underfills
90
Figure 3-12. SEM photographs of nanosilica composite materials (a) untreated-30, (b)
treated-30
91
Figure 3-13. Cross-section views of a quartz chip with no-flow underfill
93
Figure 3-14. Wetting picture of quartz chip with treated-30 underfill
93
Figure 3-15. Glass transition temperature of the silica composites
98
Figure 3-16. Glass transition temperature of the silver composites
99
Figure 3-17. Glass transition temperature of the aluminum composites
99
Figure 3-18. Glass transition temperature of the carbon black composites
101
Figure 3-19. TGA measured weight loss at a heating rate of 20 C/min under air
102
Figure 3-20. Dynamic loss moduli of the silica composites and the blank resin
104
Figure 3-21. Deconvolution of loss modulus of nanocomposite
104
Figure 3-22. Molecular structure of anhydride/epoxy polymer
105
Figure 3-23. Possible local motion of segments in anhydride/epoxy polymer
105
XIII
Figure 4-1. Electrode design of the single surface sensor used in the experiment
113
Figure 4-2. Dielectric property of the control sample after curing
114
Figure 4-3. Dielectric property of the epoxy/silica micron-composite after curing
115
Figure 4-4. Dielectric property of the epoxy/silica nanocomposite after curing
115
Figure 4-5. DEA multi-frequency experiment results of nanocomposite sample (not all
the temperature listed)
118
Figure 4-6. Shift factors of TTS for nanocomposite sample
118
Figure 4-7. Master curves of loss factor for three samples after obtained by TTS shifting
119
Figure 4-8. Moisture absorption of three materials as aging time
120
Figure 4-9. Loss factor and ionic conductivity of the three samples at 1 Hz
122
Figure 4-10. Loss factor of three samples after curing, (a) 1Hz; (b) 1000Hz
124
Figure 4-11. Loss factor of three samples after aging under humidity, (a) 1Hz; (b)
1000Hz
125
Figure 4-12. Loss factor of three samples after drying (a) 1Hz; (b) 1000Hz
126
Figure 5-1. Molecular structure of two hardeners used in the experiment
129
Figure 5-2. Elastic and viscous modulus as a function of frequency for

nanosilica/anhydride mixture
134
Figure 5-3. Elastic and viscous moduli as a function of frequency for nanosilica/amine
mixture
134
Figure 5-4. Steady-shear viscosity as a function of shear stress for nanosilica in two
hardeners
136
Figure 5-5. Schematic representations of two possible scenarios that can occur in the case
of silica particles dispersed in a liquid.
136
Figure 5-6. Van der Waals potential between silica particles in different hardeners
139
Figure 5-7. Glass transition temperatures of silica composites with different hardeners 141
Figure 6-1. Molecular structure of photo-initiator
146
Figure 6-2. Scheme of real-time FITR setup
147
XIV
Figure 6-3. Scheme of photo-DSC setup
148
Figure 6-4. Light transmittance of two kinds of nanosilica in ethanol solution
151
Figure 6-5. TEM picture of the 20 nm colloidal silica
152
Figure 6-6. UV absorbance of pure epoxy, nanosilica and photo-initiator
154
Figure 6-7. Influence of nanosilica on the UV absorption of photo-initiator in epoxy 154

Figure 6-8. Reaction mechanism of cationic photo-polymerization
156
Figure 6-9. FTIR absorption of epoxy
157
Figure 6-10. Peak intensity changes of pure epoxy with different UV exposure times (the
arrow direction represents the time increases)
158
Figure 6-11. Relationship of integrated band area of epoxide peak and UV exposure time
159
Figure 6-12. Heat flow of underfill after UV exposure
162
Figure 6-13. Heat flow of UV-initiated underfill during thermal heating
162
Figure 6-14. Photo-DSC curves of the nanocomposite with different filler loading
163
Figure 6-15. DSC measured heat flow in an isothermal experiment
164
Figure 6-16. Polymerization rate versus time for the photo-polymerization of

nanocomposite underfill
166
Figure 6-17. Conversion versus time for the photo-polymerization of nanocomposite

underfill
166
Figure 6-18. Absorbance of underfill with different filler loading
168
Figure 6-19. Light transmittance of photo-cured nanocomposite with different filler

loading (particle average size: 20 nm)
171
Figure 6-20. Light transmittance of composite with the particle volume fraction, fp
(particle average size: 8m)[140]
171
Figure 6-21. Comparison of light transmittance and the particle volume fraction for
composite with different silica size
172
Figure 6-22. DSC Tg of the nanocomposite after photo-curing followed thermal curing
173
Figure 6-23. TGA graphs of the photo-cured nanocomposites
XV
174
Figure 6-24. Coefficient of thermal expansion of the nanocomposite with various filler
loading
175
Figure 6-25. DMA curves of the photo-cured nanocomposites
176
Figure 6-26. Tan delta peak temperature (DMA Tg) of the photo-cured nanocomposites
177
Figure 6-27. Comparison of composite modulus between the theoretical prediction and
experimental measurement
179
Figure 6-28. TEM picture of nanocomposite after photo-curing
180
Figure 6-29. A plot of load vs. displacement in a nanoindentation experiment
181
Figure 6-30. Hardness of nanocomposite films after photo-curing
181
Figure 6-31. Double ball redistribution uses two solder balls for each I/O, one being
encapsulated in epoxy. (Source: Fraunhofer IZM/Technical University of Berlin)
183
Figure 6-32. Wafer level process with laser ablation method to open the microvia on
underfill
183
Figure 6-33. Proposed wafer process with novel photo-curable nanocomposite
185
Figure 6-34. Molecular structure of SU-8, gamma-butyrolactone and propylene glycol

monomethylether acetate (PGMEA)
188
Figure 6-35. Flow chart of photolithography process for SU-8 nanocomposite
189
Figure 6-36. Photo-defined pattern of SU-8 nanocomposite containing 40 wt% nanosilica

189
Figure 7-1. Silica surface grafting of imidazolium salt as a catalyst
196
Figure 7-2. Surface initiation of epoxy curing reaction.
196
Figure 7-3. Poyhedral oligosilsesquioxane (POSS) structure
197
Figure 7-4. Synthesis route of POSS-containing underfill
199
Figure 7-5. Chemical structure of cyanate ester monomer
200
XVI
SUMMARY
Underfill material is a special colloidal dispersion system with silicon dioxide
particles in the organic liquid. It is used to improve the reliability of integrated circuits
(IC) packaging in the microelectronics. In order to successfully synthesize the
nanocomposite underfill meeting the requirements of the chip package, it is necessary to
have a fundamental understanding of the particle stability in the non-aqueous liquid and
the relationship between materials properties and interphase structure in the composite.
The results of this thesis contribute to the knowledge of colloidal dispersion of
nanoparticles in organic liquid by systematically investigating the effects of particle size,
particle surface chemistry and surface tension, and liquid medium polarity upon the
rheological and thermal mechanical properties of underfill materials. The relaxation and
dielectric properties studies indicate that the polymer molecular chain motion and
polarization in the interphase region can strongly influence the material properties of
nanocomposite, and so a good interaction between particle and polymer matrix is key.
With this study, a potential nanocomposite underfill can be synthesized with low
viscosity, low thermal expansion, and high glass transition temperature. The excellent
transmittance of nanoparticles leads to further investigation of their ability as reinforcing
filler in the photo-curable polymer.
XVII
CHAPTER 1.
INTRODUCTION
1.1. Electronic Packaging and Flip-Chip Technology

1.1.1. Packaging technology development
After the first transistor was invented in Bell Lab in 1947, semiconductor
technology has proceeded from the big, high cost single transistor to highly integrated
circuits(IC), and will continue to develop toward the low cost and high functions of the
electronic products. Today, the electronic industry is the largest and most pervasive
manufacturing industry in the developed world, which has brought profound impact onto
our life.
From the silicon chip to the final products, electronic packaging acts as the key
bridge for the transforms based on the following four major functions: 1) providing an
electrical path to power the circuits, 2) distributing signals onto and off the IC chip, 3)
dissipate the heat generated by the circuits, and 4) supporting and protecting the IC chip
from hostile environments[1].
Figure 1-1 shows the hierarchy of electronics packaging[2]. From the bare chip
fabricated from the silicon wafer, to the final product ready for use, the whole system can
be divided into three levels of the packaging. The first level packaging provides the
interconnection between an IC chip and a module. There are at least three popular
methods for interconnecting the chips on the substrates (either to the module or the
board): 1) face-up wire boding, 2) face up tape-automated bonding (TAB), 3) flip-chip
technology. Second level packaging provides the interconnection between the module to
the printed wiring board (PWB) or a card, which could be realized by the pin through
hole (PTH) technology, or surface mount technology(SMT). Third-level packaging
mainly is the process to put second-level packages onto a motherboard. With the
-1-
requirements towards low-cost, miniaturization and high performance for the current
semiconductor devices, the bare IC chips can be connected to the integrated board using
flip-chip technology directly[3], which is called flip-chip on board (FCOB) or direct chip
attach (DCA).
Figure 1-1. Scheme of electronic packaging hierarchy
Today the electronic assembly and packaging are limiting factors in both cost and
performance for electronic systems. The International Technology Roadmap for
Semiconductor (ITRS) has predicted the main trends in the semiconductor industry
spanning across 15 years into the future. Table 1-1 shows some trends in the
microelectronic manufacturing[4]. The most frequently cited trend is so-called scaling
down, e.g. the ability for industry to exponentially decrease the minimum feature size
used to fabricate integrated circuits. It can be seen that the feature size of IC fabrication
-2-
already shrinks into nanometer scale and will keep decreasing down to 20nm.
Correspondingly, the total I/O number and power on each chip are continuously
increasing. This has resulted in the acceleration of innovation in design concepts,
packaging architectures, materials, manufacturing processes and systems integration
technologies. Specifically, with the smaller and smaller pitch size (distance between the
metal contacting pads on chip), the high-density, high performance method is needed to
connect the IC to the substrate.
Table 1-1. Trend in the microelectronic manufacturing
1.1.2. Flip-chip technology

Flip chip is the first level IC packaging approach in which the active side (with
integrated circuit) of the silicon chip is faced down and connected to the substrate or
printed wire board (PWB)[5].
Figure 1-3 shows a general scheme of the flip-chip
package. The active sides of the chips are bumped with eutectic tin/lead, high lead, or
lead-free solders. After a thermal reflow process, the solder can melt and wet on the
metal contact pad of the substrate, and form the electrical and mechanical connections
-3-
between the IC and the substrate after cooling down. Compared to the conventional wire
bonding technology (Figure 1-2) where the active side of the silicon chip is faced up and
interconnection is made by drawing gold, silver or copper wires from the peripheral edge
of the chip to the substrate, flip chip has many advantages. Since the full area of the chip
surface can be used for interconnection, the input/output density is much higher. It can
provide the shortest possible leads, lowest inductance, smallest device footprint, and
lowest profile. Since the interconnections on the chip can be finished in a one-time
thermal treatment, flip chip avoids the tedious process for individual wires as in wire
bonding.
Figure 1-2. General configuration of wire bonding package
Figure 1-3. General configuration of flip-chip package
-4-
The concept of flip-chip was demonstrated about 40 years ago[6] by IBM on a

ceramic substrate, which was then called Controlled Collapse Chip Connection (C4).
Although the ceramic substrate has low coefficient of thermal expansion (CTE) that
matches the CTE of silicon chip, it entails high temperature and expensive process, as
well as the high dielectric constant that aggravates the signal delay. Recently, the desire
for low cost and mass production has led to increased use of organic substrate. Organic
substrate is favored in terms of its low dielectric constant and low cost. But the high CTE
difference between the organic substrate and the silicon chip exerts great thermal stress
on the solder joints during temperature cycling. This thermal stress is proportional to the
Distance to the Neutral Point (DNP). The larger the chip, the higher the stress, hence, the
worse the solder joint fatigue life. So the organic substrate was inapplicable to flip-chip
technology until underfill was invented in the late 80s[7].
1.2. Underfill materials and nano size filler
1.2.1. Overview of underfill materials
In the early stage of flip-chip technology, the substrates were limited to the highcost ceramic or silicon materials because the great concerns of the thermo-mechanical
fatigue life of the solder joints. Table 1-2 shows the CTE of major materials used in the
flip-chip packaging. Obviously, the CTE mismatch between chip and organic substrate is
much higher than that between solder and ceramic board, which can cause significant
stresses in the solder joints during the product use and leading to fatigue failure.
Therefore, the low-cost organic substrate such FR-4 board and polyimide could not be
used extensively until the reliability issue of solder joints can be solved. Some improved
methods such as optimizing bump distribution design and joint geometry[8, 9], using
highly strong solder composition[10], or matching the CTE of circuit board to that
of
silicon[11], have been explored. However, since they are expensive processes and
-5-
provide limited improvement, above methods are still not fully satisfactory. The socalled underfill, which is placed under the chip to fill the gap between the chip and
substrate, was discovered and became one of most innovative development to enable the
use of low-cost organic substrate in flip-chip packages.
Table 1-2. Coefficient of thermal expansion of major materials in flip-chip packaging
Materials
Silicon
Solder
Alumina
FR-4 board
Polyimide
Epoxy
Silicon dioxide
CTE (ppm/C)
2.5
18-22
6.9
16
45
55-75
0.5
Application in flip-chip
Chip or substrate
interconnects
Ceramic substrate
Organic substrate
Flexible organic substrate
Underfill polymer
Filler of underfill
Underfill is a liquid encapsulant, usually based on un-cured epoxy resin monomer

heavily filled with SiO2 (fused silica). It can be applied to the assembly before or after
chip reflow. Then the liquid underfill can be thermally cured to form a cross-linked
network and converted to a thermoset polymer. With the highly filling of inorganic filler,
the cured underfill shows high modulus, low CTE matching that of the solder joint, as
well as good adhesion to mechanically couple the chip to substrate to restrain most of the
lateral movement between two interfaces. Thermal stresses on the solder joints are
redistributed among the chip, underfill, substrate and all the solder joints, instead of
concentrating on the peripheral joints. The hardened underfill can reduce the solder strain
level to 0.10-0.25 of the strain in joints which are not encapsulated[12, 13], and increase
the fatigue life of the solder by a factor of 10-100. Besides dissipating the thermal stress,
the underfill also provide the environmental protections to the solder joints as the
encapsulant. With the superior advantage mentioned above, underfill products are now
available that deliver on the promise of providing the reliability required for 2nd
generation flip chip on organic platforms. Millions of flip chips are now being assembled
-6-
on FR4 and BT laminate for a wide range of products like cellular phones, pagers, disk
drives, memory modules and much more.
1.2.2. Underfill classifications
The development of underfill technology is always driven by the advances of the
flip-chip technology and advanced in the both directions of underfilling processes and
underfill materials. Generally, the development of the underfilling process pushes the
development of new underfill materials. According to the different processing
procedures, the underfill can be dividend into capillary underfill, molded underfill, noflow underfill and wafer level underfill[14].
The capillary underfill (conventional underfill) is the most mature and
predominant underfill technology in industry manufacturing. It relies on capillary forces
to draw liquid underfill into the gap between the IC and the substrate, as shown in Figure
1-4[14]. Currently, this method faces many problems due to its intrinsic weakness. The
incomplete capillary flow can cause voids and non-homogeneity in the resin/filler system.
The curing of the underfill takes a long time in the oven, consuming additional
manufacturing time. The flux cleaning and flux residue incompatibility create the voiding
problems in the packaging. Decreasing bump pitch and chip height, and increasing bump
density and chip size will eventually push the limits of capillary flow underfill materials.
In order to address the problems associated with conventional underfill and satisfy
the needs of future generations of products, there are several alternative underfill
technology options have been invented. One method is to combine the process of
underfilling and transfer molding into one step and creates the molded underfill[15].
Molded underfill can be applied to the FCIP via a transfer molding process, and it not
only fills the gap between the chip and the interposer/substrate but also encapsulates the
whole chip. In order to easily flow through the gap between chip and substrate, the
molded underfill requires smaller filler size than conventional molding compound. On the
-7-
other hand, the injection-molding process allows high filler loading materials with low
CTE, high modulus and low moisture absorption to be used[16]. Therefore, the molded
underfill can be used in the packages with smaller bump pitch and die standoff height, for
that the capillary underfill usually could not work well.
Chip
Flux
Heated
Bond pad
Solder ball
Board
Fluxing Dispensing
Chip Placement
Solder Reflow
Heated
Underfill
Solvent spray
Flux Cleansing
Underfill Dispensing
Underfill Cure
Figure 1-4. Conventional capillary flow underfill process
No-flow Unde rfill

Heated
Underfill Dispensing
Chip Placement
Solder Reflow &

Underfill Cure
Figure 1-5. No-flow underfill process

Besides the molded underfill, another process called no-flow underfill was also
invented to solve the limitations of capillary underfill. The idea of no-flow underfill was
-8-
first patented by Pennisi et al. in 1992 [17], and the first working material developed was
patented by C.P. Wong and S.H. Shi in 2001 [18]. In the no-flow underfill process, as
shown in Figure 1-5, the underfill material is dispensed on the substrate before the chip is
assembled onto it. After chip placement, the whole assembly is subjected to the solder
reflow. The underfill materials are cured simultaneously during the reflow process.
Sometimes the subsequent post-curing is also needed after assembly to fully cross-link
the underfill. This technique simplifies the flip-chip underfill process by eliminating
separate flux application and cleaning steps before assembly, and avoiding underfill
capillary flow. Thus, the no-flow underfill process can greatly improve the flip-chip
production efficiency.
Although no-flow underfill eliminates the capillary flow and combines fluxing,
solder reflow and underfill curing into one step, it still has some inherent disadvantage,
including individual underfill dispensing step and not totally transparent to standard
surface mount technology (SMT). An improved concept, wafer level underfill, was
invented to improve no-flow underfill and achieve low cost and high reliability[19, 20]. In
this process, the underfill is applied onto a wafer using a proper method, such as printing
or coating. Then the underfill is partially cured to B-stage and the wafer is diced into
single chips. The individual chips will go for further chip assembly and reflow by
standard SMT equipment with no additional process steps required. During the reflow,
the pre-applied material will melt first in order to allow solder wetting and then cure as
solid underfill. This innovative wafer level process eliminates the underfilling and curing
step for each individual die during assembly, and makes the direct die attach process truly
transparent to the assembly line.
Table 1-3 compares the four different underfill processes. It is indicated that not
only the materials chemistry and rheology are different, but also the process and
application steps are varied for these underfills. Therefore, a successful underfill
-9-
approach needs close collaborations between the materials suppliers and assembly
designers.
Table 1-3. Underfill classification
Name
Dispense Stage
Wafer level
After IC fabrication
underfill
and before wafer dicing
No-flow
Before chip assembly
underfill
and reflow
Molded
After chip assembly
underfill
and reflow
Capillary
underfill
After chip assembly

and reflow
Application
Location
On the wafer
Fluxing
ability
Yes
Material Form
On the substrate
Yes
semi-solid (after
B-stage)
liquid
Between chip and

substrate,
overmolding the chip
Between chip and
substrate
No
solid
No
liquid
1.2.3. Composition of epoxy underfill

Different kinds of materials can be used as underfills. However, most underfills
are based on epoxy. The material system is generally composed of an epoxy resin
monomer or epoxy mixture, a curing agent, a catalyst, SiO2 filler, and other necessary
additives depending on the specific application, such as fluxing agent, toughening agent,
adhesion promoter, dispersant agent, etc.
Epoxy resin
The organic compound that contains oxirane groups can be called as epoxy. The
commonly used epoxy resin monomers can be classified in three large groups: diglycydyl
ether type[21], cycloaliphatic type[22] and epoxy novolac resin[23]. The selection of
base epoxy resins is of critical importance to a successful underfill since the many desired
material properties such as viscosity, toughness, and moisture uptake were mainly
determined by the base epoxy resins. With the different polymerization degree and
molecular weight, the epoxy before curing can be low viscosity liquid, high viscosity
- 10 -
liquid, semi-solid and solid. Depending on the application method, the epoxy resin can be
modified with solvent to adjust the viscosity for underfill application. Table 1-4 lists the
epoxy resins which have been use in the thesis study.
Table 1-4. List of Epoxy resin used in this study
Molecular structure
O
CH2 O
CH3
C
CH3
CH2 O
CH2
CH2
O
O CH2
Synonym
EPON 828
O
O CH2
EPON862
O
EPON SU-8
CH2 ]
Name
diglycidyl ether
of bisphenol-A
epoxy resin
diglycidyl ether
of bisphenol-F
epoxy resin
Epoxy phenol
novolac resin
CH3
C
CH3
CH2
ERL4221
cycloaliphatic
epoxy resin
O
O
Curing agent
Although epoxy resin can be initiated by a catalytic initiator and cross-linked by
homo-polymerization, it is necessary to use a curing agent, also known as hardener, to
promote the cross-linking reaction or curing of epoxy resins in the practical application in
order to obtain good material properties. Many organic compounds, including amines,
acid anhydrides, and phenol-formaldehydes, have been used as curing agents[24].
Table 1-5 shows the curing agents used in the thesis study. These three curing
agents can cure epoxy resins through polyaddition reactions by the active hydrogen[25].
For underfill application, the decision among curing agents should consider the viscosity
and flow ability, curing mechanism, gelation behavior[26], wetting ability to the metal
- 11 -
before curing, as well as the chemical structure and material properties after curing. The
compatibility of curing agents to the SiO2 is also an important issue because the filler is
the largest composition in the underfill materials. The polarity and hydrophilicity of
curing agents will influence the filler dispersion and the rheology of the underfill.
Table 1-5. List of curing agents used in this study
Molecular structure
OH
OH
H2 C
OH
Synonym
LBR-6
Name
Phenolic resin
HMPA
Methyl tetra hydro

phthalic anhydride
DETDA
Diethyl toluene
diamine
CH2
n
O
NH2
H2N
H2N
or
NH2
Catalyst
Latent catalyst is another component of critical importance in a successful
underfill since the pot-life, curing temperature and time, and processability of an underfill
is mainly determined by the latent catalysts. To provide convenience to the end user, the
underfill materials are usually formulated as a one-pot composition, by which user just
need dispense the materials without further materials processing such as mixing the
catalyst before usage. This one-pot formulation brings great challenge to the materials
shipping and storage process. Those catalysts in the epoxy which provide an efficient rate
of curing at high temperature are generally not stable enough to be stored for any
- 12 -
appreciable periods. The catalyst tends to gel the epoxy resin prematurely at normal room
temperature, or at temperatures which may be encountered during storage. Thus, it has
been necessary to ship and store the epoxy formulations under the frozen environment,
usually -40C, to prevent the polymerization reaction before material application[27]. For
the no-flow underfill and wafer level underfill, the successful formation of solder joints is
dependent on the curing kinetics of the underfill, which should maintain low degree of
reaction at the solder melting point. Latent catalyst is the key to control over the curing
temperature.
OH
O
R'NH2 + CH2 CHR
R'NH CH2 CHR
Figure 1-6. Curing mechanism of primary amine

For the underfill formulations with primary and secondary amine curing agents,
the catalyst is usually not necessary because these amines contain the active hydrogen
which can add to the epoxy group. Generally, primary and secondary amines are used at
mix ratios that provide one amine active hydrogen for each epoxy group, i.e. the
stoichiometric amount. Figure 1-6 illustrates the initial step which involves the primary
amine reaction. This is followed by the resulting secondary amine adding to another
epoxy group. Sometimes the catalyst also can be added to cure the epoxy more
effectively. Tertiary amines are usually used as a catalyst with trace amount in the
formulation.
There are at least four categories of latent catalysts that have been investigated in
patents and literature. They are: (1) imidazoles and their derivatives[28-30]; (2)
quaternary phosphonium compounds[31]; (3) metal acetylacetonates[32]; (4) some
photoliable onium salts[33]. Table 1-6 lists the catalyst used in this study.
- 13 -
Table 1-6. Lists of catalysts used in this study

Molecular structure
N
N
N
H
CN
CH2 CH2N
N HOOC
C11H23
COOH
Synonym
2E4MZCN
Name
1-cyanoethyl-2-ethyl-4methylimidazole
C11Z-CNS
1-(2-Isocyano-ethyl)-2undecyl-1H-imidazole
COOH
1.2.4. Filler in the underfill
Figure 1-7. Scheme of underfill flow and possible filler clog between chip and substrate
Among the material components in the underfill formulation, filler plays an
important role in reducing overall CTE of the underfill material, minimizing moisture
uptake, and eventually improving device reliability. As a general rule of thumb, the
maximum filler particle size should be less than one third of the gap height between chip
and substrate[34]. Otherwise, the probability of particles getting trapped, shown in Figure
1-7, is very high. Currently, the flip-chip gap size has reached to 50 micron and will
target to 15 micron in the future [35]. The shrinking gap in the flip-chip package
continues to demand the underfill material with smaller and smaller particle size. The
- 14 -
filler size has been quickly decreased to below single digital micron in diameter in most
advanced underfill formulation. The nanosilica ranging in diameter from 20nm to 550nm
has been investigated as the filler for underfill application[36-39].
Another important phenomenon associated with filler is filler particle settling.
Filler settling could occur at different stages of underfill processing, such as during
dispensing, after dispensing, and during curing, resulting in a non-homogeneous filler
content distribution along the Z direction of underfill layer. As many properties of
underfill are function of filler content, filler settling modulates Tg, CTE, toughness and
adhesion of underfill[40]. Severe filler settling can cause cracks, and deteriorate the
potential reliability performance of the underfill materials[41, 42].
The underfill material is a fluid with very small Reynolds number (<<1)[43], and
the flow pattern is in the creeping-flow regime. Therefore, the filler settling velocity ( Vt )
can be described by the Stokes equation:
Vt =
( s f ) g
18 f
ds
Equation 1-1
Where the s is the filler density, g is the acceleration of gravity, d s is filler diameter,
f and f are fluid density and viscosity, respectively.

Stokes equation indicates that the underfill viscosity and filler size are two
controlling parameters for the filler settling velocity. Because the underfill viscosity is
usually temperature-dependent, and the curing temperature is much higher than the
dispensing temperature, it was found that the filler settling mainly occurred during the
underfill curing process and was negligible during the filling process[43]. For an underfill
with a giving viscosity, the filler settling speed could be minimized by reducing the filler
particle size especially when the particle size less than 1 um in diameter[35].
- 15 -
Figure 1-8. Micron size silica entrapped between the solder and the contact pads
Figure 1-9. Optical microscope picture of the flip-chip assembly with nanocomposite noflow underfill
The reducing of filler size is important for underfill processes to address the
concerns about the shrinking of pitch size and gap height, and filler settling problems.
Moreover, the fine size filler is also the most critical factor to ensure the solder wetting
and interconnect formation during a no-flow process. Since the no-flow underfill is
applied to the substrate prior to the placement of the IC chip, conventional micron size
- 16 -
fillers have a great probability of being entrapped between the solder bumps on the chip
and the contact pads on the substrate[44], as shown in Figure 1-8. The trapped fillers
prevent solder wetting on contact pads and thus significantly reduce the solder joint yield
and the electrical continuity[45]. Thus, the micron size fillers have to be replaced by the
nano size fillers in the no-flow process. It was found that silica in the size region of 100
to 150 nm were less likely to be trapped in the flip-chip assembly and a nanocomposite
no-flow underfill with 50 wt% silica filler of 120 nm size has been demonstrated to offer
both high solder joint yield and package reliability in air-to-air thermal cycling test[46].
Figure 1-9 shows an optical microscope image of the flip-chip assembly with the
nanocomposite no-flow underfill.
The wafer level underfill also faces to many challenges regarding the filler
addition. Besides the similar problems related with the solder wetting as no-flow
underfill, the vision recognition in the wafer level underfill process becomes a new issue
because the wafer is covered with underfill[20]. During the pick-and-place vision process
of wafer level packaging, the solder bumps are used as locating features in place of
fiducials. In the presence of underfill, the apparent size and shape of the bumps may be
altered. Especially, the highly filled underfill with large filler particles makes the bumps
almost invisible due to the light scattering. In addition, the three-dimensional topography
of an uncoated bumped die aids vision recognition with the formation of shadows around
the bumps. These shadows enhance the cameras ability to sense the bright bump against
the darkness of the shadows. With a coated die, the topography is flattened, further
complicating the recognition step. The superior optical transparency of nanocomposite
underfill will be helpful to solve this problem[38]. It was found that the underfills with
nanosilica were almost as transparent as pure epoxy in visible region (400~700 nm)
because the filler has a particle size smaller than the wavelength of the visible light.
The nano size fillers show superior properties in the underfill applications and
have the potential to solve the problems associated with the large size filler, such as
- 17 -
clogging flow in the fine pitch package, filler setting down, hindering the solder wetting,
and disturbing the vision recognition during assembly. Nanocomposite underfill will
provide significant reliability improvement for the large-area flip-chip packages.
However, challenges remain and some fundamental problems need to be addressed
before the successful implementation of the nanocomposite underfill to meet the
requirements of low cost, high yield and high reliability for flip-chip assembly.
The filler particle dispersion becomes a dominated factor for the nanocomposite
underfill application. The nanoparticle filled underfill can be considered as a solid-liquid
colloidal dispersion with silica as colloidal phase and epoxy monomer as the liquid
medium. The degree of dispersion, the interaction between the dispersed phase and the
dispersion medium, and the interaction between the particles can influence the materials
properties. The silica colloidal stability in the liquid underfill means the particles have no
tendency to aggregate. However, the filler dispersed state is not the lowest energy
condition; in other words, there is natural tendency for particles to aggregate. To maintain
the silica stable in the underfill, we must use the correct conditions to disperse the particle
and overcome the particle-particle attractive force.
1.3. Particle dispersion
1.3.1. Energy state of particle in the medium
The colloidal domain of matter can be broadly defined as particles of size ranging
from 1 nm to 1m. The stability of dispersion against flocculation is highly dependent
upon
the
total
interaction
energy
between
particles.
If
the
particles
can
thermodynamically seek the lowest free-energy state, a stable colloidal suspension can be
formed. The net, or total, interaction energy is obtained by the summation of the
individual attractive and repulsive energy terms. DLVO (Derjaguin-Landau-Verwey-
- 18 -
Overbeek) theory considers the case in which the attractive energy is due to van der
Waals interactions balanced by the repulsive energy. Now, we will analyze the factors
which influence the silica colloids in the underfill systems.
Figure 1-10. Energy diagram of particle surface with distance

1.3.2. Attractive force
Without any stabilization effects, the particles which are dispersed in a liquid can
attract with each other naturally and tend to fairly unstable[47]. This attraction is
attributed to the fact that the atoms or molecules forming the dispersed particles can
generate transient dipoles because electron density moves about a molecule
probabilistically. There is a high chance that the electron density will not be evenly
distributed throughout a non-polar molecule. This leads to a so-called London-van der
Waals attraction between molecules. The interaction energy between two particles in
vacuum can be found by summing the attractive energies of atom pairs over all atoms of
both particles, which is given by:
- 19 -
r
Equation 1-2
12d
Where A is the Hamaker constant, d is the distance between two particles, and r is the
= A
diameter of the particle.

