STUDY ON THE NANOCOMPOSITE UNDERFILL FOR FLIP-CHIP APPLICATION

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 I STUDY ON THE NANOCOMPOSITE UNDERFILL FOR FLIP-CHIP APPLICATION Approved by: Dr. C. P. Wong, Advisor School of Materials Science and Engineering Georgia Institute of Technology Dr. Karl Jacob School of Polymer, Textile and Fiber Engineering Georgia Institute of Technology Dr. Z. John Zhang School of Chemistry and Biochemistry Georgia Institute of Technology Date Approved: November 8, 2006 II Dr. Rigoberto Hernandez School of Chemistry and Biochemistry Georgia Institute of Technology Dr. Boris Mizaikoff School of Chemistry and Biochemistry Georgia Institute of Technology 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. Wong’s 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 TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES SUMMARY CHAPTER 1. INTRODUCTION III V IX XI XVII 1 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 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 Stöber 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 V 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 CHAPTER 3. MATERIAL PROPERTIES CHARACTERIZATION OF THE NANOCOMPOSITE UNDERFILL AFTER CURING 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. RESULTS AND DISCUSSIONS 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 THE HARDENER EFFECTS TO COLLOIDAL SILICA 51 51 52 53 53 57 64 71 73 73 73 74 75 77 77 79 81 83 86 91 92 94 95 96 97 108 108 108 112 113 114 114 116 119 122 127 128 128 5.1. EXPERIMENT 5.1.1. Materials VI 5.1.2. Dynamic rheology 5.1.3. Dielectric constant of liquid sample 5.2. RESULTS AND DISCUSSIONS 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 PACKAGING 143 145 145 146 146 151 151 152 155 155 157 160 164 170 170 173 174 175 176 179 180 182 182 185 186 190 190 195 195 197 199 200 201 6.6.1. 6.6.2. 6.6.3. CHAPTER 7. Novel wafer level packaging process Advantages of photo-curable nanocomposites Pattern formation with photo-curable nanocomposite CONCLUSIONS AND SUGGESTED WORK 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 AUTHOR’S AWARDS, PATENTS, AND PUBLICAITONS VII REFERENCE 207 VIII LIST OF TABLES Table 1-1. Trend in the microelectronic manufacturing 3 Table 1-2. Coefficient of thermal expansion of major materials in flip-chip packaging 6 Table 1-3. Underfill classification 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 10 11 12 14 23 35 40 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 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 56 58 74 89 119 129 135 138 144 157 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 Table 6-5. Light absorptivity and components concentration in the nanocomposite underfill Table 6-6. TGA measured for various filler loadings in the nanocomposites Table 6-7. Materials Constant 167 167 174 179 X 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 2 4 4 8 8 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 Figure 1-11. Electrostatic force in the dispersion system with ionic strength 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 Figure 1-14. Viscosity definition model Figure 1-15. Viscosity of concentrated suspensions 19 21 21 21 26 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: 3µm, theoretical calculation is based on the Equation 1-8) 30 Figure 2-1. Synthesis of fumed silica Figure 2-2. TEM picture of fumed silica structure[56] 36 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 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 2-12. Proposed mechanism for the silane reaction onto the glass slides 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 2-18. Nanosilica dispersion with amino-silane treatments and with sonication 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 2-23. Physical water decreases and silanol groups condense [83] Figure 2-24. Re-absorption of physical water below 400°C[83] Figure 2-25. Irreversible elimination of adjacent silanol group [83] Figure 2-26. Weight loss of silica at different drying condition Figure 2-27. Weight loss of nanosilica with different surface modification Figure 2-28. Viscosity of nanocomposite underfills XII 43 43 45 46 48 49 54 55 59 59 60 61 62 63 63 64 65 67 67 68 70 70 71 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 Figure 3-3. Glass transition temperatures of composite underfills by DSC Figure 3-4. Viscosity of silica filled composite underfills Figure 3-5. Effect of filler size on the UV-Vis spectra of the composite underfills Figure 3-6. CTE of silica filled composite underfills Figure 3-7. Dynamic moduli of composite underfills with untreated nanosilica Figure 3-8. Comparison of dynamic moduli of composite underfills with different nanosilica 80 80 81 82 83 85 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 Figure 3-11. Density measurement for silica filled composite underfills 89 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 Figure 3-14. Wetting picture of quartz chip with treated-30 underfill Figure 3-15. Glass transition temperature of the silica composites Figure 3-16. Glass transition temperature of the silver composites Figure 3-17. Glass transition