EFFECTS OF CARBON NANOTUBE (CNT) DISPERSION AND INTERFACE CONDITION ON THERMO-MECHANICAL BEHAVIOR OF CNTREINFORCED VINYL ESTER by Seyed Morteza Sabet A Dissertation Submitted to the Faculty of The College of Engineering and Computer Science In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy Florida Atlantic University Boca Raton, FL May 2016 Copyright 2016 by Seyed Morteza Sabet ii ACKNOWLEDGEMENTS I wish to thank all those who supported me in the completion of this study and made the way of research smooth for me. First of all, I would like to convey my sincere gratitude to my advisor and committee chair, Dr. Hassan Mahfuz, Department of Ocean and Mechanical Engineering, who provided me with fundamental advices. He always supported me throughout my research with dedication, patience and encouragement. I also gratefully thank my committee members; Dr. Javad Hashemi, Department of Ocean and Mechanical Engineering, for his valuable comments towards the perfection of my dissertation and his unique support in my academic career; Dr. Andrew C. Terentis, Department of Chemistry and Biochemistry, for his insightful directions in our discussions and for his kindness by allowing me to utilize research facilities to the best of his capabilities; Dr. Gary Salivar, Department of Ocean and Mechanical Engineering, for his valuable recommendations and suggestions; and Dr. Francisco Presuel-Moreno, Department of Ocean and Mechanical Engineering, for always being ready to passionately answer my questions. I would like to extend my thanks to Dr. Ali Zilouchian, Associate Dean for Academic Affairs of the College of Engineering for his valuable supports and key advices. A warm thank to Dr. Leif Carlsson for our technical discussions. I would also like to send a heartfelt acknowledgment to my friends at Florida Atlantic University who made all these years vey joyful and memorable. iv Finally, and most importantly, I could not have accomplished this without the support of my family. I would like to thank my wonderful wife, who has been a constant source of patience, concern, and inspiration during this journey. She has made countless sacrifices to help me get to this point. Along with her, I want to acknowledge my son, who has never known his dad as anything but student, and has been a great source of love and relief from scholarly endeavor. From the bottom of my heart, I wish to thank my beloved parents for their love, prayers, caring, and sacrifices for educating and preparing me for my future. Also I express my thanks to my favorite brother, and my dear sisters, who have always loved me unconditionally and provided me with the best support throughout my life. v ABSTRACT Author: Seyed Morteza Sabet Title: Effects of Carbon Nanotube (CNT) Dispersion and Interface Condition on Thermo-Mechanical Behavior of CNT-Reinforced Vinyl Ester Institution: Florida Atlantic University Dissertation Advisor: Dr. Hassan Mahfuz Degree: Doctor of Philosophy Year: 2016 In fabrication of nanoparticle-reinforced polymers, two critical factors need to be taken into account to control properties of the final product; nanoparticle dispersion/distribution in the matrix; and interfacial interactions between nanoparticles and their surrounding matrix. The focus of this thesis was to examine the role of these two factors through experimental methodologies and molecular-level simulations. Carbon nanotubes (CNTs) and vinyl ester (VE) resin were used as nanoparticles and matrix, respectively. In a parametric study, a series of CNT/VE nanocomposites with different CNT dispersion conditions were fabricated using the ultrasonication mixing method. Thermomechanical properties of nanocomposites and quality of CNT dispersion were evaluated. By correlation between nanocomposite behavior and CNT dispersion, a thermomechanical model was suggested; at a certain threshold level of sonication energy, CNT vi dispersion would be optimal and result in maximum enhancement in properties. This threshold energy level is also related to particle concentration. Sonication above this threshold level, leads to destruction of nanotubes and renders a negative effect on the properties of nanocomposites. In an attempt to examine the interface condition, a novel process was developed to modify CNT surface with polyhedral oligomeric silsesquioxane (POSS). In this process, a chemical reaction was allowed to occur between CNTs and POSS in the presence of an effective catalyst. The functionalized CNTs were characterized using TEM, SEM-EDS, AFM, TGA, FTIR and Raman spectroscopy techniques. Formation of amide bonds between POSS and nanotubes was established and verified. Surface modification of CNTs with POSS resulted in significant improvement in nanotube dispersion. In-depth SEM analysis revealed formation of a 3D network of well-dispersed CNTs with POSS connections to the polymer. In parallel, molecular dynamics simulation of CNT-POSS/VE system showed an effective load transfer from polymer chains to the CNT due to POSS linkages at the interface. The rigid and flexible network of CNTs is found to be responsible for enhancement in elastic modulus, strength, fracture toughness and glass transition temperature (Tg) of the final nanocomposites. vii DEDICATION To my wife, Zahra To our son, Parsa To my parents EFFECTS OF CARBON NANOTUBE (CNT) DISPERSION AND INTERFACE CONDITION ON THERMO-MECHANICAL BEHAVIOR OF CNTREINFORCED VINYL ESTER LIST OF TABLES ........................................................................................................... xiii LIST OF FIGURES ......................................................................................................... xiv CHAPTER 1. INTRODUCTION ....................................................................................... 1 CHAPTER 2. LITERATURE REVIEW ............................................................................ 6 2.1 Carbon Nanotube.................................................................................................. 6 2.2 Polymer Matrix .................................................................................................... 8 2.2.1 Molecular Structure ...................................................................................... 8 2.2.2 Glass Transition ............................................................................................ 9 2.2.3 Vinyl Ester Thermosetting Polymer ........................................................... 11 2.3 Ultrasonication ................................................................................................... 12 2.3.1 Acoustic Cavitation Mechanism ................................................................. 13 2.3.2 Physical Properties of Liquid Medium ....................................................... 14 2.3.3 Intensity of Irradiation ................................................................................ 15 2.3.4 Mixing Time ............................................................................................... 16 2.4 Studies on the Dispersion of CNTs in VE Resin ............................................... 16 2.5 CNT/Polymer Interface ...................................................................................... 17 2.5.1 Experimental Characterization of the Interface .......................................... 18 2.5.2 Computational Characterization of the Interface ........................................ 19 2.5.3 Modification of the Interface ...................................................................... 20 2.5.4 Polyhedral Oligomeric Silsesquioxane (POSS) .......................................... 21 2.5.5 POSS-Functionalization of CNTs ............................................................... 21 2.5.6 Studies on the Incorporation of POSS-Treated CNTs in Polymer ............. 22 ix CHAPTER 3. RESEARCH MOTIVATION .................................................................... 23 3.1 Optimization of Nanocomposite Synthesis Process ........................................... 23 3.2 Improvement of the CNT/VE Interface ............................................................. 23 3.3 Molecular-Level Investigation of the Interface .................................................. 25 CHAPTER 4. MATERIALS, SYNTHESIS AND FABRICATION ............................... 26 4.1 Materials ............................................................................................................. 26 4.1.1 Vinyl Ester .................................................................................................. 26 4.1.2 Carbon Nanotubes ....................................................................................... 27 4.1.3 POSS Nanostructures .................................................................................. 28 4.1.4 Chemicals .................................................................................................... 29 4.2 Surface Modification of MWCNTs by POSS .................................................... 30 4.3 Preparation of Nanocomposites ......................................................................... 32 4.3.1 Nanocomposite Fabrication – Phase 1 ........................................................ 32 4.3.2 Nanocomposite Fabrication – Phase 2 ........................................................ 35 CHAPTER 5. EXPERIMENTATION.............................................................................. 36 5.1 Characterization of Nanomaterials ..................................................................... 36 5.1.1 Transmission Electron Microscopy ............................................................ 36 5.1.2 Scanning Electron Microscopy of CNTs .................................................... 36 5.1.3 Atomic Force Microscopy .......................................................................... 36 5.1.4 Raman Spectroscopy................................................................................... 38 5.1.5 Fourier Transform Infrared Spectroscopy .................................................. 38 5.1.6 Physical Stability in Tetrahydrofuran ......................................................... 39 5.2 Testing of Vinyl Ester and its Nanocomposites ................................................. 40 5.2.1 Nanoindentation .......................................................................................... 40 5.2.2 Three-Point Bending ................................................................................... 43 5.2.3 SEM of Fracture Surfaces ........................................................................... 44 5.2.4 Optical Microscopy..................................................................................... 45 5.2.5 Differential Scanning Calorimetry .............................................................. 45 5.2.6 Thermogravimetric Analysis ...................................................................... 46 x CHAPTER 6. EFFECTS OF ULTRASONICATION ENERGY..................................... 48 6.1 Introduction ........................................................................................................ 48 6.2 Flexural Behavior ............................................................................................... 48 6.3 The 0.25 wt% MWCNT/VE Nanocomposites ................................................... 51 6.4 The 0.5 wt% MWCNT/VE Nanocomposites ..................................................... 58 6.5 Nanoindentation Curves ..................................................................................... 62 6.6 Thermal Properties ............................................................................................. 66 CHAPTER 7. SURFACE MODIFICATION OF CARBON NANOTUBES .................. 71 7.1 Introduction ........................................................................................................ 71 7.2 Investigation of Chemical Bonding between CNTs and POSS ......................... 71 7.3 Electron Microscopy Examinations of CNTs .................................................... 74 7.4 AFM Studies of CNTs ....................................................................................... 78 7.5 Physical Stability of CNTs ................................................................................. 81 7.6 Thermal Behavior of Nanomaterials .................................................................. 83 CHAPTER 8. EFFECTS OF INTERFACE MODIFICATION ....................................... 85 8.1 Introduction ........................................................................................................ 85 8.2 Microstructural Studies ...................................................................................... 86 8.2.1 Optical Microscopy of Thin Nanocomposite Coatings .............................. 86 8.2.2 SEM Studies of Fracture Surfaces .............................................................. 89 8.3 Mechanical Properties ...................................................................................... 102 8.4 Thermal Properties ........................................................................................... 110 8.4.1 Differential Scanning Calorimetry (DSC) Results ................................... 110 8.4.2 Thermogravimetric Analysis (TGA) Results ............................................ 113 CHAPTER 9. THEORETICAL STUDY OF THE EFFECT OF INTERFACE ............ 117 9.1 Introduction ...................................................................................................... 117 9.2 Preparation of Molecular Models ..................................................................... 118 9.2.1 VE Model .................................................................................................. 118 9.2.2 CNT/VE Models ....................................................................................... 119 9.3 Force Fields ...................................................................................................... 122 xi 9.4 MD Simulation Details..................................................................................... 122 9.5 Simulation of Unidirectional Tension Tests..................................................... 124 9.6 Results and Discussion ..................................................................................... 126 CHAPTER 10. SUMMARY AND FUTURE WORK ................................................... 130 10.1 Summary .......................................................................................................... 130 10.2 Recommendations for Future Work ................................................................. 132 APPENDIXES ................................................................................................................ 134 Appendix A. Force Field ............................................................................................. 135 Appendix B. FORTRAN Code for Energy Minimization of MD Models ................. 139 Appendix C. JAVA Code Used for Making Single-Walled CNT Model ................... 150 REFERENCES ............................................................................................................... 155 xii TABLES Table 1: The physical and mechanical properties of Derakane 8084 vinyl ester resin at room temperature. ........................................................................................... 26 Table 2: Technical specifications of the as-received MWCNTs in this study. ................. 28 Table 3: The physical properties of POSS-NH2 hybrids. ................................................. 29 Table 4: Different sonication parameters examined in this study. .................................... 34 Table 5: Plasticity index of VE and its nanocomposites................................................... 63 Table 6: Raman data from the characteristic bands for the MWCNTs............................. 74 Table 7: The mechanical properties of VE and its nanocomposites. .............................. 104 Table 8: Details of the equilibrated molecular models for MD simulation. ................... 124 Table 9: The mechanical properties of materials from MD simulation. ......................... 129 xiii FIGURES Figure 1: The chiral vector and chiral angle in a graphite plane [7]. .................................. 7 Figure 2: (a) Zigzag, (b) armchair, and (c) chiral carbon nanotubes [7]. ........................... 8 Figure 3: The schematic representation of (a) thermoplastic polymer and (b) thermoset polymers [7]. .................................................................................... 9 Figure 4: The different behavior of amorphous and crystalline polymer melts during a cooling cycle. .................................................................................................. 11 Figure 5: Chemistry of a vinyl ester resin (asterisks denote the reaction sites at both ends) [7]............................................................................................................. 12 Figure 6: Schematic representation of ultrasonic cavitation and implosion [44]. ............ 14 Figure 7: Molecular structure of polyhedral oligomeric silsesquioxanes (POSS) [78]. .................................................................................................................. 21 Figure 8: The attachment of a COOH group to the surface of CNT. ................................ 27 Figure 9: The molecular structure of POSS-NH2 hybrids. ............................................... 29 Figure 10: The successive steps used for the surface modification of MWCNTs with POSS (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet)........... 31 Figure 11: The prepared setup for ultrasonication process (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). .......................................................... 33 Figure 12: The basic principles of AFM technique. ......................................................... 37 Figure 13: The Raman spectrometer used in this study (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). ......................................................... 38 xiv Figure 14: The FTIR-ATR spectrometer used for the experimental nanomaterials (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). .................... 39 Figure 15: The schematic load-displacement curve in a nanoindentation test [97]. ......... 41 Figure 16: The schematic illustration of contact geometry through nanoindentation test using a Berkovich indenter [97]. .............................................................. 42 Figure 17: The nanoindentation testing setup for experimental materials (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). ................ 43 Figure 18: The 3-point bending setup used in this study (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). ......................................................... 44 Figure 19: The DCS apparatus used for thermal characterizations in this study (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). ................... 46 Figure 20: The TGA equipment used for thermal analysis of experimental materials (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). .................... 47 Figure 21: The effect of sonication conditions on the flexural behavior of composites containing (a) 0.25 wt% and (b) 0.5 wt% CNTs. ...................... 49 Figure 22: SEM micrographs of the fracture surface of (a) the neat VE polymer and (b) nanocomposite fabricated with 0.5 wt% CNT content. .............................. 51 Figure 23: The effect of sonication energy on the elastic modulus of composites synthesized with 0.25 wt% CNT content. ..................................................... 52 Figure 24: SEM micrographs of the fracture surface of composite with 0.25 wt% CNT content sonicated at 70% amplitude for 3 minutes (10 kJ). ................... 54 xv Figure 25: SEM micrographs of the fracture surface of composites with 0.25 wt% CNT content sonicated at 50% amplitude for (a) 9 min (20 kJ) and (b) 18 min (40 kJ). ................................................................................................ 55 Figure 26: SEM micrographs of the fracture surface of composites with 0.25 wt% CNT content sonicated at 70% amplitude for (a, b) 18 min (60 kJ) and (c, d) 24 min (80 kJ). ....................................................................................... 56 Figure 27: The effect of sonication parameters on the elastic modulus of composites synthesized with 0.5 wt% CNT content. ......................................................... 59 Figure 28: SEM micrographs of the fracture surface of composites with 0.5 wt% CNT content sonicated at 50% amplitude for (a) 4 min (10 kJ), (b, c) 7 min (15 kJ), and (d, e) at 70% amplitude for 7 min (20 kJ)............................ 60 Figure 29: Representative load-displacement curves derived from nanoindentation tests. ................................................................................................................ 62 Figure 30: Nanoindentation creep deformation at maximum load of 800 μN. ................. 64 Figure 31: (a) modulus and (b) hardness variations along the indentation array.............. 66 Figure 32: The DSC heating curves of the neat VE polymer and its nanocomposites (at heating rate of 10 ºC min-1)........................................................................ 67 Figure 33: The glass transition temperature of the composites with (a) 0.25 wt% and (b) 0.5 wt% CNT content as a function of total sonication energy. ......... 69 Figure 34: The thermo-mechanical model suggested for CNT/VE system. ..................... 70 Figure 35: FTIR spectra of nanomaterials in this study. ................................................... 72 Figure 36: Raman spectra of MWCNTs before and after POSS-treatment. ..................... 74 Figure 37: (a) SEM image and (b) EDS spectrum of as-received MWCNT-COOH. ...... 75 xvi Figure 38: (a) SEM image and (b) EDS spectrum of POSS-treated MWCNTs. .............. 76 Figure 39: TEM images of the as-received MWCNT-COOH. ......................................... 77 Figure 40: TEM images of the POSS-functionalized MWCNTs. .................................... 78 Figure 41: AFM topography image of POSS-treated carbon nanotubes. ......................... 79 Figure 42: AFM amplitude image of as-received MWCNT and the corresponding height profile measurements. .......................................................................... 80 Figure 43: AFM amplitude image of POSS-modified MWCNT and the corresponding height profile measurements................................................ 81 Figure 44: Physical stability testing of CNTs in THF. ..................................................... 82 Figure 45: TGA thermograms of the nanomaterials. ........................................................ 84 Figure 46: Optical images of POSS/VE coatings at different POSS concentrations. ....... 87 Figure 47: Optical images of CNT/VE coatings at different CNT concentrations. .......... 88 Figure 48: Optical images of CNT-POSS/VE coatings at different nanotube concentrations. .............................................................................................. 89 Figure 49: SEM images of the fracture surface of the neat vinyl ester polymer. ............. 90 Figure 50: SEM images of the fracture surface of the 0.25 wt% POSS/VE specimen..... 91 Figure 51: SEM images of the fracture surface of the 0.5 wt% POSS/VE specimen....... 92 Figure 52: SEM images of the fracture surface of the 1.0 wt% POSS/VE specimen....... 93 Figure 53: SEM images of the fracture surface of the 0.25 wt% CNT/VE specimen. ..... 94 Figure 54: SEM images of the fracture surface of the 0.5 wt% CNT/VE specimen. ....... 95 Figure 55: SEM images of the fracture surface of the 1.0 wt% CNT/VE specimen. ....... 96 Figure 56: SEM images of the fracture surface of the 0.25 wt% CNT-POSS/VE specimen. ........................................................................................................ 98 xvii Figure 57: SEM images of the fracture surface of the 0.5 wt% CNT-POSS/VE specimen. ....................................................................................................... 99 Figure 58: SEM images of the fracture surface of the 1.0 wt% CNT-POSS/VE specimen. ..................................................................................................... 100 Figure 59: The schematic representation of the revealed fracture mechanism for CNT-POSS/VE nanocomposites. ................................................................ 101 Figure 60: Flexural stress-strain curves of the POSS/VE system. .................................. 103 Figure 61: Flexural stress-strain curves of the CNT/VE system. ................................... 103 Figure 62: Flexural stress-strain curves of the CNT-POSS/VE system. ........................ 104 Figure 63: The variations of fracture strain with nanofiller concentration. .................... 106 Figure 64: The variations of elastic modulus with nanofiller concentration. ................. 108 Figure 65: The variations of flexural strength with nanofiller concentration. ................ 109 Figure 66: DSC thermograms of the VE polymer and its nanocomposites. ................... 112 Figure 67: The glass transition temperature as a function of nanofiller concentration. ............................................................................................ 112 Figure 68: TGA and DTG heating curves for POSS/VE system. ................................... 114 Figure 69: TGA and DTG heating curves for CNT/VE system. .................................... 114 Figure 70: TGA and DTG heating curves for CNT-POSS/VE system. ......................... 115 Figure 71: The onset temperature as a function of nanofiller concentration. ................. 116 Figure 72: The decomposition temperature of VE and its nanocomposites. .................. 116 Figure 73: The molecular models of (a) VE resin molecule, (b) styrene molecule and (c) cross-linked VE fragment. ................................................................ 119 xviii Figure 74: The molecular models of (a) pristine CNT, (b) carboxylic acid functionalized CNT, (c) POSS-NH2 and (d) POSS-functionalized CNT. ........................................................................................................... 121 Figure 75: Computational cells of (a) the neat VE and (b) the nanocomposite with CNT-COOH after complete equilibration. ................................................... 124 Figure 76: Simulated stress-strain graphs of (a) VE polymer and its nanocomposite models containing (b) pristine CNT, (c) CNT-COOH, and (d) POSSfunctionalized CNT. ...................................................................................... 128 Figure 77: A schematic representation of valence interactions including (a) stretch, (b) angle, (c) torsion, and (d) inversion terms. ............................................... 