When the particles are immersed in a medium (e.g. silica filler in the liquid epoxy
monomer for the underfill case), an effective Hamaker constant is substituted in the
Equation 1-2 in order to determine the effective van der Waals interaction energy, which
is given by:
Aeff = A11 + A22 2 A11 A22
Equation 1-3
Where A is the Hamaker constant for particle-particle interaction in vacuum, D is the

Hamaker constant for medium-medium interaction in vacuum.
The effective Hamaker constant Aeff is always positive for particles of type 1
immersed in medium of type 2. Therefore, the van der Waals interaction energy is always
negative (i.e. attractive energy). The higher Hamaker constant, the stronger attractive
interaction occurs between particles. Because of this attraction, dispersed particles
consisting of equal material will form aggregations (flocs), unless there are factors that
retard the aggregation formation. Such factors will be discussed in next section.
1.3.3. Repulsive force
Electrostatic stabilization:
The particles immersed in the liquid can develop an electrical surface charge due
to the interactions of the ions in the liquids, especially in an aqueous system with ions.
With absorbing the ions on the surface, the colloidal particles are charged (Figure 1-11).
In order to maintain overall electro-neutrality of the system, the charged particles are also
accompanied by a surrounding cloud of ions of opposite sign (called counter ions) in the
medium. The formation of charge on the oxide surface is usually controlled by two
important processes[48].
- 20 -
Figure 1-11. Electrostatic force in the dispersion system with ionic strength
OH
Silica
Particle
Silica
Particle
OH
OH
O
O
OH2
OH
Silica
Particle
OH
nH (aq)
Silica
Particle
+ nH (aq)
OH2
OH
OH2
OH
Silica
Particle
OH
+ nOH (aq)
Silica
Particle
OH
+ H2O
Figure 1-12. Adsorption-dissociation mechanism of ions on the silica surface in the

aqueous medium
Figure 1-13. Steric stabilization of particles by adsorbed polymer
- 21 -
(1) Adsorption of water molecules on the oxide followed by dissociation of the

hydrolytic product formed. The fully hydroxylated solid surface has an amphoteric
character and is positively or negatively charged depending on the pH. Figure 1-12 shows
the charging mechanism of silicon dioxide in the aqueous medium.
(2) Adsorption of referentially released ion species from the soluble salts. Salts
such as AgI, CaF2, BaSO4 can be added into solution and ions can form after hydration.
Steric stabilization:
For polymer adsorbed or anchored to the particle surface at large surface
coverage, a polymer-mediated force that can extend 10-100nm will arise due to polymerpolymer interactions across the gap between particles. These can be either repulsive or
attractive depending on the solvent quality and the extent of the polymer excluded
volume effect. Such force, arising from the proximity and overlapping of polymer chains,
have been referred to as steric forces.
In a word, the DLVO theory demonstrated that the stability of charged colloidal
systems was governed by the competition between the attractive van der Waals forces,
and the repulsive electrostatic forces and polymer chain mediated forces. Both of these
forces are long range in nature, extending to a few tens of nanometers[49]. Many welldefined analytical forms have been developed to characterize the particle separation
characteristic and flocculation phenomena.
1.3.4. Filler stabilization in underfill
Electrostatic stabilization and steric stabilization are the most common methods to
stabilize the colloidal particles in the medium. Nevertheless, underfill materials which
consist with the epoxy monomers and silica colloidal particles could not use these two
mechanisms to stabilize the filler particle and achieve good dispersion, because of the
special composition of underfill. Unlike the common colloidal dispersion such as
- 22 -
pigment, polyelectrolyte, or polymer solution, underfill material are quite unique

colloidal dispersion system.
Table 1-7. Bulk resistivity of underfill formulation (before curing)
name
chemical
Epoxy monomer
Curing agent
Bisphenol A
amine
anhydride
Bisphenol A -amine
Bisphenol A -anhydride
water
Underfill
De-ionized water
Bulk resistivity
(ohm-cm)
2.67E+10
2.31E+09
1.66E+08
7.72E+09
1.15E+09
1.80E+07
Firstly, underfill is a non-aqueous/non-solvent system, and the ionic concentration

inside the underfill is limited to extremely low level in order to avoid current leakage and
break-down during the application in the electronic device. Therefore, it is not possible to
form ions or charge on the silicon dioxide particle with the neutral surroundings of epoxy
monomer. Typical particle surface charge densities in aqueous dispersions may be ~0.2
C/m2, but they can be 2 or 3 orders of magnitude lower in the non-aqueous systems[50].
Table 1-7 lists the resistivity of underfill and its components. Comparing to water, these
liquids are quite insulating, and the ion concentration is much lower. From the colloid
point of view, sufficient ions are needed to charge the particle surface and maintain the
overall colloidal stability of the system from a charge stabilization mechanism, which is
lacking in the underfill systems.
Secondly, the silica particles in the underfill system only contact with epoxy
monomer and other small organic molecules (such as curing agents) whose molecular
weight is usually less than 400 g/mol. The organics absorbed on the silica surface are not
long enough to form steric interaction and stabilize the particle. The overlapping and
entanglement of polymer long chains, as in a polyelectrolyte-colloidal system stabilized
by the long chain polymer surfactant, will not happen in the underfill system. It is found,
- 23 -
in fact, that a number of inorganic particles dispersed in the non-aqueous system could
not be stabilized by non-ionic surfactant[51]. Hence, the steric mechanism could not
provide sufficient protection to the colloidal silica particles in the underfill.
As we mentioned before, the DLVO theory is used to describe the surface energy
potential on the particle surface and depict the equilibrium between attractions and
repulsions as a function of the separation distance between two particles. However, the
traditional DLVO treats the liquid medium, in which the particles are immersed, as a
structureless continuum. This approach may be valid when the separation distance
between the colloidal particles is great. When two surfaces of particles approach closer
than a few nanometers, this theory fails to describe interactions between the hydrophilic
particles. Aggregation of hydrophobic colloids has been observed to occur in the
presence of electrostatic repulsion, whereas hydrophilic colloids are stable even without
electrical charge on the surface[48].
For nanosilica underfill system, the filler size has fallen within a few tens of
nanometers, and the distance between each other is very close at the high filler loading.
The non-DLVO forces come into play when the particles are too close to be separated.
For the case of nanosilica dispersed in the underfill, the surfaces of silica particles are
surrounded by the epoxy monomer (and hardener), which act as liquid phase in the
colloidal dispersion. Within the short range, the thin layer of epoxy liquid is not
continuous to the bulk epoxy phase, but has a discrete structure that differs significantly
from that of the bulk phase. Some people refer this region as the interphase, while these
short-distance interactions are usually referred to as solvation forces or as a structure
force. If the solvent (organic liquid) can spread and have good wetting on the particle
surface, the nanoparticle can be stabilized by solvation force. In the other word, if
interaction between the particle surface and liquid is much stronger than that of the
particle and particle, it is possible to stabilize the nanosilica and form a stable colloidal
dispersion. In this scenario, questions like, how the physical and chemical properties of
- 24 -
the solvent and the particle surface affect the free-energy balance and phase behaviors of
the colloidal nanoparticles, and how the properties of stabilized colloidal dispersion
compares with the flocculated one, need to be answered.
1.4. Impact of nanoparticles on the rheology
The application and flow of underfill is governed by its rheology properties. For
the nanosilica filled underfill, the rheology is highly dependent with the filler dispersion
condition. In another word, we also can use the rheological measurements to assess the
state of dispersion in suspensions. Rheology is the science dealing with flow and
deformation of materials. The rheological behavior of particle/liquid systems is important
in most processing operation, including powder and batch preparation, materials
transport, coating and deposition, and shape forming. For underfill application, the
rheology properties are the key to determine the processing condition for underfill flow.
Rheological properties are highly dependent upon the physical structure of the
particle/liquid systems. Structure is governed by factors such as the particle size and
shape distributions, solid/liquid volume ratio, and interparticle forces. Rheological
measurements can often be used to deduce information about the state of particulate
dispersion in the suspension.
1.4.1. Definition of viscosity
Consider a model situation in which a liquid is confined between two parallel

plates (Figure 1-14). One plate is movable and one is held stationary. The plates are
separated by distance y. A force, F, acts on the top, movable plate of area, A, in a
tangential direction, so that the plate moves sideways with a velocity, , relative to the
bottom, stationary plate. The layers of liquid also move in a sideways direction. The top
layer moves with the greatest velocity (i.e. ) and the bottom layer moves with the
- 25 -
smallest velocity(i.e. zero). However, the velocity gradient or shear rate, , is constant
and given by :
d
dy
Equation 1-4
The shear stress, , acting on the top plate is given by:
F
A
Equation 1-5
The viscosity, ,is defined as the ratio of the shear stress to the shear rate:
=
dy
Equation 1-6
Figure 1-14. Viscosity definition model

As the proportionality factor between shear stress and shear rate, the viscosity of a
liquid is an index of the resistance to fluid flow (or alternatively, the rate of energy
- 26 -
dissipation in a flowing liquid depends on the viscosity). The presence of particle in a

liquid results in perturbations of the streamlines during laminar shear flow and, therefore,
an enhanced rate of energy dissipation. In other words, a larger shear stress is required to
maintain the same shear rate as in the pure liquid. Hence, the presence of particles results
in an increased viscosity for the suspension relative to the pure liquid.
1.4.2. Einstein Equation for dilute suspension
Einstein derived an equation which relates the viscosity of a particle/liquid

suspension to the viscosity of the liquid and the volume fraction of solids in the
suspension[52]. It was assumed that particles are: 1) spherical, 2) rigid, 3) uncharged, 4)
very low in the concentration (i.e., hydrodynamic interactions between particles are
ignored), 5) small compared to the dimensions of the container (i.e., wall effects are
ignored), and 6) large compared to the size of liquid molecules (i.e. the liquid medium is
treated as a continuum). Furthermore, the flow rate of the liquids should be low (i.e.,
laminar flow under Stokesian conditions). Under these conditions:
= 0 (1 + 2.5)
Equation 1-7
Where is suspension viscosity, 0 is pure liquid suspension, and is the volume

fraction of solid/particles
This viscosity of colloidal dispersion is the low shear limiting behavior so that the
spatial arrangement of the particles is not perturbed by the shear rate.
1.4.3. Kreigher-Dougherty Equation for concentrated suspension
Although the Einstein equation provides a simple way to estimate the viscosity of
a filler/liquid system, it is only useful for the ideal situation with all the conditions
mentioned above should be satisfied. However, in practical systems, the flow and
viscosity of filler-filled medium are affected by both the decreased spaces between the
particles and particle-particle interactions, which result in attraction and structure
- 27 -
formation. For highly concentrated suspensions with high filler loading, there is not a
rigorous hydrodynamic equation which can describe all the situations accurately due to
the difficulties in handling multibody interactions. Many empirical or semi-empirical
equations have been proposed. The Kreigher-Dougherty Equation[53], as below:
[ ] m
2.5 m
)
0 (1
)
m
m
Where m is the maximum packing fraction of filler in the system
= 0 (1
Equation 1-8
The effect of fillers concentration on suspension viscosity is illustrated

schematically below. Deviations from the Einstein relationship occur at low filler
loading, as the suspension viscosity increases rapidly with increasing solids content. At
high filler loading, (i.e. m ), the particle-particle interlocking occurs and the viscosity
become infinite. The filler loading at which a rigid particulate structure develop is
strongly dependent upon the particle characteristics (shape and size distribution) and the
nature of the interparticle forces.
Relative viscosity
Agglomerated
Dispersed
1
0
Einstein Eq.
A
max
D
max
Volume fraction of filler

Figure 1-15. Viscosity of concentrated suspensions
- 28 -
1.4.4. Particle size effect to viscosity
The viscosity of colloidal dispersion systems is also influenced greatly by the

colloidal particle size. As we know, the colloidal particle in the liquid medium could
move around with a random motion, which is called Brownian motion, due to the thermal
motion and collision. As the filler size decreasing, the colloidal forces between the
particles become significant and the stress required to move them relative to each other
increases. This means that the viscosity at zero-shear condition is controlled by Brownian
term. In order to calculate the maximum packing density, the particle size need be
replaced by the effective radius which is the collision radius of the particles during a
Brownian motion encounter. Due to the increase in the excluded volume of the smaller
size particle, the maximum volume fraction ( m ) is also modified. Hence, Equation 1-8
should be revised for colloidal dispersion with nanoparticle [54]:
0 (1
Where 'm = 0.495(
2.5 'm
)
'm
Equation 1-9
2a 3
) , is the radius of the particle, r0 is the closest distance between
r0
particle centers, or the value of the effective hard sphere diameter.

Russel et al. has calculated the r0 for a colloidal dispersion system which is
equilibrated by the electrostatic repulsive force. It is found that the small particles have
large excluded volume. Based on the revised Kreigher-Dougherty Equation, Figure 1-16
calculates and plots the viscosity increasing versus filler volume fraction for different
filler size. According to this theoretical calculation, the viscosity of dispersion system
with 500nm filler is only about 3 times higher than that of the pure liquid (assuming
volume fraction 0.25). Viscosity increases to 7 times higher for 200nm filler and reaches
to infinite for the 100nm filler. It is shown that the viscosity is very sensitive to the filler
size. The high shear will helpful to reduce the viscosity by the shear-thinning force.
- 29 -
Although this modeling work is not the same as the piratical cases, and the particle size
dependence for other dispersion systems may be different, there are many common
origins so that we can see the general trends for the filler size effect to viscosity.
Figure 1-16. Calculated viscosity at low shear rate as a function of particle diameter: (1)
100nm; (2) 200nm; (3) 300nm; (4) high-shear limit
Viscosity of composite (Pa.s)
1000
nano silica
micron silica
Theoretical calculation
800
600
400
15
10
5
0
0.0
0.1
0.2
0.3
0.4
Filler volume fraction

Figure 1-17. Viscosity of underfill with silica filler (nanosilica: 100nm; micron silica:
3m, theoretical calculation is based on the Equation 1-8)
- 30 -
We know that filler size reduction can lower the maximum packing density due to
the fast Brownian motion; therefore, the viscosity for nanoparticle filled dispersion is
inevitably higher according to Equation 1-9. Moreover, the particles with nanoscale size
possess potentially large excess interfacial free energies due to the high specific surface
area. The Van der Waals attractive potentials would lead to bare (un-stabilized)
nanoparticles having very strong net attractive interaction comparing to the large
particles. The overall effect is that in the absence of methods to lower the interfacial
excess free energy (interfacial tension) and influence the attractive potential, irreversible
flocculation of particles occurs.
Therefore, the reduction of particle size causes
extremely difficulty for particle dispersion and liquid flow, especially for the underfill
systems in which the filler loading is usually very high.
1.5. Research objectives
Underfills are composite materials mainly based on epoxy chemistry and silica
reinforcement. Most of the underfill applications have strict requirements for the
rheological properties of the materials. The nanocomposite underfill presents a novel
solution to the filler entrapment issue in the no-flow underfill process and potentially can
provide significant reliability improvement for the large-area flip-chip packages.
However, challenges remain and some fundamental problems need to be addressed
before the successful implementation of the nanocomposite underfill to meet the
requirements of low cost, high yield and high reliability for flip-chip assembly. The key
issue in nanocomposite underfill and chemical processing is to find the stability
mechanism of colloidal particle in the polymer dispersion, where the traditional DLVO
mechanism by electrostatic repulsion and steric repulsion is absent. The present thesis
will pursue the following research objectives:
(1) To investigate the nanoparticle stability and dispersion in the underfill by
particle surface modification. Silane coupling agent with different functional groups will
- 31 -
be characterized of their ability to alter the filler surface tension and change the liquid
wetting. Solder wetting test will be performed to prove experimentally that the nanosilica
can be used as filler in underfill and will not increase significantly the viscosity if the
optimal modification condition can be found.
(2) To measure the mechanical, thermo-mechanical, and physical properties of the
no-flow underfill with nanosilica. Apart from the impacts on viscosity of underfill, the
filler-filler interaction and filler-polymer interaction also affect the materials properties
due to the high interface area in the nanocomposite. Experiments will be performed to
assure that nanocomposites possess critical material properties for a successful no-flow
underfill process.
(3) To characterize the interphase properties of nanocomposites. Filler
agglomeration and poor filler-polymer compatibility result in deterioration of material
properties such as a lower Tg, higher moisture absorption, and increased viscosity[38].
Despite the large amount of research on the nanocomposites in recent years, the lack of
effective methods or techniques to characterize the interphase property has restricted our
understanding of the nanocomposites properties. The different structure and mobility of
interphase from the continuous bulk polymer phase will be detected, and the effect of
microscopic interphase properties will be studied on the macroscopic material properties
including Tg, dielectric, etc.
(4) To identify the liquid medium properties of the underfill materials. Control
over the interaction between colloidal fillers can be achieved not only by chemical or
physical modification of filler surface properties, but also by tuning the solvent (underfill
liquid) in which colloidal particles are dispersed. By changing the hardener compounds,
we easily can adjust the underfill liquid. A suitable underfill matrix will be found for
nanosilica dispersion.
(5) To study the cationic photo-polymerization of epoxy reinforced by the
nanosilica. The nanoscale size gives the filler unique optical properties and makes it
- 32 -
possible to add filler into the photo-sensitive epoxy without blocking the UV absorption.
The novel photo-curable material with good thermal-mechanical properties can by
synthesized and studied. Understanding the effects of nanoparticles on curing kinetics,
reaction mechanism, and bulk material properties can provide guidelines for the design
and process of photo-curable nanocomposite for microelectronic packaging technology.
- 33 -
CHAPTER 2.
NANOSILICA SYNTHESIS AND MODIFICATION
The silicon dioxide particles (silica) are the most important composition in the
underfill to reduce the thermal expansion of epoxy polymer and reduce the cost of
material. As the reduction of flip-chip feature size, the filler inside the underfill is also
forced to shrink to the nanometer region. The synthesis of large scale nanosilica will be
different from micron size silica. How to add the nanosilica into the epoxy matrix without
significantly increasing the viscosity brings great challenges for the material development.
A fundamental understanding is needed for the surface chemistry of filler, filler
dispersion, solid-liquid interaction and composite liquid viscosity, etc, which will be
investigated in this Chapter.
2.1. Silica synthesis
2.1.1. Pyrogenic silica
The name of silica comprises a large class of products with the general formula
SiO2. Some silica is a natural material, such as flint and quartz. Most of the silica used in
the industry is the synthetic amorphous silica. There are many different synthesis
methods to prepare silica, mainly divided into two categories: wet chemical route to
prepare the silica in the liquid phase and dry chemical route to form silica at high
temperature. A more profound discussion of the wet chemical route will be given below.
Here we are talking about the dry chemical route with pyrogenic condition.
There are two kinds of pyrogenic silica, fumed silica and fused silica. The fused
silica is made by the fusion of high purity sand in electric arc or plasma arc furnace at
temperatures of around 2000C. It is the noncrystalline form of quartz. As a typical
glassy state material, it lacks long range order in its atomic structure. But the highly
cross-linked three dimensional structures give rise to its high use temperature and low
- 34 -
thermal expansion coefficient. Table 2-1 lists the physical, mechanical, thermal and
electrical properties of quartz and fused silica. Fused silica has comparable properties as
the crystal quartz in the terms of thermal conductivity, mechanical and electrical
properties, and furthermore, it has much lower coefficient of thermal expansion (CTE),
which is around 0.5 ppm/K. Therefore, the fused silica particles have been used as filler
in the underfill or encapsulant in the microelectronics application in order to reinforce the
epoxy and reduce the thermal expansion of the composite.
Table 2-1. Physical, mechanical, thermal and electrical properties of silica
Material
Density (g/cm3)
Thermal conductivity (Wm-1 K)
Thermal expansion coeff. (10-6 K-1)
Tensile strength (MPa)
Compressive strength (MPa)
Poisson's ratio
Fracture toughness (MPa)
Melting point (C)
Modulus of elasticity (GPa)
Thermal shock resistance
Permittivity (') *
Tan ( x 104) *
Loss factor ('') *
Dielectric field strength (kV/mm) *
Resistivity (m) *
* Dielectric Properties at 1 MHz 25C
Quartz
2.65
1.3
12.3
55
2070
0.17
1830
70
Excellent
3.8-5.4
3
0.0015
15.0-25.0
1012-1016
Fused silica
2.2
1.4
0.5
110
690-1380
0.165
0.79
1830
73
Excellent
3.8
15.0-40.0
>1018
The fused silica formed by arc or plasma shows a greater variation in particle size.
The primary particles do not form chains but form dense, hard, non-microporous
secondary particles in the micron meter range. In the real application, fused silica needs
to be sieved to get well-controlled filler size. The silica used in the epoxy reinforcement
and thermal stability improvement for underfill and encapsulant applications are the
discrete spherical particles with defined shape and geometry, which is usually the fused
silica with micron range size. For even finer particle sizes, fused silica will be replaced
- 35 -
by the nanosilica synthesized by the sol-gel process by precipitating and drying the
colloidal silica particle. A more profound discussion of this process will be given below.
Fumed silica is another kind of widely used silica additive, which is also
synthesized from the pyrogenic condition. Flame is used to burn the SiCl4 with hydrogen
and oxygen, and following reactions take place (Figure 2-1):
2H2 + O2
2 H2O
SiO2 + 4 HCl
SiCl4 + 2 H2O
2H2 + O2 + SiCl4
SiO2 + 4 HCl
Figure 2-1. Synthesis of fumed silica

Although with the same chemical composition, the fumed silica has quite
different structure and application from the participated silica or fused silica. The fumed
silica has a random network structure formed by SiO4 tetrahedrons, which are highly
disordered. Even with heated up to 1000C, the fumed silica dose not change its
morphology and will not crystallize neither. The precipitated silica differs considerably
from fumed silica, which are completely crystallized after only 20 minutes at 1000C[55].
Visually, the fumed silica is identified as a loose, bluish-white powder with about
98vol% air. The tapping density of fumed silica is around 0.05-0.12g/cm3, which is very
less than the silica density (2.2g/cm3). The primary particles of fumed silica are
extremely small, around 7-40 nm. They contact with each other by hydrogen bonds or
van deer Waals force and build up a loose, non-isolated network. Figure 2-2 shows the
TEM picture of the fumed silica.
With this special structure, the fumed silica, also known by their brand name Cabo-sil (Cabot) or Aerosil (Degussa), is usually used as a thixotropic additive which
when dispersed into the epoxy resin or other organic system increases viscosity, imparts
thixotropic behavior and adds anti-sad and anti-settling characteristics before and during
- 36 -
the potlife or curing of an epoxy systems. Although the primary size of fumed silica is
small, it could not be used as added filler for epoxy reinforcement and CTE reduction
because of its loose structure.
50nm
Figure 2-2. TEM picture of fumed silica structure[56]

2.1.2. Sol-gel method
The sol-gel process is a wet chemical method to make ceramic or glass materials.
The sol is made of solid particles of a diameter of few hundred nanometers, suspended in
a liquid phase. In a typical sol-gel process, the precursor is subjected to a series of
hydrolysis and condensation (polymerization) reactions to form a colloidal suspension,
then the particles either condense in a new phase, the gel, in which a solid macromolecule
is immersed in a solvent, or grow up to large particle. For the silica sol-gel synthesis, the
alkoxysilane such as tetramethoxysilane (TMOS) or tetroethoxysilane (TEOS) are used
as starting compounds.
The alkoxysilane can hydrolyze into monomer Si(OH)4 in the polar solvent such
as ethanol. In order to get a rapid and complete hydrolysis, a base or acid catalyst and
trace amount water may be used. The reaction occurs by a nucleophilic attack of the
oxygen contained in water to the silicon atom[57]. For the basic catalyzing reaction, the
- 37 -
leaving alkoxy group is rapidly protonated. After forming a pentacoordinate intermediate,

and a hydroxyl group is formed on the silicon atom (Figure 2-3 A).
The four alkoxy
groups on the Si atom will be hydrolyzed one by one with enough reaction time.
In the condensation step, the silicic acid Si(OH)4 molecules condense to form
siloxane bonds, with release of water. The condensation reaction may also occur between
the alkoxysilane and the silanol group, releasing an alcohol (Figure 2-3 B). The asformed siloxane molecule can be considered as the monomer which can further cross-link
to difference structure. The condensation reaction may also be acid or base catalyzed.
A: hydrolysis
BH
Si OR + H2O + B
OH Si
OR
Si OH + B + R
OH
B: condensation
Si OH +
Si OH
Si O Si
+H2O
Si OH +
Si OR
Si O Si
+ ROH
Figure 2-3. Reaction process of sol-gel method for silica generation (with basic catalyst)
The hydrolysis and condensation occurs simultaneously. The relative rate of both
processes determines the products structure[58, 59]. In acidic conditions, hydrolysis is
faster than condensation. The rate of condensation slows down with increasing number of
siloxane linkages around a central silicon atom. This leads to weakly branched polymeric
network and further evolution will give the gel structure. In the basic condition, on the
contrary, condensation is accelerated relative to hydrolysis. The rate of condensation
increases with increasing number of siloxane linkage. Thus, highly dense network with
ring structure are formed and gives sol structure. In the initial state, the primary particle
size is limited to about 2 nm. Under the basic environment, the particles are charged by
- 38 -
ionization so the aggregation and bridge formation between particles are limited. The size
growth occurs by monomer deposition and Ostwald ripening. Figure 2-4 illuminates the
growth process and structure control method. In a word, it can generally be said that solgel derived silicon oxide networks, under acid-catalyzed conditions, yield primarily linear
or randomly branched polymers which entangle and form additional branches resulting in
gelation. On the other hand, silicon oxide networks derived under base-catalyzed
conditions yield more highly branched clusters which do not interpenetrate prior to
gelation and thus behave as discrete clusters.
Figure 2-4. Polymerization behavior of silica [60]

For the formed three-dimensional gel networks is called can be dried to hydrogel
if solvent is water, or alcogel with alcohol as solvent. The corresponding drying products
are called as xerogel and aerogel, respectively. These gel materials are highly porous and
extremely low density, therefore are not suitable to be used as reinforcement particles in
the polymer matrix.
- 39 -
In the conventional underfill material, the silica filler is usually the fused silica.
However, as mentioned above, the synthesis of fused silica does not have good sizecontrol, and the size of the commercially available fused silica in the US market is
generally larger than 1 m. Most of the nanosilica is synthesized by sol-gel method
developed by Stber et al. [61]. This involves the condensation of TEOS in alcoholic
solution of water and ammonia and participation of the silica sol. The nanosilica
mentioned later is defined to the Stber silica with submicron diameter.
2.1.3. Size control of nanosilica by Stber method
Experiments
Table 2-2. Ingredients for sol-gel synthesis silica

sample
A01
A02
Ethanol(ml)
88
88
Water (ml)
11.4
10.2
NH4OH(ml)
0.6
1.8
TEOS(ml)
3.6
3.6
Concentration of ammonia
0.1
0.3
(mol/l)a
a: calculation based on the ethanol volume
A03
88
8.4
3.6
3.6
0.6
A04
88
0
12
3.6
2.0
It has been known that with the Stber method, the concentration of the alkoxy
silane, water and base catalyst as well as the reaction temperature control the resulting
particle size and distribution [62]. We did the silica synthesis according to the reference
method. The tetroethoxysilane (TEOS) was used as starting compound; ammonium
hydroxide was the base catalyst. The concentration of TEOS was kept constant (0.15
mol/l). The concentration of ammonia was changed by adjusting the ratio between it and
water. The detailed ingredients for reaction are listed in the Table 2-2. After adding the
TEOS into the solution of ammonia, water in the ethanol, the mixture was slowly stirred.
In most case, the solution became opaque after about half an hour which indicated the
- 40 -
growing of silica. The solution was then evaporated to get rid of the ammonia and most
of ethanol under the reduced pressure by the rotary evaporator. After several time rinsing,
the synthesized silica was kept into the ethanol as colloidal particles.
The colloidal solution of silica was dropped on the glass substrate. After drying
the solvent, the sample was coated with Au and then characterized by the Scanning
Electron Microscopy (SEM, Hitachi S-800). Figure 2-5 shows the shape of assynthesized silica. All the silica particles are solid and hard particles, with no crosslinking structure and pores on the surface of particles. In the basic condition, the
condensation reaction is accelerated relative to hydrolysis during the sol-gel reaction,
therefore the rate of condensation increases with increasing number of siloxane linkage.
Thus, a large, more ramified network of siloxane polymerization was formed and finally
grow up to a spherical sol.
Size characterization
The size distribution of the silica colloids were also characterized by a laser
particle analyzer (Coulter, LS13 320). Ethanol was used as the flowing liquid. Figure 2-6
shows the particle size distribution profiles of four samples. Sample A04 has a wide size
distribution which has the highest ammonia concentration. In order to obtain the
condition for nano-size silica synthesis, the average size and standard deviation of silica
measured from the laser particle analyzer is plotted with the concentration of ammonia,
as shown in Figure 2-7. It shows that with careful adjustment of the concentration of base
catalyst, the silica from sol-gel process can be easily controlled below 100nm.
As we have discussed before, the silica with size below 100nm is good for noflow underfill or other applications requiring fine particle size. For industry application,
the mass production of nanosilica with well size and shape control is needed. Our
experiments here indicated the possibility to synthesize nanosilica with narrow size
- 41 -
distribution and fine particle diameter. After scale up, the source for nanosilica filler will
not become the bottleneck for its applications in the electronics packaging.
(a) A01
(b) A02
(c) A03
(d) A04
Figure 2-5. SEM picture of silica synthesized with different ammonia concentration (a)
0.1M, (b) 0.3M, (c) 0.6M, (d) 1.0M (magnification: 20,000)
- 42 -
25
Volume (%)
20
A01
A02
A03
A04
15
10
5
0
0.01
0.1
10
Particle size (um)

Figure 2-6. Particle size distribution of as-synthesized silica
Silica size (um)
0.6
0.5
A02
0.4
A03
A04
0.3
0.2
0.1
A01
0
0
0.5
1.5
Ammonia concentration (mol/l)
Figure 2-7. Relation between particle size and ammonia concentration
- 43 -
2.2. Surface modification of silica by silane

2.2.1. Contact angle and surface wetting
When the filler is added into a polymer for reinforcement, it is expected that the
filler component which is strong and stiff should bear most of the load or stress applied to
the system which the polymer which is of low strength, fairly tough and extensible
should effectively transmit the load to the filler. In order that the load transfer takes place
effectively, the matrix must have sufficiently high cohesive and interfacial shear strength.
Thus, apart from the filler and the polymer, it is the inevitable region between them,
namely, the interphase which plays a vital role in the fabrication and subsequent behavior
of the filled polymer systems. The interphase is that region separating the filler from the
polymer and comprises the area in the vicinity of the interface[63, 64]. It is synonymous
with the word interfacial region but different from the term interface which would be
the contacting surface where two materials meet. Good material properties for a
composite can be achieved only by uniform and efficient stress transfer through a strong
interfacial interaction between the filler and the polymer matrix. To achieve that, the
filler should be present in the polymer and its surface is wetted by the polymer either
because of the inherent polymer affinity for the filler or because the filler is appropriately
surface treated to provide this affinity.
The wettability of liquid underfill polymer on the solid particle surface is an
important parameter which affects many technological processes such as solid-liquid
dispersion, the interfacial interaction between two phases composite microstructure. The
wettability properties of solids or minerals are assessed quantitatively by a number of
experimental and empirical techniques. One of the quantifying parameters mentioned
above is the contact angle of liquid on the solid surface. Suppose a drop of liquid is
placed on a perfectly smooth solid surface, and these phases are allowed to come to
equilibrium with the surrounding vapor phase.
- 44 -
LV
vapor
liquid
solid
SL
SV
Figure 2-8. Relationship between interfacial tension and contact angle
As shown in Figure 2-8, contact angle ( ) is defined as the angle between the
surface and the tangent plane to the drop profile at the point of contact with the surface.
Contact angle measurement is the most convenient and rapid method to probe the
surface constitution of a solid. It can sense the force of monolayer (5-10 ), thus it is
extremely surface sensitive. Contact angle is related to the surface tension of the solid
surface ( SV , some time is also written as S ), the liquid surface ( LV , or written as L ),
and the interface between solid and liquid ( SL ). Their relationship is given by Youngs
equation:
L cos( ) = S SL
Equation 2-1
The wetting condition can be described by the contact angle of liquid drop. When
the contact angle is greater than 90, s - sl is negative, and the surface asperities cannot
be filled by the advancing liquid, and this situation is non-wetting. When is between 0
and 90, s - sl is smaller than the surface tension of the liquid, thus liquid film can not be
formed spontaneously. Nevertheless, s - sl is still positive, so there will be a decrease in
free energy on converting the solid-vapor interface to a solid-liquid interface, and the
substrate is wettable to this liquid. In the context of water, a wettable surface may also be
- 45 -
termed hydrophilic and a non-wettable surface hydrophobic. This situation in which three
phases meet at a common line of contact is called partial wetting, and it is characterized
by the existence of a contact angle as shown in Figure 2-8.
When the contact angle of a liquid on a solid is 0, liquid film can be formed by
the decreasing of the total free energy (l + sl - s < 0). This situation is called complete
wetting. The liquid can spontaneously spread on the solid surface drawing by the extra
surface tension, and form a macroscopically thick layer of the liquid phase on the solid
phase.
For the particle/liquid suspension, such as underfill system, the dry powders of
silica are immersed into the liquid polymer medium. If the solid is complete wetted by
the liquid, that is called immersional wetting[65]. If the contact angle between solid and
liquid is not zero, i.e., the solid is only partially wettable in the liquid, the air bubbles will
be entrained on the surface of the particles. Figure 2-9 indicates the immersion process.
silica
silica
underfill liquid
Figure 2-9. Wetting phenomenon of silica filler in the underfill
- 46 -
So the wetting of liquid to the particle external surface can be describe by the
spreading coefficient: S SL :
S SL = S SL L
Equation 2-2
The spontaneous wetting of the powder by the liquid can only happen when
the S SL > 0 . Therefore, the higher surface tension of solid, the better chance to for the
liquid to spread on the surface. In another word, the high surface tension of solid is
preferred in term of liquid wetting.
2.2.2. Silane coupling agent
As we discussed in the Chapter 1, the nanosilica is need for no-flow underfill and
wafer level underfill to solve the filler hindrance problem during reflow process
associated with the micron size filler. However, the nano-size silica produced by sol-gel
method has a large surface area covered by silanol groups. This hydrophilic surface does
not process good compatibility with the polymer resin, and therefore the silica cannot be
wetted very well by the resin. On the contrary, the silica particles with hydrophilic
surface easily adhere to each other through hydrogen bonding and form irregular
agglomerations. The agglomerations of the nanosilica can form a network through the
whole polymer matrix and occlude liquid polymer in their interparticle voids, thereby
affecting the rheology of the composite underfill and giving a significant rise to the
viscosity as filler loading increases. The high viscosity of the no-flow underfill not only
makes underfill dispensing difficult, but prevents the chip from collapsing and forming
solder joints during the solder reflow process as well. Moreover, the presence of filler
agglomerations will decrease the maximum filler loading[66], resulting in an inferior
thermal mechanical performance.
In order to decrease the viscosity of underfill and to increase the extent of filler
loading, it is therefore necessary to reduce the degree of agglomeration and improve the
- 47 -
wetting ability of underfill liquid to silica surface. In fact, it is said the fully
disagglomerated fillers in a liquid polymer do not appreciably increase the viscosity[60].
For nano-size filler, the mechanical mixing/dispersion methods such as high speed
shearing or milling are not effective to break down the agglomerations because the
electrostatic forces holding the particles together are stronger than the shear force created
by the velocity gradient. In such circumstances, chemical treatment of nanoparticle
surface is necessary to achieve better compatibility and dispersion of the filler in epoxy
resin, as such if improves the rheological behavior of the nanosilica filled underfill.
Silane coupling agents are often used to treat the silica filler due to their unique
structure. The general formula for a silane coupling agent typically shows the two classes
of functionality (Figure 2-10). X is a hydrolysable group typically alkoxy. Following
hydrolysis, a reactive silanol group is formed, which can react with the silanol groups on
silica surface. The R group is non-hydrolysable organic group that imparts desired
characteristics. The modification process is described as a hydrolysis and condensation
reaction between the silane coupling agents and the silica surface in a polar medium. The
bonding between the silane and the silica surface removes the surface silanol groups and
changes the surface tension of silica particle. As shown in Figure 2-11, the ideal result of
surface treatment is to reduce the filler-filler interaction and to achieve the homogenous
of the nano-size silica in the polymer.
R
organofunctional
group
CH2
Si
linker
Silicon
atom
X3
hydrolyzable
group (alkoxy)
Figure 2-10. General structure of silane coupling agents
- 48 -
Surface
Modification
Figure 2-11. Scheme of surface modification for nano-size filler

There are many factors affecting the result of silica surface treatment, including
the type of coupling agent, concentration of coupling agent, treatment time and predisperse method. The treatment by different silane may give variable surface energy,
partition characteristics and chemical properties on the solid surface due to different
functionality. Since these factors might interact with each other in determining the final
result of the surface treatment, a design of experiment (DOE) is needed to achieve the
optimal conditions for the treatment. The purpose of this chapter is to investigate the
optimal condition for nanosilica surface treatment in order to formulate nanocomposite
no-flow underfill with low viscosity and good filler dispersion. The results of nanosilica
modification were also studied using characterization techniques including FTIR, TGA
and contact angle measurement. The research will be helpful to elucidate the structureproperty relationship of the silica nanocomposite underfills.
2.3. Experiment
2.3.1. Material
Silica nanoparticles (SiO2 powder, around 90nm average diameter, S.D. of 20nm)
were commercially available and used as received. Five kinds of silane coupling agents
were used to modify the SiO2 surface (Table 2-3). A mixture of de-ionized (DI) water and
absolute ethanol was used as the medium for silica treatment, and formic acid was used to
adjust the pH value of the medium.
- 49 -
Table 2-3. Chemistry structure of silane coupling agents