temperature of the aluminum composites Figure 3-18. Glass transition temperature of the carbon black composites Figure 3-19. TGA measured weight loss at a heating rate of 20 °C/min under air Figure 3-20. Dynamic loss moduli of the silica composites and the blank resin Figure 3-21. Deconvolution of loss modulus of nanocomposite Figure 3-22. Molecular structure of anhydride/epoxy polymer Figure 3-23. Possible local motion of segments in anhydride/epoxy polymer 93 93 98 99 99 101 102 104 104 105 105 XIII Figure 4-1. Electrode design of the single surface sensor used in the experiment Figure 4-2. Dielectric property of the control sample after curing 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 113 114 115 115 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 Figure 4-9. Loss factor and ionic conductivity of the three samples at 1 Hz Figure 4-10. Loss factor of three samples after curing, (a) 1Hz; (b) 1000Hz Figure 4-11. Loss factor of three samples after aging under humidity, (a) 1Hz; (b) 1000Hz Figure 4-12. Loss factor of three samples after drying (a) 1Hz; (b) 1000Hz Figure 5-1. Molecular structure of two hardeners used in the experiment Figure 5-2. Elastic and viscous modulus as a function of frequency for nanosilica/anhydride mixture 120 122 124 125 126 129 134 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 Figure 6-2. Scheme of real-time FITR setup 146 147 XIV Figure 6-3. Scheme of photo-DSC setup Figure 6-4. Light transmittance of two kinds of nanosilica in ethanol solution Figure 6-5. TEM picture of the 20 nm colloidal silica Figure 6-6. UV absorbance of pure epoxy, nanosilica and photo-initiator 148 151 152 154 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 Figure 6-9. FTIR absorption of epoxy 156 157 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 Figure 6-13. Heat flow of UV-initiated underfill during thermal heating Figure 6-14. Photo-DSC curves of the nanocomposite with different filler loading Figure 6-15. DSC measured heat flow in an isothermal experiment Figure 6-16. Polymerization rate versus time for the photo-polymerization of nanocomposite underfill 162 162 163 164 166 Figure 6-17. Conversion versus time for the photo-polymerization of nanocomposite underfill 166 Figure 6-18. Absorbance of underfill with different filler loading 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: 8µm)[140] Figure 6-21. Comparison of light transmittance and the particle volume fraction for composite with different silica size 168 171 171 172 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 174 XV 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 Figure 6-29. A plot of load vs. displacement in a nanoindentation experiment Figure 6-30. Hardness of nanocomposite films after photo-curing 180 181 181 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 Figure 7-2. Surface initiation of epoxy curing reaction. Figure 7-3. Poyhedral oligosilsesquioxane (POSS) structure Figure 7-4. Synthesis route of POSS-containing underfill Figure 7-5. Chemical structure of cyanate ester monomer 196 196 197 199 200 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 Bond pad Board Solder ball Heated Fluxing Dispensing Chip Placement Solder Reflow Heated Solvent spray Underfill 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 Application Location On the wafer On the substrate Between chip and substrate, overmolding the chip Between chip and substrate Fluxing ability Yes Yes No No Material Form semi-solid (after B-stage) liquid solid liquid 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 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 O CH2 O Synonym EPON 828 EPON862 O O CH2 O CH2 O CH2 CH2 O O CH2 ] CH3 C CH3 O CH2 O EPON SU-8 n [ Name diglycidyl ether of bisphenol-A epoxy resin diglycidyl ether of bisphenol-F epoxy resin Epoxy phenol novolac resin O ERL4221 O 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 OH H2 C CH2 n Synonym LBR-6 Name Phenolic resin O HMPA Methyl tetra hydro phthalic anhydride O O DETDA NH2 H2N Diethyl toluene diamine 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 -40°C, 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. O R'NH2 + CH2 CHR OH 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 Synonym 2E4MZCN Name 1-cyanoethyl-2-ethyl-4methylimidazole CN CH2 CH2N N HOOC C11H23 COOH COOH C11Z-CNS 1-(2-Isocyano-ethyl)-2undecyl-1H-imidazole 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 = (ρ s − ρ f ) g 18η f ds Equation 1-1 Where the ρ s is the filler density, g is the acceleration of gravity, d s is filler diameter, ρ f and η 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 camera’s 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 1µm. 