136 xix CHAPTER 1. INTRODUCTION Since the discovery of carbon nanotubes (CNTs) [1], numerous studies have been conducted to explore their exceptional mechanical, chemical, electrical, and thermal properties. Within a short period of time, CNTs have generated considerable interest in both science and engineering fields due to their potential applications. The incorporation of nanotubes into polymer matrices expands their utilization in a broad range of applications such as structural composites [2,3], energy storage devices, sensors and actuators [4,5]. The exceptional properties of CNTs originates from the combination of their dimension, structure and topology. The nano-scale dimensions can provide CNTs with a large surface area (50–1315 m2 g-1) [6]. Due to their hollow structure, the density of CNTs is significantly lower than that of most nanomaterials. The long conjugated carbon structure provides a high thermal stability, excellent mechanical properties, high thermal conductivity, and good electrical conductivity. Their axial elastic modulus is reported to be more than 1 TPa, which is close to that of diamond and 3–4 times higher than that of carbon fibers. They are thermally stable up to 2800 ºC in vacuum; their thermal conductivity is about twice that of diamond and their electric conductivity is 1000 times higher than that of copper [2, 7]. Large-scale engineering composite structures often use vinyl ester (VE) as a primary resin system because of its low cost and low viscosity, making it suitable for most of synthesizing processes [8]. Commercial vinyl esters are typically a mixture of a vinyl ester monomer and 40-60% styrene monomer. Polymerization transpires free radical 1 growth chain from the one styrene reactive site and the several reactive vinyl end groups of the vinyl monomer [9,10]. Cured vinyl ester resin offers numerous attractive properties including high strength, toughness, tensile elongation and hydrophobicity [11,12]. To date, a small number of researches have focused on CNT incorporation in VE matrix through different fabrication processes. However, it is necessary to understand the molecular-scale interactions between CNTs and VE polymer matrix. It is generally accepted that there are three main requirements in a nanoparticlepolymer system for an effective reinforcement: large aspect ratio of nanoparticles, uniform dispersion of nanoparticles through the matrix, and effective interfacial stress transfer [13]. Majority of CNTs possess large aspect ratio (typically >1000) [14], which is much higher than that of other nanofillers. As a result, large interfacial areas are available for load transfer between CNTs and polymer matrix. Thus, CNTs are expected to have significant influence in the mechanical properties of polymeric materials, compared to nanofillers with lower aspect ratio. A homogeneous dispersion of CNTs can further produce large interfacial area between nanoparticles and the host polymer. Moreover, with a better dispersion, the stress distribution over the bulk material would be more uniform. The latter prevents the local stress concentration, thus enhancing the deformability of the composite product. It is well-stablished that when improvement of mechanical properties is of major concern, the load transfer mechanisms between CNTs and the matrix at their interface become essential. The strength and stiffness of the material is directly related to the characteristics of these interfaces. An efficient load transfer from the polymer matrix to the nanotubes is required to take advantage of CNTs. 2 Because of fibrous shape, excellent mechanical properties and large specific surface area of carbon nanotubes, CNT-reinforced polymer nanocomposites can be expected to show considerably improved mechanical properties compared with the neat matrix material. However, there are some important technical challenges in the processing and development of nanocomposites that keep the CNT reinforcement far from achieving its theoretical potential. These challenges are to be overcome in order to realize the full potential of nanocomposites. Some of these issues are as follows. Non-uniform dispersion: Due to the van der Waals (vdW) interaction, the nanotubes tend to aggregate to form bundles or ropes and further agglomerate when dispersed in the polymer matrix. The high surface area of CNTs also results in a high viscosity of the nanotube/polymer mixture particularly when fabricating composites with higher concentration of nanotubes, which makes the dispersion of nanotubes extremely difficult. In addition to the uniform dispersion, it is critical to obtain a desired alignment of the nanotubes to control the properties of the resulting nanocomposites. So, a number of attempts have led to achievement of well-dispersed and well-aligned CNT/polymer composites by means of various methods such as ultrasonication [14,15], the use of surfactants [16–18] and processing agents [19,20]. Some solvents such as aniline [21] and dimethylformamide (DMF) [22] can stabilize the CNT suspensions better than other surfactants. Ultrasonication is a common technique to break up the CNT bundles in lowviscosity resins using the acoustic cavitation mechanism. The cavitation involves the formation, expansion and implosion of numerous microscopic bubbles, which result in the generation of attenuated waves [23]. Although the amount of energy released by each individual bubble is minimal, the cumulative energy can be high enough to induce powerful 3 shock waves in the molecules of the resin. The production of these shock waves promotes the peeling off individual CNTs located at the outer part of the nanotube bundles or agglomerates, and thus results in the separation of individualized nanotubes from the bundles. Cavitation intensity and the efficiency of the sonication process depends on the equipment used as well as the physical properties of the resin [24,25]. Weak interfacial bonding: Another challenging issue is the interfacial bonding between the nanotubes and polymer matrix in the composites. Because of their structural features, it is really difficult to properly wet the pristine CNTs by polymer materials. The strong sp2 bonding between carbon atoms of nanotube wall makes the CNTs chemically inactive [26]. Consequently, the nanocomposite product fails to demonstrate its anticipated behavior. Functionalization of CNTs is a promising method to improve the load transfer from the host polymer to the nanotubes. The two main methods of CNT functionalization are non-covalent and covalent modifications. Non-covalent modifications utilize vdW forces and π-π interactions by adsorption of aromatic compounds, surfactants, polymers or biomolecules by CNTs. Non-covalent modifications do not disrupt the natural configuration of CNTs with the cost of chemical stability, and is prone to phase separation, dissociation in between two phases, in the solid state. Covalent modification, on the other hand, attaches a functional group onto the carbon nanotube. The functional groups can be attached onto the side wall or ends of the carbon nanotube [27]. Although covalent modifications are very stable, the bonding process disrupts the sp2 hybridization of the carbon atoms. In fact, the hybridization state changes from sp2 to sp3, leading to a loss of conjugation [28]. The covalent modification provides the flexibility to modify the CNT surface to fit the desired environment. Such modification allows a more compliant interface 4 between the nanohybrid and the surrounding polymer that translates into enhanced thermal and mechanical properties of the eventual nanocomposites. The overall properties of nanocomposites cannot be precisely understood by simple micromechanics rules that can be applied to traditional polymeric composites. Theoretically and computationally predicting the mechanical behavior of CNT/polymer composites will be critically important before they are used in real structural applications [13,29,30]. Molecular dynamics (MD) simulation is a useful tool to investigate the influence of CNT incorporation on the structure and dynamics of polymers in molecularscale. Technically, a combination of experimental characterization and computational method should be used for the future development of nanocomposites. In this study, the inclusion of CNTs into a VE matrix will be considered, and the effects of dispersion conditions and interface modification between CNTs and polymer will be investigated. Molecular dynamics simulation method will be performed to estimate theoretically the mechanical properties of CNT-VE systems. 5 CHAPTER 2. LITERATURE REVIEW 2.1 Carbon Nanotube Carbon nanotubes (CNTs) are quasi one-dimensional carbon allotropes with cylindrical shape, which were first prepared by Iijima [1,31]. Generally, CNTs are produced in two forms; single-walled CNTs (SWCNTs) and multi-walled CNTs (MWCNTs). SWCNT is a seamless hollow cylinder and can be visualized as formed by rolling a sheet of graphite layer, whereas MWCNT consists of a number of concentric SWCNTs stacked together with π-π interactions. The structure of a CNT depends on how the graphite sheets is rolled up and is characterized by its chirality or helicity, which is defined by the chiral angle and the chiral vector (Fig. 1). The chiral vector is written as, πΆβ = ππ1 + ππ2 (1) where a1 and a2 are unit vectors in a two-dimensional graphite sheet and (n, m) are called chirality numbers. Both n and m are integers and they define the way the graphite sheet is rolled to form a nanotube. Nanotubes with π ≠ 0, π = 0 are called the zigzag tubes (Fig. 2a), and nanotubes with π = π ≠ 0 are called armchair tubes (Fig. 2b). In zigzag tubes, two opposite C–C bonds of each hexagon are parallel to the tube’s axis, whereas in the armchair tubes, the C–C bonds of each hexagon are perpendicular to the tube’s axis. If the C–C bonds are at an angle with the tube’s axis, the tube is called a chiral tube (Fig. 2c). The chiral angle, 0° ≤ π ≤ 30° , is defined as the angle between the zigzag direction and the chiral vector, and is given by 6 1 3 ⁄2 π π = π‘ππ−1 [2π+π] (2) The diameter of the nanotube is also given by π= π0 √3 π √π2 + ππ + π2 (3) where a0 is the C–C bond length, which is equal to 1.42 Å [7]. Figure 1: The chiral vector and chiral angle in a graphite plane [7]. 7 Figure 2: (a) Zigzag, (b) armchair, and (c) chiral carbon nanotubes [7]. 2.2 Polymer Matrix 2.2.1 Molecular Structure A polymer is defined as a long chain of molecules containing one or more repeating units of atoms. In solid state, these molecules are frozen in space either in the amorphous or semicrystalline fashion. Polymers are divided into two main categories: thermoplastics and thermosets. In a thermoplastic polymer, individual molecules are not chemically joined together as shown in Fig. 3a. The interaction between molecules are defined by weak secondary bonds or intermolecular forces, such as van der Waals or hydrogen bonds. In a thermoset polymer, on the other hand, the molecules are chemically joined together by 8 cross-linking process. Consequently, a rigid and three-dimensional network structure is formed (Fig. 3b). Once the cross-links are formed through the polymerization reaction (also called curing), the thermoset polymer cannot be melted by the application of heat. Instead, a glass transition phenomenon occurs in the molecular structure of the polymer [7,32,33]. Figure 3: The schematic representation of (a) thermoplastic polymer and (b) thermoset polymers [7]. 2.2.2 Glass Transition The glass transition is the reversible transition in amorphous materials from a hard and relatively brittle state into a rubber-like state. Despite the massive change in the physical properties of a material through its glass transition, the transition is not itself a phase transition of any kind; rather it is a phenomenon extending over a range of temperature. Upon cooling or heating through this glass-transition range, the material also 9 exhibits a smooth step in the coefficient of thermal expansion (CTE) and in the specific heat, with the location of these effects again being dependent on the history of the material. The thermal expansion, heat capacity, elastic modulus, and many other properties of inorganic materials show a relatively sudden change at this glass transition [34]. In the study of polymers and their applications, it is important to understand the concept of the glass transition temperature, Tg. As the temperature of a polymer drops below Tg, it behaves in an increasingly brittle manner. As the temperature rises above the Tg, the polymer becomes more rubber-like. Thus, knowledge of Tg is essential in the selection of materials for various applications. This behavior can be understood in terms of the structure of glassy materials which are formed typically by substances containing long chains, networks of linked atoms or those that possess a complex molecular structure [35,36]. To become more quantitative about the characterization of the liquid-glass transition phenomenon and Tg, it is noted that in cooling an amorphous material from the liquid state, there is no abrupt change in volume such as occurs in the case of cooling of a crystalline material through its freezing point, Tf. Instead, at the glass transition temperature, there is a change in slope of the curve of specific volume vs. temperature, moving from a low value in the glassy state to a higher value in the rubbery state over a range of temperatures. These different behaviors are presented in Fig. 4 for a crystalline polymer (graph 1) and an amorphous polymer (graph 2). 10 Figure 4: The different behavior of amorphous and crystalline polymer melts during a cooling cycle. 2.2.3 Vinyl Ester Thermosetting Polymer Vinyl ester (VE) is a resin produced by the esterification of an epoxy resin with an unsaturated carboxylic acid. The chemical structure of a VE resin molecule is shown in Fig. 5. A linear molecular chain is formed with few unsaturated ester functional linkages located at the ends of the chains. The presence of a few ester groups in addition to Steric Shielding effects increases the ability to withstand moisture absorption, hydrolytic attack, and corrosive environment. This makes the VE as one of the most common matrix materials for marine applications and chemical plant structures, e.g., fiberglass reinforced plastic tanks and vessels. The addition of tough epoxy backbone into the long molecule of VE as well as the formation of a flexible network after curing procedure provide a flexible nature for VE polymer. Therefore, the cured VE can withstand high-energy impacts and repeated flexing without developing cracks. This makes VE as an appropriate option for ballistic part (shock loading) and armor applications. Moreover, low cost and low viscosity of VE resin makes it suitable for most of synthesizing processes [7]. 11 Figure 5: Chemistry of a vinyl ester resin (asterisks denote the reaction sites at both ends) [7]. 2.3 Ultrasonication To realize the promising potentials of CNTs, it is important to achieve highly- exfoliated CNT bundles as well as homogeneous distribution of nanotubes followed by their integration into the molecular structure of polymeric matrix [14,22,37]. Particularly, the dispersion of CNTs helps determine the optimal reinforcing effect. Carbon nanotubes tend to aggregate in a matrix due to their small size and the strong van der Waals interaction between nanotubes [38–40]. Thus, a uniform dispersion of CNTs is still one of the challenges in research associated with CNT-reinforced composites that must be resolved. Ultrasonication is a common technique to break up the CNT bundles in low-viscosity liquids and resins using the acoustic cavitation mechanism. In this technique, ultrasonic frequencies (>20 kHz) are usually used. There are two major methods to introduce ultrasonic energy into liquids; ultrasonic bath, and ultrasonic horn or wand. The ultrasonic 12 wand oscillates at fixed frequency and creates a conical field of high energy. As only the fluid within the field is affected, repeated circulation through the conical zone is needed to treat all the particles. In a bath sonicator, the liquid transfers the sonic energy from the transducer to the sample. Generally, bath sonication is utilized where low energy is required, since they dissipate a high fraction of sonic energy. Ultrasonication tools with low frequencies (20–100 kHz) and high power (100–5000 W) are commonly used for CNT dispersion [23]. 2.3.1 Acoustic Cavitation Mechanism Cavitation can be defined as the phenomena of the formation, expansion, and implosion of numerous micro bubbles occurring in extremely small interval of time (i.e., milliseconds). The cavitation mechanism is schematically shown in Fig. 6. When a liquid is subjected to an intense ultrasonic field, numerous bubbles of gas are formed under negative pressure and grow in successive cycles. The high pressure on the newly expanded bubbles can compress them and increase the temperature of their containing gas until the bubble collapses on itself. This implosion results in the generation of attenuated waves. Although the amount of energy released by each individual bubble is minimal, the cumulative energy can be high enough to induce powerful shock waves in the molecules of the liquid [41–43]. The production of these shock waves promotes the peeling off individual CNTs located at the outer part of the nanotube bundles or agglomerates as mentioned earlier. This phenomenon results in the separation of individualized nanotubes from the bundles [40]. 13 Figure 6: Schematic representation of ultrasonic cavitation and implosion [44]. 2.3.2 Physical Properties of Liquid Medium For cavitation to occur in a liquid, it has to overcome the natural cohesive forces present in the medium. Any increase in these forces will tend to increase the threshold pressure, and hence the energy required to generate cavitation. In highly viscous liquids, severe attenuation of the sound intensity occurs, and the active cavitation zone gets reduced substantially. As the viscosity of material increases, its ability to transmit vibrations decreases. Therefore, only a small fraction of the total volume of the liquid in the immediate vicinity of the ultrasound source experiences the effects of cavitation. The intensity of the ultrasound, I, at any distance, d, from the source is given by πΌ = πΌ0 expβ‘(−πΌπ) (4) 14 where α is the absorption or attenuation coefficient and I0 is the ultrasound intensity at the tip of the ultrasound source. The relation between intensity I and ultrasonication power is as follows: πΈ πΌ=π΄ (5) β where E is the sonication power and Ah is the surface area of the sonicator probe. As is seen in Eq. 4, the intensity of ultrasound reaching the interface through a given liquid drops exponentially with the distance it travels from the source. The attenuation coefficient α for a given liquid is dependent on the liquid-phase physicochemical properties and the characteristics of the ultrasound itself, as expressed in the following: πΌ= 8π 2 ππ 2 (6) 3ππΆ 3 where ρ is the density of the liquid, μ is the viscosity of the liquid, C is the speed of the sound in the liquid, and f represents the ultrasound frequency. Knowledge of the attenuation coefficient can be used to determine the active volume in a sonicator for a given ultrasound characteristic, and the properties of the liquid used in the system [44–46]. 2.3.3 Intensity of Irradiation The intensity of irradiation is defined as the power dissipation into the system per unit area of irradiation, and hence can be changed either by changing the power dissipation, or the area of irradiation. If the power dissipated into the system is increased, the collapse pressure usually decreases. The fraction of the total energy supplied to the system utilized for cavitation process can be determined from the concept of cavitational efficiency. Considering the acoustic horn system, the intensity is given by 15 π2 ππΆ πΌ0 = 8π2 π2 (7) where ρ and C are the density and speed of sound in the liquid, respectively as explained before; and f and a are the ultrasound frequency and amplitude, respectively [47]. 2.3.4 Mixing Time Several studies have been performed to discover the effects of ultrasonication duration on morphology and structure of CNTs. Generally, nanotubes with higher aspect ratio can easily get damaged or broken up into shorter segments by ultrasonication [48]. Studies on MWCNTs show that during the first few minutes of ultrasonication, a rapid CNT length reduction occurs, while the rate of length reduction levels out with further sonication [24]. However, a prolonged sonication may change the structural features of CNTs. It is found that after prolonged sonication, the graphene layer of CNTs is destroyed, and the nanotubes are converted into amorphous carbon nanofibers [49]. In MWCNTs, excessive ultrasonication has also been found to "peel" graphene layers [48,50]. 2.4 Studies on the Dispersion of CNTs in VE Resin To date, few researches have focused on CNT dispersion in VE matrix through different synthesizing processes. Fan et al. [51,52] examined the influence of different processing techniques on the dispersion of CNTs in a VE resin. They reported that acid oxidized nanotubes can be dispersed homogeneously in the matrix because nanotube length is reduced during the acid reflux. A 3-roll milling technique was employed to disperse 0.1, 0.25, 0.5 and 1 wt% MWCNTs in a VE resin [8], and the electrical properties of the resulting materials were evaluated. An electrical percolation threshold below 0.1 wt% 16 nanotubes was observed. Based on this observation, a new application of CNTs is proposed as in-situ sensors for detecting deformation and damage in advanced naval composites. Seyhan et al. [53,54] successfully employed a 3roll milling technique to disperse CNTs into a VE-polyester hybrid resin. It was revealed that nanocomposites with MWCNT-NH2 possess larger modulus due to the contribution of NH2 groups in chemical interactions at the CNT-polymer interface. They also suggested that the application of electric field can encourage the trapped radicals to move [55]. This alters the free radical polymerization of VE, and thus enhances the glass transition temperature. However, the studies on ultrasonication processing of CNT/VE composites are very limited. Gryshchuk et al. [56] investigated the inclusion of 0.5, 1 and 2 wt% MWCNTs in VE hybrid systems using bath sonication for 15 min (35 kHz). A significant effect of MWCNTs on the electrical conductivity of pure polymer was reported. Poor nanotube dispersion in the VE system was observed due to nano-scale entanglement. Minor increases in the elastic properties and the fracture toughness at varying nanotube concentrations were reported. Carbon nanotubes with lower aspect ratio were suggested for highly cross-linked polymers like VE. 2.5 CNT/Polymer Interface The role of interface between CNTs and polymer is the key factor for determining many properties of composite materials. Specifically, in order to enhance the mechanical properties, the external load has to be efficiently transferred from the matrix to the nanotubes. Three main mechanisms of load transfer are micromechanical interlocking, vdW bonds between the nanotube and the matrix, and chemical bonding at the interface 17 [13]. For pristine CNTs, weak vdW attractions between the matrix and the nanotubes is mainly responsible for slightly higher interfacial shear stress [57]. Studies have shown that the molecular level entanglement of CNTs and surrounding polymer chains can control the strength of the interface. On the other hand, there is a direct correlation between the surface chemistry of the nanotubes and the interfacial interactions [22]. It should be noted that a weak adhesion, and as a result, a poor load transfer can cause debonding at the interface, while very strong interactions between CNTs and polymer would lead to the brittleness of the nanocomposite. Thus, modification of the CNT/polymer interface is very crucial for the composite properties. 2.5.1 Experimental Characterization of the Interface Many experimental studies have been performed to investigate the relationship between the interfacial adhesion and the load transfer. Transmission electron microscopy (TEM) studies showed a significant adhesion of SWCNTs to an epoxy polymer [58]. A nano-pullout experiment using atomic force microscopy (AFM) by Wagner and coworkers measured a separation stress of 47 MPa which was required to pull out a single nanotube from polyethylene [59]. The stress transfer can also be investigated by Raman spectroscopy. It has been found that the characteristic peaks of CNTs shift to lower wave numbers when strain is applied to the composite [60]. This shift corresponds to the graphite structural distortion of the nanotube wall, indicating stress transfer between the CNTs and polymer [61]. 18 2.5.2 Computational Characterization of the Interface In recent years, researchers employed the computational modeling techniques to study the role of interface. Since the diameter of the nanotubes is at the nanometer scale, the interactions are highly dependent on the local molecular structure and bonding [62]. Therefore, molecular dynamics (MD) simulations have been used to investigate molecular interactions at the nanotube/polymer interface. MD simulation of single CNT pullout test is a common method to characterize the interface under tensile loading condition. Liao and Li [63] simulated the CNT pullout in a CNT/polystyrene nanocomposite. They found an interfacial shear stress of 160 MPa, which demonstrates a strong interface. Gou et al. [64,65] have examined the molecular interactions between a SWCNT and epoxy during curing reaction. Based on the pull-out simulations, the interfacial shear stress was calculated to be up to 75 MPa, indicating that there could be an effective stress transfer from the epoxy matrix to the nanotube. The influence of nanotube chirality has also been studied [66]. For SWCNTs with similar molecular weight, diameter, and length, nanotubes with larger chiral angles achieved higher bonding energy, and the armchair nanotubes were suggested to be the best choice for the reinforcement of epoxy. Recently, Li et al. [67] have modeled CNT/polyethylene interface with vdW interactions. They have explored the effects of nanotube length, diameter and wall number through pullout testing simulations. It was found that interfacial properties are independent of nanotube length, but they are proportional to the nanotube diameter. 