Chemical Name
Triethoxy (methyl)
silane
Abbreviation
CA-ME
Molecular Structure
Functional group
Methyl group
CH3
O
Si
O
Vinyltrimethoxysila
ne
CA-VI
Vinyl group
O
Si
-aminopropyltriethoxysilane
CA-NH
Amino group
H2N
H2N
O
Si
-glycidoxypropyltrimethoxysilane
CA-EP
O
Epoxide group
O
Si
2.3.2. Surface tension measurement after treatment
The effects of silane functional group were identified by measuring the surface
tension of SiO2 after treatment. A glass slide was processed by immersing into
ethanol/water/silane (95/5/0.5 by weight ratio) solution overnight. After rinsing and
drying, these glass slides were used as substrate to mimic the silica particle surface. The
three-liquid-probe method was used to characterize the contact angle of three liquids on
the SiO2 substrate. Two kinds of silane will be chosen for particle surface modification
according to the surface tension measurements.
- 50 -
2.3.3. Surface modification of nanosilica
According to the design of experiment, four independent variables were

investigated, including the type of coupling agents (two), concentration of coupling agent
in the solution (high: 1wt%, and low: 0.5wt%), treatment time (6 and 48 hours) and pretreatment method (sonication or stirring). 5g DI water, 95g ethanol and 10g nanosilica
were mixed with ultrasonicator or magnetic stirring plate for 30 minutes. Then a specific
amount of coupling agent was added into the mixture. The pH value was adjusted to
around 4 by formic acid. The mixture was refluxed at 90C for a prescribed duration. The
detailed experimental conditions for each sample were listed in Table 2-6. After cooling
down, the transparent nanosilica colloidal solution was obtained and was used for further
characterizations.
2.3.4. Particle characterization
Particle distribution
The particle (or particle agglomeration) size and distribution of nanosilica colloid
were characterized by a laser particle analyzer (Coulter, LS13 320) after silica surface
treatment. Ethanol was used as the flowing liquid.
TEM
To prepare samples for Transmission Electron Microscopy (TEM) analyses, a
drop of nanosilica colloid solution was placed on a copper grid coated with carbon
membrane. The solvent was evaporated at room temperature, leaving the nanosilica on
the grid. These samples were examined with a TEM (JEOL 100C, 100kV). Based on the
results of particle size analysis and TEM, an optimal condition under which a monodispersion of the nanosilica in the solution was selected as a general experimental method
to prepare treated nanosilica.
- 51 -
FTIR
To study the functionalization of silane coupling agents on nanosilica surfaces, a
Fourier-Transform Infrared (FTIR) Spectroscope was used (Nicolet, Magna IR 560). The
nanosilica obtained after the modification was first washed by ethanol and then spuncoated on the gold-coated glass substrate. After drying at 110C for 6 hours, the silicacovered substrates were scanned at a reflectance mode in FTIR for 320 times in the
wavelength range from 400 to 4000 cm-1. The background was collected using a goldcoated glass.
TGA
The Thermal Gravimetric Analyzer (TGA) was used to measure surface
absorption of nanosilica before and after surface treatment. The nanosilica powder, which
had been dried on 110C for 6 hours, was heated under nitrogen from room temperature
to 1000 C at a ramping rate of 20 C/min. The weight loss of the nanosilica powder as a
function of temperature was recorded.
Contact angle
One purpose of silane treatment is to replace the surface silanol groups with silane
functional groups and to change the silica surface from a hydrophilic nature to a
hydrophobic nature, as well as to increase the surface tension of SiO2. A goniometer
(Rame-Hart, NRL100-00) was used to measure the contact angles of the water and epoxy
liquid on silane modified surface. A glass slide, which has been processed following the
same procedure as the silica particle modification, was used as the substrate. The test was
performed at the room temperature.
2.3.5. Underfill composite preparation and characterization
The composite underfills were prepared by combining EPON828 and silane

treated nanosilica solution in a 250mL single-neck flask and then by subjecting the
- 52 -
mixture to a vacuum stripping on a rotation evaporator. After most of the solvent was
removed, the mixture was transferred to a beaker and dried in a vacuum at 90C
overnight to remove all the residue solvent. HMPA and catalyst were added into the
obtained viscous mixture and stirred to form a homogenous composite underfill. For the
comparison, the as-received untreated nanosilica powder was directly added into the
underfill and mixed by a high-speed blender. The viscosity of these composite underfills
at room temperature was studied using a stress rheometer (TA Instruments, AR1000N) in
a steady flow mode. A cone-and-plate geometry was used.
2.4. Results and discussion
2.4.1. Surface tension measurement of silicon dioxide after treatment
Silane coupling agents have been used to improve the interface between different
subjects and increase the adhesion in the composite materials for a long time. By treating
the solid surface with silane, the surface can acquire specific surface energy and chemical
properties. The type of silane functional group plays an important role for the treatment
effect. Among many types of silane, we choose four typical silanes to characterize their
effect to alter the surface properties of silicon dioxide. They are name as CA-ME with
methyl group, CA-VI with vinyl group, CA-NH with amino group, and CA-EP with
epoxide group.
The underfill before curing is a colloidal dispersion liquid with silica particle as
solid phase and epoxy monomer as the liquid phase. In order to have good interaction
between the silica phase and underfill, the filler surface must have high surface tension in
order to get positive spreading coefficient (as shown in Equation 2-2). The contact angle
has been offered as the best wettability indicator when polymer fluids and smooth
surfaces are used[67]. For a review of this method refer the paper of Anderson [68-70].
However, it is virtually impossible to measure contact angles on the single nanoparticles
- 53 -
with current technology. An alternative method used by some people is to make a

compressed cake of particles and then measure the contact angle of liquid droplet on the
cake surface. The reproducibility of this method is poor, and it may present a particular
problem for low angles because the liquid could diffuse into the cake[65]. Therefore, in
our experiments, the flat silicon dioxide substrate was used to replace the silicon dioxide
particles. The substrates were treated as the similar way as the particles with different
silanes, and then contact angle of liquid can be measured on the flat surface. Although
results obtained by this way may not as same as on the nanoparticle surface, some
common trend of silane treatment to surface tension of solid can be obtained and these
results will be helpful to the silane selection.
Figure 2-12. Proposed mechanism for the silane reaction onto the glass slides
After silane treatment, the silanol group on the solid surface will be replaced by
the silane functional group. A schematic of the silane reaction mechanism [71, 72] is
shown in Figure 2-12, where R is terminated functional group of the silane. Zisman etc.
had developed an empirical characteristic, called critical surface tension ( c ) to roughly
determine the wetting [73]. A series of liquids with know surface tension are used as
- 54 -
probes. These experiments have to be conducted on a flat, non-porous sample of that

solid. A small droplet of each liquid is placed onto the surface. One measures the angle of
contact at the solid-liquid-air contact line. The angle is drawn through the liquid phase.
One plots the cosine of the angles of contact versus the surface tension of each liquid.
The critical surface tension equals the surface tension at which the plotted line intersects
1.0. Schematically, the resulting Zisman plot looks like Figure 2-13. The liquid with a
surface tension below the critical surface tension of a substrate will wet the surface, i.e.,
show a contact angle of 0 degree. If the surface tension of liquid is higher than the c of
the substrate, the spontaneous spreading of the liquid will not occur on this substrate.
1
cos()
0.8
0.6
0.4
0.2
0
20
30
40
50
60
70
80
Surface tension (mJ/m2)
Figure 2-13. Zisman plot to determine the critical surface tension

In our experiments, three liquids, Diiodo methane (CH2I2), water and Ethylene
glycol (EG) are used as testing liquids. Their surface tension ( L ) is 50.8, 72.8 and 47.9
mJ/m2, respectively. Table 2-4 indicates the contact angle of different liquids on the
silicon dioxide substrate with silane treatment. Each value is the average value of fine
- 55 -
measurements. The critical surface tension on the silicon dioxide surface with different
silane treatment can be calculated according to Figure 2-13.
Table 2-4. Contact angles (degree) of three probe liquids and epoxy on SiO2 surfaces at
different treatment conditions
Substrate type
Treatment Condition
CH2I2
H2O
EG
SiO2
CA-ME
40.9
42.5
35.8
SiO2
CA-VI
40.0
57.3
23.0
SiO2
CA-NH
24.2
47.4
30.3
SiO2
CA-EP
16.5
56.8
23.3
Table 2-5. Critical surface tension of SiO2 surfaces with different silane treatment
Substrate type
Treatment
Functional
Critical surface tension of
silane
group
treated surface (mJ/m2)
SiO2
CA-ME
Methyl
24.3
SiO2
CA-VI
Vinyl
29.4
SiO2
CA-NH
Amino
36.3
SiO2
CA-EP
Epoxide
47.6
From Table 2-5, it can be seen that the surface with epoxide silane treatment has
the highest c . The amino silane takes the second place. Vinyl and methyl silane treated
surface has lower c than the other two. It seems the silane with polar functional group is
more effective to change the c and benefit the wetting for liquid. The surface tension of
pure Bisphenol A monomer ( l ) is about 42.5 mJ/m2, which is lower than c of the
epoxide silane treated surface. Therefore the epoxide silane seems to be most promising
candidate to treat the surface. Nevertheless, our experiment used the flat glass to replace
- 56 -
silica particle in order due to the technical difficulty. As we mentioned before, the results
are of more qualitative interest than the quantitive number. Two polar silanes (amino
silane and epoxide silane) will be both used to silica particle surface characterization, and
the effect of treatment will be evaluated by the particle dispersion.
From the molecular level, the surface tension can be interpreted as some
molecular interactions[74], such as: (1) Hydrogen bonding: hydrogen atoms serve as
bridges linking together two atoms of high electron-negativity. If these two atoms are in
separate molecules, so the molecules themselves are mutually attracted by these bonds.
(2) Permanent dipole interactions: polar molecules have relatively positive and negative
regions. Regions of opposite charge on different molecules result in an attraction between
these molecules. (3) London forces: deformable electron clouds in adjoining molecules
distort one another, resulting in an instantaneous polarity with accompanying attraction
between the molecules involved. The polarizability of a molecule is a measure of its
tendency to display this effect. Therefore, a high polar function group of silane is
preferred for the silica surface modification in term of epoxy wetting.
2.4.2. Optimal experimental conditions for nanosilica modification
The surface modification of silica can be carried on in either aqueous solvent or

organic solvent. If modification is performed in fully dry conditions, the absolute dry
solvent is needed, which is usually dehydrated toluene[75]. This method is not suitable
for large scale production due to the cost and toxicity of solvent. Hence the aqueous
solvent, such as water/ethanol mixture, is the most commonly used one. The reaction
temperature of surface modification was chosen according to literature [76]. Other
experiment parameters, such as silane type and concentration, reaction time and pretreatment method may play complex roles to the treatment effect. Therefore, the DOE
was carried out to investigate the effect of the experiment on the silica surface
modification. The average particle size after treatment and the standard deviation (S.D.)
- 57 -
measured by laser particle size analyzer are listed in the Table 2-6Error! Not a valid
bookmark self-reference..
Table 2-6. DOE of modification condition

Sample
No.
001
002
003
004
005
006
007
008
009
010
011
Coupling
agent
No
No
CA-EP
CA-NH
CA-NH
CA-EP
CA-EP
CA-NH
CA-NH
CA-EP
CA-EP
Conc.
of CA
No
No
Low
High
Low
High
Low
High
Low
High
Low
Treat
time (h)
No
No
48
48
48
48
48
6
6
6
6
Dispersion before
treatment
Stirring 30min
Sonication 30min
Stirring 30min
Sonication 30min
Sonication 30min
Sonication 30min
Sonication 30min
Sonication 30min
Sonication 30min
Sonication 30min
Sonication 30min
Ave. size S.D.

(m)
(um)
9.66
3.70
8.16
3.81
3.80
4.87
4.73
3.79
7.03
5.19
0.55
1.78
0.11
0.05
4.88
3.62
5.21
3.72
7.23
3.97
5.46
4.18
Effect of pre-treatment to the particle dispersion

Figure 2-14 compared four nanosilica colloidal solutions w/wo ultra-sonication
and silane treatment. The particle size distribution showed that the ultra-sonication
treatment was essential to ensure a good surface modification by the silane coupling
agent in obtaining a mono-dispersion of the nanosilica. Sample #003 was treated with the
same condition as #007 except that it was not pre-dispersed by sonication. However,
large silica agglomerations with size around 10um were found in sample #003. As for
sample #001 and #002 which were only dispersed by physical method, it is obvious that
large size agglomerations were very difficult to disperse without chemical treatment.
- 58 -
10
001:stirring, untreated
002:sonication, untreated
003:stirring, treated
007:sonication, treated
Volume (%)
0
0.01
0.1
10
100
1000
10000
particle diameter (um)
Figure 2-14. Nanosilica dispersion with different pre-treatments

10
006:high conc., long time

007:low conc., long time
010:high conc., short time
011:low conc., short time
Volume (%)
0.01
0.1
10
100
1000
10000
particle diameter(m)
Figure 2-15. Nanosilica dispersion with epoxy-silane treatments
- 59 -
Effect of treatment duration and silane concentration to the particle dispersion

The size distribution of nanosilica treated by epoxy silane CA-EP is shown in
Figure 2-15. Among them, sample #007 is mono-dispersed with an average particle size
of 0.11m, which was almost equal to the individual particle size provided by
manufacturer, i.e., the experimental condition to treat #007 sample was the optimal
treatment method and can achieve nanosilica mono-dispersion. The results favored a long
treatment time and low silane concentration to achieve better dispersion, as indicated in
Figure 2-16. This can be explained by the silane reaction mechanism in Figure 2-17. The
alkoxy groups of silane were firstly hydrolyzed by the water in the solution to form
silanol groups. The added acid can catalyze this reaction though a nucleophilic attack at
the silicon atom of silane. Compared with the hydrolysis, the silane condensation reaction
was slower in acidic conditions than in basic conditions[77]. So a long reaction time was
necessary for the completion of the condensation reaction between silane and the
nanosilica surface.
Particle average size (um)
high conc.
low conc.
10
0.1
6
Treatment time (hrs)
48
Figure 2-16. Average size of nanosilica with different treatment conditions
- 60 -
OR
Si
OR
OR
organofunctional silane
hydrolysis of silane
+
3H2O
Y
Si
OH
+ 3ROH
OH
OH
Y
Y
Si
HO
HO O
H
O
Si
Si
HO
HO
HO
OH
Si
condensation of silnaol
on silica surface
HO
Si
Si
OH
Si
Si
O
O
Si
O
O
Si
form a stable organic

functionalization on silica surface
nanosilica surface
Figure 2-17. Reaction mechanism of silane treatment to nanosilica surface

The as-formed silanol group from silane can either react with the silanol groups
on silica surface to form a stable functionalization on the silica surface, or with
neighboring hydrolyzed silane molecules to form siloxane by further polymerization [78].
So the silane concentration should be very dilute to keep the highly reactive hydrolyzed
molecules from reacting with one another. From the results of our experiment, a silane
concentration of 0.5wt% in the solution is preferred.
Effect of silane type to the particle dispersion
The particle dispersion of nanosilica modified by amino silane CA-NH at
different experimental conditions is shown in Figure 2-18. Compared to sample #002 that
was dispersed by sonication but without any chemical treatment, the effect of amino
silane treatment was not obvious and the bimodal distribution of nanosilica was only
slightly changed. So the type and functionality of the silanes played an important role in
the experiment. The TEM pictures in Figure 2-19 also shows that the amino silane hardly
had any effect in breaking the particle agglomerations, while epoxy silane effectively
- 61 -
broke the agglomerations and achieved particle mono-dispersion (Figure 2-20 & Figure
2-21). The reason that epoxy silane is superior to amino silane in term of nanosilica
modification is related on the hydrophilicity of their functional groups. Compared with
epoxy group, the amino group is still hydrophilic, as shown in the following contact angle
test. On the other hand, the amino group in the silane is easily to form hydrogen bonding
with silanol groups on the silica surface[79], which also influence the proceeding of
reaction between silica and hydrolyzed silane. After all, the filler-filler interaction
between nanosilica surfaces cannot be eliminated by amino silane treatment, so as the
irregular agglomerations of nanosilica.
10
004:high conc., long time

005:low conc., long time
008:high conc., short time
009:low conc., short time
002:sonication, untreated
Volume (%)
0
0.01
0.1
10
100
1000
10000
Particle diameter (um)
Figure 2-18. Nanosilica dispersion with amino-silane treatments and with sonication
- 62 -
Figure 2-19. Dispersion of #004 (amino silane treated)
Figure 2-20. Dispersion of #007 (epoxy silane treated)
- 63 -
Figure 2-21. Dispersion of #007 (enlarged)

Based on the results of the DOE, we can summarize the optimal experimental
conditions to achieve nanosilica mono-dispersion as nanosilica treated by epoxy
functional silane with a long reaction time and low silane concentration assisted with
pretreatment by sonication. An acidic condition, probably the formic acid is the most
suitable, is also necessary in order to prevent the intermolecular condensation of silane.
Silica modified by this condition was then used for further characterization. Nanosilica
without treatment was also used for comparison.
2.4.3. Characterizations of treated nanosilica
FTIR
The surface chemistry of the nanosilica was studied using Fourier Transformed
Infrared Spectroscopy (FTIR). Figure 2-22 shows the FTIR spectra of three nanosilica
samples. The silane functional groups are not easy to detect due to the strong and broad
peaks of Si-O (1020-1250 cm-1) and OH (3300-3700 cm-1). However, the intensity of
the absorption peak between 3300 and 3700 cm-1 related to the presence of silanol group
was lower in the treated silica than that in the untreated silica. Compared to the CA-NH
treated silica, CA-EP treated sample shows even lower intensity of silanol group,
indicating a more complete surface condensation reaction. In addition, the vibration peak
- 64 -
of NH is also detected between 1500 and 1700 cm-1 in CA-NH treated silica. From the
FTIR analyses, it can be ascertained that the functional groups are successfully
introduced onto the nanosilica surface.
Figure 2-22. FTIR spectra of nanosilica with different surface modification

TGA
The surface moisture absorption of particles is closely related with the surface
chemistry and total surface area of the silica particles. Micron-size silica particles used in
the underfill are usually synthesized by the pyrogenic method at very high temperature;
therefore moisture absorption is almost negligible due to the limited silanol groups and
less surface area. The nanosilica synthesized by the Stber method in the solution is
covered by the silanol groups on the surface. The hydrogen-bonded moisture and the
hydroxyl groups on the nanosilica particle surface can be characterized by the TGA
according to the weight change of the particles versus temperature. By measuring TGA,
valuable information can be obtained for water absorption properties and chemical
residue change on the particle surface.
- 65 -
Some researchers have investigated the dehydration of the silica gel materials[8082]. Although the synthesis methods for nanosilica particles silica gel are different, a
similar drying mechanism is expected for them due to the similar chemical composition
on the surface. It is known that the weight loss of silica during heating is coming from
both the physically absorbed water and the chemically absorbed water. As temperature
increasing, the water molecules on the silica surface are dehydrated step by step.
(1) Reversible weight loss below 400C. During this stage, the physical absorbed
water on the particle surface is eliminated by the evaporation. The surface vicinal
hydroxyls groups (Si-O-H) can condense and release water starting at about 170C[80,
83], as shown in Figure 2-23. This process is completely reversible. Some decomposition
of organic residuals may also happen during this stage. Hair [84]found that at 400C, no
more than half of the surface hydroxyl groups had been desorbed and that the most of the
remaining surface hydroxyl groups were adjacent to each other and therefore situated for
preferential water adsorption at low temperature(Figure 2-24).
(2) Irreversible weight loss above 400C. During this stage, the adjacent silanol
groups continue to condense and form siloxane link or isolated silanol group (Figure
2-25). After high temperature annealing, these isolated silanol groups are found to have
no interaction with water molecules[85]. As temperature above 800C, the viscous flow
occurs on the silica surface and isolated silanol groups can react with each other, bridging
the particles tighter.
- 66 -
Figure 2-23. Physical water decreases and silanol groups condense [83]
Figure 2-24. Re-absorption of physical water below 400C[83]
- 67 -
Figure 2-25. Irreversible elimination of adjacent silanol group [83]

Figure 2-26 shows the TGA measurements for silica particles with different
drying conditions. The micron size silica almost has no weight loss during heating, e.g.
no moisture absorption or organic residue on the surface. Nevertheless, the nanosilica
synthesized from sol-gel method shows obvious weight loss after heating. It can be seen
that the nanosilica without drying had a weight loss (about 2 wt%) before 200C, which
related to the elimination of
physically absorbed water on the surface[86]. It also
displayed a second weight loss (about 4 wt%) at 500 C, which is due to the chemically
bonded water and residual organics, typically methanol or ethanol, from the sol-gel
process of the nanosilica synthesis. The heating process could not remove the surface
residue and moisture effectively. Nanosilica after drying at 120C or 200C for 6 hours
still show similar weight loss behavior to those without drying, because the weight loss
during heating under 400C is reversible. With large surface area, nanosilica can easily
pick up the moisture again in a typical laboratory environment.
- 68 -
If an ultra high
temperature was used to dry the particles (1000C), the nanosilica show no physically
absorbed water any more (the blue line in Figure 2-26). Some weight loss continues to
occur above 400C, which is related with the condensation between adjacent silanol
groups. However, as mentioned above, the high temperature heating for SiO2 particle
could cause viscous flow and bridging between particles and increase the particle size.
Therefore, physically drying method is not favored in order to solve the moisture
absorption problems on the nanosilica surface.
The effects of chemical surface on the nanosilica were characterized by TGA, too.
It can be seen from Figure 2-27 that the untreated nanosilica had weight loss in two steps;
first one about 2wt% from physical absorbed water, second one about 4wt% from
chemically absorbed water. On the other hand, the nanosilica almost didnt pick up water
(<0.3 wt%) after silane surface treatment because the surface hydrophilicity had been
changed. The even lower weight loss of CA-EP treated nanosilica before 200C also
shows that the CA-EP is more effective than CA-NH at preventing the silica surface from
absorbing water. The treated nanosilica begins to lose weight continuously with
temperature increasing, which was contributed to the debonding and degradation of
grafted silane functional group on the surface.
- 69 -
Figure 2-26. Weight loss of silica with different drying condition
Figure 2-27. Weight loss of nanosilica with different surface modification
- 70 -
2.4.4. Viscosity of nanocomposite no-flow underfill
pure underfill
CA-NH treated
Viscosity (Pa.s)
1000
CA-EP treated
untreated
100
10
1
0.1
0.1
10
100
Shear rate (1/s)

Figure 2-28. Viscosity of nanocomposite underfills
The as-received untreated fillers and the silane-treated nanosilica fillers were
incorporated into EPON862/HMPA mixture to formulate a no-flow underfill. The filler
loading for composite underfill was in the 30wt%. The rheology of the composite
underfill with 30wt% filler loading was studied using a stress rheometer. It has been
found in Figure 2-28 that the treatment of the nanosilica with epoxy group terminated
silane significantly reduced the viscosity of the underfill. In addition, the viscosity of the
nanocomposite was less shear-rate dependent. Similar to the pure epoxy resin, the
underfill with CA-EP treated nanosilica showed a Newtonian behavior while the ones
with untreated or CA-NH treated silica showed shear-thinning behavior. The rheology of
the underfill with CA-EP treated nanosilica indicated that the filler-filler interaction was
significant reduced by the surface treatment. The reduced filler-filler interaction was
manifested by the mono-dispersion of the silica filler as shown in the particle size
analysis and TEM analysis. In addition, the high critical surface tension of the solid after
epoxide silane treatment showed better compatibility with the epoxy resin and helped the
- 71 -
wetting of the epoxy resin on the nanosilica surface. The results from the rheology study
are consistent with the previous results from the silica surface treatment and
characterizations.
- 72 -
CHAPTER 3.
MATERIAL PROPERTIES
CHARACTERIZATION OF THE NANOCOMPOSITE
UNDERFILL AFTER CURING
Nano-size silica is known to agglomerate easily due to its high surface area and
hydrophilic surface property. As such, the viscosity of the nanocomposite underfill tends
to be high, which makes it impossible to achieve high filler loading with good flow
ability of liquid underfill. Chemical treatment of nanoparticle surface is necessary to
improve the compatibility between the nano size filler and the resin, so as to decrease the
viscosity of the underfill. In Chapter 2, we discussed the surface modification process for
nanosilica and reduced the viscosity of uncured underfill.
In this chapter, the nanosilica were treated by the optimal modification method
and added into the underfill formulation. The optical, physical, thermal mechanical
properties and curing behaviors of these materials were studied in comparison to the
control samples with micron-size silica filler or with un-treated nanosilica. The objective
of this study was to elucidate the filler-filler and filler-polymer interactions in the
composites and to achieve an in-depth understanding of the effect of the filler size and
surface chemistry on the underfill material properties.
3.1. Experiments
3.1.1. Materials
The same nanosilica as in 2.3.1 was used as-received or treated with silane
additives. For comparison, conventional silica with a 3m average diameter was also
used as filler. The epoxy used was diglycidyl ether of Bisphenol-A type (EPON828, from
Shell Chemicals with an average molecular weight of 377). The hardener was hexahydro4-methylphthalic anhydride (HMPA, from Lindau Chemicals). 1-(2-isocyano-ethyl)-2-
- 73 -
undecyl-1H-imidazole from Shikoku Chemicals was used as a latent catalyst for epoxy
curing. -glycidoxypropyl-trimethoxysilane (CA-EP) was used as the silica modification
compound. All of these chemicals were used as received. Table 3-1 shows the chemical
structure of organic materials.
Table 3-1. Chemicals used in the underfill formulations
Chemical name
Poly(bisphenol A-coepichlorohydrin)
Designation
EPON828
Hexahydro-4methylphthalic
anhydride
HMPA
1-(2-isocyano-ethyl)2-undecyl-1Himidazole
-aminopropyltriethoxysilane
C11Z-CNS
Chemical structure
H2C
CH3
C
CH3
HC H2C O
O
O CH2
CH CH2
O
H3C
C
O
C
O
CN
CH2 CH2 N
HOOC
COOH
COOH
C11H23
CA-EP
O
O
Si
3.1.2. Underfill composite preparation
The base polymer formulation was prepared by mixing EPON828 and HMPA
with a weight ratio of 1:0.75. After stirring the polymer mixture for 10 minutes, the
catalyst, with 1wt% based on the polymer mixture, was added into polymer liquid and
stirred for another 30 minutes until a homogenous polymer solution was achieved.
Silica nanoparticles were modified and functionalized by an optimized treatment
procedure
in
which
-glycidoxypropyl-trimethoxysilane
was
used
as
the
functionalization reagent, as studied in 2.3.3. The transparent silica colloidal solution in

ethanol was obtained. The treated nanosilica was mixed with epoxy by the solvent
transfer method. Firstly, most of the solvent in the nanosilica colloidal solution was
removed by the rotary evaporator. Then, the epoxy monomer was added into the
- 74 -
remaining solution. After epoxy completely dissolved into the silica colloidal solution,
the residue solvent inside the mixture was completely dried by the vacuum oven for
overnight. By this procedure, we can avoid the agglomeration of treated silica during
drying process and always keep the good dispersion of silica in the liquid medium. At
last the anhydride hardener and imidazole catalyst were added and stirred to form a
homogenous composite underfill.
For the control samples with micron size filler or un-treated nano size filler, the
dry powder without modification was added into the liquid underfill formulation directly
and the mixture was sonicated for 30 minutes using a Sonicator (Misonix 3000) at a
power of 450 W.
All the samples were named according to their filler nature and loading. For
example, treated-30 means an underfill sample with 30wt% surface-treated nanosilica,
and 3micron-30 means that with 30wt% micron silica. The filler loading of the
composite was 5%, 10%, 20%, 30% and 40% in weight percent.
3.1.3. Underfill composite characterization
The viscosity of the underfills at room temperature was studied using a stress
rheometer (TA Instruments, AR1000N) in a steady flow mode. A cone-and-plate
geometry was used. A liquid underfill sample was dispensed on the plate before the run
and the experiments were conducted with a stepped shear rate from 0.01 to 100 s-1.
Light absorption measurements of underfills were made on a UV-Visible
spectrophotometer (Beckman Du520). To prepare the sample, the liquid composite
underfill was coated on the quartz glass slide with a thickness of 25 micron. Then the
specimen was put into the chamber of the UV-Vis spectrophotometer and scanned from
350nm to 750 nm which included the visible region. The absorption of nanosilica
colloidal solution was measured in the quartz cuvette.
- 75 -
The curing behavior and glass transition temperature of the epoxy composites
were characterized by a modulated Differential Scanning Calorimeter (DSC, TA
Instruments, Model 2920). A sample of approximately 10mg was sealed in a hermetic
aluminum pan. A dynamic scanning experiment was conducted with a ramp rate of
5C/min, from ambient temperature to 300C, to obtain the curing heat flow diagram of
the composite. The cured sample was left in the DSC cell and cooled to room
temperature. Then the sample was reheated to 200C at 5C/min to obtain another heat
flow diagram. The initial temperature of the heat flow step of the second diagram is
defined as the DSC glass transition temperature (DSC Tg).
In order to evaluate bulk properties of the composite samples after curing, the
liquid filler-polymer mixture was poured into an aluminum dish and cured in the
convection oven at 150C for one hour and 180C for another two hours.
A Dynamic Mechanical Analyzer (DMA, TA Instruments, Model 2980) was used
to measure the dynamic moduli and glass transition temperature of the composites. The
cured sample was cut into a strip with dimensions of about 1862 mm. The test was
performed in a single cantilever mode. The temperature was increased from room
temperature to 250C at a heating rate of 3C/min, while the storage modulus (E), loss
modulus (E) and tan were calculated by the pre-installed software. The DMA Tg was
determined by the corresponding peak of the loss modulus (E) curve.
The coefficient of thermal expansion (CTE) of the cured sample was measured on
a Thermo-Mechanical Analyzer (TMA, TA Instruments, Model 2940). The dimensions
of the sample were about 552 mm. The sample was heated in the TMA furnace at
5C/min from room temperature to 200C. The CTE before the Tg is defined as 1 and
after the Tg as 2.
The density was measured by using the Archimedes approach[81, 87], where the
cured sample was weighed in air and then in water using a balance with a reproducibility
- 76 -
of better than 0.5 micrograms. The average of measurements of at least 3 specimens was
reported for each sample.
The cured underfill was subjected to temperature /humidity aging at 85C and
85% RH. The sample was taken out of the aging chamber and the increased weight due to
moisture uptake was recorded daily.
The dispersion of the nanosilica in the cured composite materials was observed
using a Scanning Electron Microscopy (SEM Hitachi S800). A preliminary wetting test
was performed using a eutectic SnPb solder bumped quartz chip and a Cu substrate. The
bumps were area-array distributed; the diameter of the bumps was 75 um. The no-flow
underfill was first dispensed on the Cu board, followed by the placement of the quartz
chips on the underfill. Then the assembly was subjected to the standard eutectic SnPb
solder reflow process. The wetting of the solder on the copper board was observed using
an optical microscope.
3.2. Results and Discussions
3.2.1. Anhydride epoxy polymerization mechanism
The imidazole compound is a commonly used catalyst for epoxy polymerization.

In our reaction, a 1-(N)-substituted imidazole was used as the Lewis base catalyst. After
the reaction starts, the lone-pair electron at the 3-N position of catalyst reacts with an
anhydride molecule to open the anhydride ring[24]. An internal salt (betaine) is formed at
the anhydride and will act as the initiator of cure. The anhydride with the active center
then reacts with the epoxide to open the epoxy ring and to transfer the reactive center to
epoxy. The continuation of the curing reaction relies on the transfer of the reaction center
between the anhydride and the epoxy. Once the reaction is initiated, the molecules grow
quickly until a large network is formed. Therefore, the reaction mechanism of
epoxy/anhydride curing is similar to that of the chain growth polymerization.
- 77 -
O
4
C
O
R1
2
R
R2
O
R2
O
N
H2C O
N
O CH2
R1
+
C
H2C
CH
CH
CH2
O
O
O
R2
R2
O
C
H2C
C
O
CH
CH2 O C
H2C
CH
CH
CH
R1
R1
C O
C
O
O-
O
-
CH2 O C
O
C
C
O
Figure 3-1. Reaction scheme of anhydride/epoxy polymerization with imidazole catalyst
- 78 -
3.2.2. Curing Behaviors and Tg of composite underfills
The curing behaviors of composite underfills were characterized by DSC dynamic

heating experiments. The curing profiles of an underfill with 30 wt% silica, including
micron silica, untreated nanosilica and treated nanosilica, are shown in Figure 3-2. For
comparison, the underfill without silica is also included in the figure. As can be seen
from Figure 2, the micron size fillers did not affect the curing process of the epoxy resin.
However, the presence of the nanosilica could inhibit the curing reaction, especially at
the late stage of curing. Both treated-30 and untreated-30 samples had a delayed tail in
the curing curves. For the development of nanocomposite underfills, the cure inhibition
effect caused by nanosilica could bring a negative effect to the underfill process since it
needs longer post-cure time.
Figure 3-3 shows the DSC Tg of silica composite underfills. It can be observed
from the graph that the addition of nanosilica can decrease the Tg of the composite
appreciably while the micron-size fillers does not have a significant effect on the Tg of
the composites. The Tg of untreated-40 sample was almost 40 C lower than that of the
pure epoxy. On the other hand, silane treatment on the fillers could increase the Tg of the
nanocomposites by several degrees at high loading levels. The Tg depression in
nanocomposites might be due to two reasons. First, nano-size filler can inhibit the curing
reaction as mentioned before, resulting in a lower crosslinking density in the epoxy
matrix and therefore a lower Tg. Second, nanosilica has a much larger surface area and
interfacial interaction with epoxy matrix compared to the micron silica. The large fillerresin interface creates extra free volume leading to a lower Tg. The silane treatment on
the silica surface improves the compatibility between the fillers and the resin, which
decreases the free volume at the interface and therefore slightly increases Tg.
- 79 -
Figure 3-2. Curing behaviors of base underfills and composite by DSC
3micron
untreated
treated
140
130
Tg (C)
120
110
100
90
80
0
10
20
Filler loading (wt%)
30
40
Figure 3-3. Glass transition temperatures of composite underfills by DSC

- 80 -
3.2.3. Rheological and optical behavior of composite underfills
Figure 3-4 shows the viscosity of the composite underfills at room temperature as
a function of filler loading with 3micron, nano-treated and nano-untreated silica fillers.
As can be seen from the figure, the addition of nanosilica filler can significantly increase
the underfill viscosity, especially at high filler loading. The viscosity of the underfill with
50wt% nano-size filler exceeded 500 Pas, more than 100 times higher than the one with
50wt% micron-size filler. This brings a great challenge to process and apply the
nanosilica filled composite underfills in no-flow underfill assembly. Results also showed
that the surface treatment on the nanosilica surface would enhance the compatibility
between the filler and the epoxy matrix and lower the viscosity of the composite material.
The underfill with surface treated nanosilica had a significantly lower viscosity at high
filler loading than the one without surface treatment. The silica surface modification is an
effective way to reduce the viscosity of the nanosilica composite underfill and to improve
their processing compatibility.
550
3micron
untreated
treated
Viscosity (Pa.s)
500
20
10
0
0
10
20
30
40
50
Figure 3-4. Viscosity of silica filled composite underfills
- 81 -
The light absorption property of composite underfills with different filler size is
shown in Figure 3-5. The transmittance of light in the visible region decreased greatly
with the 3micron silica fillers while composite underfills with nanosize silica were
almost as transparent as pure epoxy in visible region (400~700 nm). Since no-flow
underfill is pre-applied on the substrate before chip placement, an optically transparent
liquid is desired for vision recognition during the chip placement process. The underfill
filled with micron-size silica becomes opaque at high filler loading due to light scattering.
This introduces difficulty in the chip placement since the underfill covers the bonding
pads on the substrate. Nanosilica filled underfill, on the other hand, is transparent to
visible light because the filler has a particle size smaller than the wavelength of the
visible light. The nanosilica composite underfills provide unique optical properties for
chip assembly application.
blank
nano -30
nano -40
3um-30
3um-40
100
90
Transmittance (%)
80
70
60
50
40
30
20
10
0
350
450
550
650
Wavelength (nm)
750
Figure 3-5. Effect of filler size on the UV-Vis spectra of the composite underfills
- 82 -
3.2.4. Thermal mechanical properties
3micron
untreated
treated
CTE(ppm/C)
80
70
60
50
40
0
10
20
30
40
Figure 3-6. CTE of silica filled composite underfills

The primary purposes of loading silica filler into no-flow underfill are to reduce
the coefficient of thermal expansion (CTE) and to increase the elastic modulus. These
two thermal mechanical properties are critical parameters to the thermomechanical
reliability of a flip-chip package[1]. Figure 3-6 shows the CTE value of composite
underfills as a function of filler loading. For all three kinds of filler, the CTE decreased
almost linearly as the filler loading increased. There was no obvious difference in CTE
value for underfill with untreated and silane treated nanosilica. It was found that the CTE
of the composite not only depended in the silica loading, but also the silica size[88].
Some literature showed that the CTE of composite underfills could reduce from 40
ppm/C to 26 ppm/C as the filler size decreased from 30 micron to 7 micron at 70wt%
filler loading. This experiment also indicated that the CTE of underfill with nanosilica is
smaller than that with 3micron filler at the same loading level. This is because the
interface of the filler particles and the resin matrix constricts the expansion of the epoxy
- 83 -
matrix. Therefore, an increase in the constriction of the matrix due to increased surface
area of the nanosilica allows a decrease in the expansion of the matrix.
Figure 3-7 shows the dynamic moduli of underfill with untreated nanosilica
measured by DMA. The Tg of the underfill measured by peak temperature of loss
modulus showed the same trend as observed in the DSC experiment; it decreased as filler
loading increased. However, at low filler loading (5, 10, and 20 wt%), the loss modulus
showed a secondary relaxation process below the glass transition, possibly related to the
surface relaxation at the resin-filler interface. As the filler loading increased, this
secondary relaxation overlapped with the glass transition. In other words, with the
increase of the interface between the filler and the resin, the surface relaxation became
dominant and reduced the Tg of the composite. Figure 8 shows the comparison of
dynamic moduli for underfills with the treated and untreated 20wt% nanosilica. Unlike
the untreated-20 nanosilica composite, the secondary relaxation below the glass transition
was not observed in the loss modulus of the treated-20. It is possible that the silane
treatment alters the interfacial properties and improves the compatibility of the resin-filler
interface.
- 84 -
Figure 3-7. Dynamic moduli of composite underfills with untreated nanosilica
Figure 3-8. Comparison of dynamic moduli of composite underfills with different

nanosilica
- 85 -
3.2.5. Moisture absorption and density measurement
3 micron
nano-untreated
nano-treated
Moisture uptake (wt%)
5
4
3
2
1
0
0
10
20
30
40
Filler loading (w t%)
(a)
3 micron
nano-untreated
nano-treated
6
5
4
3
2
1
0
0
10
15
20
25
(b)
- 86 -
30
35
40
3 micron
nano-untreated
nano-treated
7
6
5
4
3
2
1
0
0
10
15
20
25
30
35
40
35
40
(c)
3 micron
nano-untreated
nano-treated
7
6
5
4
3
2
1
0
0
10
15
20
25
30
(d)
Figure 3-9. Moisture uptake evaluations for underfill with different silica: (a) 24h (b) 48h
(c) 72h (d) 96h
Silica composite underfills were also characterized in term of moisture
absorption. The cured underfills were subjected to temperature/humidity aging at 85C
- 87 -
and 85% RH. With the assumption that all the weight increasing during the 85C/85%
RH aging was due to the water absorption in the polymer, the moisture uptake was
normalized to polymer weight percentage. Figure 3-9 shows the evaluation of moisture
uptake in underfills with aging time. Micron-size filler did not alter the moisture
absorption behavior of the polymer matrix. However, the nano-size filler increased the
moisture absorption due to the additional free volume at the interface. As filler loading
increased, the effect became more prominent.
Generally, moisture absorption processes in polymer composites can be described
by Ficks second law of diffusion, which can be expressed as[89]:
Mt
8
= 1 2
M
D(2n + 1) 2 2 t
1
exp
2
h2
n = 0 ( 2n + 1)
Equation 3-1
Where M t and M are the moisture content at time t and the equilibrium or maximum
moisture content, respectively. D is the diffusion coefficient and h is the sample
thickness.
At short times (i.e., the initial absorption), Equation 3-1 can be reduced to:
Mt
Dt
= 4( 2 ) 0.5
M
h
Equation 3-2
Equation 3-2 can be rewritten as:

M t = kt 0.5
Equation 3-3
Where k is a rate constant relating to the diffusion coefficient.