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 Φ = −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 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 Equation 1-3 Where A is the Hamaker constant for particle-particle interaction in vacuum, D is the 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 O Silica Particle OH OH Silica Particle O O + nH (aq) OH OH2 Silica Particle OH OH + nH (aq) Silica Particle OH2 OH2 OH Silica Particle O OH OH + nOH (aq) Silica Particle O O + 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 Bulk resistivity (ohm-cm) 2.67E+10 2.31E+09 1.66E+08 7.72E+09 1.15E+09 1.80E+07 Epoxy monomer Curing agent Underfill De-ionized water Bisphenol A amine anhydride Bisphenol A -amine Bisphenol A -anhydride water 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, ν , 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. ν ) and the bottom layer moves with the - 25 - smallest velocity(i.e. zero). However, the velocity gradient or shear rate, γ , is constant and given by : γ = dν dy Equation 1-4 The shear stress, τ , acting on the top plate is given by: τ= F A Equation 1-5 The viscosity, η ,is defined as the ratio of the shear stress to the shear rate: η= dν τ =τ dy γ 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: η = η 0 (1 + 2.5Φ) Equation 1-7 Where η is suspension viscosity, η 0 is pure liquid suspension, and Φ 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: Φ −[η ]Φ m Φ − 2.5Φ m ) ≈ η 0 (1 − ) Φm Φm Where Φ m is the maximum packing fraction of filler in the system η = η 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. Φ 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 A Φ max Einstein Eq. D Φ 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 ( Φ m ) is also modified. Hence, Equation 1-8 should be revised for colloidal dispersion with nanoparticle [54]: η ≈ η 0 (1 − Φ −2.5Φ 'm ) Φ 'm Equation 1-9 Where Φ 'm = 0.495( 2a 3 ) , α 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 1000 Viscosity of composite (Pa.s) 800 600 400 nano silica micron silica Theoretical calculation 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: 3µm, 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 2000°C. 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 (ε') * Tan (δ x 104) * Loss factor (ε'') * Dielectric field strength (kV/mm) * Resistivity (Ωm) * * Dielectric Properties at 1 MHz 25°C 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 SiCl4 + 2 H2O 2H2 + O2 + SiCl4 2 H2O SiO2 + 4 HCl 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 1000°C, 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 1000°C[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 Si OR + H2O + B BH OH Si OR Si OH + B + R B: condensation Si OH + Si OH + Si OH Si OR Si O Si Si O Si OH +H2O + 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 µm. Most of the nanosilica is synthesized by sol-gel method developed by Stöber 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 Stöber silica with submicron diameter. 2.1.3. Size control of nanosilica by Stöber 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 Stöber 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 20 Volume (%) 15 10 5 0 0.01 A01 A02 A03 A04 0.1 1 10 Particle size (um) Figure 2-6. Particle size distribution of as-synthesized silica 0.6 Silica size (um) 0.5 0.4 0.3 0.2 0.1 0 0 0.5 1 1.5 2 Ammonia concentration (mol/l) A01 A02 A03 A04 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 - γ LV vapor liquid θ solid γ SL γ SV Figure 2-8. Relationship between interfacial tension and contact angle As shown in Figure 2-8, contact angle ( θ ) 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 Å), thus it is extremely surface sensitive. Contact angle is related to the surface tension of the solid surface ( γ SV , some time is also written as γ S ), the liquid surface ( γ LV , or written as γ L ), and the interface between solid and liquid ( γ SL ). Their relationship is given by Young’s equation: γ L cos(θ ) = γ S − γ 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°, γs - γsl is negative, and the surface asperities cannot be filled by the advancing liquid, and this situation is non-wetting. When θ is between 0° and 90°, γs - γsl is smaller than the surface tension of the liquid, thus liquid film can not be formed spontaneously. Nevertheless, γs - γsl 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 (γl + γsl - γs < 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 = γ S − γ SL − γ 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 linker n Si 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 O Vinyltrimethoxysila ne CA-VI O Si O O Vinyl group γ-aminopropyltriethoxysilane CA-NH H2N Amino group H2N O Si O O γ-glycidoxypropyltrimethoxysilane CA-EP O O Si O O O Epoxide group O O 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 90ºC 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 110ºC 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 110ºC 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 90ºC 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. H H 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 ( γ 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 γ c of the substrate, the spontaneous spreading of the liquid will not occur on this substrate. 