19 2.5.3 Modification of the Interface Surface modification of CNTs may be performed for a variety of purposes, such as: (a) improving CNT dispersion in the polymer matrix, (b) providing a better CNT adhesion to the polymer matrix, (c) increasing CNT solubility in different solvents, and (d) joining of CNTs to form network structures [19,22,27,68]. The main approaches for the surface treatment of CNTs are divided into non-covalent and covalent functionalization. Noncovalent modifications involve the adsorption of a variety of surfactants and molecules by the surface of CNTs, which usually occurs through vdW interactions. These molecules mainly entangle the CNTs and improve interfacial attraction [69]. Specifically, the surfactants are capable of causing the counterbalance of vdW attractions between individual nanotubes in order to prevent their re-aggregation, and even match the polarity to the polymer matrix [17,70]. The other end of surfactants usually bond to additional functional groups, which are capable of interacting with polymer molecules, forming a chemical bridge between CNTs and the matrix [69]. Covalent functionalizations, on the other hand, involve the chemical attachment of the functional groups onto the nanotube surface. Studies have shown that the interfacial adhesion and shear strength can be significantly increased by covalently bonding the carbon atoms of CNT surface to the polymer through functional groups [71–73]. The covalent bonds can be produced by direct attachment of functional groups to the nanotubes in a single-step reaction and/or by a multiple-step process [74]. Some common functional groups used for covalent modification of CNTs are amine groups [75], carboxylic groups [26], epoxide groups [76], and dicarboxylic acid peroxide groups [77]. 20 2.5.4 Polyhedral Oligomeric Silsesquioxane (POSS) Polyhedral oligomeric silsesquioxanes (POSS) are nanostructures with the empirical formula RSiO1.5, where R may be a hydrogen or an inorganic functional group such as alkyl, acrylate, hydroxyl or epoxide unit [78]. Fig. 7 shows the molecular structure of POSS, which consists of both organic and inorganic matter with an inner core of inorganic silicon and oxygen and an outer layer of organic constituents. The average diameter of POSS nanostructures are in the range of 1 to 3 nm [79]. Recently, these organicinorganic hybrid materials are becoming the focus of many studies due to their simple chemistry, ease of handling and the enhanced mechanical properties, thermal stability, and flame retardation that they can provide [80,81]. Particularly, the organic groups attached to the POSS cubic cage can be used as reaction sites for further functionalization, making POSS an ideal modification reagent [78]. Therefore, POSS can be a good candidate for covalent functionalization of CNTs. Figure 7: Molecular structure of polyhedral oligomeric silsesquioxanes (POSS) [78]. 2.5.5 POSS-Functionalization of CNTs To date, various strategies have been employed to introduce covalent bonding between POSS and CNTs. The conventional functionalization consists of a three-stage 21 reaction: oxidation of nanotubes by acid treatment, amidation through acylation with thionyl chloride under controlled atmosphere, and CNT/POSS bonding through a reflux reaction [82–84]. Yadav et al. [85] employed a “click” chemistry reaction to functionalize MWCNTs. The method involves relatively complex chemical reactions in three steps in which two steps need to be performed under controlled atmosphere. Recently, Sun et al. [86] prepared a core–shell hybrid material consisting of crosslinked POSS-coated MWCNTs by direct in situ free-radical polymerization under nitrogen atmosphere. 2.5.6 Studies on the Incorporation of POSS-Treated CNTs in Polymer Recently, the functionalization of MWCNTs with POSS reagents has been found to significantly improve the mechanical, thermal and electrical properties of MWCNTpolymer nanocomposites [87–92]. Li et al. [87] reported an improved dispersion of MWCNTs throughout a poly(L-lactide) polymer after POSS-treatment of nanotubes, resulting in a higher tensile modulus, yield strength and electromagnetic interference shielding effectiveness. Sun et al. [88] studied the effect of POSS-coating of MWCNTs on the electrical properties of MWCNT-poly(vinylidene fluoride) composites. The dielectric properties of composite was enhanced by controlling the thickness of the POSS layer. The incorporation of a small amount of POSS-MWCNT hybrid into an epoxy polymer was found to improve the thermal properties of neat polymer [89]. The grafting of POSS-CNT onto carbon fiber (CF) surface and the interfacial properties between CF and epoxy matrix was also studied [90–92]. Significant improvements in the shear strength, storage modulus, impact toughness and service temperature were reported. 22 CHAPTER 3. RESEARCH MOTIVATION 3.1 Optimization of Nanocomposite Synthesis Process Literature review given in Chapter 2 has shown that although the ultrasonication technique has been used in the past to disperse CNTs in the VE matrix, this process has not been optimized. Most studies have primarily approached the development of CNT/VE nanocomposites with different types and concentrations of nanotubes, while the role of CNT dispersion quality has remained mostly uninvestigated. There has not been a thorough investigation of the effect of sonication conditions, i.e. duration and amplitude on the CNT dispersion in CNT/VE nanocomposites. Therefore, our goal was to systematically and methodically investigate this effect. In this regard, ultrasonication mixing method was employed to synthesize CNT/VE nanocomposites containing two CNT concentrations, namely 0.25 and 0.5 wt%. A variety of sonication parameters was selected to produce different nanotube dispersion states. The correlation between nanotube dispersion conditions and the thermal and mechanical properties of the composites was utilized as an effective tool to assess and optimize the synthesizing process. 3.2 Improvement of the CNT/VE Interface CNT-POSS nanohybrid is a new nano-scale hybrid material which can be used as an effective reinforcement in polymer matrices. The idea behind the development of this nanohybrid is to modify the surface of CNTs by coating them with POSS. Such modification would allow a more compliant interface between the nanohybrid and the 23 surrounding polymer, which translates into enhanced thermal and mechanical properties of the eventual nanocomposites. To date, a limited number of techniques have been developed to prepare CNT-POSS nanohybrids. These strategies have mainly focused on the grafting of nanotubes through complex chemical reactions under a controlled atmosphere. The setup preparation and providing the required chemicals for such a complex reactions would be costly and time consuming. Our goal in this thesis was to stablish a simple, versatile, and effective technique to chemically functionalize MWCNTs with POSS nanostructures. The approach was based on covalent bonding between MWCNTs and POSS through the formation of amide bonds through a simple chemical reaction. We employed advanced microscopic and spectroscopic tools to verify the formation of amine bonds between MWCNTs and POSS. To understand the role of interface, we prepared different composite systems by infusing as-received POSS and CNTs as well as POSS-treated CNTs into VE polymer using ultrasonication. Three nanoparticle concentrations of 0.25, 0.5 and 1.0 wt % was considered, while the ultrasonication parameters were kept constant for all systems. There are two distinct benefits in using the CNT-POSS nanohybrid as reinforcements that come through the interface; Firstly, the chemical treatment used for the grafting allows formation of amide bonds between POSS and nanotube, permits smooth load transfer, and thus carries a higher load. Secondly, the presence of POSS as an extra phase between nanotube and the polymer would allow sliding at the interface and hence increase fracture toughness of the nanocomposites. 24 3.3 Molecular-Level Investigation of the Interface In order for nanostructured materials to be applied in advanced materials design, it is necessary to have a quantitative understanding of their deformation behavior. A recent focus of the Materials Genome Initiative is to build the understanding of the material properties through a bottom-up approach, which basically employs the assembly of atoms and molecules in molecular level. In this study, the molecular structure of CNT/VE system was established through MD simulation. The mechanical properties of the VE polymer and its nanocomposites with different CNT surface conditions were evaluated. In summary, the primary objective of this study is to investigate the role of CNT/polymer interface and develop a novel technique to modify the interfacial properties. In this regard, the approach of this study was as follows: ο· Fabrication of CNT/VE nanocomposites using the as-received nanotubes under a variety of ultrasonication parameters. ο· Investigation of CNT dispersion conditions and thermal-mechanical behavior of composite materials. ο· Correlation between the dispersion conditions and the material properties in order to find the optimized synthesizing parameters. ο· Surface functionalization of the as-received CNTs with POSS-NH2 nanostructures and characterization of the CNT-POSS nanohybrids. ο· Fabrication of new nanocomposites with POSS-treated nanotubes, and determination of the structure-property relationship. ο· Theoretical prediction of elastic properties of CNT/VE composites through MD simulation. 25 CHAPTER 4. MATERIALS, SYNTHESIS AND FABRICATION 4.1 Materials 4.1.1 Vinyl Ester Derakane 8084 epoxy vinyl ester resin is an elastomer modified resin designed to combine high strength with an improved toughness. The physical and mechanical properties of Derakane 8084 VE resin after clear casting at room temperature (RT) are listed in Table 1. Table 1: The physical and mechanical properties of Derakane 8084 vinyl ester resin at room temperature. Property Value Unit Density 1.14 g cm-3 Volume shrinkage 8.2 % Heat distortion temperature 82 ºC Glass transition temperature 115 ºC Tensile modulus 2900 MPa Tensile elongation at break 8–10 % Flexural modulus 3300 MPa Flexural strength 130 MPa This grade of VE also offers significant adhesive strength with superior wear resistance. The improved toughness of this thermoset polymer provides the components 26 with better impact resistance and less cracking due to cyclic temperature and pressure fluctuations. Derakane 8084 resin is also an appropriate choice where superior chemical resistance across a broad range of acids, bases and organic chemicals is required [93]. Derakane 8084 VE resin was procured from Ashland Composites (Dublin, Ohio, USA). The as-received VE contains ~40 wt% styrene, as a reactive solvent, and has a nominal dynamic viscosity of 360 centipoise. 4.1.2 Carbon Nanotubes Multi-walled carbon nanotubes (MWCNTs) were provided by Sigma-Aldrich (St. Louis, MO, USA). According to the manufacturer, these nanotubes were produced by the catalytic carbon vapor deposition (CCVD) process and then surface modified with > 8% carboxylic acid (COOH) functional groups. Fig. 8 schematically shows the attachment of a COOH group to the CNT surface. Technical specifications of the as-received nanotubes (MWCNT-COOH, product number: 755125) are presented in Table 2. The aspect ratio (AR) of nanotubes is defined as the ratio of average length to average diameter. From values of Table 2, the AR of nanotubes is ~158, which is a relatively low AR and the MWCNTs in this study is considered as short nanotubes. Figure 8: The attachment of a COOH group to the surface of CNT. 27 Table 2: Technical specifications of the as-received MWCNTs in this study. 4.1.3 Property Value Unit Average diameter 9.5 nm Average length 1.5 μm Carbon purity > 80.0 % Metal oxide < 5.0 % COOH content > 8.0 % POSS Nanostructures POSS nanostructures are an emerging class of new materials that hold significant promise in polymer composite field. The organic constituents on the outer surface of these molecules can make them compatible with many polymers. These materials are being designed with good physical properties of ceramics and excellent choice of functional group chemical reactivity associated with organic chemistry. In the current study, aminopropylisobutyl POSS (product number: AM0265), which is also called POSS-NH2, was purchased from Hybrid Plastics, Inc. (Hattiesburg, MS, USA). The molecular structure of POSS-NH2 is shown in Fig. 9. POSS-NH2 is a hybrid molecule with an inorganic silsesquioxane at the core, organic isobutyl ((CH3)2CHCH2–) groups attached to seven corners of the cage, and an aminopropyl (NH2CH2CH2CH2–) group attached to the eighth corner. The physical properties of the as-received POSS are presented in Table 3. 28 Figure 9: The molecular structure of POSS-NH2 hybrids. Table 3: The physical properties of POSS-NH2 hybrids. 4.1.4 Property Data Unit Appearance White powder - Formula weight 874.58 - Density 1.16 g cm-3 Thermal stability (5% weight loss) 221 ºC Chemicals All the required chemicals for the surface modification of MWCNTs and synthesis of nanocomposites were procured from Sigma-Aldrich (St. Louis, MO, USA), VWR International (Radnor, PA, USA), and Fisher Scientific International (Hampton, New Hampshire, USA). 29 4.2 Surface Modification of MWCNTs by POSS POSS-functionalization of CNTs was performed in the Department of Chemistry and Biochemistry at Florida Atlantic University. The successive steps of modification process are illustrated in Fig. 10. The POSS-functionalized nanotubes were prepared as follows. In a 100 mL round-bottom flask, aminopropylisobutyl POSS (POSS-NH2, 1g) and dicyclohexylcarbodiimide (DCC) catalyst (50 mg) were added to a suspension of carboxylic acid functionalized MWCNTs (MWCNT-COOH, 50 mg) in anhydrous tetrahydrofuran (THF, 25 mL). The mixture was sonicated in an ultrasonic bath for 20 min, and then refluxed at 68 ºC for 48 h. After completing the condensation reaction, the product was cooled down to room temperature. THF was then removed using a rotary evaporator (Buchi EL130 Rotavapor), and the resultant solid was heated at 120 ºC for 8 h. To remove the unreacted POSS as well as DCC, the solid product was dissolved in 50 mL of THF and sonicated for 5 minutes in an ultrasonic bath to facilitate removal of possibly entrapped DCC molecules in MWCNT bundles. It was then poured into 300 mL methanol, and vacuum filtered through a 0.22 μm polyvinylidene fluoride (PVDF) membrane. This process was repeated five times for complete removal of DCC and unreacted POSS. The filtered material was next dried in a vacuum oven for 12 h at 80 ºC to get the final product. 30 Figure 10: The successive steps used for the surface modification of MWCNTs with POSS (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). 31 The main idea in the functionalization of MWCNT-COOH with POSS-NH2 is to chemically bond the amine (NH2) group of POSS with carboxylic acid group of MWCNTs. An appropriate chemical reaction with a dynamic environment needs to be chosen. During reflux reaction, raw materials can react in a boiling solvent. Performing reflux at high temperature in addition to the flow of solvent molecules due to boiling can facilitate the interaction between nanoparticles. Magnetic stirring was also utilized during reflux to better mix the particles with a magnetic bar. Another important parameter is the selection of an appropriate catalyst. DCC is an effective catalyst between amine and carboxyl groups [94]. It is believed that a successful reaction in presence of DCC would form the C–N covalent bonding between MWCNTs and POSS. 4.3 Preparation of Nanocomposites In the current study, fabrication of nanocomposites was performed in two separate phases. In first phase, a number of composite materials were produced with a variety of fabrication parameters to investigate the optimization of synthesizing process. In the next phase, different nanoparticles (including POSS, as-received CNTs and POSSfunctionalized CNTs) were incorporated in VE resin using the same synthesis parameters for all composites in order to study the role of interface. 4.3.1 Nanocomposite Fabrication – Phase 1 The as-received MWCNTs with > 8% COOH functionality were considered as nanofillers. Nanocomposite specimens with different CNT contents namely 0, 0.25 and 0.5 wt% were fabricated as follows. Measured amount of CNT powder was dispersed in the 32 VE resin first by mechanical stirring and then by ultrasonication. The total mass of the mixture was 15 g. The sonication was carried out in a Sonics Vibra Cell liquid processor (Ti-horn, frequency of 20 kHz). In order to avoid a rise in the temperature during sonication, the mixing beaker was kept inside a 6 ºC water bath, Fig. 11. Figure 11: The prepared setup for ultrasonication process (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). A variety of mixing times and amplitudes were examined to determine the optimal sonication parameters. The sonication times, amplitudes and the corresponding total sonication energy are shown in Table 4. The sonication processor continuously displays the actual amount of power that is being delivered to the probe. During sonication at 50% and 70% amplitude, an output power of 35-38 W and 55-57 W, respectively, was recorded. The total sonication energy was calculated by multiplying the output power by mixing time, according to Eq. 8: πππ‘ππβ‘πΈπππππ¦β‘(π½ππ’ππ) = πππ€ππ(πππ‘π‘) × πππ₯πππβ‘π‘πππ(π ππ) 33 (8) For example, during sonication of a 0.25 wt% CNT-VE mixture at 50% amplitude for 18 min, an average power of 37 W was recorded. Therefore, the total energy input would be: πΈπ‘ππ‘ = 37(πππ‘π‘) × 18 × 60(π ππ) = 39,960β‘π½ππ’ππ ≈ 40β‘ππ½ Table 4: Different sonication parameters examined in this study. Sonication Parameter MWCNT (wt%) Sample 0 0.25 0.5 Duration (min) Amplitude (%) Energy (kJ) Neat VE - - 0 0.25-10 3 70 10 0.25-20 9 50 20 0.25-40 18 50 40 0.25-60 18 70 60 0.25-80 0.5-10 24 4 70 50 80 10 0.5-15 7 50 15 0.5-20 7 70 20 After sonication, CNT/VE mixtures were degassed under vacuum for approximately 20 minutes, and then mixed with curing agents before casting into the mold. In accordance with the supplier datasheet, Methyl Ethyl Ketone Peroxide (MEKP), Cobalt Naphtenate-6% (CoNap-6%) and dimethylaniline (DMA) were used as curing agents. Subsequently, the mixtures were cured for 24 hours at room temperature and post-cured for 2 hours at 99 ºC. 34 4.3.2 Nanocomposite Fabrication – Phase 2 In this phase, three types of nanoparticles were used to fabricate nanocomposites: as-received MWCNTs, as-received POSS-NH2, and surface modified nanotubes with POSS. For each type of nanoparticle, three concentrations of 0.25, 0.5, and 1.0 wt% were considered for composite preparation. Therefore, totally ten materials (pure VE polymer as control plus nine nanocomposites) were fabricated in this phase. A similar methodology was utilized to fabricate nanocomposites, as described in Section 4.3.1. It is noted that the ultrasonication amplitude and duration were kept constant to 50% and 20 min, respectively, during synthesizing of all materials. 35 CHAPTER 5. EXPERIMENTATION 5.1 Characterization of Nanomaterials 5.1.1 Transmission Electron Microscopy Transmission electron microscopy (TEM) was employed to study the surface morphology of CNTs before and after POSS-functionalization process. TEM studies were performed using Philips CM200 microscope, operating at accelerating voltage of 200 kV. Samples for TEM were prepared by drop-casting of a suspension of CNTs in ethanol on TEM Cu grid, and were dried overnight in a desiccator. 5.1.2 Scanning Electron Microscopy of CNTs Scanning electron microscopy (SEM) examinations of the as-received and POSS- grafted CNTs were carried out using a JEOL 6330F FEG-SEM at 10 kV with the energydispersive X-ray spectroscopy (EDS). Samples were prepared by drop-casting of a suspension of nanotubes in ethanol on Cu substrate. After drying overnight and prior to SEM examination, the deposited CNTs were spatter-coated with a conductive gold layer. 5.1.3 Atomic Force Microscopy Scanning probe microscopy (SPM) is a large and growing collection of techniques for investigating the properties of materials at or near the sample surface. Atomic force microscopy (AFM) is a very-high-resolution type of SPM with demonstrated resolution on the order of fractions of a nanometer. AFM can provide high-resolution and three36 dimensional information with little sample preparation. As shown in Fig. 12, a sharp tip at the free end of a cantilever (namely a probe) is brought to near contact with the sample surface. The tip interacts with the surface, causing the cantilever to bend. A laser spot is reflected from the probe onto a position-sensitive photodiode detector. The resulting signals are processed for determining the surface features. Figure 12: The basic principles of AFM technique. In this study, an Agilent 5420 atomic force microscope (Chandler, AZ, USA) was employed to investigate the surface features of CNTs. Measurements were done in AC (tapping) mode using a silicon probe with the tip radius of < 10 nm and resonant frequency of ~192 kHz. For sample preparation, a very small amount of nanotubes was suspended in ethanol using ultrasonic bath. A drop of the suspension was deposited onto a newly cleaved mica surface, and let it completely dry in a desiccator. 37 5.1.4 Raman Spectroscopy Raman spectroscopy was performed on nanoparticles using an XploRA-1 Raman microscope (Horiba Scientific, Edison, NJ, USA), with 1mW, 532 nm laser excitation, and 10x objective beam focusing. Fig. 13 displays the Raman spectrometer used in this study. For sample preparation, some nanomaterial powder was collected inside a capillary tube. The tube was then sealed on both sides. A minimum of 10 spectra were collected from different points on each sample with 20 s data accumulation per point, and then combined by averaging. Figure 13: The Raman spectrometer used in this study (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). 5.1.5 Fourier Transform Infrared Spectroscopy Fourier transform infrared (FTIR) spectroscopic analysis of nanomaterials was carried out directly on solid samples with a Nicolet iS5 spectrometer, as seen in Fig. 14. 38 Attenuated total reflection (ATR) is a sampling technique used in conjunction with FTIR which enables samples to be examined directly in the solid state without further preparation. The spectrometer was equipped with an iD5 diamond ATR. For each type of nanomaterial, a minimum of 5 spectra from different samples were collected. Data collection was performed in high-resolution scanning mode with 32 data accumulation per sample, and then the resulting spectra were combined by averaging. Figure 14: The FTIR-ATR spectrometer used for the experimental nanomaterials (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). 5.1.6 Physical Stability in Tetrahydrofuran The stability of CNTs in polar solvents can be used as a simple physical method to qualitatively examine the degree of CNT functionalization. In this regard, equal amounts of as-received and POSS-treated CNTs were added to clean centrifuge tubes, separately. The tubes were then filled with ~6 mL THF, and nanotubes were dispersed in THF using an ultrasonic bath for 15 min. In the next step, both tubes were centrifuged at 3000 rpm. The state of CNT stability in THF was monitored continually. 39 5.2 Testing of Vinyl Ester and its Nanocomposites 5.2.1 Nanoindentation Nanoindentation is a powerful and effective way of measuring the mechanical properties such as Young's modulus and hardness of various types of materials. Recently, it has been used for measuring mechanical properties of nanocomposites. Nanofillers that play major roles in enhancing the mechanical properties of nanocomposites have dimensions on a nanometer level. Similarly, deformations in the nanoindentation test are also of nanometer magnitude. Hence, the measured properties by this method could be correlated with certain nanostructures embedded in the material. As a result, using this method enhances our perception of the nanostructure of nanocomposites. A typical loading-unloading curve of the nanoindentation test is sketched in Fig. 15. Maximum normal load (Pmax), maximum depth (hmax), final depth (hf) and contact stiffness (i.e., S = dp/dh) are the important quantities that can be measured from the loaddisplacement curve. The parameters that are used in analyzing the data are shown at the cross section of an indentation in Fig. 16. Among the mechanical properties which can be determined by a nanoindentation test, the elastic modulus (E) and the hardness (H) are very common, and can be measured by analyzing the unloading part of the load-displacement curve. The indentation hardness is given by: π»= ππππ₯ (9) π΄ where A is the projected contact area at the maximum load. The elastic modulus (E) can be calculated from the following equations: πΈπ = √π ππ 1 2 πβ √π΄ (10) 40 1 πΈπ = 1−π 2 πΈ − 1−ππ 2 (11) πΈπ where Er is the reduced modulus of indentation contact, Ei (1140 GPa) and νi (0.07) are the elastic modulus and Poisson's ratio of the diamond indenter [95], and ν = 0.35 is the Poisson's ratio of the VE polymer [96]. Figure 15: The schematic load-displacement curve in a nanoindentation test [97]. Nanoindentation tests were performed following ASTM E2546-07 [98] using a NANOVEA hardness tester (Nanovea, CA, USA), equipped with a Berkovich indenter tip as shown in Fig. 17. The indentation setup was placed inside an acrylic enclosure to reduce the noise in test data caused by air circulation. Samples for indentation tests were cut and mounted on a holder. In order to obtain a good surface finish, samples were successively polished using 6, 3, 1, and 0.25 μm diamond paste. The diamond paste was suspended in 41 MetaDi Fluid (Buehler, USA) before applying to the polishing cloth. All nanoindentation tests were performed with a maximum normal load of 800 μN and a constant loading/unloading rate of 25 μN s-1. Prior to experiments, the tip area function was calibrated using Oliver and Pharr method [95]. In order to reduce the creep effect, the maximum load was kept constant for 10 seconds. More than 50 indentations were made on each type of sample on random locations to obtain reliable results. Moreover, an array of 30 indentations over a total length of 150 μm was made on selective samples. Figure 16: The schematic illustration of contact geometry through nanoindentation test using a Berkovich indenter [97]. 42 Figure 17: The nanoindentation testing setup for experimental materials (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). 5.2.2 Three-Point Bending Room temperature mechanical properties of the neat VE polymer and its composites were measured according to ASTM D790-10 [99]. A Zwick/Roell universal tension-compression testing machine with 3-point bending setup (Fig. 18) was employed. Specimens with the width of ~12.7 mm were cut from the post-cured blocks using IsoMet1000 precision saw (Buehler, Illinois, USA). Testing was performed with a support span-to-depth ratio of 16:1 and crosshead speed of 0.5 mm min-1. The fractured specimens were collected for further study of fracture behavior and nanotube dispersion. The load (P)-deflection (D) results of the flexural tests were converted to the flexural stress (σf)-strain (εf) values using equations below [99], ππ = 3ππΏ⁄2ππ 2 (12) ππ = 6π·π ⁄πΏ2 (13) 43 where, L is span length (mm), b is specimen width (mm), and d is specimen thickness (mm). The flexural modulus (Ef) was also extracted from the slope of the linear portion of load-deflection curve (m), according to Eq. 14, πΈπ = πΏ2 π⁄4ππ3 (14) Figure 18: The 3-point bending setup used in this study (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). 