To elucidate this phenomenon clearly, the moisture absorption of underfills with
30wt% filler as a function of time before moisture saturation is shown in Figure 3-10.
The moisture absorption kinetics were also calculated according to the Equation 3-3. It
can be seen that surface treatment of nanosilica slowed the rate of moisture uptake in the
- 88 -
composites. The samples with both treated and untreated nanosilica had much higher
moisture uptake and faster absorption rate than the pure polymer and micron size
composite. The parameters of Equation 3-3 were listed in Table 3-2 for four samples,
which also indicated the different moisture diffusion kinetics for the samples.
without filler
3 micron
nano untreated
nano treated
Moisture uptake (%)
2.5
2
1.5
1
0.5
0
0
4
Time (hours)
Figure 3-10. Kinetics of moisture uptake for the samples
Table 3-2. Moisture absorption kinetics parameter

Materials
k (rate constant)
without
filler
0.1727
micron
filler
0.2319
untreated
nanosilica
0.7694
treated
nanosilica
0.6963
The experiments on moisture absorption showed that the incorporation of nanosize filler into the polymer matrix could change the free volume of polymer. In order to
verify the difference of free volume at the glassy state, a density measurement was
conducted for composite underfills. The influence of the silica content and size on the
- 89 -
density of the composite underfills is shown in Figure 3-11. It can be seen that the density
of the composites with 3micron silica was higher than that with nanosilica. This
difference was much more evident at the high filler loading. These results are consistent
with the moisture absorption and glass transition temperature experiments, i.e., the nanosize silica can change the free volume of polymer in composites, so the nanocomposites
show a lower density, higher moisture absorption and lower Tg than micron-size filled
composites. The density of the nano-treated sample was slightly higher than that of the
nano-untreated. This means that silane modification to nanosilica can improve the
compatibility of the silica and polymer phases and thus result in a more compact
composite structure and less free volume in the polymer.
1.45
3 mic ron
nano- untreated
nano- treated
1.4
Density (g/cm3)
1.35
1.3
1.25
1.2
1.15
0
10
20
30
40
Fille r lo ad in g (w t%)
Figure 3-11. Density measurement for silica filled composite underfills
- 90 -
3.2.6. Morphology
(a)
(b)
Figure 3-12. SEM photographs of nanosilica composite materials (a) untreated-30, (b)
treated-30
To investigate the dispersion of nanosilica in the epoxy matrix, the cured
composite materials were polished carefully to get a very reflective surface. The thin
- 91 -
sections of polished surface were cut and mounted on an aluminum stub using a
conductive tape and were sputter-coated with gold before the cross-section examination.
Figure 3-12 shows the SEM photographs of cured underfill with 30wt% treated and
untreated nanosilica, respectively. The nanosilica without silane treatment formed large
agglomerations and obviously phase separated from the polymer matrix, as shown in
Figure 3-12 (a), while after silane treatment, the nanosilica kept even mono-dispersion in
the polymer matrix without large aggregation, as shown in Figure 3-12 (b). This serves as
direct evidence that the silane treatment can change the silica surface properties and
improve the compatibility between silica and epoxy, so nano-treated samples have tighter
integration of the inorganic-organic phase and less nanosilica-epoxy interfacial
interaction than nano-untreated ones. The disadvantages with small size and large surface
area of nanosilica, which greatly influence the composite material properties, can be
overcome by silane treatment of the nanosilica surface, as indicated by previous
experimental results.
3.2.7. Wetting test
Some preliminary solder wetting experiments were conducted using eutectic

Sn/Pb bumped quartz chips and Cu substrate. The bumps on the chip were area-array
distributed; the height and diameter of the bumps was 75 m and 75 m, respectively.
The no-flow underfills were dispensed onto the Cu substrate and the quartz chips were
placed onto the substrate. A fluxing agent was added to the underfill to eliminate the
surface oxide on the solder. Standard eutectic Sn/Pb solder reflow profile was used with
peak temperature around 210C. Silica fillers in the size of 3 m were used at different
filler loading in the wetting test. After reflow, the testing coupon was cut and the cross
section was examined.
For the no-flow underfill process, the underfill was firstly applied on the
substrate. Therefore, the solders on the chip must penetrate the underfill layer in order to
- 92 -
melt on the substrate contact and form the interconnects. The wetting test improved that
the presence of no-flow underfill did not hinder the solder wetting if no filler or low
loading of micron size filler in the underfill, as shown in Figure 3-13 (a)-(b). With filler
loading higher than 20wt% or above, the large size silica could cause problem during
solder reflow (Figure 3-13 (c)), and the solder joint could not be formed.
(a) No filler
(b) 15 wt% micron filler
(c) 20 wt% micron filler
(d) 30 wt% nano filler
Figure 3-13. Cross-section views of a quartz chip with no-flow underfill
Figure 3-14. Wetting picture of quartz chip with treated-30 underfill

In comparison, a 30 wt% nanosilica (80~100 nm) filled no-flow underfill was also
investigated. The silica was treated by silane in order to reduce the viscosity. The
- 93 -
nanosilica did not prevent the solder wetting at a filler loading of 30 wt % (Figure
3-13(d)). Figure 3-14 shows the top view of the test coupon. The solder balls on the
quartz chip spread well on the copper board below. The incorporation of treated nanosize silica did not hinder the formation of the solder joint between the quartz chip and the
copper board. This indicates that surface modified nanosilica composite can be used as
no-flow underfill.
3.3. Glass Transition and Relaxation Behaviors of Nanocomposites
In the last section, the thermal properties of composite underfills were

investigated. It was found the unexpected glass temperature depression in the underfill
with nano size filler, which is up to 35C lower than underfill with micron size filler.
The Tg depression in nanocomposite underfill may bring problems for its application in
real component assembly because low Tg of underfill can reduce the thermal reliability.
SO it is interesting to investigate the reason for Tg depression and the solution to this
problem.
It is well known that the composite properties can change with the filler
dispersion state, geometric shape, surface property, and the particle size and distributions.
Nanocomposites show different properties compared to the bulk polymers and their
counterparts with micron size fillers due to the small size of the filler and corresponding
increase in surface area[90, 91]. The effect of the nano-fillers in polymer composites on
the glass transition behavior of the polymer matrix has been studied in different
filler/resin composites. In some cases, increases in the glass transition temperature (Tg)
were reported[92-94]. In other cases, decreases in the Tg were reported [95]. An initial
increase of Tg followed by a Tg decrease with higher filler loading was observed in
poly(styrene butylacrylate) latex/nano-ZnO composites[96]. In many cases, the
dispersion and the surface condition of the nanoparticles play an important role in the
- 94 -
change of Tg [95, 96]. It is said that a short-range highly immobilized layer about 1 nm in
length develops near the surface of the fillers. In this interaction region of the polymer
layer surrounding the particles, the conformational entropy and chain kinetics are
significantly altered[95]. As the filler size enters the nano-region, the volume fraction of
the interaction region in the nanocomposites increases with the increasing interface area
of the polymer and the nano-fillers. This becomes the basis for potentially tremendous
changes in nanocomposite properties.
This section will systematically investigate the Tg change of epoxy
nanocomposites with different nano size fillers and to interpret the polymer relaxation
behavior in term of the polymer/nanoparticle interface properties.
3.3.1. Experiments
Materials
The blank epoxy formulation was prepared by the same method as in 3.1.2. Both
nanosilica (~100nm) and micron silica (~3m) were used as filler. The nanosilica used in
this experiment was not treated by the silane. The filler loading varied from 5 wt% to 40
wt%. The micron size silica was mixed into the resin using a high-speed blender for 5
min. The nanosilica was mixed into the resin through sonication for 30 min using a
sonicator (Sonicator 3000 by Misonix) at a power of 450 W. The sample with 20wt% was
used for further investigation of the mechanical and dielectric properties. The blank
epoxy formulation was used as the control sample.
Silver composite
Silver (Ag) nanoparticles (average diameter 65 nm, from NanoPowders
Industries) and the micron size flakes (2 types of Ag flakes with average size of 1 m and
2 m, from Ferro Corporation) were treated with surfactants by the manufacturer for deagglomeration and used as received. The same blank epoxy formulation as the silver
- 95 -
composites was used for the aluminum composites. The filler loadings of the silver
composites were 68, 72, 75 wt%. The micron size silver was hand-mixed into the resin
for 10 min. The bimodal formulations used the combination of the two micron size silver
flakes at a ratio of 1:1. An addition formulation with mono-dispersed silver flakes (2 m)
was prepared at a filler loading of 72 wt%. The silver nanoparticles were hand-mixed for
30 min.
Aluminum composite
Nano aluminum (Al) particles (average size 100 nm, from AlfaAesor) and micron
size aluminum particles (average size 3 m, from AlfaAesor) were used as received
without any further surface treatment. The same blank epoxy formulation as the silica
composites was used for the aluminum composites. The filler loadings were 41 and 45
wt%. The micron size aluminum was hand-mixed into the resin for 10 min and the nano
size aluminum was hand-mixed for 30 min.
Carbon black composite
Carbon black (primary particle size around 30 nm, surface area around 950 m2/g,
from Degussa) was used as received. Same blank epoxy formulation as the silica
composites was used for the carbon black composites. The filler loadings were 2 and 5
wt%. The carbon black was mixed into the resin using sonication for 5, 30, and 60 min,
respectively.
3.3.2. Characterization
Curing and glass transition

The curing behavior and glass transition temperatures of the epoxy composites
were characterized by a modulated differential scanning calorimeter (DSC Q1000, TA
Instruments). A sample of approximately 10 mg was sealed in a hermetic aluminum pan.
- 96 -
Dynamic scanning experiments were conducted with a ramp rate of 5C/min, from
ambient temperature to 250C to obtain the curing heat flow diagram of the composite.
The cured sample was left in the DSC cell and cooled to room temperature. Then the
sample was reheated to 200C at 5C/min with a temperature modulation of 1 C/min to
obtain the second heat flow diagram. From the step change of the reversible heat flow of
the second diagram, the glass transition temperature (Tg) was determined.
Dynamic mechanical properties
The dynamic mechanical properties of the epoxy-silica composites were
evaluated after the samples were cured in a convection oven at 150C for 1 hour. The
cured samples were cut into a strip by 18 6 3 mm. A dynamic mechanical analyzer
(DMA 2980, TA Instruments) was used to measure the dynamic moduli of the samples.
The test was performed in a single cantilever mode from -120C to 200C at a heating
rate of 3C/min. Liquid nitrogen was used for the low temperature testing. The amplitude
applied to the samples in the experiments was 15 m, resulting in a strain level of around
0.5%, which is within the linear viscoelastic region of the epoxy. The storage modulus
(E) and loss modulus (E) of the samples were calculated by the pre-installed software.
The glass transition temperature also can be obtained from the peak position of the loss
modulus.
3.3.3. Results and discussion
Glass transition behavior of different nanocomposites

The effects of the filler size and the filler loading on the glass transition
temperature of the polymer composites were studied with four different types of fillers:
silica, silver, aluminum, and carbon black. The Tgs of the silica composites with nanoand micron size fillers are shown in Figure 3-15. The micron size filler did not have a
significant effect on the Tg of the composites while the nano size filler had an appreciable
- 97 -
impact. With an increase in filler loading, the nano-silica composites first showed a slight
increase in Tg and then the Tg decreased significantly with higher filler loadings.
Compared to the control sample, the 40 wt% nano-silica composite showed a Tg
depression by almost 30 C.
Similar behavior was observed in the silver and aluminum composites. As can be
seen in Figure 3-16, the Tgs of the nano-silver composites were about 20 C lower than
those of the micron size silver composites, decreasing with the increase of the filler
loading as well. Figure 3-17 shows the Tgs of the aluminum composites. The difference
in Tg caused by the filler size was not as great as the other two composites. However, Tg
depression in the nanocomposites was quite obvious.
Nano size
Micron size
140
130
Tg (C)
120
110
100
90
80
0
10
20
30
Weight Percentage (wt%)
40
50
Figure 3-15. Glass transition temperature of the silica composites
- 98 -
Nano size
Micron size bimodel
Micron size mono-dispersed
Control
130
125
120
Tg (C)
115
110
105
100
95
90
85
80
67
68
69
70
71
72
73
74
75
76
Figure 3-16. Glass transition temperature of the silver composites

Nano size
Micron size
Control
130
Tg (C)
125
120
115
110
105
40
41
42
43
44
45
46
Figure 3-17. Glass transition temperature of the aluminum composites
- 99 -
As mentioned earlier, the change of Tg in the nano filler composites has been
controversial. Some have found that the Tg of nanocomposites increases as a function of
the filler loading while others observed the opposite. The Tg of a polymer system varies
for a variety of reasons, including changes in tacticity, molecular weight, crosslinking
density and the amount of reaction residue acting as plasticizers. In this study, however,
the ingredients and curing conditions of the nanocomposites and the counter-parts with
micron size fillers were the same for all the samples. Therefore, it is expected that the
observed change in the Tg is mainly due to the properties of the fillers including the filler
size. As the filler size decreases, the interfacial area between the fillers and the polymer
matrix increases dramatically. It is possible that the increasing interfacial area can
influence the polymer chain mobility and therefore change the glass transition
temperature of the composites.
In order to elucidate the relationship between the interfacial area and the Tg of the
composites, carbon black filled epoxy resins were prepared at low filler loadings. Carbon
black is known to have very small primary particle size (1~50 nm) and easily
agglomerates to form large secondary clusters. These large clusters poorly dispersed in a
polymer matrix effectively increase the filler size and decrease the resin/filler interface.
In this study, the carbon black was added into the epoxy resin and sonicated for different
durations in order to examine the effect of the filler dispersion. As can be seen in Figure
3-18, the sample with the shortest duration of sonication had a Tg comparable or even
higher than the control sample. Further sonication helps the secondary clusters break into
small particles and increases the resin/filler interface. Accordingly, it is thought that the
Tg of the composites decreased due to the increased interfacial area.
- 100 -
1 hr sonication
30 min sonication
5 min sonication
Control
132
130
128
Tg (C)
126
124
122
120
118
116
1
3
4
Figure 3-18. Glass transition temperature of the carbon black composites

Moisture absorption of nanosilica
The further investigation was carried out on the material properties of the silica
composites. As shown in Figure 3-15, significantly Tg depression in the nano-silica
composites was observed, especially at high filler loadings. In order to study the
difference between the micron- and nanosilica fillers, thermo-gravimetric analysis (TGA)
was performed on these two types of silica. The fillers were heated under air from room
temperature to 800 C at a ramping rate of 20 C/min. Figure 3-19 shows the TGA
diagram of these two fillers. The micron size filler was stable over the temperature range
while the nano size filler underwent a weight loss during heating. The initial weight loss
of the nano filler (about 2 wt%) is likely due to the adsorbed moisture. The same fillers
were dried at 200 C for 6 hours and the TGA diagram of the dried nano fillers appeared
to be the same as the un-dried ones. The nano size fillers possess a large surface area
which makes it easy to pick up the moisture in a typical laboratory environment even
after they are dried with the above condition. The nano fillers also displayed a second
- 101 -
weight loss (about 4 wt%) at 500 C, which might be due to the chemically bonded water
and residual organics, typically methanol or ethanol, from the sol-gel process of the
nanosilica synthesis. However, these residual organics appeared to be bonded at the filler
surface and de-bonding occurred at a high temperature. Therefore, the residual organics
and bonded water would tend to remain at the filler surface even after the polymer
composite is cured at a temperature lower than 200 C. The adsorbed moisture and the
residual organics can play an important role in reducing the glass transition temperature
by assisting molecular motion and creating more free volume at the resin/filler interface.
Figure 3-19. TGA measured weight loss at a heating rate of 20 C/min under air
Dynamic mechanical of nano-silica composites
In order to investigate the relaxation process of the silica composites, dynamic
mechanical analysis was performed. The DMA offers the advantage of studying the subTg transition in the polymeric materials. Figure 3-20 shows the dynamic loss moduli of
the three samples. As can be seen in the figure, there is a significant difference in the
peak of the loss modulus at round 150 C that signifies the large-scale cooperative
- 102 -
motion of the polymeric network, typically referred to as the glass transition, or called
relaxation. This process has the highest activation energy and is generated by the
relaxation of the whole network.
For the nanocomposite, there is a secondary relaxation in the loss modulus at the
sub-Tg region (80-120 C), which is not found in the micron composite or the pure
polymer. These peaks may be related with the molecular chains movements in the
interfacial region, or called the interphase. In the inorganic-organic composite, the
presence of filler can influence the structure of the polymer matrix near the fill surface
and polymer chain mobility due to the interfacial interaction between the filler surface
and the polymer matrix[97]. These changes in local morphology may lead to the
formation of an interphase around the filler particle with properties different from that of
the bulk polymer matrix, i.e., the composite can be regarded as a three-phase material.
The increased resin/filler interface created extra free volume and therefore assisted the
large scale segmental motion of the polymer. The formation of such an interphase is
expected to influence the viscoelastic properties of polymer matrix, too. On the other
hand, the nano size filler has large surface area with hydrophilic characteristic, and the
moisture absorption is hard to remove, as shown in TGA before. The motions bonded
water on the filler surface also can contribute to the relaxation process [98]. After the
deconvolution process by the software Peakfit (version 4.11), we can separate a series
relaxation peak during this region (80-120C), as seen in Figure 3-21. Nano size filler
has much higher surface area than the micron size filler at the same filler loading;
therefore, the interfacial interaction in the nanocomposite makes the interphase more
prominent than that in the micron-composite. So the micron-composite, as well as the
pure polymer sample, did not have such a sub-Tg relaxation. From Figure 3-20, it also
can be seen that the relaxation process of polymer in the interphase region requires less
activation energy, as it happens at lower temperature of glass transition in the
nanocomposite.
- 103 -
550
nanocom posite sam ple

m icron com posite sam ple
control pure polym er
Loss modulus (MPa)
500
450
400
350
300
250
200
150
100
50
0
-100
-50
50
100
150
200
T em perature (degree C )
Figure 3-20. Dynamic loss moduli of the silica composites and the blank resin
Loss modulus (MPa)
350
300
250
200
150
100
50
0
60
80
100
120
140
160
Temperature (degree C)
Figure 3-21. Deconvolution of loss modulus of nanocomposite
- 104 -
Besides the transition in all the the samples and the sub-Tg transition in the
nanocomposite, there is another broad peak of the loss modulus in all the three samples at
around -50 C, which is usually referred to as the transition or relaxation. The origin of
the transition in polymer materials may come from the relaxation of diphenypropane
units or the dicarboxylic ester units in the cross-linked network of epoxy[97]. The
popular crankshaft mechanism involves restricted motion of the main chain requiring at
least four consecutive CH2- bonds[99], and is not likely to be the case in an
epoxy/anhydride network due to the high crosslinking density.
O
H2C
CH
CH3
C
CH3
H2C O
O CH2
CH
CH2 O C
C O
Figure 3-22. Molecular structure of anhydride/epoxy polymer
CH3
C
CH3
Figure 3-23. Possible local motion of segments in anhydride/epoxy polymer

The DMA results prove the addition of nanosized filler into the epoxy matrix can
create a significant interphase, which generates a sub-Tg transition besides the bulky
motion of epoxy network ( transition) or the segmental motion of molecule chain (
- 105 -
transition). The special morphology and structure in the interphase region of

nanocomposite also create extra free volume. As a result, the Tg of the nanocomposites
decreases with increased filler loading. However, the sub-Tg transition involved local
movement of the chain and requires much less free volume and activation energy.
Therefore the increased interface does not have a significant effect on the transition
temperature.
From the dynamic mechanical study on the silica composites, it was found that
the Tg depression of the nanocomposite is closely related to the resin/filler interphase.
Molecular dynamic simulations of the polymer melts in presence of nanoparticles have
shown that the dynamics of the polymer melts can be influenced by the
polymer/nanoparticle
interactions[100].
Strongly
attractive
polymer-nanoparticle
interactions were found to slow the dynamics relative to a pure polymer melt while nonattractive interaction enhanced the dynamics. Macroscopically, the change in the
dynamics results in a shift in the glass transition temperature. In this study, the various
fillers do not possess an attractive interaction with the epoxy resin. The dynamics of the
polymer segments are enhanced, or, in other words, there exists more free volume at the
filler/resin interface. In the case of the composites with micron size fillers, the enhanced
polymer dynamics at the interface does not noticeably reduce the Tg of the polymer
matrix due to the limiting interfacial area. However, the large filler/resin interfacial area
in the nanocomposites brings significant changes to the polymer mobility and therefore
affects the Tg of the polymer matrix appreciably. In addition, a hydrophilic surface of
some nano fillers such as silica is likely to adsorb water. The adsorbed water, together
with other bonded surface organics, that are not compatible with the polymer matrix
further enhances the polymer dynamics and reduces the Tg. On the other hand, since an
attractive polymer-filler interaction suppresses polymer dynamics, an increase in Tg can
be expected in the nanocomposites with good filler/polymer compatibility. Surface
modification is a common method to improve the filler/resin compatibility in polymer
- 106 -
composites. By replacing the hydrophilic surface groups with hydrophobic groups,

adsorbed moisture can also be reduced. The effects of surface modification of the nanosilica on the properties of the nanocomposites are current under investigation.
- 107 -
CHAPTER 4.
INFLUENCE OF INTERPHASE AND MOISTURE
ON THE DIELECTRIC SPECTROSCOPY OF EPOXY/SILICA
COMPOSITES
The underfill materials are used in the electronic packaging, where they surround
and protect the metal solder alloys. The electrical characteristics of microelectronic
devices, such as signal attenuation, propagation velocity and cross talk, are influenced by
the dielectric properties of the packaging materials. An important role of packaging
materials is to ensure the electrical insulation of the silicon chip and of circuit pins.
Ideally, a low conductivity is needed to avoid current leakage, a low dielectric constant
() to minimize the capacitive coupling effects, and a low loss factor () to reduce
electrical loss. So it is necessary to understand the dielectric behaviors of
nanocomposites.
4.1. Dielectric properties of composite materials
4.1.1. Theory and background
The dielectric properties of materials can be measured by applying a sinusoidal

voltage to samples. This electric field produces polarization within the sample, causing
oscillation that is at the same frequency as the field but with a phase angle shift (). The
phase angle shift is measured by comparing the applied voltage to the measured current.
The measured response is separated into a complex form (*) that can be separated into
capacitive and conductive components giving the real part of permittivity () and loss
factor ():
* = ' j "
Equation 4-1
For the solid epoxy materials, there are two bulk effects that influence the
dielectric property: ionic conductivity and molecular dipole orientation. So these effects
lead to:
- 108 -
'
'
' = dipole
+ ion
Equation 4-2
"
"
" = dipole
+ ion
Equation 4-3
The dipole orientation in epoxy comes from the motion of polar groups under
electric field in epoxy structure. From the classic Debye theory, the following equations
are obtained:
'
dipole
= +
"
dipole
=
s
1 + w 2 2
( s ) wt
1 + w 2 2
Equation 4-4
Equation 4-5
Where is the dielectric permittivity at infinite frequency, s is the static dielectric

permittivity at zero frequency, w is the angular frequency, and is the relaxation time.
Due to the cross-linking nature of the cured epoxy, the molecular motions are
hindered by the cross-linking network. When the dielectric measurement was taken under
a low temperature, for example, below the glass transition temperature, the molecules
lose their long range segmental mobility. The dipolar relaxation due to the segmental
movement and orientation with the electrical field cannot contribute to the dielectric
response. Under this circumstance, the relaxation time ( ) to achieve a polar orientation
can become very large. So for a glassy state epoxy, the dipolar relaxation due to long
range molecular mobility doesnt dominate the material dielectric behavior. However,
there is local relaxation processes associated with the dipolar orientation of the side
groups pending to the main polymer chain.
Another effect, ionic conductivity, comes from the motion of ions under the
electric field and is independent of applied frequency. During the synthesis of epoxy and
its composite, some chloride ions and other impurities are inevitably exist in the solid
materials and will contribute to the ionic source. Ionic conductivity is also related to the
- 109 -
polymer mobility and increases dramatically with increasing temperature. Therefore, it

usually dominates the dielectric properties at low frequency or high temperature.
Neglecting the dipole term, the loss factor can be derived as follows[101]:
"=
w 0
Equation 4-6
Where is the bulk ionic conductivity, 0 is the absolute permittivity of free space
(8.8510-12 F/m).
As the results from ionic conductivity, the interfacial effects between electrode
and epoxy polymer due to the dielectric measurement become obvious with the
increasing of conductance of the epoxy. The so-called electrode polarization can happen
by the accumulation of ion layers on the electrodes.
The effect of ionic conductivity on the dielectric behavior considering electrode
polarization can be illustrated as follows[102]:
'
ion
= C 0 Z 0 sin(
"
ion
=
n ( n +1) 2
)w
( )
2
0
n
C 0 Z 0 cos( ) w ( n +1) ( ) 2
w 0
2
0
Equation 4-7
Equation 4-8
Where is the ionic conductivity, 0 is the permittivity of the free space, C 0 is the
capacity of the cell in air, Z 0 is the constant depending on the interface of electrode and
epoxy polymer, and n is the real number alternating between 0 and 1.
For the epoxy/silica composite, on the other hand, the heterogeneity introduces an
extra degree of complexity. The dielectric property of the composites has been
investigated by various methods accounting for the interfacial relaxation. The MaxwellWagner mixing rule has been widely employed for the calculation of the dielectric
constant of a polymer/filler composite[103]. From M-W rule, the dielectric constant of a
- 110 -
mixture of plates aligned parallel to the electrodes can be considered as two capacitors in
series with the frequency dependent dielectric properties taking the similar form of the
Debye equations:
1d 2 + 2 d1
1d 2 + 2 d1
Equation 4-9
d1 + d 2
M W =
d1
Equation 4-10
d2
sM W = M W (1 + d1 d 2
1
1
2 2
d1
2
1
d2
Equation 4-11
Where d is the thickness of layer and the subscripts 1 and 2 denote the two components
of the composite.
However the M-W rule is only true when the properties of the two phases in the
mixture are similar. Bruggeman developed an equation that can represent the dielectric
property of ellipsoid particles in a different medium as following:
1

) + (1 )( 2
)=0
1 + 2
2 + 2
Equation 4-12
Where is the volume fraction of higher resistivity material.