1 0.8 0.6 0.4 0.2 0 20 cos(θ) γc 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 ( γ 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 SiO2 SiO2 SiO2 CA-ME CA-VI CA-NH CA-EP 40.9 40.0 24.2 16.5 42.5 57.3 47.4 56.8 35.8 23.0 30.3 23.3 Table 2-5. Critical surface tension of SiO2 surfaces with different silane treatment Substrate type Treatment silane Functional group Critical surface tension of treated surface (mJ/m2) SiO2 SiO2 SiO2 SiO2 CA-ME CA-VI CA-NH CA-EP Methyl Vinyl Amino Epoxide 24.3 29.4 36.3 47.6 From Table 2-5, it can be seen that the surface with epoxide silane treatment has the highest γ c . The amino silane takes the second place. Vinyl and methyl silane treated surface has lower γ c than the other two. It seems the silane with polar functional group is more effective to change the γ c and benefit the wetting for liquid. The surface tension of pure Bisphenol A monomer ( γ l ) is about 42.5 mJ/m2, which is lower than γ 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. (μm) (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 8 001:stirring, untreated 002:sonication, untreated 003:stirring, treated 007:sonication, treated Volume (%) 6 4 2 0 0.01 0.1 1 10 100 1000 10000 particle diameter (um) Figure 2-14. Nanosilica dispersion with different pre-treatments 10 8 006:high conc., long time 007:low conc., long time 010:high conc., short time 011:low conc., short time Volume (%) 6 4 2 0.01 0.1 1 10 100 1000 10000 particle diameter(µm) 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 0.11μm, 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. high conc. Particle average size (um) 10 low conc. 1 0.1 6 Treatment time (hrs) 48 Figure 2-16. Average size of nanosilica with different treatment conditions - 60 - Y OR Si OR OR organofunctional silane Y hydrolysis of silane + 3H2O Y Si OH + 3ROH OH OH Y Si HO HO O H O Si O H Si HO HO HO OH Si O OH Si condensation of silnaol on silica surface HO O Si O Si O Si O Y O Y Si O Si nanosilica surface form a stable organic functionalization on silica 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 8 004:high conc., long time 005:low conc., long time 008:high conc., short time 009:low conc., short time 002:sonication, untreated 6 Volume (%) 4 2 0 0.01 0.1 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 Stöber 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 400°C. 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 170°C[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 400°C, 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 400°C. 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 800°C, 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 400°C[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 200ºC, 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 120°C or 200°C for 6 hours still show similar weight loss behavior to those without drying, because the weight loss during heating under 400°C is reversible. With large surface area, nanosilica can easily pick up the moisture again in a typical laboratory environment. If an ultra high - 68 - temperature was used to dry the particles (1000°C), the nanosilica show no physically absorbed water any more (the blue line in Figure 2-26). Some weight loss continues to occur above 400°C, 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 didn’t 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 200ºC 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 1000 Viscosity (Pa.s) 100 10 1 0.1 0.1 pure underfill CA-NH treated CA-EP treated untreated 1 10 Shear rate (1/s) 100 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 3μm 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. γ-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 Chemical structure H2C HC H2C O O H3C CH3 C CH3 O C O C O O CH2 CH CH2 O Hexahydro-4methylphthalic anhydride 1-(2-isocyano-ethyl)2-undecyl-1Himidazole γ-aminopropyltriethoxysilane HMPA C11Z-CNS CA-EP CN CH2 CH2 N N C11H23 HOOC COOH COOH O O O Si O O 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 a γ-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 5°C/min, from ambient temperature to 300ºC, 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 200ºC at 5ºC/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 150ºC for one hour and 180ºC 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 18×6×2 mm. The test was performed in a single cantilever mode. The temperature was increased from room temperature to 250°C at a heating rate of 3ºC/min, while the storage modulus (E’), loss modulus (E”) and tanδ 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 5×5×2 mm. The sample was heated in the TMA furnace at 5ºC/min from room temperature to 200ºC. The CTE before the Tg is defined as α1 and after the Tg as α2. 