5.2.3 SEM of Fracture Surfaces After bending tests, the fracture surfaces of VE polymer and composite specimens were examined using a high-resolution SEM (Hitachi SU6600 Analytical VP FEG-SEM). Detailed observation was made on each specimen to investigate the surface roughness, CNT dispersion state, interfacial adhesion condition and active fracture mechanisms. The fracture surfaces were sputter-coated with gold prior to observation. 44 5.2.4 Optical Microscopy To better understand the state of CNT distribution in larger scale, transparent samples of CNT/VE composites were prepared for optical microscopy. Once the nanocomposite mixture was ready for casting into silicon mold (Section 4.3.2), few drops of the mixture were deposited on a glass slide. The slide was then spin-coated at 3500 rpm for 1 min. The resulting thin composite layer was allowed to cure inside a desiccator. For imaging, a Nikon Eclipse TE2000-S microscope was employed. 5.2.5 Differential Scanning Calorimetry Thermosetting polymers like VE undergo a transition from glassy behavior to rubbery behavior, which is called glass transition. This transition usually occurs over a temperature range of a few degrees. The state of the material below this range is glassy, while it turns to a rubbery state beyond this range. The decrease in the stiffness of about three orders of magnitude that accompanies this transition makes it clear that it is one of the most important parameters in characterizing the mechanical behavior of amorphous polymers [36]. In this study, a TA Instrument Q-10 differential scanning calorimetry (DSC) apparatus was used to determine the Tg of VE polymer and its nanocomposites, Fig. 19. Prior to testing, the equipment was calibrated. All DSC tests were performed under nitrogen atmosphere (50 mL min-1). For sample preparation, three samples (10-15 mg each) were prepared from each post-cured material and tested for the measurements. An aluminum empty pan was used as a reference. Each sample was heated from 0 ºC up to 200 ºC with a constant heating rate of 10 ºC min-1. After cooling down to room temperature, 45 a similar heating cycle was applied on the sample and the data were recorded. The Tg values were determined according to ASTM E1356-08 [100]. Figure 19: The DCS apparatus used for thermal characterizations in this study (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). 5.2.6 Thermogravimetric Analysis To study the thermal degradation behavior of experimental materials, thermogravimetric analysis (TGA) was performed using a Perkin-Elmer STA 6000 thermal analyzer, Fig. 20. Prior to testing, the equipment was calibrated. For each experiment, about 10–15 mg of post-cured material was put into TGA crucible. All TGA experiments were conducted under nitrogen atmosphere (50 mL min-1) from room temperature (RT) to 985 ºC 46 at a heating rate of 10 ºC min-1. The weight change of the materials as a function of temperature was recorded and analyzed. Figure 20: The TGA equipment used for thermal analysis of experimental materials (Photographed by S.M. Sabet. Copyright 2016 by S.M. Sabet). 47 CHAPTER 6. EFFECTS OF ULTRASONICATION ENERGY 6.1 Introduction In this chapter, we report a systematic examination of ultrasonication energy and the subsequent dispersion condition of nanoparticles in the VE resin. The nanoparticles were MWCNTs functionalized with COOH groups. Two nanoparticle concentrations of 0.25 wt% and 0.5 wt% with a variety of sonication amplitudes and duration (as presented in Table 4) were considered. The mechanical and thermal properties and the fracture behavior of the experimental materials were evaluated. For deeper understanding of the observed results, the nanotube dispersion quality was also examined through HR-SEM. 6.2 Flexural Behavior Typical flexural stress-strain curves of the pure polymer and its nanocomposites are illustrated in Fig. 21. As expected from the brittle nature of thermosetting polymers [7], both the neat VE and its nano-modified systems behave in a typical brittle manner with limited plastic deformation. According to the graphs, the incorporation of CNTs reduces the ductility of the pure polymer. This is more pronounced in the system with 0.5 wt% CNT reinforcement. By adding the stiff nanotubes into the polymer matrix, the regions of high CNT concentration act as local stress concentration sites, and thus promote crack initiation [53]. Consequently, the material fails before reaching the ultimate flexural strength. However, a detailed microstructural investigation of the fractured specimens has been performed to identify the involved mechanisms. 48 Figure 21: The effect of sonication conditions on the flexural behavior of composites containing (a) 0.25 wt% and (b) 0.5 wt% CNTs. 49 Figure 22 shows the representative SEM micrographs of the fracture surface of the VE polymer and a nanocomposite fabricated with 0.5 wt% CNTs. In Fig. 22a, the presence of river marks on a relatively smooth fracture surface of the pure VE specimen indicates a predominantly brittle fracture mode. The river patterns are highlighted in the figure by arrows. The fracture surface of the nanocomposite (Fig. 22b), however, shows substantial increase in the surface roughness, implying that the crack propagation was opposed or resisted by CNT reinforcement [101]. The red arrow in Fig. 22b shows the initial crack nucleation site as well as the crack propagation direction. The nucleation of the initial crack seems to have started from a CNT agglomeration at the tension side of the specimen. A narrow region (Zone I) of torn surfaces with nonlinear deformation paths surrounding the initial crack can be recognized. As the crack propagates, a new region appears on the fracture surface, Zone II, which is covered by the outwardly expanding microcracks with dimpled patterns. Detailed SEM analysis demonstrates that at this stage of the fracture, the highly-textured surface is possibly generated by sequential nucleation, growth and coalescence of microcracks. A remarkable amount of fracture energy can be dissipated through the generation of additional fracture surfaces [102]. Finally, a relatively smooth surface with river patterns can be recognized at the compression side of the specimen, Zone III. To better understand the effect of sonication parameters, the following approach was considered. Based on the linear part of the flexural stress-strain curves, the flexural modulii of the neat VE and CNT-modified nanocomposites were calculated. In addition, the average elastic modulus of each material was determined using nanoindentation testing method. The idea of two parallel studies using flexural and nanoindentation techniques is 50 to estimate the effect of CNT reinforcement in the matrix in a bulk as well as in a local context. In flexural tests, a large volume of the material is subjected to linearly varying stress field and the measured properties are indeed bulk properties. Whereas in nanoindentation tests, the modulus measurement is from the response of a small volume of the material as encountered by the indenter tip, making it very much local. It is believed that these two aspects, i.e., bulk and local, can best cover the dispersion issues. At each CNT content, the variations of the flexural strength and the modulus with sonication parameters were studied. Moreover, the microstructural evolution and the fracture behavior of the composite materials were carefully investigated by SEM analysis. Figure 22: SEM micrographs of the fracture surface of (a) the neat VE polymer and (b) nanocomposite fabricated with 0.5 wt% CNT content. 6.3 The 0.25 wt% MWCNT/VE Nanocomposites The effect of sonication parameters on the elastic modulus of nanocomposites with 0.25 wt% CNT are summarized in Fig. 23. For comparison, the measured modulus of the pure polymer is also presented. The values at zero sonication energy are referred to the neat VE modulii. The neat VE polymer possesses an average flexural and indentation modulus 51 of 2.99 and 3.5 GPa, respectively. It is clear from the results that the measured modulii from indentation tests are higher than those obtained from flexural tests. In nanoindentation tests, a highly non-uniform stress field is generated around the indenter tip due to the complex shape of the Berkovich indenter. Moreover, the nanoindentation modulus is measured from the slope of unloading curve in the load-displacement graph [98]. In contrast, in 3-point bending, a linearly varying stress field is encountered during the test and the flexural modulus is measured using global deflection of the sample. Therefore, the modulii measured from two different methods will be different. Higher modulus measured from nanoindentation is attributed to complex material plasticity during penetration of the indenter, as indicated in the literature [97,103,104]. Figure 23: The effect of sonication energy on the elastic modulus of composites synthesized with 0.25 wt% CNT content. 52 As mentioned before, a very small volume of the material is deformed during penetration of the indenter. The elastic modulus is measured from the deformation behavior of this microscopic volume. Since the chances of hitting nanotubes in this volume would be different at different locations, inherently the data will give a relatively large error bar. On the contrary, 3-point bending tests determine the modulus and strength where a large volume of the specimen is taken into consideration. As a consequence, the error bar is smaller. It is also evident that the variation of indentation modulus with sonication energy presents a similar trend to that observed for the flexural modulus. A slight drop in the modulus is found at an energy level of around 10 kJ, corresponding to the sonication for 3 min at 70% amplitude. Fig. 24 presents the fracture surface of this specimen. A poor CNT dispersion condition with separate CNT agglomerates (red arrows in Fig. 24a) can be detected in the surrounding matrix. Because of the high specific surface area (SSA) of the nanotubes as well as their high polarizability [20], the amount of energy required to separate an individual tube from the agglomerate is considerable. Thus, in the absence of strong favorable interactions between the tubes and the polymer matrix, the preferred configuration of tubes in a polymer will be in a bundled arrangement with sufficient nanotube concentration [40]. Moreover, in-depth SEM analysis reveals that the agglomerates of CNTs exhibit a relatively weak adhesion to the matrix, so that an extensive de-bonding and matrix cracking (indicated in Fig. 24b by arrows) occur around the agglomerates. The slight drop in the modulus can be attributed to the poor dispersion of nanotubes and their weak adhesion to the VE matrix. The early failure of the specimen due to the aforementioned mechanisms can also be responsible for a remarkable drop in the flexural strength of the material (Fig. 21a, 10 kJ). 53 Figure 24: SEM micrographs of the fracture surface of composite with 0.25 wt% CNT content sonicated at 70% amplitude for 3 minutes (10 kJ). The results show that sonication at 20 kJ does not influence the elastic modulus and flexural strength of the composite, comparing to those of the neat VE resin. By doubling the mixing time from 9 to 18 min at constant amplitude of 50%, and thus doubling the sonication energy to 40 kJ, the mechanical properties show an ascending trend. An improvement of about 6 and 9% is achieved in the flexural strength and modulus, respectively. Fig. 25 presents the fracture surface of specimens fabricated at the constant amplitude of 50% and sonicated at various durations. According to Fig. 25a, some individual CNTs are visible around a partially disentangled agglomerate. By comparing this micrograph with the microstructure of Fig. 24, it is evident that the dispersion of nanotubes is slightly enhanced after sonication at 50% for 9 min. Although the application of higher sonication energy leads to a better CNT dispersion state, the de-bonding of the agglomerate from the surrounding matrix is still detectable (Fig. 25a). It can be conceived that more than one mechanism with opposite effects have been activated during the deformation, resulting in the mechanical properties of the specimen to be close to those of the pure resin, see Figs. 21a and 23 (20 kJ). 54 Figure 25: SEM micrographs of the fracture surface of composites with 0.25 wt% CNT content sonicated at 50% amplitude for (a) 9 min (20 kJ) and (b) 18 min (40 kJ). As illustrated in Fig. 25b, sonication for longer time up to 18 min provides the required energy to break up the nanotube entanglements and improves the CNT dispersion state. The isolated CNTs are in both forms of broken segments (bright dots indicated by red arrows) and curled ropes (indicated by yellow arrows). The presence of small agglomerates is still obvious. Improvement of nanotube dispersion is believed to increase the required load to the failure of the specimen, resulting in an enhancement of the flexural strength (Fig. 21a, 40 kJ). Furthermore, the high degree of CNT disentanglement provides a large interfacial area between CNTs and the polymer matrix. In this condition, the welldispersed, stiff nanotubes can reduce the mobility of the surrounding polymer chains to some extent [53], leading to an increase in the elastic modulus of the material. These results are in agreement with those previously published on the effect of sonication time on the mechanical properties of different MWCNT modified polymer systems [105,106]. Further increase in the modulus (~15%, comparing to the pure resin) can also be achieved at 60 kJ, i.e., by sonication at 70% amplitude for the same duration of 18 min. However, longer sonication at 70% amplitude is found to have a detrimental effect on the 55 mechanical properties. To identify the reasons of the observed behaviors, SEM micrographs of the fracture surface of these specimens are presented in Fig. 26. Figure 26: SEM micrographs of the fracture surface of composites with 0.25 wt% CNT content sonicated at 70% amplitude for (a, b) 18 min (60 kJ) and (c, d) 24 min (80 kJ). A completely disentangled agglomerate is detected in Fig. 26a. The in-depth SEM analysis reveals the presence of isolated, well-dispersed nanotubes on the fracture surface of this specimen. The majority of the nanotubes exhibit short protruding lengths out of the matrix, implying their rupture either during prolonged sonication or during the fracture. Principally, fillers with stronger interfacial adhesion tend to be fractured under loading [107]. In fact, the carboxylic acid functional groups attached to the outer wall of MWCNTs can facilitate the formation of a relatively strong interfacial adhesion of the nanotube to the 56 surrounding matrix [108]. Consequently, these CNTs are fractured during the crack propagation. Besides, a few nanotubes possess relatively long protruding lengths, indicating that they failed by a pull-out mechanism. This mechanism usually occurs in absence of a good interfacial adhesion [101]. Once the CNT bridges the crack tip, one side of the nanotube is completely pulled out from the polymer matrix. As a result, a protruded nanotube and a hole appear on the fracture surface, indicated by oval in Fig. 26b. Another interesting observation in this micrograph is the occurrence of bifurcation mechanism. During the crack propagation, individual nanofillers can resist the propagation by this mechanism [109–111]. As demonstrated in Fig. 26b, a typical crack tip bifurcation (indicated by dotted red arrows) starts with isolated, well-dispersed nanotubes (indicated by yellow arrows). It is believed that the energy dissipation due to the crack bifurcation enhances the elastic modulus of the composite material [101]. Based on the explanation above, it is believed that a remarkable dispersion improvement in addition to an extensive crack bifurcation by individual CNTs is responsible for the observed increase in mechanical properties of the composite. Further sonication up to 24 min at constant amplitude of 70% (80 kJ, Figs. 26c and d) may have some adverse effects on the CNT dispersion state and their morphology. Generally, two effects can be identified from the SEM micrographs. Although the CNT dispersion is overall satisfactory, it seems that a re-agglomeration process occurred. Some local regions containing high density of entangled CNTs (Fig. 26d) appear within the isolated, broken nanotubes (indicated by arrows in Fig. 26c). The re-agglomeration of CNTs in thermosetting polymers has been previously reported [23,112,113]. Li et al. [113] have observed the re-agglomeration process in a CNT/epoxy system after a sequential 57 synthesizing method including CNT surface treatment, sonication and shear mixing. They have suggested that the disentangled CNTs without sufficient surface functional groups tend to re-agglomerate due to vdW and Coulomb attractions. The presence of some nanotubes without sufficient carboxylic acid functional groups was revealed earlier (Figs. 25b and 26b). These nanotubes are believed to be capable of re-bundling at over-sonication condition (with the total energy of 80 kJ), which can deteriorate the mechanical properties of the composite. Moreover, a CNT length reduction is apparent in the micrographs of the specimen sonicated for long duration at high amplitude of 70%. It is generally accepted that the high energy of sonication often results in CNT damage and degrades the structure of the nanotubes [24,40,49], which in turn could eventually reduce their length, and therefore deteriorate both the electrical and mechanical properties of the material. In this regard, Battisti et al. [114] have reported an increase in the electrical resistivity of a MWCNT/polyester system with 0.25 wt% nanotube under over-sonication condition. They proposed that at the latest stage of the sonication process, the eventual damage to the CNT structure could be responsible for the observed behavior. Based on the explanations above, one can conclude that the localized re-agglomeration of CNTs as well as the remarkable reduction of their length may play a significant role in decreasing the mechanical properties of the composite material under excessive sonication (Fig. 21a and 23, 80 kJ). 6.4 The 0.5 wt% MWCNT/VE Nanocomposites Figure 27 illustrates the variation of the elastic modulus with sonication energy for the composites fabricated with 0.5 wt% CNTs. The values at zero sonication energy is 58 associated with the neat VE modulii. The corresponding SEM images of the fracture surface of these specimens under flexural testing are presented in Fig. 28. Figure 27: The effect of sonication parameters on the elastic modulus of composites synthesized with 0.5 wt% CNT content. After sonication at a low energy level of 10 kJ (4 min at 50% amplitude), the modulus remains unaffected being close to that of the neat resin, whereas there was a slight drop in the modulus of 0.25 wt% CNT composite under the same sonication condition. Fig. 28a shows a partially disentangled agglomerate. Similar to the 0.25 wt% CNT system, the de-bonding between the agglomerate and the matrix is detectable. It is believed that the drop in the modulus is due to poor CNTs dispersion and their weak adhesion to the matrix. Weak dispersion would lead to agglomeration of CNTs, which might behave more like defects. In addition, adhesion to the matrix plays a significant role in controlling the modulus. By comparing the fracture surfaces of 0.25 and 0.5 wt% CNT samples at 10 kJ 59 (Figs. 24 and 28a), it can be realized that the de-bonding between agglomerates and the VE matrix is less pronounced in 0.5 wt% CNT composite. Less de-bonding means CNT bundles will still carry some load to maintain the modulus at 0.5 wt% concentration. Figure 28: SEM micrographs of the fracture surface of composites with 0.5 wt% CNT content sonicated at 50% amplitude for (a) 4 min (10 kJ), (b, c) 7 min (15 kJ), and (d, e) at 70% amplitude for 7 min (20 kJ). 60 However, it was observed that extended mixing up to 7 min at 50% amplitude (15 kJ), can significantly enhance the modulus– as high as ~24.1% more than that of pure resin. The corresponding microstructure of the fractured specimen exhibits the breakup of the nanotube agglomerates, shown in Fig. 28b. It is seen in Fig. 28c that CNTs are agglomerated with the VE resin to 30–50 nm diameter bundle, which is much larger than the as-received MWCNTs (~9.5 nm). This suggests that the nanotubes were held tightly to the matrix, demonstrating a strong interaction between VE matrix and the functionalized nanotubes. Introducing the next higher sonication energy up to 20 kJ– seems to have a direct effect in destroying the CNTs. The fracture surface at this condition reveals a high number of broken, individually dispersed CNTs (bright dots specified by arrows in Fig. 28d) as well as a cluster of broken nanotubes (Fig. 28e). The CNT entanglement regions are smaller than those shown in Fig. 28a, and thus the flexural strength and modulus are improved due to a relatively better dispersion state. The dispersed CNTs as seen in the SEM image can be either non-functionalized or functionalized. It is generally accepted that the functionalized CNTs can make a stronger bonding to the matrix, compared to non-functionalized ones. Therefore, the functionalized CNTs can be detected in SEM images with two main features; (a) broken segments rather than pulled-out ropes, and (b) coated by surrounding matrix and appearing like ropes with larger diameter. The percentage of functionalized CNTs may be statistically estimated by visual inspection of several SEM images from well-dispersed regions, and counting the CNTs with the two distinct features mentioned above. In this study, the average ratio of the functionalized CNTs to all nanotubes detected in SEM images has been estimated to be ~77%. 61 6.5 Nanoindentation Curves To further study the mechanical behavior, nanoindentation results of samples 0.25- 60 (that is 0.25 wt% concentration and 60 kJ sonication energy) and 0.5-15 are compared with the neat VE polymer in Fig. 29. According to Fig. 29, at higher CNT content, the loaddisplacement curve is shifted to the left and the slope of unloading curve is increased. It is known that the elastic modulus has a direct relationship with the slope of the unloading curve [98]. Therefore, a growth in stiffness will result in a higher modulus. Microstructural investigations have revealed that these nanocomposite samples possess the most efficient dispersion condition among those with the same CNT content (see Figs. 26a and 28b). As expected, a higher amount of CNTs with an efficient dispersion in the resin matrix can effectively increase the required load for the indenter tip to penetrate into the surface. Thus, the elastic modulus will increase. Figure 29: Representative load-displacement curves derived from nanoindentation tests. 62 The plasticity index (ψ) is a useful parameter to evaluate the elastic-plastic response of a material under external loading. This index can be calculated as follows: π= π΄1 −π΄2 (15) π΄1 where A1 is the area under the loading curve and A2 is the area under the unloading curve (see Fig. 15). The area A1 equals to the total spent energy during the penetration of indenter and the area A2 equals to the energy released during unloading. The difference between these two areas indicates the irreversible work during the nanoindentation experiment. For materials with viscoelastic-plastic behavior such as polymers, the plasticity index is in the range of 0 < ψ < 1. Fully-elastic and fully-plastic materials exhibit ψ = 0 and ψ = 1, respectively [97]. Table 5 presents the extracted ψ values from nanoindentation curves in Fig. 29. The plasticity index of 0.58 for pure VE polymer demonstrates a viscoelasticplastic behavior. By increasing the CNT content, the plasticity index shows a reduction, which is more pronounced at 0.5 wt% CNT concentration. This behavior can be explained as follows. An effective dispersion of CNTs between polymer molecules can limit their plastic flow during penetration of the indenter. Moreover, the presence of more CNTs in penetration area can facilitate the elastic recovery of the material during unloading. These synergetic effects would reduce the plasticity index at higher CNT concentration. Table 5: Plasticity index of VE and its nanocomposites. Sample Plasticity index (ψ) VE 0.58 0.25 wt% CNT (60 kJ) 0.57 0.5 wt% CNT (15 kJ) 0.55 63 The displacement-time data under constant indentation creep load is illustrated in Fig. 30. Addition of CNTs results in a significant decrease in creep deformation of the pure polymer. This is more pronounced at 0.5 wt% nanotube content, where ~32% reduction of creep deformation is achieved. The ability of the CNT additives to increase the creep resistance of an epoxy resin has also been reported [115]. This effect may be attributed to several factors. The confinement effect of the highly elastic CNTs can prevent the deformation of the VE network. In other words, CNTs act as the blocking sites that restrict the viscous flow of the VE when subjected to an external indentation force. Moreover, the large interfacial area as a result of effective dispersion can contribute to a significant interface adhesion between individual MWCNTs and the matrix. The latter can improve the load transfer mechanism [116], and thus results in a remarkable enhancement in the creep resistance. Figure 30: Nanoindentation creep deformation at maximum load of 800 μN. 64 In order to compare the mechanical behavior of these materials at larger scale, the modulus and hardness variations along an array of indentation are presented in Fig. 31. It is evident that the average modulus and hardness of nanocomposites are higher than that of neat polymer. Sample 0.5-15 possesses relatively uniform modulus and hardness values along the indentation array, demonstrating a good distribution of CNTs. On the other hand, for sample 0.25-60, a considerable heterogeneity in modulus and hardness values can be found in two regions of the indentation array (i.e., distance of 20–55 μm and 100–130 μm). This shows a local concentration of CNTs in these regions. A more uniform CNT distribution in 0.5 wt% CNT may be responsible for the observed improvement in elastic recovery (reduction of plasticity index) and creep resistance. 65 Figure 31: (a) modulus and (b) hardness variations along the indentation array. 