This equation is basically a statistical mixture of two complex properties, giving
values of the dielectric constant, dielectric loss and ionic conductivity of two phases
through complex algebra.
Although with the help of theoretical model we can roughly predict the dielectric
property of material, the results are influenced by many other factors which are not taken
- 111 -
into account. Therefore, experimental evaluations are needed to explore the dielectric
property of epoxy and its composites in the microelectronics application.
4.1.2. Existing dielectric study for composite material
Previous research about the dielectric property of epoxy and its composite has
found that the water absorbed into the materials has great influence to the dielectric
loss[104]. A general trend is to observe an increase in loss factor and dielectric constant
under the influence of moisture uptake[105, 106]. Experiments showed that with around
1wt% water gain in the pure epoxy, the loss factor could increase by more than 20%
while the dielectric constant could increase about 10%. Same trends were also observed
for epoxy/silica composites[107], but changes in the dielectric properties were more
severe, which was usually explained by the interfacial polarization mechanism. These
results depended on the filler density, shape, as well as its surface treatment.
Some authors proposed that the interphase region between the filler and epoxy
matrix played an important role in determining the dielectric properties of composite
materials[108, 109] , and the epoxy/silica composite can be considered as a three-phase
system, including the silica, epoxy and interphase region. This indicated that the effective
dielectric constant of composite was dependent on the dielectric constant ratio between
filler and polymer, and the degree of interaction between filler and polymer as well. With
the experimental data and molecular dipole polarization calculations, they showed that
the addition of silane coupling agent can form chemical bonding and improve the
interphase interaction in the composite, so the interphase dielectric constant can be
increased.
Although the general trend of observation is an increase in the dielectric constant
and loss factor upon water absorption, there are also few papers reporting the reverse
effect, i.e., a decrease in the loss under the influence of humidity which was especially
observed for low water concentrations[110]. It has been argued that reduction in loss due
- 112 -
to a decrease in segmental mobility (segments length increased due to the water

absorption).
In this chapter, we focus on the filler size influence to the dielectric properties of
epoxy/silica composite. Silica with nano-scale size was used as the fillers in our
experiments. A different dielectric behavior of nanocomposite is expected because of the
extremely small size and high surface area of nano size fillers.
4.1.3. Dielectric properties measurement
Resistance temperature detector (RTD)
RTD contact
Excitation contact
Ground
Response contact
RTD contact
Interdigitated
Array of electrodes
Figure 4-1. Electrode design of the single surface sensor used in the experiment
The dielectric properties of the epoxy/silica composite were measured by a
dielectric analyzer (DEA 2970, TA Instruments). The single surface sensor was used for
the DEA experiments. The electrode design of the sensor is based on a co-planar
interdigitated-comb configuration, as shown in Figure 4-1. The space between the
electrodes is measured as 100 m, and the width of each electrode is 150 m. The
uncured liquid resin was coated on the sensor surface and flowed into the spaces between
the electrodes. The sample geometry between the electrode channels can be precisely
controlled and the contact between sample and electrode was also very intimate. After
- 113 -
coating, the liquid sample was cured in a convection oven at 150C for 1 hour. The DEA
experiments were performed from 35C to 250C with a stepping temperature of 5C and
a frequency sweep from 0.01 to 100,000 Hz varying logarithmically. The dielectric
permittivity () and dielectric loss factor () were calculated by the pre-installed
software.
After the first test, all the samples were put in an 85C/85%RH temperaturehumidity aging chamber for 5 days. The aged samples with moisture gain were tested for
a second time with DEA following the same procedure. Then the samples were dried in
the convection oven at 120C for 6 hours for water desorption. The dried samples were
tested for the third time with DEA.
4.2. Results and discussions
4.2.1. Dielectric properties
Figure 4-2. Dielectric property of the control sample after curing
- 114 -
Figure 4-3. Dielectric property of the epoxy/silica micron-composite after curing
Figure 4-4. Dielectric property of the epoxy/silica nanocomposite after curing
- 115 -
Figure 4-2 to Figure 4-4 show the DEA results of the control sample and the silica
composites after curing. These three figures exhibit a similar behavior in the dielectric
relaxation as shown by the change of the loss factor with the temperature and the
frequency. Nevertheless, the transition temperature of the nanocomposite marked by the
peak of loss factor at each frequency appears to be lower than those of the control sample
and micron-composite at corresponding frequencies. This result is consistent with the
observation from DSC measurement for material thermal behaviors since the transition
temperature of the dielectric property reflects the known glass transition temperature of
the polymer[111]. It is also observed that dielectric permittivity and the loss factor of the
nanocomposite at low frequencies are much higher than those of the control sample and
micron-composite.
4.2.2. TTS shifting of dielectric loss curve
The mechanical and dielectric response of the polymeric material follows the
time-temperature superposition (TTS) principle[112, 113]. Briefly, the dielectric
permittivity of solid material is a complex function of two variables: frequency and
temperature. A complete representation should therefore comprise two three
dimensional plot of (f,T) and (f,T). These are cumbersome and are therefore seldom
employed. The prevailing method of the representation consists of plotting the frequency
dependence with temperature as a parameter or vice versa. It makes no difference
whether one is plotting the frequency response or the temperature response in the
appropriate coordinates. One can obtain an identical loss peak which moves to higher
temperature with increasing frequency, or conversely, to higher frequency with
increasing temperature. If the dielectric response curves in a log-frequency plot at
different temperatures are shifted horizontally, a master curve can be constructed to
describe the dielectric behavior at a wide time/frequency range at a particular temperature
(reference temperature). Figure 4-5 show the multi-frequency experiment result for the
- 116 -
nanocomposite sample, which has different loss factor at different temperature. One can
horizontally move the curve of one temperature to a specific temperature by a shift
factors ( T ) which follows the Williams-Landel-Ferry (WLF) equation[114]:
log T =
c1 (T Tref )
c2 + T Tref
Equation 4-13
In the current study, the reference temperature was chosen as the 150C of each
sample. Figure 4-6 shows the shift factors of TTS for nanocomposite sample, which is in
good agreement with the WLF equation. The parameters in the WLF equations for all the
three samples are shown in Table 4-1. It can be seen that these constants take similar
values for all three samples. By curve shifting, master curves were obtained for the three
samples at temperature of 150 C as shown in Figure 4-7.
- 117 -
Figure 4-5. DEA multi-frequency experiment results of nanocomposite sample (not all
the temperature listed)
4.000
3.000
shift factor
fitting curve
2.000
X-shift
1.000
0
-1.000
-2.000
-3.000
-4.000
-5.000
125.0
temperature (C)
250.0
Figure 4-6. Shift factors of TTS for nanocomposite sample

- 118 -
Table 4-1. Constant parameters of WLF equation for three samples

Sample
Control
Micron-composite
Nanocomposite
Reference temp.(C)
150
150
150
C1
5.3
5.6
6.1
C2
87.34
107.0
97.81
4.2.3. Dielectric loss in composites

7
6
control sample
micron size composite
nano size composite
5
4
Log [e'']
3
2
1
0
-1
-2
-3
-4
-5
-4
-3
-2
-1
1
2
3
4
Log [frequency (Hz)]
10
Figure 4-7. Master curves of loss factor for three samples after obtained by TTS shifting
The loss factor of polymer sample is frequency-dependent, which is divided into
two ranges by the loss factor peak (Figure 4-7). The loss peaks of the control sample and
the micron-composite are observed at similar frequency range (around 1-104 Hz for the
former, and 10-104 Hz for the latter), while the loss peak of the nanocomposite occurs at
a much higher frequency (around 103-106 Hz). By the TTS principle, it is concluded that
the loss peak of nanocomposite occurs at a much lower temperature than the pure
- 119 -
polymer or the micron-composite. This result is consistent with the DSC and DMA
measurement, which characterize the transition temperature of free volume change and
viscoelastic modulus change.
At the high frequency range after the loss peak transition(or low temperature
range below the Tg), the change of electrical field is too fast to allow the dielectric
relaxation of bulk molecules, in another word, the long range segmental movement of
polymer is frozen in the glassy state. Moreover, the anhydride cured epoxy is lack of
polar side groups in the molecular structure (as shown in Figure 3-22); other relaxation
mechanisms, such as motion of absorbed water and ionic conductivity, are also
minimized at the low temperature. Therefore, the dielectric loss of each sample keeps a
small value (<0.1), and is almost independent to the frequency.
At a low frequency range before the loss peak transition(or high temperature
range below the Tg), the loss factor has the following sequence: nanocomposite > micron
composite > pure polymer. The dielectric loss can increase several orders of magnitudes
as the frequency decreases. This is the integrative effect of water motion, ionic
conductivity and polymer network relaxation.
control
micron size
nano size
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0
Aging time (day)
Figure 4-8. Moisture absorption of three materials as aging time

- 120 -
It has been revealed that the dielectric loss of the nanocomposites is associated
with water molecules adsorbed on the silica surface[104]. The adsorbed water oriented
with the electric field and therefore increased the permittivity and the loss factor of the
composites at low frequencies. The extra free volume at the interface also assisted the
water mobility as well as the polymer mobility. Silica composites were also characterized
in term of moisture absorption. Figure 4-8 shows the moisture absorption of three
samples in our experiment. The micron-sized filler does not alter the moisture absorption
behavior of the polymer matrix, however, the nanocomposite shows a dramatic increase
in moisture uptake. This is due to several possible reasons. First of all, the nanosilica
synthesized via sol-gel reaction has the hydrophilic nature and tends to attract moisture
on the surface. Compared to the micron-composite, the nanocomposite has a much larger
interface area between the filler and the resin, and therefore, larger number of sites to
attract moisture. And finally, the increased polymer dynamics in the nanocomposite also
helps the moisture diffusion.
The loss factor in the low frequency or high temperature is also affected by the
ionic conductivity in the materials. Equation 4-14 shows the logarithm transformation of
Equation 4-6, in which the log-log plot of the loss factor to the frequency appears to be
linear with a slope of -1. This relationship is verified in the Figure 4-7Figure 4-9Figure
4-7. The loss factor for all three samples appeared to be linear to the frequency with a
slope of nearly -1 in the low frequency region, suggesting that the dielectric loss is
dominated by the DC conductivity.
"=
=
w o 2f 0
log( " ) = log(
Where f is the frequency (s-1).
- 121 -
) log( f )
2 0
Equation 4-14
Figure 4-9. Loss factor and ionic conductivity of the three samples at 1 Hz
A plot of the low frequency (1Hz) loss factor and the ionic conductivity of the
three samples in Figure 4-9 shows that the ionic conductivity of nanocomposite is much
higher than other two samples. It is possible that the addition of the nanosilica also
brought more contaminants into the system that enhanced the ionic conductivity. The solgel synthesized nanosilica contains more contaminants from the solution reaction.
It is indicated from above analysis that the nanocomposite has high water
absorption and high ionic conductivity; therefore, it displays a higher dielectric loss at
low frequency range. Among the many relaxation mechanisms, the DC conductivity is
the dominated factor of dielectric loss in the nanocomposite material.
4.2.4. Moisture influence for dielectric properties
The humidity aging influence was also investigated by comparison of loss factors
of samples with different history. Figure 4-10 and Figure 4-11 show the loss factor of the
three samples before and after aging at a humidity chamber. Before moisture absorption,
- 122 -
three samples had similar loss factor at low temperature due to the low polymer mobility.
However, the silica filled composites, regardless the filler size, have much higher loss
factor after aging in 85C/85% humidity chamber with moisture absorption than the
control sample. Although the amount of moisture absorption of the micron--composite
was similar to that of the control sample as shown in Figure 4-8, the composite materials
with heterogonous structure and interface between filler and polymer displayed the
interfacial relaxation which was enhanced with the absorbed water on the silica surface.
Therefore, the moisture absorption of the silica composite significantly increased the loss
factor.
From Figure 4-11, it is also observed that the trapped water can be desorbed as the
temperature increasing in the micron size composite, which corresponded to the decrease
of loss factor in the temperature range 70~110C. On the other hand, the decrease in loss
factor of the nanocomposite happened at a high temperature, which indicates that
desorption is more difficult with a higher surface area of the nanosilica with hydrophilic
nature that retains moisture. Even at temperatures higher than Tg of the nanocomposite
when the polymer mobility is high, the moisture desorption rate is still slow. Figure 4-12
showed the loss factor of three samples after drying in the thermal oven for 6 hours. The
dielectric property can be fully recovered after moisture desorption.
- 123 -
1.0E5
z control
10000 micron size
nano size
1000
Loss Factor
100
10
0.001
50
z
z
0.1
0.01
100
150
Temperature (C)
200
250
(a)
1.0E5
z control
10000 micron size
nano size
1000
Loss Factor
100
10
0.1
0.01
0.001
50
100
150
Temperature (C)
200
250
(b)
Figure 4-10. Loss factor of three samples after curing, (a) 1Hz; (b) 1000Hz
- 124 -
1.0E5
z control
10000 micron size
nano size
1000
Loss Factor
100
10
0.001
50
z
z
0.01
z
z
0.1
z
z
100
150
Temperature (C)
200
250
(a)
1.0E5
z control
10000 micron size
nano size
1000
Loss Factor
100
10
0.01
50
0.1
0.001
z
z
100
150
Temperature (C)
200
250
(b)
Figure 4-11. Loss factor of three samples after aging under humidity, (a) 1Hz; (b)
1000Hz
- 125 -
1.0E5
z control
10000 micron size
nano size
Loss Factor
100
10
1
0.1
0.001
50
z
z
1000
0.01
z
z
100
150
Temperature (C)
200
250
(a)
1.0E5
z control
10000 micron size
nano size
1000
Loss Factor
100
10
0.1
0.01
0.001
50
100
150
Temperature (C)
200
250
(b)
Figure 4-12. Loss factor of three samples after drying (a) 1Hz; (b) 1000Hz
- 126 -
CHAPTER 5.
THE HARDENER EFFECTS TO COLLOIDAL

SILICA DISPERSION
In the previous chapters, we have discussed that the nanosilica tends to be fairly
unstable in the underfill formulation due to the ubiquitous attractive Van der Waals
forces between them. In addition, with the large surface area of nanosilica in dispersion,
the interfacial interaction between the particle and the epoxy liquid medium tends to
dominate the equilibrium and dynamic behaviors of the nanocomposite underfill system.
These interactions can be affected either by changing the chemical and physical
properties of filler surface, or by changing the properties of liquid medium. How to
stabilize and disperse the nanosilica in the epoxy liquid medium is the key to synthesize
the nanocomposite underfill with low viscosity and good material properties.
The related studies focusing on the filler side are discussed in Chapter 2. The
particle surface has been chemically modified by the silane coupling agent. This
approach can change the surface tension of particles and improve the wetting of epoxy on
them. If the nanosilica is treated under the optimal condition, the composite underfill still
keeps a low viscosity which allows the material dispensing and chip collapsing.
However, this approach could not completely solve the problems caused by the
nanosilica, such as Tg depression in the cured materials, as discussed in Chapter 3.
In this chapter, we will focus on the liquid properties of underfill matrix. Underfill
is usually composed of liquid epoxy monomer as basic composition, and the hardener is
used to improve the crosslinking density. Since the selection of epoxy monomer is
usually limited to Bisphenol A or Bisphenol F epoxy, most of the underfill formulations
- 127 -
are adjusted by varying the hardener types. The dispersion of nanosilica in different
hardeners will be studied by the dynamic rheology method. When dispersed in the liquid
underfill medium, the nanosilica interacts to form characteristic microstructures which in
turn determine the rheological properties of the composites. So we can elucidate the
microstructure of nanocomposite underfill by using the dynamic rheological method. It is
expected to find a more efficient mechanism to stabilize the nanosilica in the nonaqueous liquid, and then correlate the solvent (hardener) effect with the colloidal
interactions excising in each system.
5.1. Experiment
5.1.1. Materials
Silica nanoparticles (SiO2 powder, around 100nm average diameter, S.D. of

20nm) are commercially available and were used as received. Two kinds of hardener with
opposite acid-basic properties were chosen. One was anhydride-type hardener HMPA
(hexahydro-4-methylphthalic anhydride, from Lindau Chemicals); another was aminetype hardener DETDA (Diethyl toluene diamine, From Lonza Chemicals Ltd). Figure 5-1
shows the chemical structure of these two hardeners.
The flow of pure organic liquid is viscous, with filler addition, it becomes
viscoelastic. To elucidate the interaction between the filler and epoxy formulation with
different hardeners, the two hardener compounds were used as liquid medium for silica
dispersion. The silica loading is 30 vol%. The mixture was stirred vigorously by a high
speed blender (Tline laboratory stirrer, from Talboys Engineering Corp.), then sonicated
for 30 minutes by using a Sonicator (Misonix 3000).
- 128 -
H2N
NH2
3,5-Diethyltoluene-2,6-diamine
(DETDA)
4-methylhexahydrophthalic anhydride
(HMPA)
Figure 5-1. Molecular structure of two hardeners used in the experiment

The material properties of the amine- or anhydride -based composites were
characterized after underfill curing. The epoxy used was diglycidyl ether of Bisphenol-A
type (EPON828, from Shell Chemicals with average molecular weight of 377). 1-(2isocyano-ethyl)-2-undecyl-1H-imidazole (C11Z-CNS) from Shikoku Chemicals was used
as a latent catalyst for HMPA formulation. Structures of these compounds are shown in
previous test. The nanosilica was added into the mixed underfill matrix with specific
loading. Other preparation and characterization process are same as in Chapter 3.
Table 5-1. Matrix composition of underfill with different hardener
Anhydride-based formulation
Amine-based formulation
EPON828
53.6 wt%
80.9 wt%
HMPA
46.1 wt%
-
DETDA
19.1 wt%
C11Z-CNS
0.3 wt%
-
5.1.2. Dynamic rheology
The rheological behaviors of the samples were characterized under both steady
and oscillatory shear using a stress rheometer (TA Instruments, AR1000N). In steady
- 129 -
shear experiments, the viscosity of the sample is measured as a function of shear stress or
shear rate. However, steady-shear measurements are not very sensitive to the
microstructure present in the systems. Dynamic rheology, performed under oscillatory
shear, is the preferred technique for detecting microstructure features. In dynamic
experiments, a low-amplitude sinusoidal deformation was applied to sample. If the
maximum strain amplitude is 0 and the temperature was held constant during the
frequency sweep, the strain can be described as:
= 0 sin t
Equation 5-1
Where is the angle frequency, t is the time.

The complex modulus of the sample will create resistance to the oscillatory shear:
= G
Equation 5-2
Within the region of linear viscoelasticity, the complex modulus can be

decomposed into an in-phase and an out-phase component:
G = G '+iG"
Equation 5-3
The in-phase component of complex modulus is defined as the elastic (or storage)
modulus G and is related to the elastic energy stored in the system on deformation. The
structure system will gain energy from the oscillatory motion as long as the motion does
not disrupt the structure. The magnitude of the storage modulus depends of the number of
interactions between the ingredients in the sample and the strength of each interaction,
which is indicated to the interactions between filler and filler or filler and organic matrix
in our study. The out-of-phase component of complex modulus gives the viscous (or loss)
modulus G. It is linked to the viscous dissipation of energy in the system due to the
- 130 -
friction between the different ingredients. To evaluate the characteristics of a viscoelastic

material, it is customary to examine the frequency spectrum showing G and G as a
function of frequency. The relative magnitudes and shapes of the G and G curves
indicate the type and extent of microstructure present. The elastic modulus, in particular,
is an important indicator of the degree of structuring in a dispersion system.
During our experiments, a cone-and-plate geometry with cone angle of 0.2 degree
was used. Before each dynamic experiment, a steady preshear was performed at a shear
rate of 1 s-1 for 60 s, followed by a 120s rest period. This procedure was necessary to
erase any previous shear histories and to ensure that the material establishes its
equilibrium structure. Experiments were done twice to ensure the accuracy of the testing.
5.1.3. Dielectric constant of liquid sample
The dielectric constant of the epoxy formulation with different hardener was
measured by the Dielectric Analyzer (DEA 2970, TA Instruments). The liquid sample
was coated on the single surface sensor and flowed into the spaces between the
electrodes. The dielectric spectrum was measured at room temperature with frequency
range from 10-2 to 104 Hz.
5.2. Results and Discussions
5.2.1. Rheology measurement
Our purpose is to probe the microstructures present in the polymer/nanosilica

composites using rheology and to explain the reasons for different rheological and
thermodynamic behaviors of the nanocomposites. We will interpret the observed
microstructures in terms of the nanosilica colloidal interactions within the organic matrix.
Note that our composites are two-phase systems, where the continuous phase is the liquid
- 131 -
organic (harder and/or epoxy), and the disperse phase is the solid silica particles of
nanometer scale.
First we describe the rheology and microstructure of nanosilica dispersed into
different hardener. From the dynamic rheology, we obtain the characteristic frequency
spectrum of the nanosilica/HMPA composite sample showing the elastic (G) and viscous
(G) moduli as a function of frequency (Figure 5-2). Two important features of the
rheological response are apparent from this figure. First, the moduli are weakly depedent
on frequency. In particular the elastic modulus (G) is independent of frequency. The
viscous modulus (G) is only slightly increased at the high frequency region. Second, the
elastic modulus significantly exceeds the viscous modulus. The dominance of G
indicates that the material behaves primarily in an elastic manner, while the frequency
independence of the modulus shows that the system behavior is unchanged over a range
of time scales. This sample exhibited rheological behavior corresponding to a weak
physical gel consisting of a three-dimensional network of physical bonds. This type of
behavior is characterized by constant or small increases in G and G (within one order of
magnitude) and no crossover point for G and G. In another words, the sample is solid
like because the fillers have strong interfacial interaction and form a flocculated system.
The filler we used is nanosilica with an average size around 100nm. As the product of
sol-gel method, the surface chemistry of the nanosilica is hydrophilic due to the presence
of hydroxyl groups on the surface. When dispersed in anhydride hardener, the silica
particles with hydrophilic surface easily adhere to each other through hydrogen bonding
and form irregular aggregations. In another words, the interfacial interaction between the
filler and filler is superior to that between the filler and organic matrix molecules.
- 132 -
Therefore, the elastic modulus G is mainly from the sum of filler-filler interactions
multiplied with the strength in the interactions, and the value of G is an indication of the
density of network bonds in the systems[115]. The formations of these structure called
flocs, which can further connected into a three-dimensional network through the whole
organic matrix, and occlude liquid organic in their interparticle voids. The flocs structure
affects the rheology of the composite and leads to a significant rise in viscosity and gellike behaviors[116].
Now let us examine the dynamic rheology of nanosilica composites in different
organic matrix. Figure 5-3 shows the frequency spectrum for a composite of nanosilica
and amine hardener. For this sample, the viscous modulus G dominates the elastic
modulus G over the entire frequency range. More over, both moduli vary significantly
with frequency. Thus, while the nanosilica/anhydride system forms a physical gel, the
nanosilica/amine system gives rise to a sol-like microstructure with few interconnections
between silica units. In other words, the hydrogen bonding between silica surfaces is
replaced by the interactions between filler and amine molecules due to their good wetting
properties. Therefore, this suspension is non-flocculated, which means that the particles
in the matrix can repel each other by strong filler-organic interactions. Table 5-2
summarizes the characteristics of dynamic rheology for two nanosilica dispersions.
- 133 -
Modulus (Pa)
10
10
10
elastic modulus (G')

viscous modulus (G")
0
10
0.01
0.1
10
Frequency (Hz)
Figure 5-2. Elastic and viscous modulus as a function of frequency for
nanosilica/anhydride mixture
2
10
Modulus (Pa)
10
10
-1
10
-2
10
Elastic modulus (G')

Viscous modulus (G")
-3
10
-2
10
-1
10
10
10
10
Frequency (Hz)
Figure 5-3. Elastic and viscous moduli as a function of frequency for nanosilica/amine
mixture
- 134 -
Table 5-2. Summary of dynamic rheology of different systems

Filler
Filler loading
Organic matrix
Dynamic rheological
behavior
Moduli change
System
interpretation
microstructure
nanosilica
30 vol%
amine
G<G
Viscous behavior dominate
Increasing over several
magnitudes
Sol like
anhydride
G>G
Elastic behavior dominate
Almost constant
No interconnections
between fillers
Strong filler-filler interaction
Gel like
The viscosity of the dispersion system was also characterized under the steady
shear stress, as shown in Figure 5-4. Nanosilica/DETDA keeps a low viscosity which
slightly decreases as the shear stress increases, while the nanosilica/HMPA shows a yield
stress, as well as shear-shinning behavior over the measurement range. In the
silica/anhydride dispersion, the shear stress around 2 Pa is a manifestation of the network
structure presented in the system, and signifies the minimum stress required to induce the
sample flow. For low stress, the silica/anhydride mixture shows considerable resistance
to flow and the corresponding viscosity are very large (> 105 Pa.s). At a high stress (i.e.,
>2 Pa), the material abruptly yields and a low viscosity plateau is observed. This shearthinning occurs because the bonds composing the network structure in the
silica/anhydride mixture are weak gel that can be disrupted by shear.
- 135 -
10
Viscosity (Pa.s)
10
silica/HMPA
silica/DETDA
10
10
10
10
10
-1
10
-1
10
10
10
10
10
Shear stress (Pa)

Figure 5-4. Steady-shear viscosity as a function of shear stress for nanosilica in two
hardeners
Sol (stable)
Silica
Gel
Figure 5-5. Schematic representations of two possible scenarios that can occur in the case
of silica particles dispersed in a liquid.
- 136 -
The microstructure of nanosilica dispersed in the two hardeners can be explained

in Figure 5-5. The nanosilica in the amine hardener is stabilized by the strong interaction
between amine liquid and silica particle. So far, we have elucidated the microstructure
features of nanosilica composites with different polymer matrix (as shown in Table 5-2).
Now we will try to unravel the colloidal interactions responsible for the microstructures
and the resulting thermodynamic properties. In the process, several issues will be
explored: 1) what are the governing interaction forces in the systems? 2) Can the trends
in rheological data be explained by changes in the intensity of these forces? 3) How are
the microstructure and interaction force related to the macrostructure behaviors of the
nanocomposites such as viscosity and glass transition temperature?
5.2.2. Van der Waals interaction
As we know, the dispersion of colloidal particles in the medium is a force balance

between an attractive force and a repulsive force. The dispersion force, the van der Waals
(vdW) force, between colloidal particles is always attractive[74]. The energy potential
between two spherical particles with identical radius of r can be described by the
following expression:
= Aeff
r
12d
Equation 5-4
Where Aeff is the effective Hamaker constant, and d is the distance between two particles.
In our system, the silica aggregates have a complex geometry and can not be
approximated as a sphere or as a flat plate, but in our analysis the geometrical details are
not important. The so-called effective Hamaker constant ( Aeff ), which considers both the
particle-particle interaction and particle-liquid interaction, is the significant parameter to
- 137 -
control the intensity of vdW. It can be calculated from the Hamaker constant of both the
particle and the medium.
Aeff = ( Am A p ) 2
Equation 5-5
Dzyaloshinskii, Lifshitz and Pitaevskii (DLP) developed a theory by which the

Hamaker constant can be calculated. Bulk dielectric properties and the refractive index of
matter are the basis for this theory, as shown below[117]:
3
( 1) 2 3h e (n 2 1) 2
A = kT
+
4
( + 1) 2 16 2 (n 2 + 1) 3 / 2
Equation 5-6
Where k is the Boltzmann constant, T is the absolute temperature, h is the Plancks

constant, e is the main electronic absorption frequency for the dielectric permittivity
(taken to be 31015 s-1for each liquid media[117]), and n is the refractive index.
This equation corresponds to the non-retarded Hamaker constant of the media in a
vacuum. From the DEA measurement we can get that dielectric constant of amine liquid
is usually less than 4.6; on the other hand, the dielectric constant of anhydride is much
higher than amine, measured as 20.9. The Hamaker constants ( A ) of two hardeners are
calculated by Equation 5-7, as shown in Table 5-3. The effective Hamaker constants of
silica/liquid/silica are calculated by Equation 5-6. Figure 5-6 plots the Van der Waals
potentials of silica surface in two hardener liquids, as calculated by Equation 5-4. The
radius of particle is 50nm. It indicates that the vdW force strength between the silica
particles in an anhydride medium is larger than that in the amine medium. This
estimation corresponds with the dynamic rheology measurements in that, particle-particle
has stronger attractive interactions in the anhydride medium.
Table 5-3. Bulk material properties for various components
- 138 -
Dielectric
Refractive
Hamaker constant, A
constanta
indexb
(10-20 J)
HMPA
4.6
1.455
6.04
DETDA
21
1.477
6.61
silica
3.82
1.463
6.22
Component
a: measured by DEA, b: obtained from sample information from vendor
silica/DETDA
silica/HMPA
Van der Waals potential, V/kT
0.5
1.5
2.5
-5
-10
-15
-20
-25
-30
Distance between particle, d (nm)
Figure 5-6. Van der Waals potential between silica particles in different hardeners
The small attractive vdW force is not the only factor contributing to the stability
of dispersion of silica in the amine medium, another possible parameter that influences
the filler-filler interaction is the repulsive force between them. S.A Khan et al.
investigated another kind of solid-liquid dispersion in a similar way to elucidate the
- 139 -
relation between colloidal dispersion and dynamic rheology[118]. The system they used
was fumed silica and polymer electrolyte based on poly(ethylene glycol). They concluded
that in such a polymer based medium, there was no significant electrostatic interactions
between silica aggregates. In another words, it is highly unlikely that the silica particles
show any appreciable surface charge because of the non-aqueous medium. Therefore the
electrostatic interaction between fillers is not the major repulsive force.
S.R.Raghavan et al. found that the solvation phenomena due to hydrogen bonding
interactions is the key aspect distinguishing the sol-like from the gel-like system[119].
According to this hypothesis, silica forms stable sols in liquids which possess sufficient
capacity to form hydrogen bonds, and conversely, silica form gels in liquids which have a
perceptibly lower hydrogen bonding capacity. Lets take a look into the amine/silica
system. The amine functional groups have an electronegative heteroatom nitrogen;
therefore, the liquid amine molecules can organize at the silica interface by forming Hbonds with silanol (Si-OH) group on the silica surface. This would lead to a solvation
layer that coats each silica unit. Thus we can envision that the solvated particles are
stabilized against coagulation by an additional repulsive force (solvation force). The
similar hydrogen bonding may also occur in the anhydride/silica system because
anhydride contains oxygen, an electronegative heteroatom. However, the hydrogen bond
strength is smaller than that in amine system.
With above analysis, we can see that the silica particles in amine medium have
smaller attractive force (vdW force) but larger repulsive force (hydrogen bonding) than
that in the anhydride medium. This explains the difference between the microstructure of
these two liquid-solid dispersion which were characterized by the dynamic rheology.
- 140 -
amine
anhydride
Glass transition temperature (degree C)
130
120
110
100
90
0
10
20
30
40
Silica loading (wt%)
Figure 5-7. Glass transition temperatures of silica composites with different hardeners
The microstructure and interaction force between the liquid-solid dispersion also
affect the microstructure behaviors of the nanocomposites after crosslinking. We
measured the glass transition temperatures of silica-epoxy nanocomposites with both
amine-type and anhydride-type hardeners. The results were shown in Figure 5-7. It is
interesting to find that the Tg of amine based composites is kept stable with the filler
loading increasing. However, the Tg of anhydride based composites decreases as the
filler loading increases. In our previous discussion, this special Tg depression phenomena
in anhydride system have been explained[120]. And we also predicted that this
phenomenon can be improved with better filler-polymer interactions. Here, we showed
the results from an amine hardener. As we mentioned before, the filler-filler interaction in
the amine type medium is not as strong as that in anhydride medium, and the filler can
form sol type dispersion in the liquid with more compatibility between filler and polymer.
Therefore, changing liquid properties is a more efficient approach than modifying the
- 141 -
filler surface properties toward good filler dispersion and composite properties. As a rule
of thumb, the polarity (dielectric constant) of dispersion liquid medium should be low,
and similar to the silica particles.
- 142 -
CHAPTER 6.
PHOTO-POLYMERIZATION OF EPOXY
NANOCOMPOSITE FOR WAFER LEVEL APPLICATION
The extremely small size of a nanosilica filler not only benefits the solder wetting
during the reflow process of no-flow underfill and wafer level underfill, but it also brings
the advantage for optical applications with the high transmittance. Based on the epoxy
nanocomposite, new photo-curable composite materials can be formulated toward the
wafer level application.
6.1. Photo-polymerization of epoxy
In the recent years, a number of investigations have demonstrated the feasibility

of photo-polymerization for polymers and their composites. These polymers can be
classified into two categories, depending on whether the polymerization proceeds by a
free radical-type or cationic-type reaction center. The first case is based on the vinyl
compounds [121] or acrylate compounds [122] initiated by a free radical reaction. Free
radical-type systems are by far the most widely used and studied in todays photo-curing
applications due to their high reactivity. However, the free radicals generated by the light
exposure can be extinguished easily by the oxygen in the air; therefore, an inert
atmosphere is usually needed during curing. The life time of a free radical is relatively
short, which stops the initiation once the light source has been removed. The second case
is based on photo-initiated cationic polymerization. A proton acid is produced by
photolysis of a triarylsulfonium (TAS) or a diaryliodomium salt to initiate the
polymerization of the epoxide ring. Cationic polymerization once initiated, may continue
to proceed without light. This process, called dark reaction, is the result of the ability of
the long-lived protonic acid or Lewis acid species to continue the polymerization. As
such, the cure process requires relatively short UV light exposure, limited to just the
- 143 -
amount of time required for the photolysis of the photo-initiator. Table 6-1 compares the
features of two photo-initiation methods.
Table 6-1. Comparison between two photo-polymerization approaches
Feature
Cure speed
Initiation
O2 sensitivity
Shrinkage
Adhesion
Post cure
Chemical resistance
Humidity sensitivity
Acid/base sensitivity
Free Radical
High
Light
Yes
Large
Moderate to good
Limited
Good
No
No
Cationic
Moderate to high
Light and heat
No
Negligible
Excellent
Strong
Moderate to good
Yes
Yes
With the advent of stable cationic photo-initiator, such as the onium salts[33], the
cationic photo-polymerization process has become increasingly important because of the
significant advantages it has over radical polymerization. Cationic photo-polymerizations
have been used to polymerize important classes of monomers such as epoxy, and the
resulting polymers exhibit excellent adhesion, abrasion resistance, and chemical
resistance. Utilizing this reaction technique, the Bisphenol-A novolac epoxy resin (EPON
SU-8) has been used as a negative near-UV photoresist in a MEMS fabrication
process[123], and polymer optical waveguide materials[124].
Previous research about the conductive adhesive has demonstrated the application
of the dual initiators including both the photo-initiator and the thermal initiator for the
cycloaliphatic epoxy based conductive adhesives[125].
The adhesive can be cured
quickly at ambient temperature by the cationic polymerization after UV exposure, and

then, the thermal initiator can cure fully the adhesive at elevated temperature and
eliminate the disadvantage of the inherent lack of transparency caused by the metal
particles in the adhesives. Nevertheless, there are a few reports of the photo-
- 144 -
polymerization of silica/epoxy composites via the cationic crosslinking reaction because

micron size silica fillers can scatter UV light and hinder the photo-polymerization
process. Nanosilica composites, on the other hand, have displayed desirable optical
properties[38, 126] and have become the best solution for photo-curable applications.
In this chapter, we discuss how photo-curable materials based on nanosilica and
epoxy are synthesized. A photo-sensitive initiator was added into the formulation that
releases cations after UV exposure and initiates the epoxy crosslinking reaction. The
photo-crosslink reaction of the epoxy makes it a negative tone photoresist. A
comprehensive study of the photo-curing behaviors of these materials was conducted to
elucidate the photo-polymerization mechanism in the composites and to achieve an indepth understanding of the effect of nanosilica on the photo-reaction of epoxy. Also
thermal and mechanical properties were determined and the data was used to optimize the
formulation of a photo-curable nanocomposite for wafer level applications.
6.2. Experiments
6.2.1. Materials
Colloidal nanosilica with average size of 20 nm was synthesized by the sol-gel

method. The size was characterized by the transmission electron microscopy (TEM).
EPON 862, a Bisphenol F epoxy, was obtained from Shell Chemicals. The sulphonium
photo-initiator, Bis[4-(diphenylsulfonio)-phenyl]-sulfide-bis-hexafluorophosphate (KI85), was obtained from Sartomer. Figure 6-1 shows the molecular structure of this photoinitiator.
- 145 -
PF6-
PF6-
S
Figure 6-1. Molecular structure of photo-initiator
6.2.2. Preparation of nanocomposites
The nanosilica was modified by the proper silane coupling agent following a
method described before[39] and then incorporated into the epoxy matrix. After surface
modification, nanosilica interacted with the polymer matrix and achieves a high filler
loading level. The nanosilica filler loadings in the epoxy matrix were 10, 20, 30 and 40
wt%. The photo-initiator KI85 (2 mol%) was added into the formulations and the
mixtures were stirred for one hour.
6.2.3. Characterization
Transmittance
The photo absorption properties of the samples were measured with a UV-visible
spectrophotometer (Beckman Du520). The absorption of the photo-initiator and the
colloidal silica was characterized in an ethanol solution. To measure the absorption of
photo-curable nanocomposites, liquid composites were spin-coated on a quartz glass
slide. Then the specimen was put into the chamber of the UV-visible spectrophotometer
and scanned.
- 146 -
Curing process
Polymerization reactions can be followed in-situ by real-time Fourier Transform
Infrared (FTIR) Spectroscopy by monitoring the decrease of the IR band characteristic of
the reactive functional group upon UV exposure[127, 128]. The sample was exposed
simultaneously to the UV beam that induces polymerization and to the IR beam that
analyzes its extent, as shown in Figure 6-2. The background spectrum was collected
using double-side polished thin silicon. The liquid samples were coated onto the silicon
wafer to form a very thin layer. The sample was exposed to UV light with a light
intensity around 20 mW/cm2 for a certain time and then put into the FTIR (by Nicolet,
Magna IR 560) chamber for the test. The spectrum of the sample was taken to observe
the change in chemical structure of the epoxy during UV curing. The wave-number of the
spectrum was from 4000 cm-1 to 400 cm-1 with a resolution of 4 cm-1. The number of
scans was 32 times.
Figure 6-2. Scheme of real-time FITR setup

The curing process of the photo-curable nanocomposite materials also was
characterized by a photo differential scanning calorimeter (photo-DSC). The photo-DSC
was performed using a Q1000 DSC (TA Instruments) equipped with the photo
calorimeter accessory (PCA). Figure 6-3 shows the scheme of the photo-DSC setup. The
wavelength of the UV light was 320~500 nm and its intensity was approximately 20
mW/cm2. Reactions were performed at room temperature in a nitrogen atmosphere.
- 147 -
Sample sizes were around 10 mg. The residual reaction heat and glass transition
temperature of the photo-curable nanocomposites after UV exposure were characterized
by the thermal DSC. A dynamic scanning experiment was conducted with a ramp rate of
5C/min, from ambient temperature to 300C. The cured sample was left in the DSC cell
and cooled to room temperature. Then the sample was reheated to 200C at 5C/min to
obtain another heat flow diagram in the modulated mode. The initial temperature of the
heat flow step of the second diagram is defined as the glass transition temperature (DSC
Tg).
Figure 6-3. Scheme of photo-DSC setup

Film sample preparation
In order to evaluate material properties of the nanocomposite samples after UV
curing, the liquid nanocomposites were cast onto an aluminum substrate with a thickness
around 50 m. UV curing of samples was performed using a UV lamp (CLE-4001, with a
long-wave UVA portion of the spectrum) for 20 min. The intensity was ~50 mW/cm2
measured by a traceable radiometer. Sample curing was performed in the air. Then, the
sample was thermally cured at 120C for 30 minutes to complete the crosslinking
reaction. Free-standing nanocomposite films were peeled off from the aluminum
substrate for further characterization.
- 148 -
Thermal mechanical properties measurement