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 85°C 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 C O O C 2 C O O C N R2 N R1 CH O R R1 N R2 N O C O N R2 N R1 + H2C O CH R CH O CH2 + N 3 2 R2 R H2C O : C O CH2 5 1 N R1 C O O O C 2 C O O C O H2C CH O CH2 O C O O C C O O H2C CH O C O C O O- R O - CH O C C O CH2 O C O 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 140 130 120 Tg (ºC) 110 100 90 80 0 10 untreated treated 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 Pa·s, 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 500 Viscosity (Pa.s) 20 10 0 0 10 20 30 40 50 Filler loading (wt%) 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 nano–size 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 80 Transmittance (%) 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 80 CTE(ppm/°C) 70 60 50 40 0 10 untreated treated 20 Filler loading (wt%) 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 5 Moisture uptake (wt%) 4 3 2 1 0 0 10 20 Filler loading (w t%) 30 40 (a) 3 micron 6 Moisture uptake (wt%) 5 4 3 2 1 0 0 5 10 15 20 25 Filler loading (wt%) 30 35 40 nano-untreated nano-treated (b) - 86 - 3 micron 7 Moisture uptake (wt%) 6 5 4 3 2 1 0 0 5 10 nano-untreated nano-treated 15 20 25 Filler loading (wt%) 30 35 40 (c) 3 micron 7 Moisture uptake (wt%) 6 5 4 3 2 1 0 0 5 10 15 20 25 Filler loading (wt%) 30 35 40 nano-untreated nano-treated (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 85°C - 87 - and 85% RH. With the assumption that all the weight increasing during the 85°C/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 Fick’s second law of diffusion, which can be expressed as[89]: Mt 8 = 1− 2 M∞ π ⎧ − D(2n + 1) 2 π 2 t ⎫ 1 ⎬ ∑ (2n + 1) 2 exp⎨ h2 n =0 ⎭ ⎩ ∞ 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∞ πh 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 2.5 Moisture uptake (%) 2 1.5 1 0.5 0 0 2 4 Time (hours) 6 8 3 micron nano untreated nano treated Figure 3-10. Kinetics of moisture uptake for the samples Table 3-2. Moisture absorption kinetics parameter Materials without filler 0.1727 micron filler 0.2319 untreated nanosilica 0.7694 treated nanosilica 0.6963 k (rate constant) 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 1.35 Density (g/cm3) 1.3 1.25 1.2 1.15 0 10 20 Fille r lo ad in g (w t%) 30 40 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 μm and 75 μm, 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 210°C. Silica fillers in the size of 3 μm 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 35°C 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 (~3µm) 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 μm and 2 μm, 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 μm) 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 μm, 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 5°C/min, from ambient temperature to 250ºC 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 200ºC at 5ºC/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 150ºC 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 -120ºC to 200°C at a heating rate of 3ºC/min. Liquid nitrogen was used for the low temperature testing. The amplitude applied to the samples in the experiments was 15 μm, 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 140 130 120 Tg (°C) 110 100 90 80 0 10 20 30 Weight Percentage (wt%) 40 50 Micron size Figure 3-15. Glass transition temperature of the silica composites - 98 - Nano size Micron size mono-dispersed 130 125 120 115 Tg (°C) 110 105 100 95 90 85 80 67 68 69 Micron size bimodel Control 70 71 72 73 Weight Percentage (wt%) 74 75 76 Figure 3-16. Glass transition temperature of the silver composites Nano size 130 Micron size Control 125 Tg (°C) 120 115 110 105 40 41 42 43 44 Weight Percentage (wt%) 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 5 min sonication 132 130 128 126 Tg (°C) 124 122 120 118 116 1 2 30 min sonication Control 3 4 Weight Percentage (wt%) 5 6 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 α 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 (80-120°C), 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 - 550 500 Loss modulus (MPa) 450 400 350 300 250 200 150 100 50 0 -100 nanocom posite sam ple m icron com posite sam ple control pure polym er -50 0 50 100 150 200 T em perature (degree C ) Figure 3-20. Dynamic loss moduli of the silica composites and the blank resin 350 Loss modulus (MPa) 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 α 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 β transition or relaxation. The origin of the β 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 C CH3 C CH3 O C C O O H2C CH O C O C O H2C O O CH2 O O CH O C C O CH2 O C O O Figure 3-22. Molecular structure of anhydride/epoxy polymer O CH3 C CH3 O O O C O C O 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 (α transition) or the segmental motion of molecule chain (β - 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 β 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 (ε’) to minimize the capacitive coupling effects, and a low loss factor (ε”) 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 (θ). The phase angle shift is measured by comparing the applied voltage to the measured current. The measured response is separated into a complex form (ε*) that can be separated into capacitive and conductive components giving the real part of permittivity (ε’) and loss factor (ε”): ε * = ε '− jε " 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 - ' ' ε ' = ε dipole + ε ion " " ε " = ε dipole + ε ion Equation 4-2 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: ' ε dipole = ε ∞ + εs −ε∞ 1 + w 2τ 2 Equation 4-4 " ε dipole = (ε s − ε ∞ ) wt 1 + w 2τ 2 Equation 4-5 Where ε ∞ is the dielectric permittivity at infinite frequency, ε s is the static dielectric permittivity at zero frequency, w is the angular frequency, and τ 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 ( τ ) to achieve a polar orientation can become very large. So for a glassy state epoxy, the dipolar relaxation due to long range molecular mobility doesn’t 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]: ε"= σ wε 0 Equation 4-6 Where σ is the bulk ionic conductivity, ε 0 is the absolute permittivity of free space (8.85×10-12 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]: ' ε ion = C 0 Z 0 sin( nπ −( n +1) σ 2 )w ( ) 2 ε0 Equation 4-7 " ε ion = σ σ nπ − C 0 Z 0 cos( ) w −( n +1) ( ) 2 wε 0 2 ε0 Equation 4-8 Where σ is the ionic conductivity, ε 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: τ= ε 1d 2 + ε 2 d1 σ 1d 2 + σ 2 d1 Equation 4-9 d1 + d 2 M ε ∞ −W = ε0 + ε1 d1 ε2 d2 Equation 4-10 1 M ε sM −W = ε ∞ −W (1 + d1 d 2 σ1 ε1 1 − ε2 σ2 σ1 d1 + ε2 ε1 σ2 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: φ( ε1 − ε ε −ε ) + (1 − φ )( 2 )=0 ε 1 + 2ε ε 2 + 2ε Equation 4-12 Where φ 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 μm, and the width of each electrode is 150 μm. 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 150ºC for 1 hour. The DEA experiments were performed from 35°C to 250°C with a stepping temperature of 5°C and a frequency sweep from 0.01 to 100,000 Hz varying logarithmically. The dielectric permittivity (ε’) and dielectric loss factor (ε”) were calculated by the pre-installed software. After the first test, all the samples were put in an 85°C/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 120°C 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 ε’(f,T) and ε”(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 ( α T ) which follows the Williams-Landel-Ferry (WLF) equation[114]: log α T = − c1 (T − Tref ) c2 + T − Tref Equation 4-13 In the current study, the reference temperature was chosen as the 150°C 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 2.000 1.000 X-shift 0 -1.000 -2.000 -3.000 -4.000 -5.000 125.0 shift factor fitting curve 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 5 4 3 Log [e''] 2 1 0 -1 -2 -3 -4 -5 -4 -3 -2 -1 0 1 2 3 4 Log [frequency (Hz)] 5 6 7 8 9 10 control sample micron size composite nano size composite 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. 3.0 2.5 Moisture uptake (wt%) 2.0 1.5 1.0 0.5 0.0 0 1 2 3 Aging time (day) 4 5 6 control micron size nano size 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. ε"= σ σ = wε o 2πfε 0 ⇒ log(ε " ) = log( σ ) − log( f ) 2πε 0 Equation 4-14 Where f is the frequency (s-1). - 121 - 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 85°C/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 70~110˚C. 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 10000 1000 Loss Factor 100 10 1 0.1 0.01 0.001 0 –––––– control – – – micron size –––– · nano size 50 100 150 Temperature (°C) 200 250 (a) 1.0E5 10000 1000 Loss Factor 100 10 1 0.1 0.01 0.001 0 50 100 150 Temperature (°C) 200 250 –––––– control – – – micron size –––– · nano size (b) Figure 4-10. Loss factor of three samples after curing, (a) 1Hz; (b) 1000Hz - 124 - 1.0E5 10000 1000 Loss Factor 100 10 1 0.1 0.01 0.001 0 –––––– control – – – micron size –––– · nano size 50 100 150 Temperature (°C) 200 250 (a) 1.0E5 10000 1000 Loss Factor 100 10 1 0.1 0.01 0.001 0 50 100 150 Temperature (°C) 200 250 –––––– control – – – micron size –––– · nano size (b) Figure 4-11. Loss factor of three samples after aging under humidity, (a) 1Hz; (b) 1000Hz - 125 - 1.0E5 10000 1000 Loss Factor 100 10 1 0.1 0.01 0.001 0 –––––– control – – – micron size –––– · nano size 50 100 150 Temperature (°C) 200 250 (a) 1.0E5 10000 1000 Loss Factor 100 10 1 0.1 0.01 0.001 0 50 100 150 Temperature (°C) 200 250 –––––– control – – – micron size –––– · nano size (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 O O O 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 EPON828 53.6 wt% 80.9 wt% HMPA 46.1 wt% DETDA 19.1 wt% C11Z-CNS 0.3 wt% - Anhydride-based formulation Amine-based formulation 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 γ was applied to sample. If the maximum strain amplitude is γ0 and the temperature was held constant during the frequency sweep, the strain γ can be described as: γ = γ 0 sin ωt Where ω is the angle frequency, t is the time. Equation 5-1 The complex modulus of the sample will create resistance to the oscillatory shear: σ = G ∗γ 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 - 10 3 Modulus (Pa) 10 2 10 1 elastic modulus (G') viscous modulus (G") 10 0 0.01 0.1 1 10 Frequency (Hz) Figure 5-2. Elastic and viscous modulus as a function of frequency for nanosilica/anhydride mixture 10 10 10 10 10 10 2 1 Modulus (Pa) 0 -1 -2 Elastic modulus (G') Viscous modulus (G") -2 -1 0 1 2 -3 10 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 nanosilica 30 vol% System interpretation microstructure amine G’G” Elastic behavior dominate Almost constant Gel like Strong filler-filler interaction 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 10 6 5 Viscosity (Pa.s) 10 10 10 10 10 10 4 silica/HMPA silica/DETDA 3 2 1 0 -1 10 -1 10 0 10 1 10 2 10 3 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: Φ = − 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 (ε − 1) 2 3hν e (n 2 − 1) 2 A = kT + 4 (ε + 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 Planck’s constant, ν e is the main electronic absorption frequency for the dielectric permittivity (taken to be 3×1015 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 - Component Dielectric constanta Refractive indexb Hamaker constant, A (10-20 J) HMPA DETDA silica 4.6 21 3.82 1.455 1.477 1.463 6.04 6.61 6.22 a: measured by DEA, b: obtained from sample information from vendor silica/DETDA 0 0 0.5 1 silica/HMPA 1.5 2 2.5 Van der Waals potential, V/kT -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. Let’s 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 130 Glass transition temperature (degree C) anhydride 120 110 100 90 0 10 20 Silica loading (wt%) 30 40 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 today’s 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 - S + PF6- S S + PF6- 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 5°C/min, from ambient temperature to 300ºC. The cured sample was left in the DSC cell and cooled to room temperature. Then the sample was reheated to 200ºC at 5ºC/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 µm. 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 120°C 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 18×6 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 250°C at a heating rate of 3ºC/min, while the storage modulus (E’), loss modulus (E”) and tanδ 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 tanδ 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 10×5 mm. The samples were heated in the TMA furnace at 5ºC/min from room temperature to 200ºC. The CTE before the Tg is defined as α1 and after the Tg as α2. 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 20ºC/min from ambient temperature to 800ºC 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 μm. 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). 100 Transmittance (%) 80 60 40 20 0 300 400 500 ~100nm silica ~20nm silica 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 ( λ 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 2.0 1.8 288nm Absorbance (a.u.) 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 300nm nanosilica pure epoxy photoinitiator 250 300 350 400 450 500 Wavelength (nm) Figure 6-6. UV absorbance of pure epoxy, nanosilica and photo-initiator 2.5 Absorbance (a.u.) 2.0 epoxy/photoinitiator with nanosilica epoxy/photoinitiator 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 BrØnsted acid. In the second step, the epoxy polymerization was initiated by the protonation of the epoxide oxygen with the BrØnsted 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 Photointiator hυ S + + PF6- + PF6- + R-H R + H PF6 Initiation O + O CH2 CH CH2 + H PF6 O+ H O CH2 CH CH2 PF6- Propagation and crosslinking O O CH2 O CH CH2 CH + O CH O+ H CH2 PF6- CH2 CH2 OH O+ H PF6 O CH2 CH CH2 O CH2 CH O O CH2 CH O+ H O CH2 CH CH2 CH2 PF6CH2 OH O O CH2 CH CH2 Figure 6-8. Reaction mechanism of cationic photo-polymerization - 156 - 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. 0.30 0.28 0.26 2925.39 2999.35 1605.85 0.24 0.22 0.20 0.18 0.16 1500.76 Absorbance 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 3520.93 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 - 916.89 846.83 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 didn’t 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 150°C 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 0.8 epoxy composite pure epoxy Normalized intensity of epoxy peak 0.6 0.4 0.2 0.0 0 photo cure thermal cure 100 Time (s) 200 300 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 30°C, 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. ΔH = ∫ H (t )dt Equation 6-1 ΔH total = ΔH photo + ΔH 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 ΔH photo during photo-polymerization. The partial reaction heat ΔH (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= ΔH (t ) ΔH total da heat flow = [M ] × dt ΔH total Equation 6-3 Rp = [M ] × Equation 6-4 Where [ M 0 ] is the initial concentration of the monomer. ΔH (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 0 20 30 40 0.01 Rp (mol / l. s) 0.008 0.006 0.004 0.002 0 0 5 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 ]( φ ε [ 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, φ is the quantum yield of active center formed by per photon absorbed, ε 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-Lambert’s law, the absorption coefficient ( ε ) can be calculated as the material constant. The equations are shown below: A = log I I0 Equation 6-6 Equation 6-7 %T = 100 × 10 − A % Beer-Lambert’s law A = ε [ PI ]L Equation 6-8 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 μm. Figure 6-18 shows the absorbance of the photo-initiator in the nanocomposite. The absorption coefficient ( ε ) 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 ](ε [ 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 underfill’s 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 - 100 Transmittance (%) 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: 8µm)[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 