6.6 Thermal Properties The results of DSC experiments on pure VE polymer and its nano-modified systems are shown in Fig. 32. The post-cured pure polymer exhibits a Tg of ~117 ºC (inset, Fig. 32), which is close to the reported value by the manufacturer (115 ºC). The variation in Tg as a function of total sonication energy for the experimental nanocomposites is shown in 66 Fig. 33. It is noticed in Fig. 33 that two distinct zones are present in each CNT content. The Tg is elevated first with increasing the sonication energy. After reaching a peak value, the Tg falls. The maximum improvement in the Tg (~13 ºC) is observed with the composite fabricated with 0.25 wt% CNT and a sonication energy of 60 kJ. Figure 32: The DSC heating curves of the neat VE polymer and its nanocomposites (at heating rate of 10 ºC min-1). This behavior can be correlated to the CNT dispersion quality. It is generally noticed that an increase in the disentanglement of CNT agglomerates may result in a higher Tg. For instance, nanocomposite samples 0.25-60 and 0.5-15 were known to have the most efficient CNT dispersion, and hence they hold the highest Tg values at 0.25 and 0.5 wt% CNT systems, respectively. As mentioned earlier, an effective dispersion of the CNTs can provide a large interfacial area between individual nanotubes and their surrounding matrix. In this condition, the nanotubes form an infiltrating network of cross-linking elements 67 [115]. The resulting network interconnects the polymer chains, thus increasing the crosslinking density and restricting the mobility of chains during polymerization [53,117,118]. Accordingly, the Tg of the resulting composite material increases. Moreover, the observed enhancement in the Tg is more pronounced for the 0.25-60 composite. One may suggest that dispersion of higher CNTs in the resin matrix could alter the rate of polymerization, and thus an effective enhancement of the Tg cannot be achieved, even at the optimum dispersion conditions. Any adverse effect on the CNT dispersion quality is found to decrease the Tg of the nanocomposite. In a 0.25 wt% CNT system, the remarkable drop in Tg at an energy level of 80 kJ can be attributed to the re-agglomeration process. It is well-established [54] that in resins polymerized via free radicals such as VE, the free radicals generated by decomposition of initiator (MEKP) can be easily entrapped within the galleries of entangled nanotubes. Therefore, one could presume that the more amount of free radicals may have probably entrapped within the galleries of larger CNT agglomerates. The latter can reduce the cross-linking density, particularly in the vicinity of the agglomerates [56], and dramatically decrease Tg of the composite material. 68 Figure 33: The glass transition temperature of the composites with (a) 0.25 wt% and (b) 0.5 wt% CNT content as a function of total sonication energy. Based on our findings in the current chapter, a thermo-mechanical model is suggested for CNT/VE nanocomposite system. The schematic representation of the model is illustrated in Fig. 34. The observed mechanical and thermal behavior of nanocomposites 69 at different sonication energies suggest that at both nanotube contents, the elastic modulus, and Tg increase with sonication energy. However, at a particular sonication energy level, CNT dispersion is optimum that provides the highest properties. This energy level has been termed as threshold energy meaning that sonication time and amplitude must correspond to this energy level to yield the most optimized dispersion, and hence the maximum enhancement in properties. Inducing sonication energies beyond the threshold level is found to have detrimental effects on the CNT structural features. It is also observed that sonication threshold energy is linked to the concentration of nanoparticles. The lower the concentration, the higher is the threshold energy level. In fact, this threshold energy is controlled by destruction of CNT structure. At higher nanotube contents, the threshold energy is reduced since a higher number of nanotubes are damaged even with a low energy input. Figure 34: The thermo-mechanical model suggested for CNT/VE system. 70 CHAPTER 7. SURFACE MODIFICATION OF CARBON NANOTUBES 7.1 Introduction One of our goals in the current study was to develop an innovative surface modification method to effectively functionalize MWCNTs with POSS. In this regard, a facile synthesizing method was designed and implemented, as described in Section 4.2. To verify the efficiency of our method, the POSS-treated nanotubes were examined using several characterization techniques. The physical, chemical and thermal properties of surface modified nanotubes were evaluated, and compared to those of the as-received CNTs and POSS nanomaterials. 7.2 Investigation of Chemical Bonding between CNTs and POSS It is well known that Raman and FTIR spectroscopy are often used for qualitative identification of materials and compounds using group frequencies and scattering intensities. However, the selection rules and relative intensities of their characteristic peaks are dissimilar, so that these spectroscopic techniques are often viewed as complementary [119]. To investigate the effects of POSS treatment on the nature of bonding between MWCNTs and POSS, both FTIR and Raman spectra were collected. Fig. 35 shows the FTIR spectra of as-received MWCNT-COOH, POSS-NH2, and POSS-treated MWCNTs. The IR spectrum of POSS-NH2 exhibits characteristic peaks of amine group at multiple frequencies. A strong peak at 1088 cm-1 indicates presence of primary amine, CN stretch. 71 In addition, peaks at 1037 cm-1 for C–N stretching and 837 cm-1 for N–H deformation vibrations [120] suggest existence of primary amine in as-received POSS-NH2. It is noted that peak at 1088 cm-1 can also be assigned to organic siloxane (Si–O–Si asymmetric stretching) that basically comes from the cage structure of POSS. A peak at 1228 cm-1 for Si–CH2 deformations, and relatively weak bands over 2952–2870 cm-1 for isobutyl C–H stretching vibrations are also detected [120]. These characteristic bands are also seen in the IR spectrum of MWCNT-POSS, demonstrating the existence of POSS molecules in the final product. Figure 35: FTIR spectra of nanomaterials in this study. 72 The IR spectra of both carbon nanotubes show a weak and broad band around 1700 cm-1 which are hardly discernible except slight bumps as observed in MWCNTCOOH (centered at 1705 cm-1) and MWCNT-POSS (centered at 1680 cm-1) samples. Since these low intensity peaks are at the low end of the range: 1725–1700 cm-1, it indicates presence of carboxyl group. But a relatively stronger absorption is observed at 1650 cm -1 which suggests conjugation with a double bond (that is carbonyl) is actually present [121]. It is also observed in Fig. 35 that after POSS-treatment, C=O band appears with higher intensity and it slightly shifts toward lower wavenumber (from 1705 cm-1 to 1680 cm-1). Similar phenomenon has been reported previously [122,123], and is correlated to the carbonyl peak with amide functionality. This confirms the covalent grafting of POSS onto the MWCNT walls through amide bonds. The Raman spectra of the as-received MWCNT-COOH and MWCNT-POSS are presented in Fig. 36. In order to compare the characteristic features of spectra, baseline correction and normalization were performed on both spectra. As seen in Fig. 36, the two main bands typical of graphite structure are present; the D-band is attributed to the disordered graphite structure or sp3 hybridized carbon of MWCNT; and the G-band is assigned to in-plane stretching of the C–C bond whose intensity indicates the structural content of the sp2-hybridized carbon atoms [124]. The Raman data from the characteristic bands are presented in Table 6. The ratio of intensities of the D and G bands (ID/IG) was also calculated. The increase in the ID/IG ratio reflects the relative degree of functionalization or defects in the nanotubes [124]. As seen in Table 6, after POSStreatment, the peak intensity ratio increases from 1.51 to 1.67. This demonstrates that the bulky POSS cages have induced distortions to CNT’s structure upon chemical grafting onto 73 the nanotube walls. Therefore, the FTIR and Raman analyses can prove the presence of a chemical bonding between CNTs and POSS. Figure 36: Raman spectra of MWCNTs before and after POSS-treatment. Table 6: Raman data from the characteristic bands for the MWCNTs. Material ID * XD (cm-1) IG * XG (cm-1) ID/IG MWCNT-COOH 275 1338 182 1579 1.51 MWCNT-POSS 360 1338 216 1575 1.67 * Arbitrary intensity unit 7.3 Electron Microscopy Examinations of CNTs The SEM images and the corresponding EDS spectra of CNTs are presented in Figs. 37 and 38. The appearance of C and O signals in the spectrum of as-received 74 MWCNTs (Fig. 37) demonstrates the presence of carboxyl groups (i.e., COOH). Moreover, indication of Si signal (Fig. 38) after the surface modification process confirms the presence of POSS nanostructures in final product. Figure 37: (a) SEM image and (b) EDS spectrum of as-received MWCNT-COOH. 75 Figure 38: (a) SEM image and (b) EDS spectrum of POSS-treated MWCNTs. TEM images of MWCNTs before and after the POSS-functionalization procedure are shown in Figs. 39 and 40. The as-received MWCNTs have 8–10 walls with a mean outer diameter of ~9.5 nm. These nanotubes possess a relatively smooth and clean wall surface, as seen in Fig. 39. The POSS-treated MWCNTs, however, exhibit a rough surface covered with some apparent particles over the nanotube surface, as indicated by arrows in 76 Fig. 40. Based on TEM observations, the size of the observed particles is between 1 to 3 nm, which is in agreement with the reported POSS average size by the manufacturer. This extra phase is presumed to be mainly associated with the grafted POSS nanostructures. It is noted that these individual particles are uniformly distributed over the nanotube surface. This is an indication of an effective reaction between the as-received CNTs and POSS in the presence of DCC catalyst during the reflux process. Figure 39: TEM images of the as-received MWCNT-COOH. 77 Figure 40: TEM images of the POSS-functionalized MWCNTs. 7.4 AFM Studies of CNTs The nano-scale characterization of CNTs was also performed using light tapping mode AFM. A representative AFM topography image of POSS-treated carbon nanotubes on mica surface is shown in Fig. 41. A few individual CNTs can be detected in the figure. The length of CNTs was measured to be around 1–1.5 μm. Generally, the topography image can only present the overall configuration of nanomaterials without details of 78 smaller surface features. To study the surface morphology of CNTs in detail, the AFM amplitude images need to be analyzed. Figure 41: AFM topography image of POSS-treated carbon nanotubes. Figure 42 and 43 illustrate the AFM amplitude images of nanotubes before and after POSS-functionalization, respectively. A section analysis software (PicoView 1.10) was used to study the surface profile of nanotubes. The as-received CNT in Fig. 42 has an average diameter of ~11 nm. This nanotube possesses a relatively smooth surface over its length. The height profile of nanotube at two different sections are presented. Both profiles have a symmetric shape and there is no significant difference between their height and width. The amplitude image of POSS-functionalized nanotube (Fig. 43), however, shows 79 a rough surface covered with some small features. The average diameter of this CNT was measured to be ~24 nm. The appearance of some apparent bulges (indicated by red arrows in Fig. 43) on the nanotube surface can be related to the presence of POSS. In this regard, the height profiles of two adjacent sections (section 1 on a smooth surface and section 2 on a bulge) are shown in Fig. 43. The non-symmetric shape of height profile due to this bulge is obvious. By carefully overlaying the height profiles, a gap of ~4 nm is detected on bulge position. This distance is very close to the size of observed particles in TEM studies, and can be another evidence to the presence of POSS in this region. Figure 42: AFM amplitude image of as-received MWCNT and the corresponding height profile measurements. 80 Figure 43: AFM amplitude image of POSS-modified MWCNT and the corresponding height profile measurements. 7.5 Physical Stability of CNTs The stability of CNT suspensions in THF was studied during centrifugation. Fig. 44 depicts the state of suspensions in several intervals. As seen, the as-received CNTs (containing COOH groups) settled immediately after centrifugation. In fact, carboxyl acid groups attached to nanotubes make them unstable in THF solvent [125], and the majority of CNTs settle after a few minutes. On the other hand, The POSS-treated nanotubes exhibit an excellent stability up to 30 min centrifugation. Afterwards, some of CNTs settle and most of them are still suspended in THF. A dark suspension can be detected after 150 min centrifuge. This indicates that POSS-modified nanotubes possess a higher degree of miscibility than as-received ones due to the presence of POSS functional groups. 81 Figure 44: Physical stability testing of CNTs in THF. 82 7.6 Thermal Behavior of Nanomaterials The thermal properties of the nanomaterials have been investigated by TGA under nitrogen atmosphere. Fig. 45 presents the thermograms from RT to 985 ºC. The study of thermal stability of functionalized MWCNTs is particularly important to show their thermal degradation properties compared with the oxidized nanotubes. As seen in Fig. 45, MWCNT-COOH exhibits two degradation steps; one from RT to ~400 ºC which is associated with the thermal desorption of water molecules and other possible impurities, as well as the loss of oxygen-containing groups on nanotube walls. Above this temperature, thermal degradation of the graphitic structure of nanotubes occurs [26]. The amount of MWCNT residue is about 82%. The decomposition of POSS occurs at ~265 ºC, and a complete weight loss is observed at temperatures over 500 ºC due to its sublimation [126]. The POSS-functionalized MWCNTs exhibit a fairly stable thermal behavior from RT to ~200 ºC. However, a rapid weight loss of 25% is detected between 200 and 550 ºC. This is followed by degradation of the nanotube structures, which results in a weight loss of 15% up to 985ºC. The amount of MWCNT-POSS residue is about 60%. Comparing the thermograms of MWCNT-COOH and MWCNT-POSS demonstrates an additional weight loss of 22% due to POSS treatment. This additional weight loss is attributed to the decomposition of POSS in the final product, indicating that by implementing the proposed synthesis procedure, a considerable amount of POSS has been grafted onto CNTs. It is also noted that the weight reduction of MWCNT-POSS at 700 and 800 ºC is about 27% (compared to 19% reported in [85]) and 30% (compared to 12% reported in [82]), respectively. This suggests an effective functionalization of MWCNTs with POSS 83 through our technique. Moreover, the entire synthesis procedure takes about 3 days, which is a relatively rapid process to achieve MWCNT-POSS product from raw materials. Figure 45: TGA thermograms of the nanomaterials. 84 CHAPTER 8. EFFECTS OF INTERFACE MODIFICATION 8.1 Introduction A promising method to control the properties of a CNT/polymer system is to modify the interface between nanotubes and the polymer. This modification may affect both the dispersion condition of CNTs and the interactions between polymer chains and nanotubes at the interface. In order to investigate these effects, we used the surface modified MWCNTs with POSS to prepare nanocomposites with 0.25, 0.5 and 1.0 wt% concentration. Reference nanocomposites (control samples) were also fabricated using asreceived MWCNTs and POSS, considering similar synthesizing parameters (i.e., ultrasonication amplitude and mixing time). We then evaluated the mechanical and thermal properties of experimental materials. The state of nanoparticle dispersion/distribution in VE matrix was observed. Also, the fracture surfaces of flexural specimens were studied through HR-SEM to examine the degree of nanoparticle exfoliation, the interfacial adhesion of nanofillers to the polymer matrix, and the modes of failure. By correlating the observed structural characteristics and material properties, we tried to make an understanding of the role of interface modification. For easy reference, nanocomposites with as-received POSS-NH2 are called “POSS/VE”, nanocomposites with as-received MWCNTs are called “CNT/VE”, and nanocomposites with POSS-modified MWCNTs are called “CNT-POSS/VE”. 85 8.2 Microstructural Studies 8.2.1 Optical Microscopy of Thin Nanocomposite Coatings Optical images of spin-coated layer of POSS/VE nanocomposites at different POSS concentrations are shown in Fig. 46. The VE resin makes a completely transparent layer after curing. With the addition of POSS, the degree of transparency is reduced in coatings and some spherical dots appear in the matrix, which can be related to a local accumulation of POSS nanostructures. The presence of a relatively large number of POSS agglomerates is obvious in the 1.0 wt% POSS nanocomposite. Optical images of nanocomposite thin layers fabricated with as-received and POSSfunctionalized CNTs are presented in Fig. 47 and 48, respectively. Generally, two different areas can be detected. The bright area corresponds to VE resin. The dark area corresponds to CNTs dispersed in the matrix. At some locations, the agglomeration of CNTs can be detected as black dots with different sizes. At large scale, the distribution of CNTs in VE matrix creates a texture. According to the figures, CNT/VE systems exhibit a coarse textured microstructure with a relatively rough arrangement of dark and bright areas. CNTPOSS/VE systems, on the other hand, show a fine textured microstructure with a finer combination of bright and dark areas. This indicates that the exfoliation of CNT aggregates improved due to POSS-functionalization of nanotubes. As a result, surface modified CNTs could disperse further into the VE resin and create a fine texture. 86 Figure 46: Optical images of POSS/VE coatings at different POSS concentrations. At 0.25 wt% concentration, the average size of agglomerates in CNT-POSS/VE system is smaller than that in CNT/VE system. Addition of nanotubes in CNT/VE system results in an increase in the size of CNT agglomerates, while the average size of agglomerates does not significantly change in nanocomposites with POSS-modified nanotubes. According to figures, the number of agglomerates is also found to increase with 87 CNT addition. However, the distribution of these agglomerates is more uniform in CNTPOSS/VE nanocomposites. Figure 47: Optical images of CNT/VE coatings at different CNT concentrations. 88 Figure 48: Optical images of CNT-POSS/VE coatings at different nanotube concentrations. 8.2.2 SEM Studies of Fracture Surfaces SEM images of the fracture surface of the VE specimen are presented in Fig. 49. The relatively smooth fracture surface with an array of river marks (Fig. 49c) at the compression side is related to the brittle fracture of VE polymer. A limited number of 89 protruded fragments are detected at the tension side of specimen, as indicated by arrows in Fig. 49b. The formation of these rough fragments in a smooth surface is an indication of limited plastic deformation of the neat VE polymer. Figure 49: SEM images of the fracture surface of the neat vinyl ester polymer. 90 Figures 50 through 52 present the SEM micrographs of the fracture surface of POSS/VE nanocomposites at different POSS concentrations. It can be seen that the surface roughness of POSS-reinforced composites are higher than that of neat polymer, and highlytextured dimple patterns are formed at the tension side of fracture surfaces. Figure 50: SEM images of the fracture surface of the 0.25 wt% POSS/VE specimen. 91 Figure 51: SEM images of the fracture surface of the 0.5 wt% POSS/VE specimen. As mentioned earlier in Chapter 6, these dimple patterns are generated during fracture propagation. With a good dispersion state, nanoparticles are expected to oppose the crack propagation more effectively [101]. In this condition, additional fracture surfaces will be generated and the resulting dimple patterns would be smaller. According to figures, 92 the average size of dimple patterns is found to increase with the addition of POSS. This can be related to a poor dispersion of POSS at higher concentrations. Figure 52: SEM images of the fracture surface of the 1.0 wt% POSS/VE specimen. SEM analysis of 0.25 wt% POSS specimen revealed the presence of nano-scale welldistributed holes, Fig. 50c. While, much larger holes containing spherical POSS 93 agglomerates are detected at 0.5 wt% (Fig. 51b) and 1.0 wt% (Fig. 52b). At 1.0 wt% POSS, extensive de-bonding of agglomerates from the matrix is obvious, as observed in Fig. 52c. Figures 53 through 55 represent the SEM images of the fracture surface of CNT/VE specimens at different nanotube concentrations. Comparing the fracture surfaces shows that with addition of as-received CNTs, the area covered with highly-textured dimple patterns decreases, and a smoother surface covered with river marks appears. This suggests that a more brittle behavior is dominant at higher CNT concentration. Figure 53: SEM images of the fracture surface of the 0.25 wt% CNT/VE specimen. 94 At 0.25 wt% concentration, some CNT agglomerates are detected on the fracture surface, as seen in Fig. 53b. However, a local disentanglement of agglomerates is identified. Some individual nanotubes in the form of broken segments (indicated by arrows in Fig. 53c) are observed around CNT agglomerates (indicated by rectangle in Fig. 53c). This partial disentanglement, and as a result a better dispersion, is believed to improve the properties of the nanocomposite. Figure 54: SEM images of the fracture surface of the 0.5 wt% CNT/VE specimen. 95 Figure 55: SEM images of the fracture surface of the 1.0 wt% CNT/VE specimen. According to Fig. 54, inclusion of 0.5 wt% as-received CNTs results in a more extensive agglomeration of nanotubes. The crack initiation from a relatively large CNT agglomerate can be seen in Fig. 54a. Several large agglomerates are also detected inside dimple patterns (indicated by arrows in Fig. 54b). Detailed SEM observation in this area has revealed that some microcracks initiated at the interface between CNT agglomerates and matrix. Fig. 54c illustrates a CNT agglomerate with outwardly expanded microcracks 96 in the root of a dimple. Moreover, the presence of some individual nanotubes around the agglomerate corresponds to a poor dispersion condition at this concentration. With further inclusion of CNTs up to 1.0 wt%, the textured area on the fracture surface is reduced to a small portion, Fig. 55a. Large accumulation of CNTs can be detected at some locations (Fig. 55b), similar to what has been observed in Fig. 47. In this area, the majority of CNTs are pulled out from the matrix, as seen in Fig. 55c. Figures 56 through 58 present the SEM images of the fracture surface of nanocomposites fabricated with POSS-modified nanotubes. The fracture surface of all specimens are highly textured, with a large number of dimpled patterns. The average size of dimples does not show considerable change with nanotube concentration. Although SEM observations have identified small local agglomerates in these specimens, the majority of fracture surface is covered with individual nanotubes. Figs. 56b, 57b and 58b show the state of CNT dispersion inside a dimple for nanocomposites with 0.25, 0.5 and 1.0 wt% CNT-POSS, respectively. A good dispersion with large number of well-dispersed CNTs can be identified. The number of individual nanotubes on the fracture surface increases with CNT concentration. Most of CNTs are in the form of broken segments, indicating their rupture during the fracture. High-resolution SEM analysis of fracture surfaces has revealed the occurrence of a new mechanism in the CNT-POSS/VE system. This mechanism can be identified in areas covered with individual CNTs, with a noticeable increase in the local surface roughness. Fig. 56c displays a region on the fracture surface with mentioned characteristics. As seen in the figure, increased roughness is due to the appearance of some “bulge” regions. These bulges are composed of few broken CNTs fully embedded in the surrounding matrix. Some 97 bulges are indicated by arrows in Fig. 56c. In this regard, a fracture mechanism is suggested. Figure 56: SEM images of the fracture surface of the 0.25 wt% CNT-POSS/VE specimen. It has been verified in Chapter 7 that after POSS-modification process, a considerable number of POSS molecules are grafted onto the surface of CNTs. With inclusion of CNT-POSS into the resin, organic functional groups attached to POSS are capable to react with polymer chains and chemically bond together [78]. Therefore, POSS 98 will act as a bonding reagent between nanotubes and their surrounding polymer. In regions with good dispersion condition, more interfacial areas are available [40], and thus, numerous bonds between CNTs and polymer can be made by POSS. This would result in the formation of CNT-reinforced polymer region in which the external load can be efficiently transferred from polymer to nanotubes through POSS linkages. Figure 57: SEM images of the fracture surface of the 0.5 wt% CNT-POSS/VE specimen. 99 Figure 58: SEM images of the fracture surface of the 1.0 wt% CNT-POSS/VE specimen. When a crack approaches this region, individual nanotubes bridge the crack tip. Since a strong interfacial adhesion is available between nanotubes and the matrix, CNTs are fractured. As a result, a large number of broken nanotubes are detected. Moreover, the flexible POSS linkage between CNTs and polymer may allow the reinforced region to deform with an increased plasticity. The presence of protruded bulges and shallow holes (in the vicinity of bulges) is related to the improved plasticity due to occurrence of this mechanism. Consequently, the fracture surface exhibits a higher surface roughness in this 100 area. The schematic representation of the aforementioned mechanism is illustrated in Fig. 59. Figure 59: The schematic representation of the revealed fracture mechanism for CNT-POSS/VE nanocomposites. SEM studies of the fracture surface of 0.5 wt% CNT-POSS/VE specimen have also revealed the occurrence of this mechanism, see Fig. 57c. At 1.0 wt% concentration, in addition to this mechanism, the occurrence of a bifurcation mechanism (Fig. 58c) was 101 identified in well-dispersed areas. The synergetic effects of these mechanisms are believed to improve the mechanical behavior of the material. 8.3 Mechanical Properties Stress-strain curves from flexural tests of nanocomposites are illustrated in Fig. 60 to 62. The bending curve of the neat VE polymer is added as a reference. The mechanical properties of materials are evaluated from flexural curves, and presented in Table 7. To determine if the resulting mechanical properties of nanocomposite materials are significantly different from those of the neat polymer, a Student’s t-test was employed. For each test, VE polymer was considered as the reference. At the threshold level of 0.05 (chosen for statistical significance), if the difference of the population means is significantly different, the corresponding mechanical property is marked with an asterisk in Table 7. Generally, addition of nanofillers in all nanocomposites shifts the fracture strain to the left. This shift is more pronounced in CNT/VE system, even at 0.25 wt% concentration (Fig. 61). This is an indication of a high number of local stress concentration sites. As revealed earlier, the presence of CNT agglomerates as stress concentration sites can promote crack initiation. Thus, the material fails at an earlier stage. 