A Dynamic Mechanical Analyzer (DMA, TA Instruments, Model 2980) was used
to measure the dynamic moduli and glass transition temperature of the nanocomposites.
The cured film was cut into a strip of dimensions about 186 mm. The accurate thickness
of the film was measured by a laser profilometer. The test was performed in the film
mode. The temperature was increased from room temperature to 250C at a heating rate
of 3C/min, while the storage modulus (E), loss modulus (E) and tan were calculated
by the pre-installed software. In order to obtain complementary measurements to the
DSC, the DMA Tg was determined by the peak temperature of the tan curve.
Thermal expansion measurement
The coefficient of thermal expansion (CTE) of the cured film was measured on a
Thermo-Mechanical Analyzer (TMA, TA Instruments, Model 2940). The dimensions of
the sample were about 105 mm. The samples were heated in the TMA furnace at
5C/min from room temperature to 200C. The CTE before the Tg is defined as 1 and
after the Tg as 2.
Thermal stability measurement
The thermal stability of the sample was characterized by a Thermogravimetric
Analyzer (TGA, TA Instruments, Model 2050). The cured samples were put into a TGA
pan. The experiment was conducted with a ramp rate of 20C/min from ambient
temperature to 800C in air.
- 149 -
Morphology
The dispersion and morphology of the nanocomposite after photo-curing were
characterized by transmission electron microscopy (TEM). The cross sections of the
nanocomposite were prepared using a Leica Microtome Nova. The samples first were
trimmed manually using a razor blade to produce a trapezoidal cross section with an area
that was approximately 0.5 mm by 0.5 mm. The samples then were mounted on the
microtome and aligned with a glass knife, which was used to trim the sample further.
Each sample was trimmed until its surface was parallel to the glass. Then the glass knife
was replaced with a diamond knife that was used make the final sections. The final
thickness of the sections was between 30-60 nm and they were mounted on a copper
TEM grid. The TEM investigations were performed on a JEOL 100CX TEM operating at
an acceleration voltage of 100 KV.
Surface hardness
In order to characterize the surface mechanical property of the nanocomposite
underfill after photo-curing, the nanoindentation was performed to measure the hardness
of the photo-cured underfill film. The samples were prepared by the same method as in
the UV test. The instrument used in the nanoindentation experiments was Nano Indentor
XP by MTS Systems Corporation. The XP/CSM tip used in the experiments allows 50
gram maximum force. The surface velocity was 10 nm/s. The indenting depth was 1 m.
The strain rate was set to be 0.05 s-1. The harmonic displacement was 2 nm at a frequency
of 45 Hz.
- 150 -
6.3. Preparation of photo-curable nanocomposites

6.3.1. Filler size of nanosilica
Previous research on silica/epoxy nanocomposites has shown that the addition of

nanosilica was small enough to not disturb the optical transparency of the composite
materials in the visible light range[36, 38]. Nevertheless, the 100 nm silica significantly
scatters the UV light at the wavelength region, which excites the cationic initiators for the
photo-polymerization of epoxy. Therefore, even smaller size silica is needed for photocuring applications. 20 nm silica can be synthesized in the colloidal form by the sol-gel
method. The transmittance of these two types of nanosilica filler in the ethanol solution is
measured by UV-visible spectroscopy as shown in Figure 6-4. It can be seen that the 20
nm silica has >95 % transmittance in the UV to visible light region, which is much better
than the 100 nm silica. Therefore, the 20 nm silica was chosen as the filler in the
composite due to its good transparency to UV light. (In this chapter, the term nanosilica
refers to 20 nm silica).
Transmittance (%)
100
80
60
40
20
~100nm silica
~20nm silica
0
300
400
500
600
700
Wavelength (nm)
Figure 6-4. Light transmittance of two kinds of nanosilica in ethanol solution
- 151 -
The morphology of the nanosilica was characterized by Transmission Electron

Microscopy (TEM). The TEM picture in Figure 6-5 shows that nanosilica has uniform
shape and size distribution. Surface modification of nanosilica by the proper silane
coupling agent through the hydrolysis and condensation reaction is commonly employed
to yield better compatibility between the modified silica filler and the polymer
matrix[129]. In this study, the liquid-phase silyation of nanosilica was performed using
the epoxy resin/ethanol solution as a solvent. This in-situ procedure avoids the reagglomeration of nanoparticles after modification which may occur during drying of
modified nanosilica prepared in common solvents[130, 131]. To enhance the reaction rate
of silane grafting, water and an organic acid as catalyst were added.
Figure 6-5. TEM picture of the 20 nm colloidal silica

6.3.2. UV absorption of compositions in the photo-curable nanocomposite
It has been reported that onium salts containing aromatic groups such as
diaryliodonium and triarylsulphonium salts are efficient photo-initiators for cationic
polymerization of epoxy. The UV-visible spectra of photo-initiator KI85 and other
compositions in the nanocomposites were characterized as shown in Figure 6-6. The UV
- 152 -
absorption maxima of sulphonium salt KI85 ( max ) is at 300 nm, which agrees with the
reference[132]. With the benzene ring structure in the epoxy molecule, the pure epoxy
strongly absorbs UV light at a peak around 288 nm. Nevertheless, the photo-initiator still
can be excited in the epoxy by the UV wavelength below 360 nm due to its wide and
strong UV absorption. Figure 6-7 showed the influence of nanosilica addition to the UV
absorption of the photo-initiator in the epoxy matrix. The two curves are almost identical
except that the absorption intensity for the photo-initiator (around 300 nm) is different.
With the nanosilica addition, the UV absorption intensity of photo-initiator in the
composite is reduced. This is expected to change the photo-curing kinetics of the
nanocomposite, as will be discussed later.
- 153 -
2.2
288nm
2.0
Absorbance (a.u.)
1.8
1.6
1.4
300nm
1.2
nanosilica
pure epoxy
photoinitiator
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
250
300
350
400
450
500
Wavelength (nm)
Figure 6-6. UV absorbance of pure epoxy, nanosilica and photo-initiator
Absorbance (a.u.)
2.5
epoxy/photoinitiator
with nanosilica
epoxy/photoinitiator
2.0
1.5
1.0
0.5
0.0
250
300
350
400
450
500
Wavelength (nm)
Figure 6-7. Influence of nanosilica on the UV absorption of photo-initiator in epoxy
- 154 -
6.4. Reaction mechanism and kinetics of photo-curable nanocomposite

6.4.1. Mechanism of cationic photo-polymerization
The cationic polymerization of epoxy is a multiple-step process. The first step is

the photolysis of initiator to generate the catalytic active center, as depicted in Figure 6-8.
In our experiments, the triarylsulfonium salt (TAS) is used as the photo-initiator. Upon
UV irradiation, TAS can undergo fragmentation by decay of the excited singlet state to
yield cations by the heterolytic cleavage of sulphur-carbon bond[133]. The cationic
species interact with a proton source, usually from the monomer or impurities (R-H), to
generate the strong Brnsted acid. In the second step, the epoxy polymerization was
initiated by the protonation of the epoxide oxygen with the Brnsted acid. Then the
polymerization will start with the active centers and the ether link will be formed by the
epoxide ring-opening reaction. Compared to the photolysis step and initiation step, the
chain propagation step has a slow polymerization rate [134].
- 155 -
Photolysis
+
h
Photointiator
PF6-
+
PF6- + R-H
+ H PF6
Initiation
O
+
O
CH2
CH
CH2
O+ H
+ H PF6
CH2
CH
PF6-
CH2
Propagation and crosslinking
CH2
CH
CH2
CH2
PF6-
O+ H
CH
CH2
CH
CH2
OH
O+ H PF6
CH2
CH
CH2
O
CH2
O
O
CH
O
CH2 CH CH2
CH2
CH2
CH
CH
CH2
O+ H
PF6-
CH2
Figure 6-8. Reaction mechanism of cationic photo-polymerization
- 156 -
CH2
OH
6.4.2. Reaction process measured by real-time FTIR
Figure 6-9 shows the FTIR spectrum of the uncured epoxy sample at room
temperature. The tentative assignments of bands observed are made according to the
epoxy molecular structure. The peak around 916 cm-1 indicates the epoxy ring absorption.
The intensity of this peak will decrease as the curing reaction proceeding. The intensity
of the peak at 2800-3200 cm-1, which indicates the absorption of the stretching of C-H
bond, was used as an internal reference. Other prominent IR bonds are hydroxyl bond
(3520 cm-1) and benzene bonds (1605, 1500, 846 cm-1). Details about the peak
assignment are listed in the Table 6-2.
846.83
0.30
0.28
0.18
1500.76
916.89
0.20
1605.85
3520.93
0.22
2925.39
0.24
2999.35
0.26
Absorbance
0.16
0.14
0.12
0.10
0.08
0.06
0.04
0.02
0.00
-0.02
-0.04
-0.06
4000
3000
2000
1000
Wavenumbers (cm-1)
Figure 6-9. FTIR absorption of epoxy

Table 6-2. Observed peaks of epoxy with FTIR
Wavenumber (cm-1)
3520
2999
2926
1605, 1500
916
846
Assignment
OH stretching vibration
Asymmetric CH2 stretching vibration
Symmetric CH2 stretching vibration
Benzene ring
Epoxide ring
1,4-substitute benzene ring
- 157 -
Epoxide peak
900
920
Figure 6-10. Peak intensity changes of pure epoxy with different UV exposure times (the
arrow direction represents the time increases)
By using real-time FTIR, the molecular structure of the polymeric materials can
be monitored in-situ by UV exposure. The change of the functional groups (epoxide ring)
can be recorded continuously be selecting the IR wavenumber where these functional
groups exhibit their characteristic absorption bands. Figure 6-10 shows the FTIR
spectrum of the epoxy curing with different UV exposure time. The intensity of the 915
cm-1 peak, where the peak indicates the absorption of the epoxide, kept decreasing as the
UV radiation time kept increasing. The intensity of the peak at 3000 cm-1, which
indicates the absorption of the stretch vibration of C-H bond, was used as an internal
reference. Figure 6-11 shows the relationship between integrated area of epoxide peaks
versus UV exposure time for pure epoxy and epoxy nanocomposite with 30 wt% filler.
The integrated area of epoxide peaks illustrated the concentration change of the
functional group. It can be seen from this figure, the concentration of epoxide group
decreased dramatically once exposing to the UV light, then slowed down until reaching
to a constant value. The intensity of epoxide peak didnt decrease to zero even after a
long UV exposure times, indicating the incomplete reaction of epoxy monomer in the
- 158 -
photo-polymerization. The concentration of the epoxide ring in the composite sample

shows a similar change as in the pure epoxy, while the final relative intensity after photocuring is lower than that in pure polymer. Therefore, the nanosilica filler in the epoxy
would not hinder the ring-opening reaction of epoxy. On the contrary, the epoxy ring
conversion is higher in the nanocomposite sample.
In order to increase the epoxide ring conversion and improve the material
properties, the photo-cured samples were thermal cured again in the 150C for 30
minutes, and the FTIR spectrum was measured again using the same testing sample.
After thermal heating, the intensity of epoxy peak in two samples reduced, as shown in
Figure 6-11. So a proper post-curing is necessary to fully cure the epoxy sample after a
cationic photo-polymerization.
1.0
epoxy composite
pure epoxy
Normalized intensity
of epoxy peak
0.8
0.6
0.4
0.2
0.0
0
thermal
cure
photo
cure
100
200
Time (s)
300
Figure 6-11. Relationship of integrated band area of epoxide peak and UV exposure time
- 159 -
6.4.3. Two-steps curing of underfill by cationic photo-polymerization
The selection of cationic polymerizable monomers and cationic photo-initiator

now is reasonably broad; however, the fundamental characterization of these reaction
systems has received relatively little attention. In contrast, free-radical polymerizations
have been studied extensively; kinetic constants for initiation, propagation, and
termination at different stages of conversion have been characterized, including
conversion-dependent effects such as auto-acceleration. One notable difference between
cationic and radical polymerizations is the rate of termination and consequently, the
active center life time. Free radicals, which are highly reactive toward one another, have
very high termination rate constants (ca. 105 L mol-1 s-1) and have correspondingly short
active center lifetime (typically less than a second). In contrast, for cationic
polymerizations, the rate of consumption of active centers is very slow and occurs by the
reaction of the propagating chain with the counterion, solvent, or other impurities present
in the reaction systems. The average cationic active center lifetime has been found to be
on the order of tens of minutes[135]. Therefore, the cationic polymerization can be
performed continuously as dark-cure or post-polymerization after the UV light is off.
The curing of the underfill samples with cationic photo-initiator is performed in
two steps. Firstly, the sample is put into photo-DSC at a fixed temperature 30C, and
exposed to constant UV illumination. The light intensity of UV is 20 mW/cm2. The
reaction is initiated by UV energy, and the typical heat flow of epoxy cationic curing
upon UV exposure is shown in Figure 6-12. After around 20 minutes, the exothermal heat
flow approaches a constant value, which indicates a slow down of the reaction. As the
proceeding of polymerization, the polymer chains build up, the system turns to be glassy
and vitrification occurs. With the diffusion limitation, the movements of monomers are
restricted and the reaction ultimately may stop at a certain conversion. Once the glass
transition temperature of the UV-cured composite reaches the sample temperature, the
- 160 -
reaction stops because the polymerization cannot proceed in the glassy state. Therefore,
there remains a certain amount of un-reacted epoxide rings in the photo-cured sample.
Nevertheless, the active species generated by the photo-initiator after UV exposure will
still stay in the sample and keeping living.
The polymerization will continue once the proper experimental condition is
provided. So the activated sample is cured again in a thermal heating process. Figure
6-13 shows the heat flow curve of underfill sample measured by thermal-DSC during
heating. The sample is exposed to the UV light in the first stage, and the photopolymerization stopped after 20 minutes of light illumination. However, heat flow, i.e.
polymerization, occurs again on the same sample when the temperature is increased even
without adding an extra thermal initiator. At the high temperature, the partially cured
sample could move again and the polymerization can continue with the action of active
propagating center. The reaction stops until all the monomers are consumed completely
and the diffusion limitation is reached again. The reaction heat during thermal scanning
also can be calculated by integrating the heat flow cure. The total reaction heat of the
epoxy and its composites during the cationic polymerization can be obtained by
combining the reaction heat during two steps.
H = H (t )dt
Equation 6-1
H total = H photo + H thermal
Equation 6-2
- 161 -
Figure 6-12. Heat flow of underfill after UV exposure
Figure 6-13. Heat flow of UV-initiated underfill during thermal heating
- 162 -
Figure 6-14. Photo-DSC curves of the nanocomposite with different filler loading
Table 6-3. Reaction heat and conversion for the nanocomposite measured by photo-DSC
and thermal-DSC
Filler
(wt%)
0
20
30
40
Ave.
Reaction heat during

photo exposure (J/g)
79.76
110.34
122.77
129.93
Reaction heat during

thermal heating (J/g)
72.87
37.45
20.92
15.97
Total reaction
heat (J/g)
152.63
147.79
143.68
145.90
147.50
An important issue in preparing the nanocomposite materials is to determine

whether the presence of the silica filler affected the photo-polymerization kinetics, with
respect to the reaction rate and the cure extent. Therefore, photo-curing experiments are
performed using the photo-DSC with both the nanocomposite and a reference sample
containing no fillers. Figure 6-14 shows the photo-DSC curves of each sample upon UV
light exposure. The reaction heat during each polymerization process is calculated and
listed in the Table 6-3. For the composite sample, the reaction heat is normalized to the
- 163 -
pure polymer by the polymer loading. The average reaction heat for the Bisphenol F
epoxy polymer during the cationic polymerization is calculated as 147.50 J/g.
6.4.4. Reaction kinetics of underfill by photo-polymerization
According to the curing profile measured by the photo-DSC, the reaction kinetics
can be calculated. The baseline of curing peak was selected according to the ending heat
flow. The area under the heat flow curve was integrated to get the total reaction
heat H photo during photo-polymerization. The partial reaction heat H (t ) at time t was
calculated by the partial integral. Then the degree of conversion (DOC, a ) and the rate of
polymerization R p at any stage (time t) can be calculated according to the following
equations:
a=
H (t )
H total
Equation 6-3
Rp = [M ]
da
heat flow
= [M ]
dt
H total
Equation 6-4
Where [ M 0 ] is the initial concentration of the monomer.
H (t )
Figure 6-15. DSC measured heat flow in an isothermal experiment

- 164 -
Figure 6-16
and Figure 6-17 show the polymerization rate and degree of
conversion, respectively, of nanocomposite underfill during the cationic photopolymerization. The detailed kinetics data are listed in Table 6-4. For each sample with
different filler loading, the profiles of the reaction rate versus time all have a general
shape that is the characteristic of cationic photo-polymerization. Immediately upon
exposure, the cationic active centers are formed and react with the surrounding monomer
molecules to form a growing polymer chain. The reaction rate increases monotonically
because of the increase in the concentration of active centers, reaches a peak at a
conversion around 0.23, then decreases, primarily due to the decrease in the
concentration of the monomers. This agrees with Decker[136] who stated that the R p in
most photo-curing systems reaches its maximum value in the 0.1-0.3 conversion range.
So the nano size filler additions into the epoxy do not change the reaction mechanism of
cationic polymerization.
However, the filler influences the reaction kinetics. As can be seen from the
Figure 6-16, the sample without filler has the highest reaction rate, while the rate quickly
decreases to a minimum level within shortest time. As filler loading increases, the
reaction rate of photo-polymerization decreases, but the reaction time, which has a
relatively high reaction rate, is longer. In another words, although the reaction rate of the
nanocomposite is not as high as that of the pure epoxy, the photo-polymerization of
composite can keep going until a high conversion ratio is reached. The overall result is
that the conversion of the composite is higher than that of the pure resin after photo
reaction. Table 6-4 shows that none of the samples reached 100% conversion after the
photo exposure process, thus a post-curing at high temperature is needed to fully cure the
samples and maintain the long-term thermal properties, such as glass transition
temperature.
- 165 -
0.012
20
30
40
Rp (mol / l. s)
0.01
0.008
0.006
0.004
0.002
0
0
10
15
20
Time (min)
Figure 6-16. Polymerization rate versus time for the photo-polymerization of

nanocomposite underfill
Figure 6-17. Conversion versus time for the photo-polymerization of nanocomposite

underfill
- 166 -
Table 6-4. Kinetics data for the photo-polymerization of nanocomposite underfill

Filler
loading
(wt%)
0
20
30
40
Degree of
conversion after
photo reaction (a)
0.52
0.75
0.85
0.90
Maximum photopolymerization rate

(Rp, mol l-1s-1)
0.0108
0.0090
0.0084
0.0066
Time of
maximum
reaction rate(s)
59.4
78.3
88.8
107.4
Conversion
at the peak
reaction rate
0.23
0.20
0.20
0.21
Table 6-5. Light absorptivity and components concentration in the nanocomposite

underfill
Filler
loading
(wt%)
0
20
30
40
Absorption
coefficient
(L mol-1cm-1)
25,356
20,612
20,492
19,854
Monomer
concentration
(mol l-1)
3.51
3.10
2.89
2.70
Photo-initiator
concentration
(mol l-1)
0.0175
0.0165
0.0144
0.0135
The kinetics of free-radical photo-polymerization have been studied extensively

based on the fast termination mechanism[137]. The rate expression for cationic photopolymerization follows in a manner analogous to those used in radical polymerization,
except the termination mode is different. Generally, the polymerization rate ( R p ) is given
by:
R p = k p [ M ](
[ PI ] I 0 10 3
2.3k t
)1 / 2
Equation 6-5
where k p and k t are the polymerization constant and the termination constant,
respectively; [M ] is the concentration of monomer, is the quantum yield of active
center formed by per photon absorbed, is the molar absorption coefficient with unit of
L mol-1 cm-1, which is from Beer-Lambert law, [PI ] is the concentration of photo-
- 167 -
initiator, and I 0 is incident light intensity. This occurs in a very thin reaction sample with
low photo-initiator concentration, which is correct for our nanocomposite film samples.
The absorption in the UV spectroscopy is the logarithm value of the light intensity
ratio before and after penetrating the sample. According to Beer-Lamberts law, the
absorption coefficient ( ) can be calculated as the material constant. The equations are
shown below:
I
I0
Equation 6-6
%T = 100 10 A %
Equation 6-7
A = [ PI ]L
Equation 6-8
A = log
Beer-Lamberts law
where A is the absorbance; I is the light intensity of the UV after absorbance, T is the
transmittance, and L is the light path length.
Absorbance (a.u.)
0wt%
20wt%
30wt%
40wt%
300
320
340
360
380
400
Wavelength (nm)
Figure 6-18. Absorbance of underfill with different filler loading
- 168 -
For the solution sample, a unit of mol/L usually is used for the sample
concentration. The length of the light path is the film thickness of the nanocomposite,
which was measured by the laser profilometer (KLA-Tencor P15) and used with unit of
m. Figure 6-18 shows the absorbance of the photo-initiator in the nanocomposite. The
absorption coefficient ( ) is calculated from the Equation 5-9 with known photo-initiator
concentration. It can be seen that as the filler loading increases of the, the absorption
coefficient of photo-initiator decreases.
Equation 5-6 shows that R p is proportional to [ M ]( [ PI ])1 / 2 . The addition of
filler into the photo-polymerization systems can dilute the monomer and photo-initiator.
Table 6-5 lists the experimental data of these three parameters for nanocomposite
underfill with different filler loading. Clearly, as the filler loading increases, the actual
concentration of monomer and photo-initiator, as well as the absorption coefficient of
photo-initiator to the incident light, decrease. The overall effect is a decrease of
polymerization rate. Therefore, the influence of filler addition to polymerization kinetics
is an integrative effect from changing the components concentration and photo-initiator
sensitivity.
On the other hand, the high monomer conversion (Table 6-4) after photo exposure
in the composite samples can also be explained by the dilution of monomer
concentration, which has a two-fold effect. First, the nanosilica with proper surface
modification provides a significant interfacial area in the matrix, resulting in a larger heat
evolved in the photo-polymerization[138]. The high conversion of composites is due to
the increased penetration of the UV beam into the sample, resulting a high intensity of the
photon flux delivered by the light exposure[139]. Second, as we mentioned before, the
stop of photo-polymerization in the cationic system is not due to the quenching of the
active centers, but by the diffusion limitations as the polymer chain builds and the
mobility decreases. Due to the separation and dilution of filler in the composite, the
- 169 -
polymer chains keep the mobility until the curing reaction proceeds to a large extent.
These two reasons probably illuminate the high conversion ratio after photo exposure in
the nanocomposite.
The characterization of photo-reaction behaviors of nanocomposite will help to
determine the changes of the reaction kinetics and the new process conditions. For the
photo-curable underfill with high filler loading, the longer UV light exposure time is
needed because of the slow polymerization rate. The high conversion after light exposure
implies that the photo-cured nanocomposite has better material properties than the pure
polymer, which is important for the underfills performance and reliability.
6.5. Material properties characterization
6.5.1. Optical properties
The optical properties of photo-cured nanocomposites are characterized by UVvisible spectroscopy. Figure 6-19 shows the transmittance of the nanocomposite film.
After photo-curing, the absorption of the photo-initiator has disappeared due to its
photolysis reaction. The photo-cured nanocomposites with 40 wt% filler still retain
excellent optical transparency (> 95%) comparable to the pure epoxy within the visible
light region. The low transmittance under 350 nm is due to the UV absorption of benzene
ring in the epoxy polymer. At relatively longer wavelengths, the particle size of the
nanometer-order filler is much smaller than the wavelength of light, thus less effect on
the light transmission reduction is observed.
- 170 -
Transmittance (%)
100
80
60
0wt%
20wt%
30wt%
40wt%
40
20
0
300
400
500
600
Wavelength (nm)
Figure 6-19. Light transmittance of photo-cured nanocomposite with different filler
loading (particle average size: 20 nm)
Figure 6-20. Light transmittance of composite with the particle volume fraction, fp
(particle average size: 8m)[140]
- 171 -
Some research has studied the transmittance of epoxy/silica composite with large
particle size[140]. They found that the light loss by traveling of the pure epoxy and the
silica particle can be neglected, because the refractive index (RI) of the particle and
matrix in the composite is are quite similar (the RI is 1.544 for epoxy and 1.542 for silica
1.). The light transmittance is governed mainly by the light scattering on the particle.
When the radius of particle is much larger than the light wavelength, the transmittance of
composite almost decreases linearly as the filler fraction increases, as shown in Figure
6-21. The filler used in our photo-curable composite only have the size of 20 nm, which
is far less than the light wavelength. Therefore, the addition of 20 nm silica has no
influence on the optical properties. The unique nanocomposite with both good optical
transparency and material thermal stability has the potential to be used in the wafer level
application or polymeric optics.
1
Transmittance
0.8
0.6
0.4
nanosize filler
micron size filler
0.2
0
0.05
0.1
0.15
0.2
0.25
0.3
Particle volume fraction
Figure 6-21. Comparison of light transmittance and the particle volume fraction for
composite with different silica size
- 172 -
Glass transition temperature (C)
6.5.2. Glass transition temperature
100
80
60
40
20
0
0
10
20
30
40
filler loading (wt%)

Figure 6-22. DSC Tg of the nanocomposite after photo-curing followed thermal curing
Glass transition temperature (Tg) of the nanocomposite materials after photocuring and thermal post-curing are measured by DSC (shown in Figure 6-22). It can be
seen that the addition of well-dispersed nanosilica have no significant influence on the
Tg. In our previous research we found that the Tg of nanocomposites could be lowered
by more than 30C if the nanosilica had not well-dispersed in the polymer matrix and had
poor interfacial interaction to the polymer matrix. In our photo-curable samples, only
epoxy polymer was used as the composite matrix, and the anhydride hardener that has
high dielectric constant was not used. Therefore, the low polarity of the matrix provides
better environment for nanoparticle dispersion. Moreover, the surface modification of
nanosilica is also carried out successfully to achieve good filler dispersion and material
properties of the nanocomposite. For these two reasons, the Tg-depression phenomenon,
which happened in the epoxy/silica nanocomposite with anhydride hardener, did not
occur in our photo-curable samples.
- 173 -
6.5.3. Thermal degradation behavior
Figure 6-23. TGA graphs of the photo-cured nanocomposites

Table 6-6. TGA measured for various filler loadings in the nanocomposites
Sample names
10
20
30
40
Filler
theoretical
10
20
30
40
(wt%)
measured
12.0
19.3
28.3
35.5
Decompos.
1st onset
336.9
337.9
341.0
338.5
346.6
temp. (C)
2nd onset
545.2
551.9
556.0
557.3
576.5
Polymers with good thermal stability are required for microelectronic packaging
applications. Figure 6-23 shows the TGA curves obtained. There is an ~4 wt% weight
decrease in the TGA curves for all the samples starting from 160C that is due to the
evaporation of photo-initiator solvent. The photo-initiator used in the experiments was
dissolved into propylene carbonate solvent which has a high boiling point of 243C.
After this small weight loss, there are two major decomposition steps starting from
around 300C and 550C, respectively. Table 6-6 shows the onset temperatures of
decomposition for each sample. It can be seen that the addition of nanosilica can improve
- 174 -
the thermal stability of pure epoxy. Due to the presence of filler, the nanocomposites
exhibit a larger amount of char formation than the pure epoxy samples. The exact filler
loading, defined as the residue in the end of TGA curve, is also listed in the Table 6-6.
6.5.4. Thermal expansion
To characterize the thermal-mechanical properties of the epoxy/silica

nanocomposite after photo-curing, the thick films were prepared by the bar-coating
method with a doctor blade on an aluminum substrate. Free-standing film of
nanocomposite is obtained by peeling films from the Al substrates. The thermal
mechanical properties are characterized using DMA and TMA in the film mode.
80
thermal curing
photo curing
CTE (m/mC)
70
60
50
40
0
10
20
30
40
Figure 6-24. Coefficient of thermal expansion of the nanocomposite with various filler
loading
The primary purpose of adding silica to the epoxy is to reduce the thermal
expansion of the epoxy[141, 142]. Generally, the coefficients of thermal expansion
(CTE) of composites are reduced with an increase of filler contents. Figure 6-24 shows
the CTE values of the nanocomposites after photo-curing. The CTE was reduced from
- 175 -
76.1 m/mC for pure epoxy to 50.2 m/mC after 40 wt% silica addition. It is expected
that with increased amounts of filler, the CTE of the composite is further reduced. The
CTE of the thermally cured sample is also shown in the figure. There is no obvious
difference between the different polymerization methods in term of material thermal
stability. The incorporation of the modified nanosilica reduces the CTE and provides
good dimensional stability for the composite films in this application.
6.5.5. Thermal mechanical properties of photo-cured nanocomposites
Figure 6-25. DMA curves of the photo-cured nanocomposites

Figure 6-25 shows the overlay of the storage modulus curves of the
nanocomposites. With the increasing filler loading, the modulus increases almost linearly
with the addition of inorganic fillers in the epoxy matrix. The DMA Tg, which is
represented by the peak temperature of the tan delta curve, is also listed in Figure 6-26.
The Tg of the nanocomposite measured by peak temperature of tan delta show the same
trend as observed in the DSC experiment, which do not change significant as filler
loading increased.
- 176 -
100
DMA Tg
80
60
40
20
0
0
10
20
30
40
50

Figure 6-26. Tan delta peak temperature (DMA Tg) of the photo-cured nanocomposites
The Mori-Tanaka method[143-145] has been used to predict the elastic properties
of two-phase composites as a function of the effective particle volume.
Then, the
effective Youngs modulus of the composite can be derived as[146]:

E = 2 [1 +
3K 2
]
2(3K + )
Equation 6-9
Where
K = K 0 {1 +
c( K1 K 0 )
}
K 0 + 3 0 (1 c)( K 1 K 0 )
Equation 6-10
= 0 {1 +
c ( 1 0 )
}
0 + 2 0 (1 c)( 1 0 )
Equation 6-11
0 =
K0
3K 0 + 4 0
Equation 6-12
0 =
3( K 0 + 2 0 )
15 K 0 + 20 0
Equation 6-13
En
3(1 2 n )
Equation 6-14
Kn =
- 177 -
n =
En
2(1 + n )
, n = 0, 1
Equation 6-15
Where E = effective bulk modulus, = effective shear modulus, and c is the filler
volume fraction. Youngs modulus and the Possions ratio of the matrix are E 0 and 0 ,
that of the particles are E1 and 1 .
The above equations show that the modulus of the composite underfill is
determined by the moduli of the epoxy matrix and the filler particles, as well as the
particle volume fraction. Theoretically, once these parameters are given, these equations
can be used as a tool to estimate the modulus of the composite underfill. Based on the
materials properties (Table 6-7), the predicted modulus of composite from Mori-Tanaka
method was compared with the measured results (Figure 6-27).
Previous research had validated the Mori-Tanaka model for the composite filled
with the micron size filler in which the agreement between the theoretically predicted and
experimentally measured modulus was excellent[88]. Nevertheless, the prediction
showed a large deviation from the experimental results of our study, as shown in Figure
13, which implied the limitation of Mori-Tenaha model for the nanocomposites. This
model assumes that only two phases exist (matrix and filler) and that they are perfectly
bonded to each other. This assumption may work well for the reinforcements of a
polymer matrix with the micron size filler, or higher. However, for nanocomposites, it
has been shown that the molecular structure of the polymer matrix is perturbed
significantly at the filler/polymer interface, and this may create a third phase,
interphase[147]. Therefore, the reinforcement of nanocomposite is not accurately
described as consisting of just two phases, thus the Mori-Tanaka model is not expected to
perform well for a nanocomposite. More research needs to be done based on the effective
interface model of the nanocomposite structure[145].
- 178 -
Table 6-7. Materials Constant
epoxy
Density
0 =1.16 g/cm3
Modulus (GPa)
E 0 =2.5GPa
Poissons ratio
0 =0.4
silica
1 =2.20 g/cm3
E1 =74GPa
1 =0.19
Exp.
Model
8
7
Modulus (GPa)
6
5
4
3
2
1
0
0
0.1
0.2
0.3
0.4
0.5
Filler volume fraction
Figure 6-27. Comparison of composite modulus between the theoretical prediction and
experimental measurement
6.5.6. Nanocomposite morphology
The TEM was used to provide the morphological information on a nanometer

scale for the photo-cured composite film. The obtained TEM image for nanocomposite
with 30 wt% filler loading is shown in Figure 6-28. Note that the nanosilica is well
dispersed in the polymer matrix after photo-curing. No obvious particle agglomerations
or clusters were observed, which indicates good compatibility between the nanosilica
surface and the polymer matrix after particle surface modification. Therefore, those
- 179 -
problems associated with filler agglomerations were eliminated. In the beginning of this
chapter, we showed the morphology of silica colloidal particles before adding into the
matrix (Figure 6-5). In that picture, the particle distribution is not as narrow as that
dispersed in the epoxy matrix.
Figure 6-28. TEM picture of nanocomposite after photo-curing

6.5.7. Surface hardness
The nanoindentation experiment is commonly used to characterize the surface

properties of the material with nanoscale spatial resolution. A hard diamond tip with
known modulus is pressed into a sample surface. The load placed on the indenter tip is
increased as the tip penetrates further into the specimen and soon reaches a value. At this
point, the load is held constant for a period and then removed to avoid viscoelastic effects
during the nanoindentation of a polymer. Figure 6-29 shows a plot of load versus
displacement during a nanoindentation testing. The area of the residual indentation in the
sample surface is measured and the hardness is defined as the maximum load divided by
the residual indentation area. Figure 6-30 shows the hardness and surface modulus of
- 180 -
nanocomposite with different filler loading after photo-curing. The hardness of the
nanocomposite materials increased linearly with the increasing of the filler loading.
8
7
Load (mN)
6
5
4
3
2
1
0
0
200
400
600
800
1000
1200
Displacement into surface (nm)
Figure 6-29. A plot of load vs. displacement in a nanoindentation experiment

0.5
Hardness (GPa)
0.45
0.4
0.35
0.3
0
10
20
30
40
50
Figure 6-30. Hardness of nanocomposite films after photo-curing
- 181 -
6.6. Application of photo-curable epoxy nanocomposite in wafer level packaging