0.8 Transmittance 0.6 0.4 0.2 0 nanosize filler micron size filler 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 - 6.5.2. Glass transition temperature Glass transition temperature (°C) 100 80 60 40 20 0 0 10 20 filler loading (wt%) 30 40 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 30°C 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 Filler (wt%) Decompos. temp. (°C) theoretical measured 1st onset 2nd onset 0 0 0 336.9 545.2 10 10 12.0 337.9 551.9 20 20 19.3 341.0 556.0 30 30 28.3 338.5 557.3 40 40 35.5 346.6 576.5 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 160°C 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 243°C. After this small weight loss, there are two major decomposition steps starting from around 300°C and 550°C, 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 70 CTE (μm/m°C) photo curing 60 50 40 0 10 20 Filler loading (wt%) 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 μm/m°C for pure epoxy to 50.2 μm/m°C 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 80 DMA Tg 60 40 20 0 0 10 20 30 40 50 Filler loading (wt%) 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. effective Young’s modulus of the composite can be derived as[146]: E = 2 μ [1 + Where K = K 0 {1 + c( K1 − K 0 ) } K 0 + 3γ 0 (1 − c)( K 1 − K 0 ) c ( μ1 − μ 0 ) } μ 0 + 2δ 0 (1 − c)( μ1 − μ 0 ) Then, the 3K − 2 μ ] 2(3K + μ ) Equation 6-9 Equation 6-10 μ = μ 0 {1 + γ0 = δ0 = Equation 6-11 K0 3K 0 + 4 μ 0 3( K 0 + 2 μ 0 ) 15 K 0 + 20 μ 0 En 3(1 − 2ν n ) Equation 6-12 Equation 6-13 Kn = Equation 6-14 - 177 - μn = En 2(1 + ν n ) , n = 0, 1 Equation 6-15 Where E = effective bulk modulus, μ = effective shear modulus, and c is the filler volume fraction. Young’s modulus and the Possion’s ratio of the matrix are E 0 and ν 0 , that of the particles are E1 and ν 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 Density ρ 0 =1.16 g/cm3 Modulus (GPa) E 0 =2.5GPa E1 =74GPa Poisson’s ratio ν 0 =0.4 epoxy silica ρ1 =2.20 g/cm3 ν 1 =0.19 Exp. 8 7 6 Modulus (GPa) 5 4 3 2 1 0 0 0.1 0.2 Model 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 6 Load (mN) 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 0.45 Hardness (GPa) 0.4 0.35 0.3 0 10 20 30 40 50 Filler loading (wt%) 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 -40°C) 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 -40°C), 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 95°C 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 120°C 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 µm 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 CH2 O O CH2 CH CH2 CH2 O O CH2 CH CH2 CH2 O O CH2 CH CH2 CH3 H3C C CH3 H3C C CH3 H3C C CH3 H3C C CH3 O O CH2 CH CH2 O O CH2 CH CH2 O O CH2 CH CH2 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 Stöber 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( γ 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 γ 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 β 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 OR OH OH + RO Si OH OR OH N N SiO2 OH OH O Si OH OH N N + RX SiO2 OH O Si OH OH N N+ R X- R: alkyl group X- : Cl-, Br-, BF4-, PF6- Figure 7-1. Silica surface grafting of imidazolium salt as a catalyst OH SiO2 OH N OH N+ R Δ X- -RX SiO2 OH O Si OH OH N N OH O- O Si OH + O O SiO2 OH OH N N+ OH O Si OH O 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 R R O O Si O Si O O O Si O Si O R R O Si Si O R Si O O Si R 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 (OSi(CH3)2H)8 R T1 1 O Si O Si O O O Si Si O O O Si Si O (OSi(CH3)2CH2CH2R1)n O Si O O Si O Si O O Si (OSi(CH3)2H)8-n R2 T2 O O Si O Si O O Si Si O O O Si Si O (OSi(CH3)2CH2CH2R1)n O Si O O Si (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 180°C, while exposure to moist environments limits this to 150°C. 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 270°C 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 C O O C N 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 AUTHOR’S 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, Georgia Institute of Technology [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 20-100μm 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 Technology Conference, pp. 761-766, 2004. 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 Conference, pp. 336-341, 2003. 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 Technology Conference, pp. 951-956, 2003. H. N. Stein, The Preparation of Dispersions in Liquids, vol. 58. New York, 1996. B. Dobiáš, 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.Schmücker, Thesis, Ruhr-Universität 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 Applied Polymer Science, vol. 90, pp. 1923-1931, 2003. 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 Technology, vol. 83, pp. 158-164, 2001. [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 Applied Polymer Science, vol. 31, pp. 1771-1784, 1986. [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 Technology, vol. 63, pp. 1671-1687, 2003. - 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 Conference, pp. 386-389, 2002. [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 -