102 Figure 60: Flexural stress-strain curves of the POSS/VE system. Figure 61: Flexural stress-strain curves of the CNT/VE system. 103 Figure 62: Flexural stress-strain curves of the CNT-POSS/VE system. Table 7: The mechanical properties of VE and its nanocomposites. Flexural modulus G/L Flexural strength G/L Fracture strain G/L Ef (MPa) (%) σf (MPa) (%) εf (%) (%) 3000 ± 27 – 130 ± 2.2 – 9.3 ± 0.6 – 0.25 wt% POSS * 3210 ± 31 7.0 134 ± 3.0 3.1 9.0 ± 0.7 -3.2 0.5 wt% POSS * 3354 ± 40 11.8 135 ± 3.5 3.8 * 6.6 ± 0.8 -29.0 1.0 wt% POSS 3072 ± 53 2.4 * 107 ± 5.3 -17.7 * 3.9 ± 1.5 -58.1 0.25 wt% CNT * 3226 ± 35 7.5 135 ± 3.2 3.8 * 6.3 ± 0.8 -32.3 0.5 wt% CNT * 3461 ± 38 15.4 128 ± 3.4 1.5 * 4.3 ± 0.8 -53.8 1.0 wt% CNT * 3384 ± 48 12.8 * 106 ± 5.1 -18.5 * 3.7 ± 1.2 -60.2 0.25 wt% CNT-POSS * 3361 ± 31 12.0 * 138 ± 2.7 6.2 8.4 ± 0.9 -9.7 0.5 wt% CNT-POSS * 3620 ± 32 20.7 * 144 ± 3.6 10.8 7.9 ± 0.8 -15.1 1.0 wt% CNT-POSS * 3745 ± 44 24.8 * 149 ± 4.8 14.6 7.5 ± 1.3 -18.9 Material Pure VE 104 Figure 63 illustrates the fracture strains as a function of nanoparticle concentration. According to the figure, the flexibility of nanocomposite systems show different variations. For POSS/VE system, a slight reduction (~3 %) in the fracture strain is detectable with 0.25 wt% POSS, compared to that of VE polymer. It is also observed that the fracture strain considerably drops by infusion of higher amounts of POSS. Maximum reduction in fracture strain occurs with nanocomposite having 1.0 wt% POSS concentration (~58 %, Table 7). Microstructural studies have confirmed a relatively good dispersion of POSS in the polymer at 0.25 wt% concentration (see Figs. 46 and 50c). Some protruded fragments appear on the fracture surface of this specimen (Fig. 50b), indicating a limited plastic deformation during fracture. The improved plasticity due to the presence of well-dispersed POSS would allow the material to deform to larger extent. As a result, the fracture strain remains close to that of the neat polymer. At higher POSS concentrations, spherical agglomerates of POSS appear in the microstructure. These agglomerates are bigger at 1.0 wt% POSS specimen (Fig. 46), and a de-bonding between agglomerates and the matrix can be seen (Fig. 52c). Agglomeration of POSS as well as the occurrence of de-bonding are believed to be the responsible for early fracture of specimens at high POSS concentrations. For CNT/VE system, the fracture strain exhibits a sudden drop with nanotube addition (up to ~54 % reduction at 0.5wt% CNT), and then reduces slightly with higher CNTs. Microstructural studies of the 0.25 wt% specimen have revealed the presence of small agglomerates with partial disentanglement (Fig 53c). With addition of CNTs, a remarkable increase in the size of agglomerates has been detected (Fig. 47). These large 105 agglomerates can be considered as stress concentration sites and promote the fracture of the specimen at early stages of loading. The CNT-POSS/VE system exhibits a more consistent behavior. There is a slight reduction in the fracture strain with CNT-POSS concentration. This can be correlated to the observed agglomerates in the structure (Fig. 48). However, nanocomposites fabricated with POSS-modified CNTs display a high level of formability, compared to that of other reference systems. For instance, nanocomposite with 1.0 wt% CNT-POSS can tolerate a flexural strain up to ~7.5 %, as shown in Table 7. As verified by SEM studies, the key factor in higher deformability of these materials is the improved nanotube dispersion due to surface modification of CNTs. Moreover, the new deformation mechanism discovered in well-dispersed areas can promote plastic deformation, and thus enhance the formability. Figure 63: The variations of fracture strain with nanofiller concentration. 106 The variations of elastic modulus with nanoparticle concentration are illustrated in Fig. 64. For nanocomposites with as-received CNT and POSS, two distinct zones are identified. The elastic modulus exhibits an improvement with addition of nanoparticles up to 0.5 wt%, and then drops. This drop is more pronounced in 1.0 wt% POSS/VE composite, where the elastic modulus remains close to that of the neat resin. The slight improvement in the modulus is correlated to the partial disentanglement of agglomerates. With higher nanoparticles, the size of agglomerates increases and weak interfaces between large agglomerates and the matrix prevent an efficient load transfer. Consequently, de-bonding occurs during deformation, which reduces the elastic modulus. On the other hand, nanocomposites fabricated with POSS-modified CNTs show a gradual increase in the elastic modulus with nanotube concentration. The maximum improvement in the flexural modulus (~25 %) occurs with 1.0 wt% CNT-POSS/VE nanocomposite, as shown in Table 7. Comparing the elastic modulus values at each concentration shows that CNT-POSS/VE system provides the highest modulus. It was indicated earlier and it is observed in this chapter that CNT-POSS nanohybrids are seen to be chemically more prone to get dispersed within the resin than the as-received nanoparticles. These well-dispersed nanotubes can bond strongly to the matrix through POSS linkages. The formation of a high number of stiff regions in the microstructure of CNT-POSS/VE specimens is responsible for such enhanced elastic modulus. 107 Figure 64: The variations of elastic modulus with nanofiller concentration. Figure 65 depicts the flexural strength of nanocomposites as a function of nanofiller concentration. A similar ascending-descending behavior is noticed for both POSS/VE and CNT/VE systems, with maximum values of flexural strength with 0.5 wt% and 0.25 wt%, respectively. The observed enhancement in flexural strength is negligible (~4 %). However, the flexural strength of CNT-POSS/VE system is higher than that of the other systems, and shows an increase with nanotube content. The flexural strength can reach up to ~150 MPa with 1.0 wt% CNT-POSS concentration, as shown in Table 7. This is an enhancement of ~15 % when compared with the neat VE. Optical microscopy studies of nanocomposites fabricated with CNT-POSS have revealed formation of a fine textured microstructure due to POSS-functionalization of nanotubes. It is seen that a finer combination of CNTs and polymer leads to an enhancement in the strength. 108 Figure 65: The variations of flexural strength with nanofiller concentration. Another important parameter in controlling the strength of nanocomposites is the state of nanoparticle dispersion. Although the as-received CNTs and POSS show a partial dispersion at low concentrations, the external load cannot be effectively transferred from polymer to these particles because of poor interfacial adhesion. As a result, the observed enhancement in strength are negligible. Due to improved interfacial adhesion of CNTPOSS with the polymer, a higher amount of load can be transferred to the well-dispersed CNTs. Consequently, nanocomposites with CNT-POSS nanohybrids provide a higher strength. A sharp drop in the flexural strength of POSS/VE and CNT/VE nanocomposites at high nanoparticle content is also detected, which can be correlated with the early rupture of the specimen due to agglomeration, as indicated earlier. 109 8.4 Thermal Properties 8.4.1 Differential Scanning Calorimetry (DSC) Results The DSC heating curves of various materials are shown in Fig. 66. Glass transition temperatures were evaluated according to the method described in Section 6.6. The variations in the Tg as a function of nanoparticle concentration are shown in Fig. 67. The value at zero concentration (117 ºC) is associated with the Tg of the neat VE polymer. As it is observed, two distinct zones are identified for each system. The Tg is elevated first with the addition of nanoparticles. After reaching a peak value at 0.25 wt%, the Tg drops with the inclusion of more nanoparticles. The maximum improvement in the Tg (~7 ºC) comes from the composite fabricated with POSS-functionalized CNTs. In nanocomposites with as-received CNTs and POSS, a reduction in Tg is noticed with addition of nanoparticles. As a result, the Tg of these nanocomposites falls below the Tg of the polymer. The maximum reduction in the Tg (~15 ºC) is found with 1.0 wt% POSS/VE nanocomposite. The observed behavior is Fig. 67 can be explained according to the quality of nanoparticle dispersion and interactions between nanoparticles and the matrix. It is generally noticed that an improvement in dispersion of nanoparticles results in a higher Tg. For instance, a relatively satisfactory dispersion state was identified for nanocomposites with 0.25 wt% nanofillers. Therefore, they hold a higher Tg value, compared to the neat polymer. With a good dispersion, large interfacial areas would be available between individual nanoparticles and their surrounding matrix. In this condition, a network of nanoparticles is created in between polymer chains, which can increase the cross-linking density and restrict the mobility of polymer chains [53,117,118]. Hence, the Tg of 0.25 110 wt% nanocomposites are higher than that of VE polymer. However, CNT-POSS nanohybrids are found more effective in enhancing the Tg than as-received CNTs. Microstructural studies in Section 8.2 have demonstrated that a more effective distribution/dispersion of CNTs occurs due to POSS-modification, which provides a more effective confinement. In addition, POSS molecules at the interface can promote the crosslinking density by reacting to polymer chains through their organic groups. Consequently, a more effective cross-linking is believed to be occurred between polymer chains, which results in a higher Tg. The observed reduction in Tg can be explained as follows. In CNT-POSS system, a higher number of agglomerates is identified with higher concentrations (Fig. 48), while their size remains at the same level. The presence of more agglomerates through the resin molecules could alter the rate of polymerization, and thus a slight reduction in the Tg is observed. Unlike the CNT-POSS system, a considerable increase in the size of agglomerates can be recognized for POSS/VE and CNT/VE system with increasing nanofiller concentration. In this condition, free radicals get entrapped within the galleries of large agglomerates as reported in reference [54]. This can reduce the cross-linking density and dramatically decrease the Tg of the composite material. 111 Figure 66: DSC thermograms of the VE polymer and its nanocomposites. Figure 67: The glass transition temperature as a function of nanofiller concentration. 112 8.4.2 Thermogravimetric Analysis (TGA) Results Thermogravimetric analysis has been carried out on the neat VE and its nanocomposites to study the effect of interfacial adhesion on the thermal stability of the polymer material. The TGA heating curves of the experimental materials are presented in Fig. 68 to 70. The corresponding derivative thermogravimetric (DTG) curve, which represents the derivative of weight with respect to time as a function of temperature, is also plotted for each material system. TGA monitors the weight loss as a function of temperature as the specimen is heated from RT to ~1000 ºC under an inert atmosphere. Two characteristic temperatures are extracted from TGA/DTG curves; onset temperature (TO), corresponds to the onset of material decomposition; and decomposition temperature (TD), which is the inflection point of the TGA curve [127], and corresponds to the maximum rate of change of mass. As the material begins to decompose, weight loss becomes more and more prominent and at a specific point, the rate of weight loss reaches a maximum. This point (i.e., TD) is usually identified as a sharp peak in the DTG curve. Generally, it can be seen that the weight loss of VE polymer and its nanocomposites occurs in two steps. The first-step decomposition exhibits a very sharp DTG peak, indicating the initiation of a rapid and vigorous decomposition feature. This behavior is corresponding to the initial breakdown of polymer backbone components [128]. The majority of material weight loss takes place in the first step. The second-step decomposition is characterized by a small and weak DTG peak, and is due to extremely slow decomposition of the remaining polymer. 113 Figure 68: TGA and DTG heating curves for POSS/VE system. Figure 69: TGA and DTG heating curves for CNT/VE system. 114 Figure 70: TGA and DTG heating curves for CNT-POSS/VE system. The variations of TO and TD with nanoparticle concentration are shown in Fig. 71 and 72, respectively. It can be seen from Fig. 71 that infusion of 0.25 wt% POSS results in a slight improvement in the TO of neat polymer. Microstructural studies revealed a good dispersion of POSS for this specimen. In this condition, POSS cages might be able to hold on to VE chains using their organic groups and thus, improve the thermal stability. A higher thermal stability is achieved by inclusion of (up to 0.5 wt%) CNT-POSS nanohybrids. In fact, presence of a rigid network of individual CNTs with POSS linkages enhances the material stability against thermal decomposition. These structural characteristics can also shift the TD of CNT-POSS/VE nanocomposites to a higher level as seen in Fig. 72. However, the agglomeration of nanoparticles is also seen to alter the TD of the polymer matrix. 115 Figure 71: The onset temperature as a function of nanofiller concentration. Figure 72: The decomposition temperature of VE and its nanocomposites. 116 CHAPTER 9. THEORETICAL STUDY OF THE EFFECT OF INTERFACE 9.1 Introduction Continuum mechanics theories are commonly used to study the mechanics of materials. However, these theories cannot predict the behavior of materials containing nanoparticles with realistic accuracy. In fact, nano-structured materials are prominently influenced by the local motions of discrete (non-continuum) set of atoms, so that their behavior can be defined by dynamic intermolecular interactions and surface forces. In this regard, Molecular dynamics (MD) simulation is introduced as a useful method to determine the molecular-level interactions of materials. In MD simulation, atoms and molecules are allowed to interact for a period of time, while the Newton's equations of motion are numerically solved with appropriate algorithms. All interactions between particles are defined by molecular mechanics force fields (see Appendix A). Therefore, MD simulations would generate information including atomic positions, velocities, and accelerations. These results can be used to evaluate different properties of the material model. As revealed earlier from our experimental studies, the surface modification of CNTs is a key factor to control the mechanical properties of polymer. This is mainly governed by molecular interactions at the CNT/polymer interface. Our goal in the current chapter is to quantitatively study the effect of CNT surface modification on the interfacial interactions during a simulated mechanical test. We prepared molecular models of VE polymer and CNT/VE nanocomposites with different nanotube surface conditions. The elastic modulus of each material model was evaluated through molecular-level 117 unidirectional tensile testing. The resulting modulus would be an appropriate tool to explain how the external load can be transferred from VE matrix to the CNT. 9.2 Preparation of Molecular Models Initial molecular models were created using the Project Aten (version 1.9) program [129]. Aten is a package with a strong graphical user interface which allows editing and manipulation of atoms and molecules in isolated and periodic systems. The program enables the user to place atoms of different species in the computational domain and to connect them with inter-atomic bonds of given order and hybridization. The atomic configurations were generated in the Aten program using the following procedure: 9.2.1 VE Model (a) A single molecule of the VE resin (73 atoms, 512.6 g mol-1) was first constructed in a cubic cell as shown in Fig. 73a. The corresponding chemical structure of VE resin (with n=1) can be seen in Fig. 5. (b) Additional molecules of the VE resin were generated following the procedure described in part (a). Twelve VE molecules were then mutually cross-linked using four styrene molecules (Fig. 73b) to create a VE polymer fragment, Fig. 73c. This fragment has a total of 932 atoms and 6560 g mol-1 weight. (c) The cross-linked VE fragment in part (b) was duplicated to create eight fragments in the computational cell. These fragments were then displaced/rotated and randomly placed in a cubic cell with side length of 85 Å. 118 (d) To obtain the equilibrated configuration, the VE model was subjected to an energy minimization process using the conjugated gradient [130] method (see Appendix B). Figure 73: The molecular models of (a) VE resin molecule, (b) styrene molecule and (c) cross-linked VE fragment. 9.2.2 CNT/VE Models (a) The atomic configuration corresponding to a pristine zigzag single-walled CNT was constructed using the DL_POLY program (Appendix C) [131], see Fig. 74a. The CNT has a total of 420 atoms with the diameter and length of 7.86 Å and 44.7 Å, respectively. 119 (b) The pristine CNT in part (a) was used to prepare the carboxylic acid functionalized CNT model with ~8 wt% COOH functional groups. In this regard, ten COOH groups were created and randomly placed next to the pristine CNT. The sp2 carbon atom of CNT (next to each COOH group) was replaced with a sp3 carbon, and then a single bond was made between this sp3 carbon and the carbon of COOH group. The resulting CNT-COOH model has a total of 460 atoms, Fig. 74b. (c) The CNT-COOH in part (b) was used for preparation of POSS-modified CNT model. The molecular structure of POSS-NH2 (Fig. 74c) corresponding to the as-received POSS in this study was created using the Aten program. In the next step, three POSS molecules were placed randomly around the CNT-COOH, so that the amine (NH2) group of each POSS was located next to one of COOH groups. Then, the –OH segment of COOH group and one of H atoms of amine group were removed, and a single bond was created between C atom of COOH and N atom of aminopropyl group. The resulting CNT-POSS model with three C–N bonds and 841 atoms is shown in Fig. 74d. (d) To obtain the equilibrated configuration, the prepared CNT models in part (a), (b) and (c) were subjected to an energy minimization process using the conjugated gradient method. (e) To make the initial configuration of CNT/VE models, the equilibrated nanotubes were separately placed at the center of a cubic cell with side length of 100 Å. The CNTs were aligned in the cell z-direction. Then, eight cross-linked VE fragments (Fig. 73c) were placed randomly around CNTs. (f) In case of nanocomposite model with pristine CNT, no covalent bonding was defined between VE fragments and nanotube. For CNT-COOH model, three covalent 120 bonds were created between CNT and VE by directly bonding the COOH group to a carbon atom at the free end of a VE molecule. For CNT-POSS model, three covalent bonds were created between POSS and VE. This was done by directly bonding a carbon atom of an isobutyl group (Fig. 74c) to a carbon atom at the free end of a VE molecule. In fact, it is assumed that in nanocomposite models containing CNT-COOH and CNT-POSS, there are equal numbers of chemical bonding between the CNT and surrounding VE matrix. Figure 74: The molecular models of (a) pristine CNT, (b) carboxylic acid functionalized CNT, (c) POSS-NH2 and (d) POSS-functionalized CNT. 121 9.3 Force Fields Once the molecular structures were developed, corresponding molecular mechanics force fields were then defined. In the current study, the polymer chains and CNTs were described by “DREIDING” force field [132], which uses appropriate bond stretching, angle bending and dihedral potentials between atoms. The non-bonded van der Waals, referred as vdW, interactions within or between VE chains were modeled using Lenard-Jones (LJ) potential [133]. The non-bonded interactions between VE chains and CNTs were also evaluated by the LJ potential. 9.4 MD Simulation Details MD simulations were performed using DL_POLY (version 4.04) simulation package obtained from Daresbury Laboratory [131]. DL_POLY is a parallel MD simulation package designed for the simulation of macromolecules, polymers, ionic systems, and solutions on a distributed memory parallel computer. All simulations were carried out at temperature and pressure of 300 K and 1 atm, respectively, with 0.5 fs time steps. To achieve a desired density comparable to the experimental condition, the unit cell size was adjusted using isothermal-isobaric NPT (constant pressure and constant temperature) ensemble for few nanoseconds depending on the molecular model and cell size. The temperature and pressure were controlled using the method of Berendsen et al. [134,135]. Integration was performed using the Verlet algorithm, which has the strength of being time-reversible [133,136]. 122 In CNT/VE models, the nanotube was modeled as a non-deformable solid inclusion. Such feature was achieved by expressing CNT as a set of “Frozen” atoms during MD simulation. From the MD point of view, the frozen atoms are defined as a set of atoms that are completely immobilized (i.e., their original coordinates remain unaltered throughout the entire simulation) [137]. Therefore, CNT alignment in the cell z-direction will be preserved. To be able to establish the stress-strain relationship for the prepared models, the equilibrium state of all systems was first obtained. From MD point of view, attainment of such a state requires fulfillment of two major criteria, i.e., to achieve energy stabilized state at a prescribed temperature, and to obtain the minimum initial stress state for the periodic box [137]. In this regard, the stable state was achieved by consecutively subjecting models to canonical (NVT) and microcanonical (NVE) ensembles for 25 ps and 5 ps, respectively. In the second step, the unit cell size was adjusted to minimize the initial stresses using NPT ensemble for 100 ps. Up to this point, simulations were done while CNTs were considered as frozen units in the cell. Next, the molecular models were further equilibrated by NVT for 100 ps to create zero-stress equilibrated structures. Once again, the CNTs were held rigid for the first 80 ps, and the constraints were then removed for the last 20 ps. The structures were considered to be equilibrated since the average total potential energy remained constant over the last 15 ps of the simulations. The final periodic cells of the neat VE and a CNT/VE nanocomposite model are shown in Fig. 75. The details of computational cells after equilibration process are presented in Table 8. 123 Figure 75: Computational cells of (a) the neat VE and (b) the nanocomposite with CNT-COOH after complete equilibration. Table 8: Details of the equilibrated molecular models for MD simulation. Model Number of atoms Cell size (Å3) Density (g cm-3) Neat VE polymer 7456 44×44×44 1.02 Pristine CNT 7876 44×44×48 1.06 CNT-COOH 7907 44×44×48 1.05 CNT-POSS 8291 45×45×48 1.04 Nanocomposite 9.5 Simulation of Unidirectional Tension Tests In this study, the static deformation approach [138] in unidirectional tensile mode was used to derive the elastic constant of each model. For each model at the equilibrated state, a uniform strain field (total of 1% engineering strain) along the z-direction was applied in 10 steps. In each step, the size of the unit cell and the atomic positions of VE chains and CNT were scaled by 0.1% strain, while keeping all other strain components fixed to zero. At the end of each step, all systems were run for 15 ps using NVT ensemble. 124 The first 10 ps run was the “equilibration” run in which the molecular structure settled at the desired thermodynamic state. The last 5 ps run was the “production” run, which was used to calculate the desired thermodynamic parameters. The resulting configuration of models at the end of each step was considered as the initial configuration for the next step. For every 10th time step during the production run, the volume averaged virial stresses were obtained using the following equation [131,137], 1 1 πΌπ½ πΌπ½ ππ ) πππ = − π ∑πΌ (ππΌ π£ππΌ π£ππΌ + 2 ∑π½≠πΌ πΉπ (16) Where V is the volume of the unit cell and equal to the sum of the atomic volumes of all the atoms. π£ππΌ is the i-component of the velocity of atom α, π£ππΌ is the j-component of πΌπ½ the velocity of atom α, πΉπ πΌπ½ and ππ is the i-component of the force acting between atoms α and β, is the j-component of separation distance between atoms α and β. The first term in Eq. 16 is associated with the contribution from kinetic energy due to thermal vibration, and the second term is related to change of potential energy due to the applied deformation in z-direction. The negative sign is used to express tensile stress as a positive quantity, as opposed to the sign conventionally used in MD simulation for compression (as a positive quantity). At each step of strain application, the recorded stress components were averaged. These stress components are considered as the corresponding stress components resulting from the applied strain field. The elastic behavior of the model was then described using continuum mechanics. The generalized constitutive relation of the equivalent continuum can be expressed by, π11 πΆ11 π { 22 } = [πΆ12 π33 πΆ13 πΆ12 πΆ22 πΆ23 πΆ13 π11 πΆ23 ] {π22 } πΆ33 π33 (17) 125 where πππ , πΆππ , and πππ are the stress, elastic constant and strain components, respectively [139]. For unidirectional tension in z-direction, the applied strain at each step is π33 = 0.1% = 0.001. The other strain components are π11 = π22 = 0 due to the constraint deformation. As a result, Eq. 17 becomes π33 = πΆ33 π33 (18) After completion of each step of unidirectional tension, the average stress (π33 ) was evaluated. For each model, the stress values and their corresponding strain values were plotted. According to Eq. 18, the elastic modulus (πΆ33 ) of material models was calculated from the slope of their stress-strain diagram. 