6.6.1. Novel wafer level packaging process
With the fast development of the electronics industry, electronic products are
becoming smaller and faster with more functionality, higher performance, and lower cost.
The technical advances have entered into the nano-scale threshold with the decreasing of
feature size and increasing I/O number in the IC chip. The demand for ever-shrinking,
performance-rich portable devices, at competitive prices, is one of the factors driving the
development of reliable electronics packaging methods. Wafer level packaging (WLP) is
a packaging technology where most or all of the packaging process steps are carried out
at the wafer level[26]. WLP can increase the silicon efficiency, reduce the packaging size
and achieve true chip scale packaging. It can also solve the Know Good Die issue
and one can package only the good die because of the wafer level packaging and burn-in
test. By the batch process, WLP can reduce significantly the cost per device. However,
like any other new technologies, the WLP technology still faces a number of challenges.
Currently the WLP technology mainly focuses on small die and low I/O devices in
consumer products; while for large die and high I/O devices in high performance
applications, the WLP has been limited due to the poor board level reliability. This is
because of the great strain in the solder bump and on the underlying surface generated by
the large bump distance from neutral point (DNP). To overcome this problem, an
encapsulate material, which usually is the epoxy/silica composite with low coefficient of
thermal expansion (CTE), is needed to protect the redistribution dielectric layer and
solder bumps on the wafer. Besides the stencil printing approach used for low I/O
devices, several new methods have been invented for high I/O devices to apply the
encapsulated layer. Fraunhofer IZM/Technical University of Berlin disclosed the double
ball method [148], where screen-printed solder balls can be covered with CTE matched
"underfill" and then this polymer layer was mechanically polished out to expose the
- 182 -
solder balls; after that, a second solder ball can be placed at these sites (as shown in
Figure 6-31). Aguila Tech[149, 150] used a high-performance resin applied on the wafer.
Then the microvias were formed by the laser ablation on the fully crosslinked resin layer
for solder bumping.
Figure 6-31. Double ball redistribution uses two solder balls for each I/O, one being
encapsulated in epoxy. (Source: Fraunhofer IZM/Technical University of Berlin)
Figure 6-32. Wafer level process with laser ablation method to open the microvia on
underfill
- 183 -
There are several problems with current methods in applying the encapsulant
layer on a wafer:
(1) Stencil printing method is not suitable for high I/O density wafer.
(2) Mechanical polishing will create dust during process, may cause delamination
of the underlayers and damage the wafer due to mechanical force.
(3) In the laser ablation method, it is difficult to control the shape and depth of
microvia, and can as well leave the polymer residue in the microvias.
(4) The thermal curing of thick encapsulated polymer layer on the wafer could
create thermal stress and wafer warpage, which become more serious for 300 mm wafer
or even larger one.
(5) All the current encapsulated materials need to be stored in frozen condition
(usually -40C) because the thermal catalyst added in the formulations can slowly initiate
the reaction even at room temperature. This increases the cost for shipping and storage of
the materials.
Here, we report on the development of a novel photo-curable material that can act
both as a photoresist and as a stress redistribution layer applied on the wafer level. In the
proposed process, this material is applied on the unbumped wafer, and then is exposed to
UV light through a mask for crosslinking. After development, the unexposed material is
removed and the bump pads on the wafer are exposed for solder bumping. The fully
cured material film is left on the wafer for protection of the bumps. Figure 6-33 shows
the proposed process.
- 184 -
Wafer w/o bumps

Apply photo-curable
underfill
Mask
Fully photo-curing
underfill
Development
Solder bump
Wafer dicing and reflow
Figure 6-33. Proposed wafer process with novel photo-curable nanocomposite

6.6.2. Advantages of photo-curable nanocomposites
The key for the new packaging concept is the development of a new photocurable material, which should not only have photo-reactivity and wafer-level processing
ability like a photoresist, but also the excellent material properties such as low CTE, high
mechanical strength and modulus, and high resistance to chemicals. The epoxy
nanocomposite developed in this chapter is the most suitable material for this process.
From the material point of view, the photo-polymerization of the encapsulant polymer is
more energy and time efficient compared to the thermal reaction. The room temperature
reaction of photo-curable materials also can avoid the distortion, thermal stress, and
warpage usually encountered in the thermal curing process, which will benefit the large
wafer (300 mm or above) in the industry. Compared to other encapsulant materials in the
electronics packaging which require special storage environments (most in -40C), this
photo-curable material is quite stable at room temperature without light exposure, and
does not need frozen conditions for shipping and storage.
- 185 -
From the process point of view, this novel material can act both as a negative tone
photoresist to form openings for solder bumping by photolithography method and also as
a stress redistribution layer permanently applied on the wafer level to improve the
packaging reliability. It is easy to form the microvias for solder bump by the standard
photolithography method but not by the mechanical polishing method or laser ablation
method. Compared to the previous methods mentioned, this material provides potential
for new wafer level packaging with easy processing. It does not require expensive
equipment such as laser machine. Application of photo-curable materials can be
combined together with the solder mask formation and will not add extra cost to industry.
6.6.3. Pattern formation with photo-curable nanocomposite
In order to apply the photo-curable nanocomposite on the wafer level for it to be

compatible to the photolithography process, the Bisphenol-F epoxy matrix for the
nanocomposite was replaced by the SU-8 epoxy. SU-8 is an epoxy-type, negative, near
UV 9365nm) photoresist that has high optical transparency and is well suited for
application where it is imaged and permanently left in place. The chemical structure of
SU-8 is shown in Figure 6-34. Because of the high functionality, SU-8 can achieve a high
degree of crosslinking. The photo-initiator is the triarylsulfonium KI85 (shown in Figure
6-1), with a fixed concentration of 10wt%. The gamma-butyrolactone (GBL) and
propylene glycol monomethylene acetate (PGMEA) were used as the solvent and
developer, respectively. The 20 nm silica colloids were modified and added into SU-8
solution following a similar procedure as described in 5.2.2.
To explore the possibility of using this novel photo-curable nanocomposite as
photoresist, a preliminary photolithography experiment was performed. The glass slide
was cleaned and used as the pattern substrate. Then the SU-8 nanocomposite solution was
spin-coated onto the substrate. The testing coupon was prebaked at 95C for 30 minutes
to evaporate GBL solvent. After drying the solvent, a non-tacky film formed on the glass
- 186 -
substrate. The back side of SU-8 composite-coated glass was exposed to UV light
through a quartz mask by a UV mask aligner (MA-6, from Karl Suss). The exposure was
at I line, with a wavelength of 365 nm.
As known from previous study, photo-
polymerization of the nanocomposite is slow; therefore a total exposure dose of 6000

mJ/cm2 was used to ensure the photoreaction. The exposed nanocomposite film was then
post-baked at 120C for 30 minutes in order to complete the acid-catalyzed cationic
polymerization. Then the patterned structure was developed in PGMEA solvent to
remove unexposed composite photoresist. Ultrasonication for a few minutes was required
to obtain good and clear pattern structure. Isopropyl alcohol (IPA) was used to rinse the
developed pattern. The flow chart of pattern fabrication is shown in Figure 6-35.
Figure 6-36 shows the preliminary result, a pattern formed by photolithography
method. The epoxy/silica nanocomposite with 40 wt% filler loading can act as a
photoresist. Round-shape openings with 200 m diameter are formed in the composite
film. The resolution for this composite photoresist is around hundred microns, which is
compatible to the current pitch-size of flip-chip fabrication. This result demonstrates the
possibility of using nanocomposite material as photoresist, as well the potential to
develop a novel wafer level process with the nanocomposite materials cured by the
cationic photo-polymerization.
- 187 -
SU8
O
O CH2 CH CH2
O
O CH2 CH CH2
CH2
H3C
CH3
O
O CH2 CH CH2
CH2
H3C
CH3
O
O CH2 CH CH2
O
O CH2 CH CH2
CH2
H3C
CH3
CH3
H3C
O CH2 CH CH2
O CH2 CH CH2
CH3
O
O CH2 CH CH2
gamma-butyrolactone
O
O
PGMEA
O
O
O
Figure 6-34. Molecular structure of SU-8, gamma-butyrolactone and propylene glycol

monomethylether acetate (PGMEA)
- 188 -
Figure 6-35. Flow chart of photolithography process for SU-8 nanocomposite
Figure 6-36. Photo-defined pattern of SU-8 nanocomposite containing 40 wt% nanosilica
- 189 -
CHAPTER 7.
CONCLUSIONS AND SUGGESTED WORK
7.1. Conclusions
Conventional flip-chip underfill is necessary for a reliable flip-chip on organic

package but is process-unfriendly and incompatible to a high production flip-chip
assembly. Many variations of the underfill have been invented to address the problem,
among which, the newly developed no-flow underfill and wafer-level underfill have
attracted much attention. No-flow underfill and wafer-level underfill can eliminate the
tedious process of the conventional underfill by combining the solder melting and
underfill solidification together. Nevertheless, the incorporation of fillers into no-flow
underfill and wafer level underfill become the bottleneck for their wide application.
Without filler, the underfill materials do not process enough thermal mechanical
properties to achieve good reliability in flip-chip packaging. However, the traditional
micron size filler in the underfill could be trapped between the solder bumps and
substrate pads because these underfills are applied first before the chip assembly and
solder joint formation. Nano size filler, on the contrary, has no influence of the solder
melting due to its extremely small size. Therefore, nanocomposite underfill is of great
interest, and the size reduction of filler should be studied carefully in order to understand
fully the influence of nanocomposite on the flip-chip packaging.
The nano size filler used in the underfill can be synthesized by the Stber method
from sol-gel reaction. Experiments show the silica with around 100 nanometer size and
good size distribution can be obtained by control of the basic catalyst concentration
during the reaction. With the large scale production, the cost will be reduced to an
accepted range for underfill materials. Due to the small diameter, nanosilica does not
form clog during the underfill flow between the gaps, has less filler settling, does not
interfere with the solder joint formation during wetting, and does not change the light
- 190 -
transparency of underfill either. Underfill is a non-aqueous colloidal dispersion system

composed with inorganic silica filler and liquid organic phase, including epoxy
monomer, hardener, catalyst, and other additives. The dispersion of nanosilica is much
more difficult than micron silica, because the surface of nanosilica synthesized by sol-gel
method is covered with polar silanol groups, and the particle-particle interactions are
strong. The formed particle agglomerations due to the strong attractive force between
filler not only increase the viscosity of underfill, but also reduce the maximum filler
loading in the epoxy matrix.
In order to decrease the viscosity of underfill and to increase the extent of filler
loading, it is necessary to reduce the degree of agglomeration and improve the wetting
ability of underfill liquid to silica surface. A silane coupling agent can modify the silica
surface and change the surface tension of the solid surface. Due to the difficulty of
measuring the surface properties and contact angle on the SiO2 particle, the SiO2 glass
surface was used to mimic the particle and characterize the effect of silane functional
groups. The critical surface tension( c ) of SiO2 after modification can be calculated by
the Zisman method and used to estimate the wetting of liquid on the solid surface.
Among the four types of silane used, the SiO2 surface treated by polar functional silane
had a higher c than that with non-polar functional silane. The surface tension of epoxy
monomer was lower than the SiO2 surface treated with epoxide functional silane, i.e.,
epoxy monomer can spread on this surface.
The optimal modification condition for silica particle modification was obtained
through a design of experiments, including the variables of silane chemistry, silane
concentration, treatment time, and pre-dispersing method. The results showed that the
nanosilica treatment by epoxide silane with a longer reaction time and lower silane
concentration assisted with pretreatment by sonication can achieve mono-dispersed silica
in a non-aqueous medium, as confirmed by particle size distribution and TEM pictures.
- 191 -
By the optimal modification condition, treated nanosilica can be added into epoxy matrix
in a high loading level, and a dispensable low viscosity can be achieved for a
nanocomposite underfill. The solder wetting experiments also confirmed that the
nanocomposite underfill did not interfere with the solder interconnection formation
during a no-flow process. Nanocomposite underfill is the most promising way to solve
the filler addition issue in no-flow and wafer level process.
The physical, optical, and thermal mechanical properties of nanocomposite
underfill were characterized after thermal curing. The mechanical properties and
dimensional stability of underfill can be increased linearly with filler. Although the
nanosilica can benefit the solder wetting and underfill transparency due to the small size,
it also had some negative effects on the underfill materials due to large surface areas and
interfacial interactions, including reducing the composite Tg, inhibiting the epoxy curing,
extremely high viscosity at high loading level, high moisture absorption, and low density.
In order to describe the property behaviors of nanocomposite, a three-phase model needs
to be developed regarding the interphase region between filler and polymer matrix. The
relaxation and dynamic of polymer chains in the interphase were different from the bulk
polymer, and so contribute to some unexpected materials properties of nanocomposite.
The glass transition temperatures of epoxy composites with micron- and nano size
fillers including silica, silver, aluminum, and carbon black were investigated using a
DSC. The Tgs of the nanocomposites showed a significant reduction compared to their
counter-parts with micron size fillers. From the results of the carbon black composites,
we found that the increased interfacial area between the fillers and the resin reduced Tg of
the composites. Thermal characterizations of the micron silica and nanosilica showed that
there was adsorbed water and bonded organics at the surface of the nano-silica, which
assisted the polymer relaxation process at the filler/resin interface in the nano-silica
composites. The dynamic mechanical properties of the silica composites were
characterized. The results showed that the nanosilica reduced the Tg of the composites but
- 192 -
did not influence the transition temperature. An extra relaxation in the nanocomposite
was found to occur at sub-Tg transition temperature and related to the different molecular
chain dynamics in the interphase. It can be concluded that the surface chemistry of the
nano-fillers and the interaction at the filler/resin interface are essential to determine the
Tg of the nanocomposites.
The dielectric property of epoxy/silica nanocomposite was studied with a focus on
the ionic contribution and the water influence to the loss factor. Experimental
measurements were conducted and showed that the nanocomposite had a higher dielectric
loss at low frequency due to enhanced ionic conductivity caused by the contaminants
from the sol-gel synthesized nanosilica. The relaxation temperature of the nanocomposite
was lower than those of the micron-composite and the control samples due to the extra
free volume at the filler-resin interface that assists the polymer mobility. The moisture
had different effects on the pure epoxy and epoxy composite in dielectric loss that can be
explained by the combined effect of ionic conductivity and interfacial interaction in
materials.
While the compatibility between nanosilica and the epoxy matrix was greatly
enhanced by silane modification of the nanosilica surface and drawbacks caused by the
incompatible interface between nanosilica and the epoxy matrix can be reduced, it could
not solve completely the depression of glass transition temperature for nanocomposite
with anhydride type hardener. In other words, an alternative solution should be found to
modify the filler surface and change the surface tension. By replacing hexahydro-4methylphthalic anhydride with diethyl toluene diamine, one can use a hardener with a
low dielectric constant in the underfill. Nano size filler can form different microstructure
with varied types of hardeners. In a non-aqueous colloidal dispersion system such as
underfill, the stability of filler could not be explained by the traditional DLVO theory
with electrostatic repulsion. The solvation force between filler and liquid medium, which
means stronger interaction between filler and liquid, is the major force to balance the Van
- 193 -
der Waals attraction between filler and filler. With an amine-type hardener, less attractive
force and more repulsive force can form in the solid-liquid dispersion. Therefore, a silicaepoxy composite based on an amine hardener show sol microstructure and good
macroscopic properties in term of glass transition temperature.
Another focus of this research is on the photo-polymerization of nanocomposite
underfill. The nanosilica show excellent optical transparency and therefore was chosen as
the filler to reinforce the photo-curable epoxy and reduce its thermal expansion. The UVvisible absorption of photo-initiator in the epoxy was not blocked by a nanosilica
addition. The photo-curing process of nanocomposites was studied by the real-time FTIR
and photo-DSC. It was found that the functional peak of the epoxy monomer did not
disappear completely after UV exposure, which required a thermally post-curing to
increase the conversion. The study of curing kinetics showed the polymerization rate of
the pure epoxy was faster than that of the composite sample during a cationic photopolymerization, because the monomer was diluted by the filler in the composite.
However, the degree of conversion after photo-curing was increased as the filler loading
increased, because the diffusion limits of polymerization occurred at high degree of
curing for a nanocomposite sample. By the fundamental study on photo-polymerization
of the nanocomposites, the reaction mechanism and polymerization kinetics were
investigated.
This study achieved a better understanding of nanosized filler on the optical
application, and provided guidelines to the design and process of photo-curable
nanocomposite materials for microelectronic packaging technology. A preliminary
photolithography process was developed with the nanocomposite photoresist. With ultrafine size of filler, the photo-curable composite still can be used to form patterns with a
hundred micron resolution. The nanoscale reinforcement to the polymer will have a deep
- 194 -
impact on the photoresist material. Previously, the photoresist was removed after pattern
formation and could not be used as structural material due to poor mechanical and
thermal mechanical properties. With the reinforcement of nanosilica, the photo-curable
nanocomposites have enough material strength and dimensional stability that they can be
used as a permanent layer to achieve different functions, which is very useful in many
fields such as electronics packaging or MEMS fabrication.
7.2. Suggested work
7.2.1. Chemical bond between filler and epoxy matrix
Underfill is a composite material with heterogeneous structure. The interactions

between two phases play an important role for the underfill viscosity, materials thermal
properties, and mechanical properties. For the nanocomposite underfill, with the total
filler surface increasing, the surface problems become more serious due to the ultra large
surface area. In the previous research, the silane coupling agent was used to modify the
filler surface and improve the wetting of epoxy matrix on the filler, as well as the
compatibility between nanosized filler and epoxy matrix. However, the modification by
changing the surface tension of filler is still limited. A more strong interaction between
filler and polymer can be achieved by the chemical bonds formation.
In the suggested work, the catalyst for epoxy polymerization can be grafted on the
silica surface with imidazole-containing silane. Imidazole is known as a very active
curing catalyst for epoxy systems. To improve the curing latency, further reaction with
the alkyl salt will convert the imidazole group into the imidazolium salt to block the
reactive nitrogen. The whole reaction process can be described in Figure 7-1. After the
incorporation of functionalized nanosilica in the epoxy, the imidazolium salt on the silica
surface will decompose at high temperature during the reflow process[151]. The resultant
active imidazole group will continue to react with epoxy molecule at the 1-N position to
- 195 -
form an adduct that contains a highly reactive alkoxide ion. This alkoxide ion will initiate
rapid anionic polymerization of epoxy resin and form crosslink structure. Figure 7-2
shows the process of epoxy polymerization. The curing temperature of epoxy is
dependent on the thermal stability of the imidazolium salt, which can be adjusted by the
counter ion and the imidazole structure. In this way, it is possible to design the chemistry
of the grafted catalyst for different solder (eutectic tin/lead and lead-free) reflow process
and to achieve complete cure after reflow. The polymerization of the epoxy starts from
the nanosilica surface so that the increases interface area will not inhibit the curing of the
epoxy resin. On the other hand, the inorganic and organic phases are integrated together
to form a network, which will improve the filler-polymer interphase property and
enhance the material properties of the composite.
SiO2
+ RX
OR
OH
OH + RO Si
OH
OR
SiO2
N+ R
X-
OH OH
O Si
OH
OH
OH
SiO2
OH
O Si
OH
OH
R: alkyl group
X- : Cl-, Br-, BF4-, PF6-
Figure 7-1. Silica surface grafting of imidazolium salt as a catalyst

OH
SiO2
OH
OH
OH
N+ R
X- -RX
O Si
OH
SiO2
O
O
OH
SiO2
OH
O-
OH
N
O Si
OH
O Si
OH
OH
N+
OH
Figure 7-2. Surface initiation of epoxy curing reaction.
- 196 -
7.2.2. Molecular level reinforcement in epoxy
Our above approach to improve the filler-polymer matrix compatibility is focused

on the (1) inorganic filler surface modification and (2) low polar epoxy matrix usage.
However, the filler still represents a difference phase from the organic matrix, and the
filler-matrix interfacial area cannot be eliminated completely. Here, we propose to
synthesize the novel epoxy grafted with silicon dioxide structure in the molecular chain
In other words, the inorganic phase for the purpose of reinforcement will be chemically
connected to the organic structure and achieve molecular level reinforcement to the
epoxy matrix, therefore the interphase region can be removed effectively, and so can the
problems associated with the poor interfacial interaction between different phases.
R
O
Si
R
O
Si
O
Si
Si
R
O
O Si
Si
Si
O
O
Si
R
T8: R= H
Q8: R= OSiH(CH3)2
Figure 7-3. Poyhedral oligosilsesquioxane (POSS) structure

This approach will be achieved by synthesis of the nano inorganic-organic
hybrids
with
polyhedral
oligosilsesquioxane
(POSS)
structure.
Octahydridosilsesquioxane (T8)[152] and octa(dimethylsiloxy) silsesquioxane (Q8)[153]

are two easily synthesized and widely used POSS structures, as shown in Figure 7-3.
They have rigid structures with a core diameter of 0.5 nm and can have a wide variety of
functional groups attached to each vertex of the core. POSS (T8 and Q8) provides a silica
- 197 -
model with a size less than 1 nm and can be further functionalized. It has been reported
that high Tg and low CTE composite could be achieved from epoxide and amine
substituted POSS derivatives[154].
In the suggested work, the functionalized POSS derivatives will be synthesized
and nano organic/inorganic hybrids will be formulated in-situ for underfill applications.
First, POSS structures will be synthesized according to literature methods, and then the
different functional groups, such as epoxide rings, anhydrides, phenols, amines, organic
acids or even latent catalytic groups, can be connected easily to POSS molecules via
hydrosilylation reactions to a series of alkenes. Each component in an underfill, including
epoxy, hardener, flux and catalyst can be represented as different functional groups. The
POSS core acts as molecular level silica for the organic matrix. Furthermore, a onecomponent underfill formulation can be designed by attaching epoxy, hardener and
catalyst groups into a single POSS core via multiple hydrosilylation steps at different
temperatures (Figure 7-4). Therefore, one easily and flexibly can manipulate hybrid
structure and achieve underfill properties within one molecule. Compared with
filler/epoxy composites, these POSS-containing molecules can form homogenous
structure after crosslinking of epoxide rings and also can eliminate the incompatibility
between filler and polymer. Another benefit for these nano inorganic-organic hybrids
may come from the excellent optical transmission due to the < 1 nm size of POSS
structure.
- 198 -
O Si
O
Si
O
O
O Si
Si
O
O
O
Si
Si
O
Si
O O
Si
O Si
O
Si
O
O
(OSi(CH3)2H)8
T1
R2
T2
O Si
O
Si
O
O
O Si
Si
O
O Si
Si
O
O
O
Si
O
Si
(OSi(CH3)2CH2CH2R1)n
Si
O
Si
O O
Si
O
(OSi(CH3)2H)8-n
(OSi(CH3)2CH2CH2R1)n
Si
O
Si
O O
Si
O
(OSi(CH3)2CH2CH2R2)8-n
R1: anhydride, phenolic, amine or organc acid

R2: epoxide
Figure 7-4. Synthesis route of POSS-containing underfill
There are many advantages to introducing a POSS structure into the underfill. A
wide variety of POSS derivatives can be synthesized due to the large number of candidate
functional groups. It is easy to control the number of functional groups on the POSS
segment, and therefore, to manipulate the crosslinking density and mechanical properties
of the POSS/epoxy hybrid. With the chemical linkage between the POSS and polymer
molecular chain, the phase separation will not occur. The POSS segment inside the
material may act as the stress-buffer structure to achieve low-stress underfill. This will
be desirable for lead-free application. Although the cost of POSS is higher than normal
silica filler, it could be within the acceptable range for the underfill material that is
targeted to the microelectronic market.
7.2.3. High performance polymer matrix
Currently epoxy resins are by far the largest segment of addition-cured

thermosetting polymers that are used to make underfill materials in the electronics
industry. Epoxies owe their wide range usage to their ease of handling and processability:
- 199 -
they can be cured at a reasonably low temperature. However, the epoxies have limited
application in some high performance requirements. Their upper temperature range for
structural performance is approximately 180C, while exposure to moist environments
limits this to 150C. For the underfill used in the electronics packaging, they may
experience several thermal reflow processes due to the multiple components assembly.
During the reflow, a short time exposure to high temperature up to 270C will impact
thereafter the performance of underfill. In circumstances where epoxies cannot be used,
some higher temperature performance polymers are needed; cyanate ester resin is one of
the alternatives. The study of the formulation, curing kinetics, and material properties of
cyanate ester is of interest in order to establish the new chemistry of underfill materials.
X
N
N
C
C
O
X: isopropylidenyl moiety or aromatic or cycloaliphatic backbone
Figure 7-5. Chemical structure of cyanate ester monomer

7.2.4. Nanocomposite polymeric optical waveguide
With the unique optical properties, the nanosilica can be used in the optical
application. The photo-curable SU-8 epoxy processes excellent optical performance and
functionality, and it is one of the most promising candidates to replace other optical
materials such as semiconductor, glass-based materials or some inorganic crystals.
Nevertheless, the reflective index of pure polymer is fixed, which is not as flexible as
optical glass for different applications. With the silica filler addition, the RI of composite
becomes tunable. We can try experiments that add filler to change the intrinsic reflective
index of pure polymer. The mechanical properties of polymer also can be improved with
inorganic filler.
- 200 -
APPENDIX A
AUTHORS AWARDS, PATENTS, AND PUBLICAITONS
A.1 Awards
[1] Advanced Publication Award, School of Materials Science and Engineering,

Georgia Institute of Technology, Feb. 2006
[2] Finalists for the 13th Motorola-IEEE/CMPT Fellowship at 55 Electronic
Component & Technology Conference, Orlando, FL, May 2005
[3] The 3rd place of student poster on the Industrial Advisory Board Meeting at
Packaging Research Center, Georgia Institute of Technology, Feb. 2003
A.2 Patents
[1] C. P. Wong, Yangyang Sun, Fei Xiao, Gusuel Yun, Kyoung-sik Moon, "
Protection of Wafer Level Packaging Devices from Electrostatic Discharge Events by
Using Novel Composite Materials", Provisional patent, Invention disclosure 3640,
[2] Zhuqing Zhang, Lingbo Zhu, Yangyang Sun, Jianwen Xu, Hongjin Jiang, C.P.
Wong, "Aligned Carbon Nanotube for Electrical and Thermal Interconnect", Provisional
patent, Invention disclosure 3302, Georgia Institute of Technology
A.3 Journal Publications
[1]
Y. Sun and C. P. Wong, "Flip-Chip Underfill for Advanced Packaging",
Encyclopedia of Smart Materials. New York: John Wiley & Sons, in press, 2007.
- 201 -
[2]
F. Xiao, Y. Sun, Y. Xiu, and C. P. Wong, "Preparation, Thermal and
Mechanical Properties of POSS Epoxy Hybrid Composites," JOURNAL OF APPLIED

POLYMER SCIENCE, accepted, 2006.
[3]
L. B. Zhu, J. W. Xu, Y. H. Xiu, Y. Y. Sun, D. W. Hess, and C. P. Wong,
"Electrowetting of aligned carbon nanotube films," Journal of Physical Chemistry B, vol.

110, pp. 15945-15950, 2006.
[4]
L. Zhu, J. Xu, Y. Xiu, Y. Sun, D. W. Hess, and C. P. Wong, "Growth and
electrical characterization of high-aspect-ratio carbon nanotube arrays," Carbon, vol. 44,

pp. 253-258, 2006.
[5]
L. Zhu, Y. Sun, D. W. Hess, and C. P. Wong, "Well-aligned Open-ended
Carbon Nanotube Architectures: A Novel Approach for Device Assembly," Nano Letters,
vol. 6, pp. 243-247, 2006.
[6]
Y. Y. Sun, Z. Q. Zhang, and C. P. Wong, "Study and characterization on
the nanocomposite underfill for flip-chip applications," IEEE Transactions on

Components and Packaging Technologies, vol. 29, pp. 190-197, 2006.
[7]
Y. Sun, Z. Zhang, and C. P. Wong, "A novel nanocomposite with photo-
polymerization for wafer level application," IEEE Transactions on Components and

Packaging Technologies, accepted, 2006.
[8]
H. Jiang, K.-s. Moon, Y. Sun, and C. P. Wong, "Tin/Indium Nanobundle
Formation from Aggregation or Growth of Nanoparticles," Journal of Nanoparticle

Research, accepted, 2006.
[9]
Y. Sun, Z. Zhang, and C. P. Wong, "Study on mono-dispersed nano-size
silica by surface modification for underfill applications," Journal of Colloid and Interface
Science, vol. 292, pp. 436-444, 2005.
[10]
Y. Sun, Z. Zhang, and C. P. Wong, "Rheology Study of Wafer Level
Underfill," Macromolecular Materials and Engineering, vol. 290, pp. 1204-1212, 2005.
- 202 -
[11]
Y. Sun, Z. Zhang, and C. Wong, "Influence of interphase and moisture on
the dielectric spectroscopy of epoxy/silica composites," POLYMER, vol. 46, pp. 22972305, 2005.
[12]
Z. Zhang, Y. Sun, L. Fan, and C. Wong, "Study on B-stage properties of
wafer level underfills," JOURNAL OF ADHESION SCIENCE AND TECHNOLOGY,

vol. 18, pp. 361-380, 2004.
[13]
Y. Sun, Z. Zhang, and C. Wong, "Development of a high curing latency
no-flow underfill with self-fluxing ability for lead-free solder interconnects," JOURNAL
OF ADHESION SCIENCE AND TECHNOLOGY, vol. 18, pp. 109-121, 2004.
[14]
Y. Sun, Z. Zhang, K. Moon, and C. Wong, "Glass transition and relaxation
behavior of epoxy nanocomposites," JOURNAL OF POLYMER SCIENCE PART BPOLYMER PHYSICS, vol. 42, pp. 3849-3858, 2004.
[15]
Y. Sun, S. Luo, K. Watkins, and C. Wong, "Electrical approach to monitor
the thermal oxidation aging of carbon black filled ethylene propylene rubber,"
POLYMER DEGRADATION AND STABILITY, vol. 86, pp. 209-215, 2004.
[16]
Y. Sun, L. Fan, K. Watkins, J. Peak, and C. Wong, "An electrical
approach to monitor wire and cable thermal oxidation aging condition based on carbon
black filled conductive polymer composite," JOURNAL OF APPLIED POLYMER
SCIENCE, vol. 93, pp. 513-520, 2004.
[17]
Y. Sun, L. Zhu, H. Jiang, and C. P. Wong, "A Paradigm of Carbon
Nanotube Interconnects in the Microelectronics," IEEE Transactions on Nanotechnology,

submitted.
[18]
Y. Sun and C. P. Wong, "Thermo-initiated Cationic Polymerization of
Epoxy-Resins by Ionic Liquids," JOURNAL OF APPLIED POLYMER SCIENCE,

submitted.
[19]
Y. Sun, H. Jiang, B. Bertram, and C. P. Wong, "Nano-Reinforcement of
Photo-Sensitive Epoxy Optical Materials," Chemistry of Materials, submitted.

- 203 -
[20]
"Rheology and Microstructure of Nanosilica Dispersions in Non-aqueous
Organic Liquid," Journal of Colloid and Interface Science, submitted.
A.4 Conference and Proceedings
[1]
Y. Sun and C. P. Wong, "Room temperature stable underfill with novel
latent catalyst for wafer level flip-chip packaging applications," 56th Electronic
Components and Technology Conference, San Diego, 2006.
[2]
Y. Sun, L. Zhu, and C. P. Wong, "A paradigm of carbon nanotube
interconnects in the microelectronics," 231st ACS National Meeting, Atlanta, GA, 2006.
[3]
J.-Y. Tsai, V. Sundaram, B. Wiedenman, Y. Sun, C. P. Wong, and R.
Tummala, "A Novel 20-100m Pitch IC-to-Package Interconnect and Assembly Process
for Pb-free Solder, Copper or Gold Stud Bumps," 56th Electronic Components and
Technology Conference, San Diego, CA, 2006.
[4]
L. Zhu, Y. Sun, J. Xu, Z. Zhang, D. W. Hess, and C. P. Wong, "Aligned
Carbon Nanotubes for Electrical Interconnect and Thermal Management," 55th

Electronic Components and Technology Conference, Orlando, FL 2005.
[5]
C. P. Wong, J. Xu, L. Zhu, Y. Li, H. Jiang, Y. Sun, J. Lu, and H. Dong,
"Recent advances on polymers and polymer nanocomposites for advanced electronic

packaging applications," 7th IEEE CPMT International Conference on High Density
Microsystem Design, Packaging and Failure Analysis, Shanghai, China, 2005.
[6]
Y. Sun, Z. Zhang, and C. P. Wong, "Development of High Performance
Photo-curable Nanocomposites for Electronics Packaging," 7th Electronics Packaging

Technology Conference Singapore, 2005.
[7]
Y. Sun, Z. Zhang, and C. P. Wong, "Photo-definable nanocomposite for
wafer level packaging," 55th Electronic Components and Technology Conference,

Orlando, FL 2005.
- 204 -
[8]
Y. Sun, Z. Zhang, and C. P. Wong, "Study of wafer level underfill
viscosity and its influence to solder wetting during reflow process," 229th ACS National
Meeting, San Diego, CA 2005.
[9]
K. S. Watkins, S. J. Morris, C. P. Wong, L. Fan, Y. Sun, D. D.
Masakowski, and W. Alvis, "An electrical Condition Monitoring Approach for wire and
Cable," 12th International Conference on Nuclear Engineering. Arlington, Virginia USA,
2004.
[10]
Y. Sun, Z. Zhang, and C. P. Wong, "Fundamental research on surface
modification of nano-size silica for underfill applications," 54th Electronic Components

and Technology Conference, Las Vegas, NV, 2004.
[11]
Y. Sun, Z. Zhang, and C. P. Wong, "Influence of nanosilica on composite
underfill properties in flip-chip packaging," 9th International Symposium onAdvanced

Packaging Materials, Atlanta, GA 2004.
[12]
Y. Sun, Z. Zhang, and C. P. Wong, "Fundamental research on surface
modification of nano-size silica for underfill applications," 9th International Symposium

on Advanced Packaging Materials, Atlanta, GA 2004.
[13]
Y. Sun and C. P. Wong, "Study and characterization on the nanocomposite
underfill for flip-chip applications," 54th Electronic Components and Technology

Conference, Las Vegas, NV 2004.
[14]
Z. Zhang, Y. Sun, L. Fan, R. Doraiswami, and C. P. Wong, "Development
of wafer level underfill material and process," 5th Electronics Packaging Technology
Conference Singapore, 2003.
[15]
Y. Sun, S. Luo, and C. P. Wong, "Electrical and mechanical properties of
carbon black filled ethylene propylene rubber during thermal oxidation aging," 53rd
Electronic Components and Technology Conference, New Orleans, LA 2003.
[16]
R. Doraiswami, S. Sankararaman, W. Kim, J. Li, Z. Zhang, P. Gupta, K.
Nakanishi, M. Borkar, R. Madhavan, V. Govind, S. Choi, A. O. Aggarwal, Y. Sun, L.