9.6 Results and Discussion The stress-strain graphs from unidirectional tension simulation on the neat VE and nanocomposite models are illustrated in Fig. 76. For each model, the elastic modulus (in zdirection) and the stress level corresponding to the applied strain of 1% are evaluated and presented in Table 9. It can be seen from Fig. 76 that all models exhibit a linear elastic behavior up to 1% strain. The experimental results from mechanical testing of the neat VE polymer and its nanocomposites has also shown this behavior, see Fig. 60 to 62. The neat VE model has a simulated elastic modulus of 3011 MPa, which is close to the reported tensile modulus for Derakane 8084 VE polymer (2900 MPa) [93]. According to Table 9, the inclusion of pristine CNT in VE matrix cannot effectively increase the elastic modulus and strength of the neat polymer. The presence of CNT as a rigid nanofiller might be the responsible for a slight enhancement in the elastic modulus. This shows that the non-bonded (vdW) interactions between the CNT and polymer chains 126 are not effective in the load transfer mechanism during tensile deformation. Similar results have been reported for polymer systems containing pristine single-walled CNTs [67,140]. By defining a limited number of chemical bonds between CNT and surrounding polymer, the mechanical properties of the VE polymer are moderately enhanced. The enhancement in the modulus and strength is about 15% and 13%, respectively. This can be correlated to the change in load transfer mechanisms by directly bonding CNTs to the neighbor VE chains. In this regard, Frankland et al. [72] demonstrated by MD simulation that the shear strength of a CNT/polymer interface with only vdW interactions could be increased by over an order of magnitude at the occurrence of covalent bonding for only 1% of CNT’s carbon atoms to the polymer matrix. However, MD simulations of a CNT/polyethylene system have shown that an excessive functionalization of SWCNTs can result in a decrease in the elastic modulus of the corresponding composites [73]. Results indicate that chemical bonding at the CNT/VE interface through POSS molecules can effectively improve the mechanical properties of the neat polymer. The elastic modulus and strength (at 1% strain) can be improved up to 4.12 GPa and 43.6 MPa, respectively. Since the number of chemical bonds are the same for nanocomposites containing CNT-COOH and CNT-POSS, the resulting mechanical behavior would be a useful measure for evaluating the load transfer at the interface. It is evident that interfacial bonding through the same number of COOH groups and POSS molecules can enhance the modulus of the neat polymer by 15% and 37%, respectively. It is suggested that with bonding through a COOH group, a limited number of VE chains at the vicinity of COOH may have a chance to directly interact with the COOH group. As a result, a small portion of the exerted load will be transferred from these VE chains to the CNT. On the other hand, 127 POSS-functionalized CNT may possess a different behavior. With several organic groups attached to the POSS, more (adjacent) VE molecules will be engaged. The rigid nature of the POSS cage can also influence the interactions at the interface. The larger molecular structure of POSS (compared to the COOH group) can also increase the interlocking effect during deformation. As a result, a more effective load transfer to the CNT is expected to occur, which cause the system to tolerate higher loads. Consequently, the elastic modulus of VE polymer is efficiently increased by providing chemical bonds through POSS nanostructures. Figure 76: Simulated stress-strain graphs of (a) VE polymer and its nanocomposite models containing (b) pristine CNT, (c) CNT-COOH, and (d) POSS-functionalized CNT. 128 Table 9: The mechanical properties of materials from MD simulation. Model Elastic modulus, C33 (MPa) Stress at 1% strain, σ33 (MPa) Neat VE polymer 3011 33.6 Pristine CNT 3054 34.3 CNT-COOH 3473 38.0 CNT-POSS 4121 43.6 Nanocomposite 129 CHAPTER 10. SUMMARY AND FUTURE WORK 10.1 Summary A thorough investigation was performed to examine the role of sonication energy in the dispersion of carbon nanotubes (CNTs) into a vinyl ester (VE) resin. The study also determines as to how dispersion condition affects thermo-mechanical properties of nanocomposites. Mechanical properties such as elastic modulus and creep deformation were estimated by three-point flexure and nanoindentation methods. Glass transition temperature (Tg) was also determined by DSC technique. It was observed that at a particular sonication energy level, CNT dispersion is optimum that provides the highest properties. This energy level was termed as threshold energy meaning that sonication time and amplitude must correspond to this energy level to yield the most optimized dispersion and hence the maximum enhancement in properties. It was observed that at this threshold energy level, the Tg and elastic modulus of the pure VE polymer can be enhanced by 13 ºC and 24%, respectively. It was also observed that sonication threshold energy is linked to the concentration of CNTs. The lower the concentration, the higher is the threshold energy level. SEM observations suggested that this threshold energy is controlled by destruction of CNT structure. At higher nanotube contents, the threshold energy is reduced since a higher number of nanotubes are damaged even with a low energy input. A simple and effective method was developed to chemically functionalize CNTs with POSS having amine (NH2) functional group. SEM and AFM analyses of POSS-coated CNTs revealed the presence of an extra phase containing O and Si in the functionalized 130 CNTs, which suggests that the nanotubes are grafted with POSS. Raman spectroscopic analysis of nanotubes showed an increase in the peak intensity ratio (ID/IG) due to POSSfunctionalization. This increase in intensity is attributed to the induced distortion in the CNT structure upon attachment of rigid POSS cages. FTIR results also confirmed covalent grafting of POSS on CNT walls through the formation of amide bonds. In addition, TGA studies suggested that the presence of POSS in CNT-POSS nanohybrids was responsible for the increased thermal stability of CNTs within the temperature range of RT–200 ºC. Such covalent grafting of POSS nanoparticles onto CNTs makes them an effective candidate as reinforcing materials for compatible polymers to enhance thermal and mechanical properties. To investigate the role of CNT interface condition, nanocomposites containing asreceived CNTs, POSS, and CNT-POSS nanohybrids were fabricated with three concentrations of 0.25, 0.5 and 1.0 wt%. Among three concentrations, CNT-POSS/VE nanocomposites turned out to be the best. Microstructural studies of these nanocomposites revealed the presence of a large number of individual nanotubes dispersed into the matrix. When viewed at a larger scale, a homogeneous distribution of CNT-POSS in VE was observed. The formation of a 3D network of individual CNTs with POSS connections to the matrix was also evident and was found to play a significant role in improving the mechanical properties of the polymer. It is also clear that the development of such a network in VE matrix can effectively limit the mobility of polymer chains. Such restriction in chain mobility thus increases the glass transition and thermal decomposition temperatures of the nanocomposites. 131 A molecular dynamics simulation (MDS) study was conducted to evaluate mechanical properties of VE polymer and CNT/VE system. Nanocomposite models with three interface conditions were developed; a pristine CNT/VE model, a CNT-COOH/VE model, and a CNT-POSS/VE model. MDS has demonstrated that interactions at the CNT/VE interface through functional groups improve mechanical properties of the polymer. This improvement is more significant with CNT-POSS/VE system. The presence of POSS linkages was found to increase elastic modulus and strength (at 1% strain) of VE polymer up to 4.12 GPa and 43.6 MPa, respectively. This indicates an effective load transfer from polymer to CNT through POSS linkages, which is in good agreement with our findings from experimental studies. 10.2 Recommendations for Future Work 1. Results have demonstrated that POSS-modification of nanotubes leads to an effective improvement in CNT dispersion condition. It is expected that at this condition, the sonication threshold energy shifts to a higher level. Therefore, a parametric study on the dispersion of CNT-POSS nanohybrids with a variety of sonication energies is suggested. 2. POSS-modification process developed in this work is based on covalent bonding between N atom of POSS and C atom of carboxylic acid (COOH) group of oxidized nanotube. In other words, grafting of POSS onto nanotube walls occurs in a selective manner. By introducing more COOH groups in CNT structure prior to POSSfunctionalization, the chance of attachment of POSS would increase. This can be achieved 132 by oxidizing nanotubes in strong acidic environments. Therefore, conducting a systematic study on oxidization of CNTs and subsequent POSS-modification is highly recommended. 3. In a thermoset polymer, Tg is directly related to the mobility of polymer chains. Inclusion of CNTs can affect Tg depending on nanotube dispersion state and interfacial interactions. With an ideal dispersion, surface condition of CNTs would be the key factor in determination of Tg. In this regard, the role of CNT/VE interface can be simulated at molecular level. Therefore, MD simulation of thermal behavior of CNT/VE system with different interactions at the interface is recommended. 4. Material behavior of the designed system needs to be scaled up theoretically to the bulk properties for engineering applications. This type of translation requires the bottom-up multi-scale modeling approach advancing from the molecular level description all the way to the continuum level. Therefore, a multi-scale modeling approach is recommended. 133 APPENDIXES 134 Appendix A. Force Field A critical point in the molecular simulations of multi-particle systems is the choice of force fields, which approximately describe the forces between the particles. In other words, the knowledge of force fields enables determination of the potential energy of a system in a given configuration. In general, the potential energy of a system of interacting particles can be expressed as a sum of the valence (or bond), πΈπ£ππππππ , cross-term, πΈππππ π −π‘πππ , and non-bond, πΈπππ−ππππ , interaction energies, with the following equation: πΈπ‘ππ‘ππ = β‘ πΈπ£ππππππ + β‘ πΈππππ π −π‘πππ + β‘ πΈπππ−ππππ (19) The valence energy commonly includes a bond stretching term, πΈππππ , a two-bond angle term, πΈπππππ , a dihedral bond-torsion term, πΈπ‘πππ πππ , an inversion (or an out-of-plane interaction) term, πΈπππ , and a Urey-Bradlay term which involves interactions between two atoms bonded to a common atom, πΈππ΅ , as πΈπ£ππππππ = β‘ πΈππππ + β‘ πΈπππππ + β‘ πΈπ‘πππ πππ + β‘ πΈπππ + β‘ πΈππ΅ (20) A schematic explanation of the first four types of valence interactions is given in Fig. 77. 135 Figure 77: A schematic representation of valence interactions including (a) stretch, (b) angle, (c) torsion, and (d) inversion terms. The cross-term interaction energy, πΈππππ π −π‘πππ , accounts for some changes in bond lengths and bond angles caused by the surrounding atoms and generally includes: stretchstretch interactions between two adjacent bonds, πΈππππ−ππππ , stretch-bend interaction between a two-bond angle and one of its bonds, πΈππππ−πππππ , bend-bend interactions between two valence angles associated with a common vertex atom, πΈπππππ−πππππ , stretchtorsion interactions between a dihedral angle and one of its end bonds, πΈπππ_ππππ−π‘πππ πππ , stretch-torsion interactions between a dihedral angle and its middle bond, πΈππππππ_ππππ−π‘πππ πππ , bend-torsion interactions between a dihedral angle and one of its valence angles, πΈπππππ−π‘πππ πππ , and bend-bend-torsion interactions between a dihedral angle and its two valence angles, πΈπππππ−πππππ−π‘πππ πππ , terms as: πΈππππ π −π‘πππ = β‘ πΈππππ−ππππ + πΈπππππ−πππππ + β‘ πΈππππ−πππππ +β‘πΈπππ_ππππ−π‘πππ πππ + β‘ πΈππππππ_ππππ−π‘πππ πππ +πΈπππππ−π‘πππ πππ + β‘ πΈπππππ−πππππ−π‘πππ πππ 136 (21) The non-bond interaction term,β‘πΈπππ−ππππ , accounts for the interactions between non-bonded atoms and includes the van der Waals energy,β‘πΈπ£ππ , the Coulomb electrostatic energy,β‘πΈπΆππ’ππππ , and the hydrogen bond energy,β‘πΈπ»−ππππ , as: πΈπππ−ππππ = β‘ πΈπ£ππ + β‘ πΈπΆππ’ππππ + β‘ πΈπ»−ππππ (22) In order to model the inter- and intra-molecular atomic interactions in CNT/polymer systems, various components of the potential energy can be defined as: πΈππππ = ∑π[πΎ2 (π − π0 )2 + πΎ3 (π − π0 )3 + πΎ4 (π − π0 )4 ] (23) πΈπππππ = ∑π[π»2 (π − π0 )2 + π»3 (π − π0 )3 + π»4 (π − π0 )4 ] (24) πΈπ‘πππ πππ = ∑∅[π1 [1 − πππ (∅ − ∅10 )] + π2 [1 − πππ (2∅ − ∅02 )] +π3 [1 − πππ (3∅ − ∅03 )]] (25) πΈπππ = ∑π₯ πΎπ₯ π 2 (26) πΈππππ−ππππ = ∑π ∑π′ πΉππ′ (π − π0 )(π ′ − π0′ ) (27) πΈπππππ−πππππ = ∑π ∑π′ πΉππ′ (π − π0 )(π ′ − π0′ ) (28) πΈππππ−πππππ = ∑π ∑π πΉππ (π − π0 )(π − π0 ) (29) 137 πΈπππ_ππππ−π‘πππ πππ = ∑π ∑∅ πΉπ∅ (π − π0 ) [π1 πππ ∅ + π2 πππ 2∅ +π3 πππ 3∅] (30) πΈππππππ_ππππ−π‘πππ πππ = ∑π′ ∑∅ πΉπ′ ∅ (π ′ − π0′ )[πΉ1 πππ ∅ + πΉ2 πππ 2∅ +πΉ3 πππ 3∅] (31) πΈπππππ−π‘πππ πππ = ∑π ∑∅ πΉπ∅ (π − π0 ) [π1 πππ ∅ + π2 πππ 2∅ +π3 πππ 3∅] (32) πΈπππππ−πππππ−π‘πππ πππ = ∑∅ ∑π ∑π′ πΎ∅ππ′ πππ ∅(π − π0 )(π ′ − π0′ )] π β‘π π π πΈπΆππ’ππππ = ∑π>π πβ‘π (34) ππ π΄ πΈπ£ππ = ∑π>π [ πππ9 − ππ π΅ππ 6 πππ (33) ] (35) where π and π ′ are the bond lengths, π the two-bond angle, ∅ the dihedral torsion angle, χ the out of plane angle, π the atomic charge, π the dielectric constant, and πππ the π − π atomic separation distance. Other parameters are the system dependent parameters, implemented into the computational program [141]. 138 Appendix B. FORTRAN Code for Energy Minimization of MD Models !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! ! dl_poly_4 subroutine for minimising the structure by using conjugate ! gradient method (CGM). ! ! Note: minimisation type and criterion: ! keymin=0 : absolute force ! keymin=1 : relative energy ! keymin=2 : absolute displacement ! ! copyright - daresbury laboratory ! author - i.t.todorov & w.smith october 2012 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Use kinds_f90 Use comms_module, Only : idnode,mxnode,gsum,gmax Use setup_module, Only : nrite,mxatms,mxcons,mxtpmf,mxpmf,engunit Use config_module, Only : natms,nlast,nfree, & lsi,lsa,lfrzn,lfree,lstfre, & weight,xxx,yyy,zzz,fxx,fyy,fzz Use rigid_bodies_module, Only : lshmv_rgd,lishp_rgd,lashp_rgd Use minimise_module Implicit None Logical, Logical, Integer, Intent( In ) :: l_str Intent( InOut ) :: relaxed,lrdf Intent( In ) :: imcon,megatm,megcon, & megpmf,megrgd,keymin Real( Kind = wp ), Intent( In ) :: min_tol,tstep,stpcfg Logical, Save :: newjob = .true. , l_rdf, l_mov Character( Len = 8 ), Save :: word Character( Len = 6 ) :: name Integer, Save :: keyopt Integer :: fail(1:8),i,j,levcfg Real( Kind = wp ), Save :: total,grad_tol,eng_tol,dist_tol,step, & eng_0,eng_min,engcon,engpmf,eng,eng0,eng1,eng2, & grad,grad0,grad1,grad2,onorm,sgn,stride,gamma ! Optimisation iteration and convergence limits Integer, Parameter :: mxpass = 1000 139 Real( Kind = wp ), Save :: min_pass,pass(1:5) Real( Kind = wp ), Allocatable :: gxx(:),gyy(:),gzz(:) ! Constraints and PMFs arrays Logical, Allocatable :: lstitr(:) Integer, Allocatable :: lstopt(:,:),listot(:) Real( Kind = wp ), Allocatable :: dxx(:),dyy(:),dzz(:) Integer, Allocatable :: indpmf(:,:,:) Real( Kind = wp ), Allocatable :: pxx(:),pyy(:),pzz(:) Real( Kind = wp ), Allocatable :: txx(:),tyy(:),tzz(:) Real( Kind = wp ), Allocatable :: uxx(:),uyy(:),uzz(:) fail=0 If (megcon > 0 .or. megpmf > 0) Then Allocate (lstitr(1:mxatms), Stat=fail(1)) If (megcon > 0) Then Allocate (lstopt(0:2,1:mxcons),listot(1:mxatms), Stat=fail(2)) Allocate (dxx(1:mxcons),dyy(1:mxcons),dzz(1:mxcons), Stat=fail(3)) End If If (megpmf > 0) Then Allocate (indpmf(1:Max(mxtpmf(1),mxtpmf(2)),1:2,1:mxpmf), Stat=fail(4)) Allocate (pxx(1:mxpmf),pyy(1:mxpmf),pzz(1:mxpmf), Stat=fail(5)) End If End If If (megrgd > 0) Then Allocate (txx(1:mxatms),tyy(1:mxatms),tzz(1:mxatms), Stat=fail(6)) Allocate (uxx(1:mxatms),uyy(1:mxatms),uzz(1:mxatms), Stat=fail(7)) End If Allocate (gxx(1:mxatms),gyy(1:mxatms),gzz(1:mxatms), Stat=fail(8)) If (Any(fail > 0)) Then Write(nrite,'(/,1x,a,i0)') 'minimise_relax allocation failure, node: ', idnode Call error(0) End If If (newjob) Then newjob = .false. ! At start no optimisation has been attempted yet keyopt = 0 ! At start the minimum energy is defined as zero eng_min = 0.0_wp 140 ! Passage accumulators ! pass(1) - cycles counter ! pass(2) - access counter ! pass(3) - average cycles ! pass(4) - minimum cycles ! pass(5) - maximum cycles pass = 0 ! total number of active particles (excluding frozen sites and massless shells) total=0.0_wp Do i=1,natms If (lfrzn(i) == 0 .and. (weight(i) > 1.0e-6_wp .or. lfree(i) == 1)) & total=total+1.0_wp End Do If (mxnode > 1) Call gsum(total) End If ! Step length for relaxation, enlarged depending on functionality step=tstep**2 If (megcon == 0 .and. megpmf == 0) Then If (megrgd == 0) Then step=10.0_wp*step Else step=5.0_wp*step End If End If If (keyopt == 0) Then ! Initial configuration energy eng_0=stpcfg ! Allocate working arrays Call allocate_minimise_arrays() ! No minimisation is yet attempted relaxed=.false. ! No RB move is yet attempted 141 l_mov=.false. ! Minimum needed for a pass for this minimisation cycle min_pass = Huge(1.0_wp) ! Avoid rdf calculation redundancy l_rdf=lrdf If (lrdf) lrdf=.false. ! Determine optimisation If (keymin == 0) Then word='force ' Else If (keymin == 1) Then word='energy ' Else If (keymin == 2) Then word='distance' End If ! Print header If (l_str .and. idnode == 0) Then Write(nrite, Fmt=*) Write(nrite,'(3(1x,a),6x,a,10x,a,10x,a,11x,a,9x,a,1p,e12.4)') & 'Minimising',word,'pass','eng_tot','grad_tol','eng_tol','dist_tol','tol=', min_tol Write(nrite,"(1x,130('-'))") End If End If ! Load original forces Do i=1,natms gxx(i)=fxx(i) gyy(i)=fyy(i) gzz(i)=fzz(i) End Do ! Minimised energy is current configuration energy eng=stpcfg ! Calculate pseudo forces and energy for constraint bonds and PMFs 142 If (megcon > 0 .or. megpmf > 0) Then lstitr(1:natms)=.false. ! initialise lstitr If (megcon > 0) Then Call constraints_tags(imcon,lstitr,lstopt,dxx,dyy,dzz,listot) Call constraints_pseudo_bonds(lstopt,dxx,dyy,dzz,gxx,gyy,gzz,engcon) eng=eng+engcon End If If (megpmf > 0) Then Call pmf_tags(imcon,lstitr,indpmf,pxx,pyy,pzz) Call pmf_pseudo_bonds(indpmf,pxx,pyy,pzz,gxx,gyy,gzz,engpmf) eng=eng+engpmf End If End If ! Average forces over all members of a RB and split torques accordingly If (megrgd > 0) Then If (lshmv_rgd) update_shared_units(natms,nlast,lsi,lsa,lishp_rgd,lashp_rgd,gxx,gyy,gzz) Call rigid_bodies_split_torque(imcon,gxx,gyy,gzz,txx,tyy,tzz,uxx,uyy,uzz) End If ! Initialise/get eng_tol & verify relaxed condition eng_tol=0.0_wp If (keyopt > 0) Then eng_tol=Abs(1.0_wp-eng2/eng) If (keymin == 1) relaxed=(eng_tol < min_tol) End If ! Current gradient (modulus of the total force) ! massless shells and frozen particles have zero forces! grad=0.0_wp Do i=1,natms grad=grad+gxx(i)**2+gyy(i)**2+gzz(i)**2 End Do If (mxnode > 1) Call gsum(grad) grad=Sqrt(grad) ! Get grad_tol & verify relaxed condition grad_tol=grad/total If (keymin == 0) relaxed=(grad_tol < min_tol) 143 Call ! Initialise dist_tol dist_tol=0.0_wp ! CHECK FOR CONVERGENCE If (.not.relaxed) Then ! Increment main passage counter pass(1)=pass(1)+1.0_wp ! min_pass = Min(min_pass,._tol) If (keymin == 0) Then min_pass = Min(min_pass,grad_tol) Else If (keymin == 1) Then If (keyopt > 0) min_pass = Min(min_pass,eng_tol) Else If (keymin == 2) Then min_pass = Min(min_pass,dist_tol) End If ! If in mxpass iterations we are not there, give up but ! allow for ten-fold boost in iteration cycle length ! for the very first MD step If (Nint(pass(2)) == 0) Then If (Nint(pass(1)) >= 10*mxpass) Then Call warning(330,min_tol,min_pass,0.0_wp) Call error(474) End If Else If (Nint(pass(1)) >= mxpass) Then Call warning(330,min_tol,min_pass,0.0_wp) Call error(474) End If End If Else Go To 100 End If If (keyopt == 0) Then 144 ! Original configuration energy eng0=eng eng1=eng eng2=eng ! Original gradient (modulus of the total force) onorm=grad grad0=grad grad2=grad ! Set original search direction Do i=1,natms oxx(i)=gxx(i) oyy(i)=gyy(i) ozz(i)=gzz(i) End Do keyopt=1 sgn=1.0_wp stride=sgn*step Else If (keyopt == 1) Then ! Line search along chosen direction eng1=eng0 eng2=eng grad1=grad2 grad2=0.0_wp Do i=1,natms grad2=grad2+oxx(i)*gxx(i)+oyy(i)*gyy(i)+ozz(i)*gzz(i) End Do If (mxnode > 1) Call gsum(grad2) grad2=sgn*grad2/onorm ! Linear extrapolation to minimum If (grad2 < 0.0_wp) Then ! BACK UP FROM THIS DIRECTION keyopt=2 stride=sgn*step*grad2/(grad1-grad2) Else ! CARRY ON IN THIS DIRECTION 145 stride=sgn*step End If Else If (keyopt == 2) Then ! Construct conjugate search vector eng1=eng2 eng2=eng gamma=(grad/grad0)**2 grad0=grad grad2=0.0_wp onorm=0.0_wp Do i=1,natms oxx(i)=gxx(i)+gamma*oxx(i) oyy(i)=gyy(i)+gamma*oyy(i) ozz(i)=gzz(i)+gamma*ozz(i) onorm=onorm+oxx(i)**2+oyy(i)**2+ozz(i)**2 grad2=grad2+oxx(i)*gxx(i)+oyy(i)*gyy(i)+ozz(i)*gzz(i) End Do If (mxnode > 1) Call gsum(onorm) onorm=Sqrt(onorm) If (mxnode > 1) Call gsum(grad2) grad2=grad2/onorm sgn=Sign(1.0_wp,grad2) grad2=sgn*grad2 keyopt=1 stride=sgn*step End If ! Move particles to their new positions accordingly If (megrgd > 0) Then ! active free particles Do j=1,nfree i=lstfre(j) If (lfrzn(i) == 0 .and. weight(i) > 1.0e-6_wp) Then xxx(i)=xxx(i)+stride*oxx(i) yyy(i)=yyy(i)+stride*oyy(i) 146 zzz(i)=zzz(i)+stride*ozz(i) dist_tol=Max(dist_tol,oxx(i)**2+oyy(i)**2+ozz(i)**2) End If End Do dist_tol=Sqrt(dist_tol)*Abs(stride) ! RB particles Call rigid_bodies_move(stride,oxx,oyy,ozz,txx,tyy,tzz,uxx,uyy,uzz,dist_tol) l_mov=.true. Else ! active particles Do i=1,natms If (lfrzn(i) == 0 .and. weight(i) > 1.0e-6_wp) Then xxx(i)=xxx(i)+stride*oxx(i) yyy(i)=yyy(i)+stride*oyy(i) zzz(i)=zzz(i)+stride*ozz(i) dist_tol=Max(dist_tol,oxx(i)**2+oyy(i)**2+ozz(i)**2) End If End Do dist_tol=Sqrt(dist_tol)*Abs(stride) End If If (mxnode > 1) Call gmax(dist_tol) If (keymin == 2) relaxed=(dist_tol < min_tol) i=Nint(pass(1)) If (l_str .and. idnode == 0) Then Write(nrite,'(1x,i23,1p,4e18.8)') i-1,eng/engunit,grad_tol,eng_tol,dist_tol If (Mod(i,25) == 0) Then Write(nrite,"(1x,130('-'))") Write(nrite,'(3(1x,a),6x,a,10x,a,10x,a,11x,a,9x,a,1p,e12.4)') & 'Minimising',word,'pass','eng_tot','grad_tol','eng_tol','dist_tol','tol=', min_tol Write(nrite,"(1x,130('-'))") End If End If 100 Continue l_x=(.not.relaxed) ! Transportation flag If (relaxed) Then 147 ! Final/Only printout i=Nint(pass(1)) If (idnode == 0) Then If (.not.l_str) Then Write(nrite, Fmt=*) Write(nrite,'(3(1x,a),5x,a,10x,a,10x,a,11x,a,9x,a,1p,e12.4)') & 'Minimised',word,'passes','eng_tot','grad_tol','eng_tol','dist_tol','tol=', min_tol Write(nrite,"(1x,130('-'))") End If Write(nrite,'(1x,i23,1p,4e18.8)') i,eng/engunit,grad_tol,eng_tol,dist_tol Write(nrite, Fmt=*) Write(nrite,"(1x,130('-'))") End If ! Collect passage statistics pass(3)=pass(2)*pass(3) pass(2)=pass(2)+1.0_wp pass(3)=pass(3)/pass(2)+pass(1)/pass(2) pass(4)=Min(pass(1),pass(4)) pass(5)=Max(pass(1),pass(5)) ! Rewind keyopt and main passage counter keyopt =0 pass(1)=0.0_wp ! Resume rdf calculations If (l_rdf) lrdf=l_rdf ! Deallocate working arrays Call deallocate_minimise_arrays() ! Dump the lowest energy configuration If (eng < eng_min) Then eng_min=eng name = 'CFGMIN' ! file name levcfg = 0 ! define level of information in file Call write_config(name,levcfg,imcon,megatm,i-1,eng_min/engunit,eng_0/engunit) End If 148 ! setup new quaternions If (l_mov) Then If (lshmv_rgd) update_shared_units(natms,nlast,lsi,lsa,lishp_rgd,lashp_rgd,xxx,yyy,zzz) Call q_setup() End If End If If (megcon > 0 .or. megpmf > 0) Then Deallocate (lstitr, Stat=fail(1)) If (megcon > 0) Then Deallocate (lstopt,listot, Stat=fail(2)) Deallocate (dxx,dyy,dzz, Stat=fail(3)) End If If (megpmf > 0) Then Deallocate (indpmf, Stat=fail(4)) Deallocate (pxx,pyy,pzz, Stat=fail(5)) End If End If If (megrgd > 0) Then Deallocate (txx,tyy,tzz, Stat=fail(6)) Deallocate (uxx,uyy,uzz, Stat=fail(7)) End If Deallocate (gxx,gyy,gzz, Stat=fail(8)) If (Any(fail > 0)) Then Write(nrite,'(/,1x,a,i0)') 'minimise_relax deallocation failure, node: ', idnode Call error(0) End If End Subroutine minimise_relax 149 Call Appendix C. JAVA Code Used for Making Single-Walled CNT Model ********************************************************************* dl_poly/java GUI class to construct fullerene CONFIG files copyright - daresbury laboratory author - w.smith 2000 ********************************************************************* */ public static GUI home; public static MakeBucky job; private static double ccbond; private static int numx,numz; private static JTextField bond,nrx,nrz; private static JButton make1,make2,close; // Define the Graphical User Interface public MakeBucky() { /* ********************************************************************* dl_poly/java GUI routine copyright - daresbury laboratory author - w.smith 2000 ********************************************************************* */ super(); setTitle("Make Fullerene"); getContentPane().setBackground(art.back); getContentPane().setForeground(art.fore); setDefaultCloseOperation(DISPOSE_ON_CLOSE); setFont(fontMain); GridBagLayout grd = new GridBagLayout(); GridBagConstraints gbc = new GridBagConstraints(); getContentPane().setLayout(grd); gbc.fill=GridBagConstraints.BOTH; // Define the Make Tube button make2 = new JButton("Tube"); make2.setBackground(art.butn); 150 make2.setForeground(art.butf); fix(make2,grd,gbc,2,0,1,1); fix(new JLabel(" "),grd,gbc,1,1,1,1); // Bond length JLabel lab1 = new JLabel("C-C Bond (A):",JLabel.LEFT); fix(lab1,grd,gbc,0,2,1,1); bond = new JTextField(8); bond.setBackground(art.scrn); bond.setForeground(art.scrf); fix(bond,grd,gbc,2,2,1,1); // Tube size - rings in x direction JLabel lab2 = new JLabel("Tube size : rings X Y",JLabel.LEFT); fix(lab2,grd,gbc,0,3,2,1); nrx = new JTextField(8); nrx.setBackground(art.scrn); nrx.setForeground(art.scrf); fix(nrx,grd,gbc,0,4,1,1); // Tube size - rings in y direction nrz = new JTextField(8); nrz.setBackground(art.scrn); nrz.setForeground(art.scrf); fix(nrz,grd,gbc,2,4,1,1); // Define the Close button fix(new JLabel(" "),grd,gbc,1,5,1,1); close = new JButton("Close"); close.setBackground(art.butn); close.setForeground(art.butf); fix(close,grd,gbc,2,6,1,1); // Register action buttons make1.addActionListener(this); make2.addActionListener(this); close.addActionListener(this); } public MakeBucky(GUI here) { 151 /* ********************************************************************* dl_poly/java GUI routine copyright - daresbury laboratory author - w.smith 2000 ********************************************************************* */ home=here; println("Activated panel for making fullerene CONFIG files"); job=new MakeBucky(); job.pack(); job.setVisible(true); ccbond=ccab; numx=8; numz=12; bond.setText(String.valueOf(ccbond)); nrx.setText(String.valueOf(numx)); nrz.setText(String.valueOf(numz)); } int bucky() { /* ********************************************************************* construction of carbon nanotube copyright - daresbury laboratory author - w.smith november 2000 ********************************************************************* */ int n,call,levels; double height,alp0,alp2,alp4,rad,xxx,yyy,zzz,ang; levels=2*(numz+1); config=new Config(); config.natms=numx*levels; config.xyz=new double[3][config.natms]; config.atoms=new Element[config.natms]; config.title="Carbon Nanotube "+BML.fmt(numx,5)+" x"+BML.fmt(numz,5); config.pbc.imcon=2; config.levcfg=0; n=0; height=ccbond*(1.5*numz+0.5); 152 alp0=2.0*Math.PI/numx; alp2=alp0/2.0; alp4=alp0/4.0; rad=0.25*Math.sqrt(3.0)*ccbond/Math.sin(alp4); for(int i=0;i<9;i++) config.pbc.cell[i]=0.0; config.pbc.cell[0]=4.0*rad; config.pbc.cell[4]=4.0*rad; config.pbc.cell[8]=height+ccbond; config.pbc.buildBoundary(config.pbc.imcon); zzz=0.5*(ccbond-height-ccbond); ang=0.0; for(int k=0;k<levels;k++) { for(int i=0;i<numx;i++) { config.atoms[n]=new Element("C_R "); config.xyz[0][n]=rad*Math.cos(ang+i*alp0); config.xyz[1][n]=rad*Math.sin(ang+i*alp0); config.xyz[2][n]=zzz; n++; } if(k%2 == 0) { ang+=alp2; zzz+=0.5*ccbond; } else { zzz+=ccbond; } } config.natms=n; config.structure=new Structure(config); cfgsav=copyConfig(config); println("Number of atoms created : "+config.natms); // write CONFIG file fname="CFGBUK."