- 205 -
Fan, V. Sundaram, M. Swaminathan, C. P. Wong, and R. R. Tummala, "Proceeding of

Advances in Fine Pitch Lead Free Assembly Process," 53rd Electronic Components and
Technology Conference, New Orleans, LA, 2003.
.
- 206 -
REFERENCE
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
R. R. Tummala, E. J. Rymaszewski, and A. G. Klopfenstein, Microelectronics

Packaging Handbook. Part II, Second ed. Norwell: Kluwer Academic Publishers,
1997.
J. H. Lau, Chip on Board Technologies for Multichip Modulues. New York: Van
Nostrand Reinhold, 1994.
R. R. Tummala and V. Madisetti, "System on Chip or System on Package," IEEE
Design and Test of Computer, vol. 4, pp. 48, 1999.
"Assembly and Packaging," International Technology Roadmap for
Semiconductors, 2004.
C. P. Wong, S. J. Luo, and Z. Q. Zhang, "Microelectronics - Flip the chip,"
Science, vol. 290, pp. 2269, 2000.
L. F. Miller, "Controlled Collapse Reflow Chip Joining," IBM Journal of
Research and Development, vol. 13, pp. 239, 1969.
F. Nakano, T. Soga, and S. Amagi, "Resin-Insertion Effect on Thermal Cycle
Resistivity of Flip-Chip Mounted LSI Devices," presented at Proceeding of the
International Society of Hybrid Microelectronics Conference, 1987.
R. Satoh, M. Ohshima, H. Komura, I. Ishi, and K. Serizawa, "Development of a
New Micro-Solder Bonding Method for VLSI," IEPS, pp. 455, 1983.
N. Matsui, S. Sasaki, and T. Ohsaki, "VLSI Chip Interconnection Technology
Using Stacked Solder Bumps," The Proceedings of the 37th Electronic
Components and Technology Conference, pp. 573, 1987.
L. S. Goldmann, R. J. Herdizk, N. G. Koopman, and V. C. Marcotte, "Lead
Indium for Controlled Collapse Chip Jointing," The Proceedings of the 27th
Electronic Components and Technology Conference, pp. 25, 1977.
S. E. Greer, "Low Expansivity Organic Substrate for Flip-Chip Bonding,"
Proceedings of the 28th Electronic Components and Technology Conference, pp.
166, 1978.
Y. Tsukada, "Surface Lamilar Circuit and Flip-Chip Attach Packaging," The
Proceedings of the 42nd Electronic Components and Technology Conference, pp.
22, 1992.
B. Han and Y. Guo, "Thermal Deformation Analysis of Various Electronic
Packaging Products by Moire and Microscope Moire Interferometry," Journal of
Electronic Packaging, vol. 117, pp. 185, 1995.
Z. Zhang and C. P. Wong, "Recent Advances in Flip-Chip Underfill: Materials,
Process, and Reliability," IEEE Transactions on Advanced Packaging, vol. 27, pp.
515-524, 2004.
C. C. Kooi, C. Y. Yan, V. A. Rudge, L. S. Shing, and E. Then, "Underfilling Flip
Chip with Transfer Molding Technologies," Proceeding of Electronics Packaging
Technology Conference, pp. 625-629, 2004.
X. Zhang, T. Y. Tee, H. S. Ng, J. Teysseyre, S. Loo, S. Mhaisalkar, F. K. Ng, C.
T. Lim, X. Du, E. Bool, W. Zhu, and S. Chew, "Comprehensive Hygro-thermomechanical Modeling and Testing of Stacked Die BGA Module with Molded
Underfill," Proceeding of Electronics Packaging Technology Conference, pp.
196-200, 2005.
- 207 -
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
R. Pennisi and M. Papageorge, "Adhesive and Encapsulant Material with Fluxing

Properties." US Patent 5,128,746, 1992.
C. P. Wong and S. H. Shi, "No-Flow Underfill of Epoxy Resins, Anhydride,
Fluxing Agent and Surfactant." US Patent 6,180,696, 2001.
S. H. Shi, T. Yamashita, and C. P. Wong, "Development of the wafer level
compressive-flow underfill process and its required materials," presented at Proc.
49th Electronic Components Technology Conf., San Diego, CA,, 1999.
R. W. Johnson, Q. Wang, F. Ding, Z. Hou, L. Crane, H. Tang, G. Shi, R. Zhao, J.
Danvir, and J. Qi, "Wafer-Applied Underfill: Flip-chip Assembly and
Reliability," IEEE Transactions on Electronics Packaging Manufacturing, vol. 27,
pp. 101-108, 2004.
J. J. Stevens, Epoxy Resin Technology. New York: Interscience, 1968.
"Cycloaliphatic Epoxide Systems," Report Issued by Union Carbide Chemicals &
Plastics Technology Corporation, 1989.
"Physical Properties Guide for Epoxy Resins and Related Products," Technical
Report Issued by Resolution Performance Products, 2001.
B. Ellis, "Chemistry and Technology of Epoxy Resins." Glasgow: Blackie
Academic & Professional, an imprint of Chapman & Hall, 1993.
S. Luo, "Study on Adhesion of Underfill Materials for Flip Chip Packaging," Ph
D thesis, 2002.
Z. Zhang, Y. Sun, L. Fan, and C. P. Wong, "Study on B-stage properties of wafer
level underfills," Journal of Adhesion Science and Technology, vol. 18, pp. 361380, 2004.
Y. Sun and C. P.Wong, "Room Temperature Stable Underfill with Novel Latent
Catalyst for Wafer Level Flip-Chip Packaging Applications," Proceeding of 56th
Electronic Components and Technology Conference, pp. 1905-1910, 2006.
M. L. Kaplan, A. L. Wayda, and A. M. Lyons, "Lanthanide Imidazole Complexes
as Latent Curing Agents for Epoxy-Resins," Journal of Polymer Science Part aPolymer Chemistry, vol. 28, pp. 731-740, 1990.
J. M. Barton, G. J. Buist, I. Hamerton, B. J. Howlin, J. R. Jones, and S. U. Liu,
"Preparation and Characterization of Imidazole-Metal Complexes and Evaluation
of Cured Epoxy Networks," Journal of Materials Chemistry, vol. 4, pp. 379-384,
1994.
F. Ricciardi, W. A. Romanchil, and M. M. Joullie, "1,3-dialkylimidazole Salts as
Latent Catalysts in the Curing of Epoxy Resins," Journal of Polymer Science,
Polymer Letter Edition, vol. 21, pp. 633, 1983.
M. Kobayashi, F. Sanda, and T. Endo, "Application of phosphonium ylides to
latent catalysts for polyaddition of bisphenol A diglycidyl ether with bisphenol A:
Model system of epoxy-novolac resin," Macromolecules, vol. 32, pp. 4751-4756,
1999.
Z. Zhang and C. P. Wong, "Study on metal chelates as catalysts of epoxy and
anhydride cure reactions for no-flow underfill applications," Polymers for
Microelectronics and Nanoelectronics, vol. 874, pp. 264-278, 2004.
Y. Yagci and I. Reetz, "Externally stimulated initiator systems for cationic
polymerization," Progress in Polymer Science, vol. 23, pp. 1485-1538, 1998.
- 208 -
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
P. Fine, B.Cobb, and L. Nguyen, "Flip Chip Underfill Flow Characteristics and
Prediction," IEEE Transactions on Components and Packaging Technologies, vol.
23, pp. 420-427, 2000.
J. Wang and T. Chen, "The Effect of Flow Properties on Fillers Settling of
Underfill in the Flip Chip Package," Proceeding of Electronic Components and
Y. L. Liu, C. Y. Hsu, W. L. Wei, and R. J. Jeng, "Preparation and thermal
properties of epoxy-silica nanocomposites from nanoscale colloidal silica,"
Polymer, vol. 44, pp. 5159-5167, 2003.
S. Rubinsztajn, D. Buckley, J. Campbell, D. Esler, E. Fiveland, A. Prabhakumar,
and D. Sherman, "Development of novel filler technology for no-flow and wafer
level underfill materials," Journal of Electronic Packaging, vol. 127, pp. 77-85,
2005.
Y. Y. Sun, Z. Q. Zhang, and C. P. Wong, "Study and characterization on the
nanocomposite underfill for flip chip applications," IEEE Transactions on
Components and Packaging Technologies, vol. 29, pp. 190-197, 2006.
Y. Sun, Z. Zhang, and C. P. Wong, "Study on mono-dispersed nano-size silica by
surface modification for underfill applications," Journal of Colloid and Interface
Science, vol. 292, pp. 436-444, 2005.
T. Chen, J. Wang, and D. Lu, "Emerging Challenges of Underfill for Flip Chip
Application," Proceeding of Electronic Components and Technology Conference,
pp. 175-179, 2004.
L. Fan, Z. Zhang, and P. C, Wong, "Effect of Filler Settling of Underfill
Encapsulant on Reliability Performance," Proceeding of International Symposium
on Advanced Packaging Materials, pp. 218-223, 2001.
M. Ranjan, L. Gopalakrishnan, K. Srihari, and C. Woychik, "Die Cracking in Flip
Chip Assemblies," Proceeding of Electronics Components and Technology
Conference, pp. 729-733, 1998.
Y. Huang, D. Bigio, and M. G. Pecht, "Fill Pattern and Particle Distribution of
Underfill Material," IEEE Transactions on Components and Packaging
Technologies, vol. 27, pp. 493-498, 2004.
C. K. Lam and C. Lee, "Assembly and Reliability Performance of Flip Chip with
No-flow Underfills," Proceeding of Electronics Packaging Technology
S. H. Shi and C. P. Wong, "Recent Advances in the Development of No-Flow
Underfill Encapsulants - A Practical Approach towards the Actual Manufacturing
Application," IEEE Transactions on Electronics Packaging Manufacturing, vol.
22, pp. 331-339, 1999.
K. Gross, S. Hackett, W. Schultz, and W. Thompson, "Nanocomposite Underfills
for Fip-chip Applications," Proceeding of the 53th Electronic Components and
H. N. Stein, The Preparation of Dispersions in Liquids, vol. 58. New York, 1996.
B. Dobi, X. Qiu, and W. v. Rybinski, Solid-Liquid Dispersions. Solid-Liquid
Dispersions, vol. 81. New York: Marcel Dekker Inc, 1999.
R. S. Farinato and P. L. Dubin, Colloid-Polymer Interactons. From Fundamentals
to Practice. New York: John Wiley & Sons, 1999.
- 209 -
[50]
[51]
[52]
[53]
[54]
[55]
[56]
[57]
[58]
[59]
[60]
[61]
[62]
[63]
[64]
[65]
[66]
[67]
R. A. Williams, Colloid and Surface Engineering: Applications in the Process

Industries. Oxford: Butterworth-Heinemann, 1992.
H. Koelmans and J. T. G. Overbeek, "Stability and Electrophoretic Deposition of
Suspensions in Non-Aqueous Media," Discussions of the Faraday Society, pp. 5263, 1954.
E. W. J. Mardles, "Viscosity of suspensions and the Einstein equation," Nature,
vol. 145, pp. 970-970, 1940.
I. M. Krieger and T. J. Dougherty, "A Mechanism for Non-Newtonian Flow in
Suspensions of Rigid Spheres," Transactions of the Society of Rheology, vol. 3,
pp. 137-152, 1959.
J. Goodwin, Colloids and Interfaces with Surfactants and Polymers. An
Introduction. Hoboken, NJ: John Wiley & Sons, 2004.
M.Schmcker, Thesis, Ruhr-Universitt Bochum, 1988.
R. Bode, H. Ferch, and H. Fratzscher, "Basic Characteristics of Aerosil Fumed
Silica," Technical Bulletin Fine Particles - Degussa, vol. 11.
F. D. Osterholz and E. R. Pohl, Silane and Other Coupling Agents. Utrecht, the
Netherlands: VSP, 1992.
J. Y. Ying, J. B. Benziger, and A. Navrotsky, "Structural Evolution of Alkoxide
Silica-Gels to Glass - Effect of Catalyst pH," Journal of the American Ceramic
Society, vol. 76, pp. 2571-2582, 1993.
J. Y. Ying, J. B. Benziger, and A. Navrotsky, "Structural Evolution of Colloidal
Silica-Gels to Glass," Journal of the American Ceramic Society, vol. 76, pp.
2561-2570, 1993.
R. K. Iler, The Chemistry of Silica. New York: John Wiley & Sons, 1979.
W. Stober, A. Fink, and E. Bohn, "Controlled Growth of Monodisperse Silica
Spheres in Micron Size Range," Journal of Colloid and Interface Science, vol. 26,
pp. 62, 1968.
G. Bogush, M. Tracy, and C. F. Zukoski, "Preparation of Monodisperse Silica
Particles: Control of Size and Mass Fraction," Journal of Non-Crystalline Solids,
vol. 104, pp. 95, 1998.
C. Gurumurthy, E. J. Kramer, and C. Y. Hui, "Controlling interfacial
interpenetration and fracture properties of polyimide/epoxy interfaces," Journal of
Adhesion, vol. 82, pp. 239-266, 2006.
J. L. Lenhart, J. P. Dunkers, J. H. van Zanten, and R. S. Parnas, "Characterization
of sizing layers and buried polymer/sizing/substrate interfacial regions using a
localized fluorescent probe," Journal of Colloid and Interface Science, vol. 257,
pp. 398-407, 2003.
J. Lyklema, Fundamentals of Interface and Colloid Science, vol. III. San Diego,
CA: Academic Press, 2000.
A. V. Shenoy, Rheology of Filled Polymer Systems. Dordrecht: Kluwer
Academic, 1999.
Y. C. Araujo, P. G. Toledo, V. Leon, and H. Y. Gonzalez, "Wettability of SilaneTreated Glass Slides as Determined from X-Ray Photoelectron Spectroscopy,"
Journal of Colloid and Interface Science, vol. 176, pp. 485-490, 1995.
- 210 -
[68]
[69]
[70]
[71]
[72]
[73]
[74]
[75]
[76]
[77]
[78]
[79]
[80]
[81]
[82]
[83]
[84]
W. G. Anderson, "Wettability Literature Survey .3. The Effects of Wettability on

the Electrical-Properties of Porous-Media," Journal of Petroleum Technology,
vol. 38, pp. 1371-1378, 1986.
W. G. Anderson, "Wettability Literature Survey .2. Wettability Measurement,"
Journal of Petroleum Technology, vol. 38, pp. 1246-1262, 1986.
W. G. Anderson, "Wettability Literature Survey .1. Rock-Oil-Brine Interactions
and the Effects of Core Handling on Wettability," Journal of Petroleum
Technology, vol. 38, pp. 1125-1144, 1986.
P. Silberzan, L. Leger, D. Ausserre, and J. J. Benattar, "Silanation of Silica
Surfaces - a New Method of Constructing Pure or Mixed Monolayers," Langmuir,
vol. 7, pp. 1647-1651, 1991.
R. Maoz and J. Sagiv, "On the Formation and Structure of Self-Assembling
Monolayers .1. A Comparative Atr-Wetability Study of Langmuir-Blodgett and
Adsorbed Films on Flat Substrates and Glass Microbeads," Journal of Colloid and
Interface Science, vol. 100, pp. 465-496, 1984.
H. W. Fox and W. A. Zisman, "The Spreading of Liquids on Low Energy
Surfaces .1. Polytetrafluoroethylene," Journal of Colloid Science, vol. 5, pp. 514531, 1950.
P. C. Hiemenz and R. Rajagopalan, Principles of Colloid and Surface Chemistry,
Third ed. New York: Marcel Dekker, Inc, 1997.
S. Kang, S. I. Hong, C. R. Choe, M. Park, S. Rim, and J. Kim, "Preparation and
characterization of epoxy composites filled with functionalized nanosilica
particles obtained via sol-gel process," Polymer, vol. 42, pp. 879-887, 2001.
C. Kneuer, M. Sameti, E. G. Haltner, T. Schiestel, H. Schirra, H. Schmidt, and C.M. Lehr, "Silica Nanoparticles Modified with Aminosilanes as Carriers for
Plasmid DNA," International Journal of Pharmaceutics, vol. 196, pp. 257-261,
2000.
M. W. Daniels, J. Sefcik, L. F. Francis, and A. V. McCormick, "Reactions of a
trifunctional silane coupling agent in the presence of colloidal silica sets in polar
media," Journal of Colloid and Interface Science, vol. 219, pp. 351-356, 1999.
E. M. Petrie, Handbook of Adhesives and Sealants. New York: McGraw-Hill,
1999.
A. S. Piers and C. H. Rochester, "Infrared Study of the Adsorption of 1Aminopropyltrialkoxysilanes on Silica at the Solid/Liquid Interface," Journal of
Colloid and Interface Science, vol. 174, pp. 97-103, 1995.
S. H. Wang, Sol-Gel Derived Silica Optics. Gainesville, FL: PhD thesis,
University of Florida, 1988.
R. S. Mcdonald, "Surface Functionality of Amorphous Silica by Infrared
Spectroscopy," Journal of Physical Chemistry, vol. 62, pp. 1168-1178, 1958.
J. A. Hockey and B. A. Pethica, "Surface Hydration of Silicas," Transactions of
the Faraday Society, vol. 57, pp. 2247-&, 1961.
L. L. Hench and S. H. Wang, "The Sol-Gel Glass Transformation of Silica,"
Phase Transitions, vol. 24-6, pp. 785-834, 1990.
M. L. Hair, Infrared Spectroscopy in Surface Chemistry. New York: Dekker,
1967.
- 211 -
[85]
[86]
[87]
[88]
[89]
[90]
[91]
[92]
[93]
[94]
[95]
[96]
[97]
[98]
[99]
J. H. Anderson and K. A. Wickersheim, "Near Infrared Characterization of Water

and Hydroxyl Groups on Silica Surfaces," Surface Science, vol. 2, pp. 252-260,
1964.
V. Dugas and Y. Chevalier, "Surface hydroxylation and silane grafting on fumed
and thermal silica," Journal of Colloid and Interface Science, vol. 264, pp. 354361, 2003.
M. Celina, K. T. Gillen, J. Wise, and R. L. Clough, "Anomalous aging
phenomena in a crosslinked polyolefin cable insulation," Radiation Physics and
Chemistry, vol. 48, pp. 613-626, 1996.
J. Qu and C. P. Wong, "Effective elastic modulus of underfill material for flipchip applications," IEEE Transactions on Components and Packaging
Technologies, vol. 25, pp. 53-55, 2002.
Y. Z. Wan, Y. L. Wang, F. G. Zhou, G. X. Cheng, and K. Y. Han, "Threedimensionally braided carbon fiber-epoxy composites, a new type of materials for
osteosynthesis devices. II. Influence of fiber surface treatment," Journal of
Applied Polymer Science, vol. 85, pp. 1040-1046, 2002.
Y. Wang and N. Herron, "Nanometer-Sized Semiconductor Clusters - Materials
Synthesis, Quantum Size Effects, and Photophysical Properties," Journal of
Physical Chemistry, vol. 95, pp. 525-532, 1991.
M. Kawasumi, N. Hasegawa, M. Kato, A. Usuki, and A. Okada, "Preparation and
mechanical properties of polypropylene-clay hybrids," Macromolecules, vol. 30,
pp. 6333-6338, 1997.
J. Q. Pham, C. A. Mitchell, J. L. Bahr, J. M. Tour, R. Krishanamoorti, and P. F.
Green, "Glass transition of polymer/single-walled carbon nanotube composite
films," Journal of Polymer Science Part B-Polymer Physics, vol. 41, pp. 33393345, 2003.
Y. M. Cao, J. Sun, and D. H. Yu, "Preparation and properties of nano-Al2O3
particles/polyester/epoxy resin ternary composites," Journal of Applied Polymer
Science, vol. 83, pp. 70-77, 2002.
W. D. Hergeth, U. J. Steinau, H. J. Bittrich, G. Simon, and K. Schmutzler,
"Polymerization in the Presence of Seeds .4. Emulsion Polymers Containing
Inorganic Filler Particles," Polymer, vol. 30, pp. 254-258, 1989.
B. J. Ash, L. S. Schadler, and R. W. Siegel, "Glass transition behavior of
alumina/polymethylmethacrylate nanocomposites," Materials Letters, vol. 55, pp.
83-87, 2002.
M. N. Xiong, G. X. Gu, B. You, and L. M. Wu, "Preparation and characterization
of poly(styrene butylacrylate) latex/nano-ZnO nanocomposites," Journal of
A. Dufresne and C. Lacabanne, "A Study of Interphase in Model Composites Via
Low-Temperature Thermo Stimulated Creep Spectroscopy," Polymer, vol. 34, pp.
3173-3178, 1993.
M. Fois, A. Lamure, P. Guuinic, and C. Lacabanne, "TSC study of secondary
dielectric relaxations in a polyepoxy matrix composite," Journal of Applied
Polymer Science, vol. 66, pp. 135-139, 1997.
R. N. Haward and R. J. Young, The Physics of Glassy Polymers. London:
Chapman & Hall, 1997.
- 212 -
[100] S. C. Glotzer and W. Paul, "Molecular and mesoscale simulation methods for
polymer materials," Annual Review of Materials Research, vol. 32, pp. 401-436,
2002.
[101] E. P. M. Williams, J. C. Seferis, C. L. Wittman, G. A. Parker, J. H. Lee, and J. D.
Nam, "Dielectric nanocapacitance effect of thermoset polymeric matrix
nanocomposites," Journal of Polymer Science Part B-Polymer Physics, vol. 42,
pp. 1-4, 2004.
[102] D. Kranbuehl, S. Delos, E. Yi, J. Mayer, T. Jarvie, W. Winfree, and T. Hou,
"Dynamic Dielectric Analysis - Nondestructive Material Evaluation and Cure
Cycle Monitoring," Polymer Engineering and Science, vol. 26, pp. 338-345, 1986.
[103] H. T. Vo and F. G. Shi, "Towards model-based engineering of optoelectronic
packaging materials: dielectric constant modeling," Microelectronics Journal, vol.
33, pp. 409-415, 2002.
[104] P. Myslinski and Z. Lazowski, "The Effect of Silica Surface Modification on the
Dielectric-Properties of Silica-Filled Epoxy Compounds," Materials Chemistry
and Physics, vol. 33, pp. 139-144, 1993.
[105] T. J. Bosma, T. Lebey, V. Pouilles, I. Chenerie, B. Ai, and N. Rieux, "The Use of
Mixing Laws for Modeling the Climatic Aging of Dielectric Materials," Journal
of Physics D-Applied Physics, vol. 28, pp. 1180-1186, 1995.
[106] I. D. Maxwell and R. A. Pethrick, "Dielectric Studies of Water in Epoxy-Resins,"
Journal of Applied Polymer Science, vol. 28, pp. 2363-2379, 1983.
[107] P. Gonon, A. Sylvestre, J. Teysseyre, and C. Prior, "Combined effects of humidity
and thermal stress on the dielectric properties of epoxy-silica composites,"
Materials Science and Engineering B-Solid State Materials for Advanced
[108] M. G. Todd and F. G. Shi, "Validation of a novel dielectric constant simulation
model and the determination of its physical parameters," Microelectronics
Journal, vol. 33, pp. 627-632, 2002.
[109] M. G. Todd and F. G. Shi, "Characterizing the interphase dielectric constant of
polymer composite materials: Effect of chemical coupling agents," Journal of
Applied Physics, vol. 94, pp. 4551-4557, 2003.
[110] J. D. Reid, W. H. Lawrence, and R. P. Buck, "Dielectric-Properties of an EpoxyResin and Its Composite .1. Moisture Effects on Dipole Relaxation," Journal of
[111] J. Brandrup, E. H. Immergut, and E. A. Grulke, Polymer Handbook, Fourth ed:
John Wiley & Sons, 2003.
[112] A. K. Jonscher, "Dielectric relaxation in solids," Journal of Physics D-Applied
Physics, vol. 32, pp. R57-R70, 1999.
[113] W. Brostow and R. Corneliussen, Failure of Plastics. Hanser-SPE: MunichVienna-New York, 1986.
[114] N. G. McGrum, B. E. Read, and G. Williams, Anelastic and Dielectric Effects in
Polymeric Solid. New York: Dover, 1967.
[115] J. D. Ferry, Viscoelastic Properties of Polymer, Third ed. New York: John Wiley
& Sons, 1980.
[116] S. A. Khan and N. J. Zoeller, "Dynamic Rheological Behavior of Flocculated
Fumed Silica Suspensions," Journal of Rheology, vol. 37, pp. 1225-1235, 1993.
- 213 -
[117] J. Israelachvili, Intermolecular and Surface Forces, Second ed. San Diego:
Academic Press, 1991.
[118] S. R. Raghavan, M. W. Riley, P. S. Fedkiw, and S. A. Khan, "Composite polymer
electrolytes based on poly(ethylene glycol) and hydrophobic fumed silica:
Dynamic rheology and microstructure," Chemistry of Materials, vol. 10, pp. 244251, 1998.
[119] S. R. Raghavan, H. J. Walls, and S. A. Khan, "Rheology of silica dispersions in
organic liquids: New evidence for solvation forces dictated by hydrogen
bonding," Langmuir, vol. 16, pp. 7920-7930, 2000.
[120] Y. Y. Sun, Z. Q. Zhang, K. S. Moon, and C. P. Wong, "Glass transition and
relaxation behavior of epoxy nanocomposites," Journal of Polymer Science Part
B-Polymer Physics, vol. 42, pp. 3849-3858, 2004.
[121] K. K. Baikerikar and A. B. Scranton, "Photopolymerizable liquid encapsulants for
microelectronic devices," Polymer, vol. 42, pp. 431-441, 2001.
[122] F. Bauer, H. Ernst, D. Hirsch, S. Naumov, M. Pelzing, V. Sauerland, and R.
Mehnert, "Preparation of Scratch and Abrasion Resistant Polymeric
Nanocomposites by Monomer Grafting onto Nanoparticles, Part 5,"
Macromol.Chem.Phys., vol. 205, pp. 1587-1593, 2004.
[123] J. Zhang, K. L. Tan, G. D. Hong, L. J. Yang, and H. Q. Gong, "Polymerization
optimization of SU-8 photoresist and its applications in microfluidic systems and
MEMS," Journal of Micromechanics and Microengineering, vol. 11, pp. 20-26,
2001.
[124] K. K. Tung, W. H. Wong, and E. Y. B. Pun, "Polymeric optical waveguides using
direct ultraviolet photolithography process," Applied Physics a-Materials Science
& Processing, vol. 80, pp. 621-626, 2005.
[125] H. Bayer, "Cationic cure of epoxy resins and UV options for conductive
adhesives," in Conductive adhesives for electronics packaging, J. Liu, Ed.:
Electrochemical Publications LTD, 1999.
[126] T. Naganuma and Y. Kagawa, "Effect of particle size on the optically transparent
nano meter-order glass particle-dispersed epoxy matrix composites," Composites
Science and Technology, vol. 62, pp. 1187-1189, 2002.
[127] C. Decker and K. Moussa, "Real-Time Monitoring of Ultrafast Curing by UvRadiation and Laser-Beams," Journal of Coatings Technology, vol. 62, pp. 55-61,
1990.
[128] C. Decker and K. Moussa, "Kinetic-Study of the Cationic Photopolymerization of
Epoxy Monomers," Journal of Polymer Science Part a-Polymer Chemistry, vol.
28, pp. 3429-3443, 1990.
[129] M. Q. Zhang, M. Z. Rong, S. L. Yu, B. Wetzel, and K. Friedrich, "Effect of
particle surface treatment on the tribological performance of epoxy based
nanocomposites," Wear, vol. 253, pp. 1086-1093, 2002.
[130] H.-J. Glasel, F. Bauer, H. Ernst, M. Findeisen, E. Hartmann, H. Langguth, R.
Mehnert, and R. Schubert, "Preparation of scratch and abrasion resistant
polymeric nanocomposites by monomer grafting onto nanoparticles, 2
Characterization
of
radiation-cured
polymeric
nanocomposites,"
Macromol.Chem.Phys., vol. 201, pp. 2765-2770, 2000.
- 214 -
[131] M. Fujiwara, K. Kojima, Y. Tanaka, and R. Nomura, "A simple preparation

method of epoxy resin/silica nanocomposite for T-g loss material," Journal of
Materials Chemistry, vol. 14, pp. 1195-1202, 2004.
[132] J. V. Crivello and J. H. W. Lam, "Complex Triarylsulfonium Salt Photoinitiators
.1. The Identification, Characterization, and Syntheses of a New Class of
Triarylsulfonium Salt Photoinitiators," Journal of Polymer Science Part a-Polymer
Chemistry, vol. 18, pp. 2677-2695, 1980.
[133] J. L. Dektar and N. P. Hacker, "Photochemistry of Triarylsulfonium Salts,"
Journal of the American Chemical Society, vol. 112, pp. 6004-6015, 1990.
[134] J. V. Crivello and S. S. Liu, "Photoinitiated cationic polymerization of epoxy
alcohol monomers," Journal of Polymer Science Part a-Polymer Chemistry, vol.
38, pp. 389-401, 2000.
[135] E. W. Nelson, J. L. Jacobs, A. B. Scranton, K. S. Anseth, and C. N. Bowman,
"Photo-Differential Scanning Calorimetry Studies of Cationic Polymerizations of
Divinyl Ethers," Polymer, vol. 36, pp. 4651-4656, 1995.
[136] C. Decker, "Kinetic Study and New Applications of UV Radiation Curing,"
Macromolecular Rapid Communication, vol. 23, pp. 1067-1093, 2002.
[137] G. Odian, Priciples of Polymerization, Fourth ed. Hoboken, NJ: John Wiley &
Sons, 2004.
[138] F. M. Uhl, S. P. Davukuri, S.-C. Wong, and D. C. Webster, "Polymer Films
Possessing Nanoreinforcements via Organically Modified Layered Silicate,"
Chem. Mater., vol. 16, pp. 1135-1142, 2004.
[139] C. Decker, "Light-induced Crosslinking Polymerization," Polymer International,
vol. 51, pp. 1141-1150, 2002.
[140] Y. Kagawa, H. Iba, M. Tanaka, H. Sato, and T. Chang, "Fabrication and
optical/thermal properties of glass particle-epoxy optically transparent
composites," Acta Materialia, vol. 46, pp. 265-271, 1997.
[141] T. J. Pinnavaia and G. W. Beall, Polymer-clay nanocomposite. New York: John
Wiley & Sons, 2000.
[142] S. Kang, S. I. Hong, C. R. Choe, M. Park, S. Rim, and J. Kim, "Preparation and
characterization of epoxy composites filled with functionalized nanosilica
particles obtained via sol-gel process," Polymer, vol. 42, pp. 879-887, 2001.
[143] T. Mura, Micromechanics of Defects in Solids. New York: Martinus Nijhoff,
1987.
[144] T. Mori and K. Tanaka, "Average Stress in Matrix and Average Elastic Energy of
Materials with Misfitting Inclusions," Acta Metallurgica, vol. 21, pp. 571-574,
1973.
[145] G. M. Odegard, T. C. Clancy, and T. S. Gates, "Modeling of the mechanical
properties of nanoparticle/polymer composites," Polymer, vol. 46, pp. 553-562,
2005.
[146] J. Qu, "Effects of slightly weakened interfaces on the overall elastic properties of
composite materials," Mechanics of Materials, vol. 14, pp. 269-281, 1993.
[147] G. M. Odegard, T. S. Gates, K. E. Wise, C. Park, and E. J. Siochi, "Constitutive
modeling of nanotube-reinforced polymer composites," Composites Science and
- 215 -
[148] P. Garrou, "Wafer Level Packaging Has Arrived," Semiconductor International,

vol. 10, 2000.
[149] R. B. Burress, M. A. Capote, Y.-J. Lee, H. A. Lenos, and J. F. Zamora, "A
Practical, Flip-Chip, Multi-layer Pre-Encapsulation Technology for Wafer-Scale
Underfill," Proceeding of Electronic Components and Technology Conference,
pp. 777-781, 2001.
[150] R. V. Burress, M. A. Capote, Y.-J. Lee, H. A. Lenos, and J. F. Zamora, "A Novel,
Wafer-Scale Technology for Addressing Process and Cost Obstacles Associated
with Underfilling FCOB," Proceeding of Electronic Components and Technology
[151] W. H. Awad, J. W. Gilman, M. Nyden, R. H. Harris, T. E. Sutto, J. Callahan, P.
C. Trulove, H. C. DeLong, and D. M. Fox, "Thermal Degradation Studies of
Alkyl-Imidazolium Salts and Their Application in Nanocomposites,"
Thermochimica Acta, vol. 409, pp. 3-11, 2004.
[152] A. Provates and J. G. Matisons, "Silsesquioxanes: Synthesis and Applications,"
Trends Polym. Sci., vol. 5, pp. 327-333, 1997.
[153] R. M. Laine, "Nanobuilding Blocks Based on the [OSiO1.5]x (x =6, 8, 10)
Octasilsesquioxanes," J. Mater. Chem., vol. 15, pp. 3725-3744, 2005.
[154] R. A. Basheer, R. M. Laine, S. Sulaiman, C. M. Brick, D. B. Workman, and A.
Chaudhuri, "Functionalized silsesquioxanes as potential no-flow underfill
materials," presented at 229th ACS National Meeting, San Diego,CA, United
States, 2005.
- 216 -

Sun Yangyang 200612 PHD

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Sun Yangyang 200612 PHD

Enviado por

Direitos autorais:

Formatos disponíveis

STUDY ON THE NANOCOMPOSITE

UNDERFILL FOR FLIP-CHIP APPLICATION

Georgia Institute of Technology

STUDY ON THE NANOCOMPOSITE UNDERFILL FOR FLIP-CHIP

Dr. Rigoberto Hernandez

Dr. Karl Jacob

Dr. Boris Mizaikoff

Dr. Z. John Zhang

work is funded by National Science Foundation through Packaging Research Center of

NANOSILICA SYNTHESIS AND MODIFICATION

CHAPTER 3. MATERIAL PROPERTIES CHARACTERIZATION OF THE

THE HARDENER EFFECTS TO COLLOIDAL SILICA

CHAPTER 6. PHOTO-POLYMERIZATION OF EPOXY NANOCOMPOSITE FOR

Novel wafer level packaging process

CONCLUSIONS AND SUGGESTED WORK

Table 1-2. Coefficient of thermal expansion of major materials in flip-chip packaging 6

Table 1-4. List of Epoxy resin used in this study

Table 1-5. List of curing agents used in this study

Table 1-6. Lists of catalysts used in this study

Table 1-7. Bulk resistivity of underfill formulation (before curing)

Table 2-1. Physical, mechanical, thermal and electrical properties of silica

Table 2-2. Ingredients for sol-gel synthesis silica

Table 2-3. Chemistry structure of silane coupling agents

Table 2-6. DOE of modification condition

Table 3-1. Chemicals used in the underfill formulations

Table 3-2. Moisture absorption kinetics parameter

Table 4-1. Constant parameters of WLF equation for three samples

Table 5-1. Matrix composition of underfill with different hardener

Table 5-2. Summary of dynamic rheology of different systems

Table 5-3. Bulk material properties for various components

Table 6-1. Comparison between two photo-polymerization approaches

Table 6-2. Observed peaks of epoxy with FTIR

Table 6-4. Kinetics data for the photo-polymerization of nanocomposite underfill

Table 6-5. Light absorptivity and components concentration in the nanocomposite

Table 6-7. Materials Constant

Figure 1-2. General configuration of wire bonding package

Figure 1-3. General configuration of flip-chip package

Figure 1-4. Conventional capillary flow underfill process

Figure 1-5. No-flow underfill process

Figure 1-6. Curing mechanism of primary amine

Figure 1-12. Adsorption-dissociation mechanism of ions on the silica surface in the

Figure 1-13. Steric stabilization of particles by adsorbed polymer

Figure 1-14. Viscosity definition model

Figure 1-15. Viscosity of concentrated suspensions

Figure 2-2. TEM picture of fumed silica structure[56]

Figure 2-4. Polymerization behavior of silica [60]

Figure 2-7. Relation between particle size and ammonia concentration

Figure 2-8. Relationship between interfacial tension and contact angle

Figure 2-9. Wetting phenomenon of silica filler in the underfill

Figure 2-10. General structure of silane coupling agents

Figure 2-11. Scheme of surface modification for nano-size filler

Figure 2-13. Zisman plot to determine the critical surface tension

Figure 2-14. Nanosilica dispersion with different pre-treatments

Figure 2-15. Nanosilica dispersion with epoxy-silane treatments

Figure 2-16. Average size of nanosilica with different treatment conditions

Figure 2-17. Reaction mechanism of silane treatment to nanosilica surface

Figure 2-19. Dispersion of #004 (amino silane treated)

Figure 2-20. Dispersion of #007 (epoxy silane treated)

Figure 2-21. Dispersion of #007 (enlarged)

Figure 2-22. FTIR spectra of nanosilica with different surface modification

Figure 2-24. Re-absorption of physical water below 400C[83]

Figure 2-25. Irreversible elimination of adjacent silanol group [83]

Figure 2-26. Weight loss of silica at different drying condition