+String.valueOf(numbuk); if(!config.configWrite(fname)) return -1; println("Radius of tube (A) is: "+BML.fmt(rad,20)); numbuk++; // Draw structure if(!editor.isVisible()) editor.showEditor(); 153 editor.pane.restore(); return 0; } public void actionPerformed(ActionEvent e) { /* ********************************************************************* dl_poly/java GUI routine copyright - daresbury laboratory author - w.smith 2000 ********************************************************************* */ int call; String arg = (String)e.getActionCommand(); if (arg.equals("C60")) { ccbond=BML.giveDouble(bond.getText(),1); call=bucky(); } else if (arg.equals("Tube")) { ccbond=BML.giveDouble(bond.getText(),1); numx=BML.giveInteger(nrx.getText(),1); numz=BML.giveInteger(nrz.getText(),1); call=tube(); } else if (arg.equals("Close")) { job.dispose(); } } } 154 REFERENCES [1] S. IIJIMA, Helical Microtubules of Graphitic Carbon, Nature. 354 (1991) 56–58. doi:10.1038/354056a0. [2] E.T. Thostenson, Z. Ren, T.-W. Chou, Advances in the science and technology of carbon nanotubes and their composites: a review, Compos. Sci. Technol. 61 (2001) 1899–1912. doi:10.1016/S0266-3538(01)00094-X. [3] R.H. Baughman, A. a Zakhidov, W. a de Heer, Carbon nanotubes--the route toward applications., Science. 297 (2002) 787–792. doi:10.1126/science.1060928. [4] C. Li, E.T. Thostenson, T.-W. Chou, Sensors and actuators based on carbon nanotubes and their composites: A review, Compos. Sci. Technol. 68 (2008) 1227– 1249. doi:10.1016/j.compscitech.2008.01.006. [5] M.A. Cullinan, M.L. Culpepper, Carbon nanotubes as piezoresistive microelectromechanical sensors: Theory and experiment, Phys. Rev. B - Condens. Matter Mater. Phys. 82 (2010). doi:10.1103/PhysRevB.82.115428. [6] a. Peigney, C. Laurent, E. Flahaut, R.R. Bacsa, a. Rousset, Specific surface area of carbon nanotubes and bundles of carbon nanotubes, Carbon N. Y. 39 (2001) 507– 514. doi:10.1016/S0008-6223(00)00155-X. [7] P.K. Mallick, Fiber-reinforced composites: materials, manufacturing, and design, 2008. doi:10.1016/0010-4361(89)90651-4. [8] E.T. Thostenson, S. Ziaee, T.W. Chou, Processing and electrical properties of carbon nanotube/vinyl ester nanocomposites, Compos. Sci. Technol. 69 (2009) 801– 155 804. doi:10.1016/j.compscitech.2008.06.023. [9] S. Pham, P.J. Burchill, Toughening of vinyl ester resins with modified polybutadienes, Polymer (Guildf). 36 (1995) 3279–3285. doi:10.1016/00323861(95)99426-U. [10] G. Chigwada, P. Jash, D.D. Jiang, C.A. Wilkie, Fire retardancy of vinyl ester nanocomposites: Synergy with phosphorus-based fire retardants, Polym. Degrad. Stab. 89 (2005) 85–100. doi:10.1016/j.polymdegradstab.2005.01.005. [11] S. Taillemite, R. Pauer, Bright future for vinyl ester resins in corrosion applications, Reinf. Plast. 53 (2009) 34–37. doi:10.1016/S0034-3617(09)70223-1. [12] V.P. McConnell, Vinyl esters get radical in composite markets, Reinf. Plast. 54 (2010) 34–38. doi:10.1016/S0034-3617(10)70215-0. [13] J.N. Coleman, U. Khan, W.J. Blau, Y.K. Gun’ko, Small but strong: A review of the mechanical properties of carbon nanotube-polymer composites, Carbon N. Y. 44 (2006) 1624–1652. doi:10.1016/j.carbon.2006.02.038. [14] Z. Spitalsky, D. Tasis, K. Papagelis, C. Galiotis, Carbon nanotube-polymer composites: Chemistry, processing, mechanical and electrical properties, Prog. Polym. Sci. 35 (2010) 357–401. doi:DOI 10.1016/j.progpolymsci.2009.09.003. [15] R. Andrews, M.. Weisenberger, Carbon nanotube polymer composites, Curr. Opin. Solid State Mater. Sci. 8 (2004) 31–37. doi:10.1016/j.cossms.2003.10.006. [16] X. Gong, J. Liu, S. Baskaran, R.D. Voise, J.S. Young, Surfactant-assisted processing of carbon nanotube/polymer composites, Chem. Mater. 12 (2000) 1049–1052. doi:10.1021/cm9906396. [17] L. Vaisman, H.D. Wagner, G. Marom, The role of surfactants in dispersion of 156 carbon nanotubes, Adv. Colloid Interface Sci. 128-130 (2006) 37–46. doi:10.1016/j.cis.2006.11.007. [18] Y.L. Zhao, J.F. Stoddart, Noncovalent functionalization of single-walled carbon nanotubes, Acc. Chem. Res. 42 (2009) 1161–1171. doi:10.1021/ar900056z. [19] J. Jancar, J.F. Douglas, F.W. Starr, S.K. Kumar, P. Cassagnau, A.J. Lesser, et al., Current issues in research on structure-property relationships in polymer nanocomposites, Polymer (Guildf). 51 (2010) 3321–3343. doi:10.1016/j.polymer.2010.04.074. [20] M. Rahmat, P. Hubert, Carbon nanotube-polymer interactions in nanocomposites: A review, Compos. Sci. Technol. 72 (2011) 72–84. doi:10.1016/j.compscitech.2011.10.002. [21] S. Bose, R.A. Khare, P. Moldenaers, Assessing the strengths and weaknesses of various types of pre-treatments of carbon nanotubes on the properties of polymer/carbon nanotubes composites: A critical review, Polymer (Guildf). 51 (2010) 975–993. doi:10.1016/j.polymer.2010.01.044. [22] P.C. Ma, N.A. Siddiqui, G. Marom, J.K. Kim, Dispersion and functionalization of carbon nanotubes for polymer-based nanocomposites: A review, Compos. Part A Appl. Sci. Manuf. 41 (2010) 1345–1367. doi:10.1016/j.compositesa.2010.07.003. [23] Y.Y. Huang, E.M. Terentjev, Dispersion of carbon nanotubes: Mixing, sonication, stabilization, and composite properties, Polymers (Basel). 4 (2012) 275–295. doi:10.3390/polym4010275. [24] J. Hilding, E. a. Grulke, Z. George Zhang, F. Lockwood, Dispersion of Carbon Nanotubes in Liquids, J. Dispers. Sci. Technol. 24 (2003) 1–41. doi:10.1081/DIS- 157 120017941. [25] S. V. Ahir, Y.Y. Huang, E.M. Terentjev, Polymers with aligned carbon nanotubes: Active composite materials, Polymer (Guildf). 49 (2008) 3841–3854. doi:10.1016/j.polymer.2008.05.005. [26] V. Datsyuk, M. Kalyva, K. Papagelis, J. Parthenios, D. Tasis, A. Siokou, et al., Chemical oxidation of multiwalled carbon nanotubes, Carbon N. Y. 46 (2008) 833– 840. doi:10.1016/j.carbon.2008.02.012. [27] L. Meng, C. Fu, Q. Lu, Advanced technology for functionalization of carbon nanotubes, Prog. Nat. Sci. 19 (2009) 801–810. doi:10.1016/j.pnsc.2008.08.011. [28] S.W. Kim, T. Kim, Y.S. Kim, H.S. Choi, H.J. Lim, S.J. Yang, et al., Surface modifications for the effective dispersion of carbon nanotubes in solvents and polymers, in: Carbon N. Y., 2012: pp. 3–33. doi:10.1016/j.carbon.2011.08.011. [29] A. V. Desai, M.A. Haque, Mechanics of the interface for carbon nanotube-polymer composites, Thin-Walled Struct. 43 (2005) 1787–1803. doi:10.1016/j.tws.2005.07.003. [30] B.A. Rozenberg, R. Tenne, Polymer-assisted fabrication of nanoparticles and nanocomposites, Prog. Polym. Sci. 33 (2008) 40–112. doi:10.1016/j.progpolymsci.2007.07.004. [31] S. Iijima, T. Ichihashi, Single-shell carbon nanotubes of 1-nm diameter, Nature. 363 (1993) 603–605. doi:10.1038/363603a0. [32] J. Brandrup, E. Immergut, E.A. Grulke, Polymer handbook, John Wiley Sons, Inc. 12 (1990) 265. doi:10.1016/0168-3659(90)90108-6. [33] E. Saldívar-Guerra, E. Vivaldo-Lima, Handbook of Polymer Synthesis, 158 Characterization, and Processing, 2013. doi:10.1002/9781118480793. [34] L.H.H. Sperling, Introduction to Physical Polymer Science: Fourth Edition, 2005. doi:10.1002/0471757128. [35] R.J. Fleming, An introduction to polymer physics, Physicist. 40 (2003) 26. doi:10.1119/1.1533063. [36] M.T. Shaw, W.J. MacKnight, Introduction to Polymer Viscoelasticity: Third Edition, 2005. doi:10.1002/0471741833. [37] M.T. Byrne, Y.K. Gun’ko, Recent advances in research on carbon nanotubepolymer composites, Adv. Mater. 22 (2010) 1672–88. doi:10.1002/adma.200901545. [38] Y. Sabba, E.L. Thomas, High-concentration dispersion of single-wall carbon nanotubes, Macromolecules. 37 (2004) 4815–4820. doi:10.1021/ma049706u. [39] X.L. Xie, Y.W. Mai, X.P. Zhou, Dispersion and alignment of carbon nanotubes in polymer matrix: A review, Mater. Sci. Eng. R Reports. 49 (2005) 89–112. doi:10.1016/j.mser.2005.04.002. [40] B.P. Grady, Recent developments concerning the dispersion of carbon nanotubes in polymers, Macromol. Rapid Commun. 31 (2010) 247–257. doi:10.1002/marc.200900514. [41] E. Christopher, Brennen, Cavitation and bubble dynamics, 1995. doi:10.1017/CBO9781107338760. [42] A. Sesis, M. Hodnett, G. Memoli, A.J. Wain, I. Jurewicz, A.B. Dalton, et al., Influence of acoustic cavitation on the controlled ultrasonic dispersion of carbon nanotubes, J. Phys. Chem. B. 117 (2013) 15141–15150. doi:10.1021/jp410041y. 159 [43] S.L. Peshkovsky, A.S. Peshkovsky, Shock-wave model of acoustic cavitation, Ultrason. Sonochem. 15 (2008) 618–628. doi:10.1016/j.ultsonch.2007.07.006. [44] L. Pena-Paras, Dispersion of Carbon Nanotubes in Vinyl Polymer Composites, Rice University, 2010. [45] B.F.J. Fuchs, Ultrasonic Cleaningβ―: Fundamental Theory and Application What is “ Ultrasonicsβ―?,” Shock. (2002) 1–14. [46] J. Chandrapala, C. Oliver, S. Kentish, M. Ashokkumar, Ultrasonics in food processing, Ultrason. Sonochem. 19 (2012) 975–983. doi:10.1016/j.ultsonch.2012.01.010. [47] G. Zhang, P. Zhang, B. Wang, H. Liu, Ultrasonic frequency effects on the removal of Microcystis aeruginosa, Ultrason. Sonochem. 13 (2006) 446–450. doi:10.1016/j.ultsonch.2005.09.012. [48] K.L. Lu, R.M. Lago, Y.K. Chen, M.L.H. Green, P.J.F. Harris, S.C. Tsang, Mechanical damage of carbon nanotubes by ultrasound, Carbon N. Y. 34 (1996) 814–816. doi:10.1016/0008-6223(96)89470-X. [49] K. Mukhopadhyay, C.D. Dwivedi, G.N. Mathur, Conversion of carbon nanotubes to carbon nanofibers by sonication [2], Carbon N. Y. 40 (2002) 1373–1376. http://www.scopus.com/inward/record.url?eid=2-s2.0-0036642805&partnerID=40. [50] E.A. Zaragoza-Contreras, E.D. Lozano-Rodríguez, M. Román-Aguirre, W. Antunez-Flores, C.A. Hernández-Escobar, S.G. Flores-Gallardo, et al., Evidence of multi-walled carbon nanotube fragmentation induced by sonication during nanotube encapsulation via bulk-suspension polymerization, Micron. 40 (2009) 621–627. doi:10.1016/j.micron.2009.02.007. 160 [51] Z. Fan, K.T. Hsiao, S.G. Advani, Experimental investigation of dispersion during flow of multi-walled carbon nanotube/polymer suspension in fibrous porous media, Carbon N. Y. 42 (2004) 871–876. doi:10.1016/j.carbon.2004.01.067. [52] Z. Fan, S.G. Advani, Characterization of orientation state of carbon nanotubes in shear flow, Polymer (Guildf). 46 (2005) 5232–5240. doi:10.1016/j.polymer.2005.04.008. [53] A.T. Seyhan, F.H. Gojny, M. TanoΗ§lu, K. Schulte, Rheological and dynamicmechanical behavior of carbon nanotube/vinyl ester-polyester suspensions and their nanocomposites, Eur. Polym. J. 43 (2007) 2836–2847. doi:10.1016/j.eurpolymj.2007.04.022. [54] A.T. Seyhan, M. TanoΗ§lu, K. Schulte, Tensile mechanical behavior and fracture toughness of MWCNT and DWCNT modified vinyl-ester/polyester hybrid nanocomposites produced by 3-roll milling, Mater. Sci. Eng. A. 523 (2009) 85–92. doi:10.1016/j.msea.2009.05.035. [55] H. Yurdakul, A.T. Seyhan, S. Turan, M. Tano??lu, W. Bauhofer, K. Schulte, Electric field effects on CNTs/vinyl ester suspensions and the resulting electrical and thermal composite properties, Compos. Sci. Technol. 70 (2010) 2102–2110. doi:10.1016/j.compscitech.2010.08.007. [56] O. Gryshchuk, J. Karger-Kocsis, R. Thomann, Z. K??nya, I. Kiricsi, Multiwall carbon nanotube modified vinylester and vinylester - based hybrid resins, Compos. Part A Appl. Sci. Manuf. 37 (2006) 1252–1259. doi:10.1016/j.compositesa.2005.09.003. [57] L.S. Schadler, S.C. Giannaris, P.M. Ajayan, Load transfer in carbon nanotube epoxy 161 composites, Appl. Phys. Lett. 73 (1998) 3842–3844. doi:10.1063/1.122911. [58] O. Lourie, H.D. Wagner, Transmission electron microscopy observations of fracture of single-wall carbon nanotubes under axial tension, Appl. Phys. Lett. 73 (1998) 3527–3529. doi:10.1063/1.122825. [59] A.H. Barber, S.R. Cohen, S. Kenig, H.D. Wagner, Interfacial fracture energy measurements for multi-walled carbon nanotubes pulled from a polymer matrix, Compos. Sci. Technol. 64 (2004) 2283–2289. doi:10.1016/j.compscitech.2004.01.023. [60] V.G. Hadjiev, G.L. Warren, L. Sun, D.C. Davis, D.C. Lagoudas, H.J. Sue, Raman microscopy of residual strains in carbon nanotube/epoxy composites, Carbon N. Y. 48 (2010) 1750–1756. doi:10.1016/j.carbon.2010.01.018. [61] C.A. Cooper, R.J. Young, M. Halsall, Investigation into the deformation of carbon nanotubes and their composites through the use of Raman spectroscopy, Compos. Part A Appl. Sci. Manuf. 32 (2001) 401–411. doi:10.1016/S1359-835X(00)00107X. [62] G.M. Odegard, T.S. Gates, K.E. Wise, C. Park, E.J. Siochi, Constitutive modeling of nanotube-reinforced polymer composites, Compos. Sci. Technol. 63 (2003) 1671–1687. doi:10.1016/S0266-3538(03)00063-0. [63] K. Liao, S. Li, Interfacial characteristics of a carbon nanotube-polystyrene composite system, Appl. Phys. Lett. 79 (2001) 4225–4227. doi:10.1063/1.1428116. [64] Z. Liang, J. Gou, C. Zhang, B. Wang, L. Kramer, Investigation of molecular interactions between (10, 10) single-walled nanotube and Epon 862 resin/DETDA curing agent molecules, Mater. Sci. 162 Eng. A. 365 (2004) 228–234. doi:10.1016/j.msea.2003.09.032. [65] J. Gou, B. Minaie, B. Wang, Z. Liang, C. Zhang, Computational and experimental study of interfacial bonding of single-walled nanotube reinforced composites, Comput. Mater. Sci. 31 (2004) 225–236. doi:10.1016/j.commatsci.2004.03.002. [66] Q. Zheng, Q. Xue, K. Yan, X. Gao, Q. Li, L. Hao, Influence of chirality on the interfacial bonding characteristics of carbon nanotube polymer composites, J. Appl. Phys. 103 (2008). doi:10.1063/1.2844289. [67] Y. Li, Y. Liu, X. Peng, C. Yan, S. Liu, N. Hu, Pull-out simulations on interfacial properties of carbon nanotube-reinforced polymer nanocomposites, Comput. Mater. Sci. 50 (2011) 1854–1860. doi:10.1016/j.commatsci.2011.01.029. [68] P. Bilalis, D. Katsigiannopoulos, A. Avgeropoulos, G. Sakellariou, Non-covalent functionalization of carbon nanotubes with polymers, RSC Adv. 4 (2014) 2911. doi:10.1039/c3ra44906h. [69] N. Nakashima, T. Fujigaya, Fundamentals and Applications of Soluble Carbon Nanotubes, Chem. Lett. 36 (2007) 692–697. doi:10.1246/cl.2007.692. [70] C. Hu, Non-Covalent Functionalization of Carbon Nanotubes with Surfactants and Polymers, J. Chinese Chem. Soc. 56 (2009) 234–239. doi:10.1002/jccs.200900033. [71] S.C. Chowdhury, T. Okabe, Computer simulation of carbon nanotube pull-out from polymer by the molecular dynamics method, Compos. Part A Appl. Sci. Manuf. 38 (2007) 747–754. doi:10.1016/j.compositesa.2006.09.011. [72] S.J. V Frankland, A. Caglar, D.W. Brenner, M. Griebel, Molecular simulation of the influence of chemical cross-links on the shear strength of carbon nanotube-polymer interfaces, J. Phys. Chem. B. 106 (2002) 3046–3048. doi:10.1021/jp015591+. 163 [73] Y.D. Kuang, X.Q. He, Young’s moduli of functionalized single-wall carbon nanotubes under tensile loading, Compos. Sci. Technol. 69 (2009) 169–175. doi:10.1016/j.compscitech.2008.09.044. [74] D. Tasis, N. Tagmatarchis, A. Bianco, M. Prato, Chemistry of carbon nanotubes, Chem. Rev. 106 (2006) 1105–1136. doi:10.1021/cr050569o. [75] L. Sun, G.L. Warren, J.Y. O’Reilly, W.N. Everett, S.M. Lee, D. Davis, et al., Mechanical properties of surface-functionalized SWCNT/epoxy composites, Carbon N. Y. 46 (2008) 320–328. doi:10.1016/j.carbon.2007.11.051. [76] S. Wang, R. Liang, B. Wang, C. Zhang, Reinforcing polymer composites with epoxide-grafted carbon nanotubes., Nanotechnology. 19 (2008) 085710. doi:10.1088/0957-4484/19/8/085710. [77] J. Zhu, H. Peng, F. Rodriguez-Macias, J.L. Margrave, V.N. Khabashesku, A.M. Imam, et al., Reinforcing epoxy polymer composites through covalent integration of functionalized nanotubes, Adv. Funct. Mater. 14 (2004) 643–648. doi:10.1002/adfm.200305162. [78] S.W. Kuo, F.C. Chang, POSS related polymer nanocomposites, Prog. Polym. Sci. 36 (2011) 1649–1696. doi:10.1016/j.progpolymsci.2011.05.002. [79] S.H. Phillips, T.S. Haddad, S.J. Tomczak, Developments in nanoscience: Polyhedral oligomeric silsesquioxane (POSS)-polymers, Curr. Opin. Solid State Mater. Sci. 8 (2004) 21–29. doi:10.1016/j.cossms.2004.03.002. [80] J. Hu, Y. Zhu, H. Huang, J. Lu, Recent advances in shape-memory polymers: Structure, mechanism, functionality, modeling and applications, Prog. Polym. Sci. 37 (2012) 1720–1763. doi:10.1016/j.progpolymsci.2012.06.001. 164 [81] W. Zhang, A.H.E. Müller, Architecture, self-assembly and properties of welldefined hybrid polymers based on polyhedral oligomeric silsequioxane (POSS), Prog. Polym. Sci. 38 (2013) 1121–1162. doi:10.1016/j.progpolymsci.2013.03.002. [82] G.X. Chen, H. Shimizu, Multiwalled carbon nanotubes grafted with polyhedral oligomeric silsesquioxane and its dispersion in poly(l-lactide) matrix, Polymer (Guildf). 49 (2008) 943–951. doi:10.1016/j.polymer.2008.01.014. [83] B. Zhang, Y. Chen, J. Wang, W.J. Blau, X. Zhuang, N. He, Multi-walled carbon nanotubes covalently functionalized with polyhedral oligomeric silsesquioxanes for optical limiting, Carbon N. Y. 48 (2010) 1738–1742. doi:10.1016/j.carbon.2010.01.015. [84] Y. Tang, J. Gou, Y. Hu, Covalent functionalization of carbon nanotubes with polyhedral oligomeric silsequioxane for superhydrophobicity and flame retardancy, Polym. Eng. Sci. 53 (2013) 1021–1030. doi:10.1002/pen.23338. [85] S.K. Yadav, S.S. Mahapatra, H.J. Yoo, J.W. Cho, Synthesis of multi-walled carbon nanotube/polyhedral oligomeric silsesquioxane nanohybrid by utilizing click chemistry., Nanoscale Res. Lett. 6 (2011) 122. doi:10.1186/1556-276X-6-122. [86] D. Sun, Q. Li, G.X. Chen, Preparation of core-shell structured carbon nanotubesilsesquioxane hybrids by a direct free-radical reaction, Mater. Lett. 120 (2014) 90– 93. doi:10.1016/j.matlet.2014.01.046. [87] Q.F. Li, Y. Xu, J.S. Yoon, G.X. Chen, Dispersions of carbon nanotubes/polyhedral oligomeric silsesquioxanes hybrids in polymer: The mechanical, electrical and EMI shielding properties, J. Mater. Sci. 46 (2011) 2324–2330. doi:10.1007/s10853-0105077-0. 165 [88] D. Sun, Z. Zhou, G.-X. Chen, Q. Li, Regulated Dielectric Loss of Polymer Composites from Coating Carbon Nanotubes with a Cross-Linked Silsesquioxane Shell through Free-Radical Polymerization, ACS Appl. Mater. Interfaces. 6 (2014) 18635–18643. doi:10.1021/am503633t. [89] C.M. Damian, C.C. Ciobotaru, S. a. Garea, H. Iovu, Effect of POSS-NH2 functionalization of MWNTs on reinforcing properties in epoxy nanocomposites, High Perform. Polym. 25 (2013) 566–575. doi:10.1177/0954008313475831. [90] R.L. Zhang, C.G. Wang, L. Liu, H.Z. Cui, B. Gao, Polyhedral oligomeric silsesquioxanes/carbon nanotube/carbon fiber multiscale composite: Influence of a novel hierarchical reinforcement on the interfacial properties, Appl. Surf. Sci. 353 (2015) 224–231. doi:10.1016/j.apsusc.2015.06.156. [91] F. Zhao, Y. Huang, L. Liu, Y. Bai, L. Xu, Formation of a carbon fiber/polyhedral oligomeric silsesquioxane/carbon nanotube hybrid reinforcement and its effect on the interfacial properties of carbon fiber/epoxy composites, Carbon N. Y. 49 (2011) 2624–2632. doi:10.1016/j.carbon.2011.02.026. [92] H. Mahfuz, F. Powell, R. Granata, M. Hosur, M. Khan, Coating of Carbon Fiber with Polyhedral Oligomeric Silsesquioxane (POSS) to Enhance Mechanical Properties and Durability of Carbon/Vinyl Ester Composites, Materials (Basel). 4 (2011) 1619–1631. doi:10.3390/ma4091619. [93] A.P. Materials, Ashland Performance Materials DERAKANE ® 8084 Epoxy Vinyl Ester Resin DERAKANE 8084 epoxy vinyl ester resin is an elastomer modified resin designed to offer increased USE, 1 (2011) 4–7. [94] L. Lloyd, Handbook of industrial 166 catalysts, 2008. doi:10.1016/S1351- 4180(08)70251-4. [95] W.C. Oliver, G.M. Pharr, Measurement of hardness and elastic modulus by instrumented indentation: Advances in understanding and refinements to methodology, J. Mater. Res. 19 (2004) 3–20. doi:10.1557/jmr.2004.19.1.3. [96] R. Poveda, N. Gupta, M. Porfiri, Poisson’s ratio of hollow particle filled composites, Mater. Lett. 64 (2010) 2360–2362. doi:10.1016/j.matlet.2010.07.063. [97] M.M. Shokrieh, Nanoindentation nanocomposites, M.R. Hosseinkhani, and nanoscratch Polym. M.R. Naimi-Jamal, investigations Test. 32 on H. Tourani, graphene-based (2013) 45–51. doi:10.1016/j.polymertesting.2012.09.001. [98] ASTM E2546-07, Standard Practice for Instrumented Indentation Testing, in: ASTM International, West Conshohocken, PA, 2007. doi:10.1520/E2546-07. [99] ASTM D790-10, Standard Test Methods for Flexural Properties of Unreinforced and Reinforced Plastics and Electrical Insulating Materials, in: ASTM International, West Conshohocken, PA, 2010. doi:10.1520/D0790-10. [100] ASTM E1356-08, Standard Test Method for Assignment of the Glass Transition Temperatures by Differential Scanning Calorimetry, in: ASTM International, West Conshohocken, PA, 2014. doi:10.1520/E1356. [101] M. Shtein, R. Nadiv, N. Lachman, H. Daniel Wagner, O. Regev, Fracture behavior of nanotube-polymer composites: Insights on surface roughness and failure mechanism, Compos. Sci. Technol. 87 (2013) 157–163. doi:10.1016/j.compscitech.2013.07.016. [102] M.G. Kim, J.B. Moon, C.G. Kim, Effect of CNT functionalization on crack 167 resistance of a carbon/epoxy composite at a cryogenic temperature, Compos. Part A Appl. Sci. Manuf. 43 (2012) 1620–1627. doi:10.1016/j.compositesa.2012.04.001. [103] M. Iijima, T. Muguruma, W.A. Brantley, I. Mizoguchi, Comparisons of nanoindentation, 3-point bending, and tension tests for orthodontic wires, Am. J. Orthod. Dentofac. Orthop. 140 (2011) 65–71. doi:10.1016/j.ajodo.2009.11.015. [104] R. Martinez, L.R. Xu, Comparison of the Young’s moduli of polymers measured from nanoindentation and bending experiments, MRS Commun. 4 (2014) 1–5. doi:10.1557/mrc.2014.19. [105] G. Gkikas, N.M. Barkoula, A.S. Paipetis, Effect of dispersion conditions on the thermo-mechanical and toughness properties of multi walled carbon nanotubesreinforced epoxy, Compos. Part B Eng. 43 (2012) 2697–2705. doi:10.1016/j.compositesb.2012.01.070. [106] A. Montazeri, M. Chitsazzadeh, Effect of sonication parameters on the mechanical properties of multi-walled carbon nanotube/epoxy composites, Mater. Des. 56 (2014) 500–508. doi:10.1016/j.matdes.2013.11.013. [107] R.M. JONES, Mechanics of composite materials, 1999. doi:10.1007/BF00611782. [108] S.H. Liao, M.C. Hsiao, C.Y. Yen, C.C.M. Ma, S.J. Lee, A. Su, et al., Novel functionalized carbon nanotubes as cross-links reinforced vinyl ester/nanocomposite bipolar plates for polymer electrolyte membrane fuel cells, J. Power Sources. 195 (2010) 7808–7817. doi:10.1016/j.jpowsour.2009.10.020. [109] P.C. Ma, J.K. Kim, B.Z. Tang, Effects of silane functionalization on the properties of carbon nanotube/epoxy nanocomposites, Compos. Sci. Technol. 67 (2007) 2965– 2972. doi:10.1016/j.compscitech.2007.05.006. 168 [110] N. a. Siddiqui, R.S.C. Woo, J.-K. Kim, C.C.K. Leung, A. Munir, Mode I interlaminar fracture behavior and mechanical properties of CFRPs with nanoclayfilled epoxy matrix, Compos. Part A Appl. Sci. Manuf. 38 (2007) 449–460. doi:10.1016/j.compositesa.2006.03.001. [111] P. Karapappas, A. Vavouliotis, P. Tsotra, V. Kostopoulos, A. Paipetis, Enhanced Fracture Properties of Carbon Reinforced Composites by the Addition of Multi-Wall Carbon Nanotubes, J. Compos. Mater. 43 (2009) 977–985. doi:10.1177/0021998308097735. [112] A.K. Chakraborty, T. Plyhm, M. Barbezat, A. Necola, G.P. Terrasi, Carbon nanotube (CNT)-epoxy nanocomposites: A systematic investigation of CNT dispersion, J. Nanoparticle Res. 13 (2011) 6493–6506. doi:10.1007/s11051-0110552-3. [113] J. Li, P.C. Ma, W.S. Chow, C.K. To, B.Z. Tang, J.K. Kim, Correlations between percolation threshold, dispersion state, and aspect ratio of carbon nanotubes, Adv. Funct. Mater. 17 (2007) 3207–3215. doi:10.1002/adfm.200700065. [114] A. Battisti, A.A. Skordos, I.K. Partridge, Monitoring dispersion of carbon nanotubes in a thermosetting polyester resin, Compos. Sci. Technol. 69 (2009) 1516–1520. doi:10.1016/j.compscitech.2008.05.012. [115] M. Tehrani, M. Safdari, M.S. Al-Haik, Nanocharacterization of creep behavior of multiwall carbon nanotubes/epoxy nanocomposite, Int. J. Plast. 27 (2011) 887–901. doi:10.1016/j.ijplas.2010.10.005. [116] J. Yang, Z. Zhang, K. Friedrich, A.K. Schlarb, Creep resistant polymer nanocomposites reinforced with multiwalled carbon nanotubes, Macromol. Rapid 169 Commun. 28 (2007) 955–961. doi:10.1002/marc.200600866. [117] M. Abdalla, D. Dean, D. Adibempe, E. Nyairo, P. Robinson, G. Thompson, The effect of interfacial chemistry on molecular mobility and morphology of multiwalled carbon nanotubes epoxy nanocomposite, Polymer (Guildf). 48 (2007) 5662–5670. doi:10.1016/j.polymer.2007.06.073. [118] J. Kathi, K.-Y. Rhee, J.H. Lee, Effect of chemical functionalization of multi-walled carbon nanotubes with 3-aminopropyltriethoxysilane on mechanical and morphological properties of epoxy nanocomposites, Compos. Part A Appl. Sci. Manuf. 40 (2009) 800–809. doi:10.1016/j.compositesa.2009.04.001. [119] S. Lefrant, M. Baibarac, I. Baltog, Raman and FTIR spectroscopy as valuable tools for the characterization of polymer and carbon nanotube based composites, J. Mater. Chem. 19 (2009) 5690. doi:10.1039/b821136a. [120] G. Socrates, Infrared and Raman Characteristic Group Frequencies: Tables and Charts, 3rd ed., John Wiley & Sons Ltd, New York, 2001. [121] J. Coates, Interpretation of infrared spectra, a practical approach, in: R.A. Meyers (Ed.), Encycl. Anal. Chem., John Wiley & Sons Ltd, 2000: pp. 10815–10837. [122] S. Majeed, V. Filiz, S. Shishatskiy, J. Wind, C. Abetz, V. Abetz, Pyrene-POSS nanohybrid as a dispersant for carbon nanotubes in solvents of various polarities: its synthesis and application in the preparation of a composite membrane, Nanoscale Res. Lett. 7 (2012) 296. doi:10.1186/1556-276X-7-296. [123] L. Valentini, M. Cardinali, J.M. Kenny, M. Prato, O. Monticelli, A photoresponsive hybrid nanomaterial based on graphene and polyhedral oligomeric silsesquioxanes, Eur. J. Inorg. Chem. (2012) 5282–5287. doi:10.1002/ejic.201200260. 170 [124] M.S. Dresselhaus, G. Dresselhaus, R. Saito, A. Jorio, Raman spectroscopy of carbon nanotubes, Phys. Rep. 409 (2005) 47–99. doi:10.1016/j.physrep.2004.10.006. [125] Y.-P. Sun, K. Fu, Y. Lin, W. Huang, Functionalized carbon nanotubes: properties and applications., Acc. Chem. Res. 35 (2002) 1096–104. doi:10.1021/ar010160v. [126] D.B. Cordes, P.D. Lickiss, F. Rataboul, Recent Developments in the Chemistry of Cubic Polyhedral Oligosilsesquioxanes, Chem. Rev. 110 (2010) 2081–2173. doi:10.1021/cr900201r. [127] J.D. Menczel, R.B. Prime, Thermal Analysis of Polymers: Fundamentals and Applications, 2008. doi:10.1002/9780470423837. [128] H. Mahfuz, S. Zainuddin, M.R. Parker, T. Al-Saadi, V.K. Rangari, S. Jeelani, Enhancement of strength and stiffness of epoxy-based composites using nanoparticle infusion and high magnetic fields, Mater. Lett. 61 (2007) 2535–2539. doi:10.1016/j.matlet.2006.09.065. [129] T.G.A. Youngs, Aten - An application for the creation, editing, and visualization of coordinates for glasses, liquids, crystals, and molecules, J. Comput. Chem. 31 (2010) 639–648. doi:10.1002/jcc.21359. [130] R. Fletcher, C.M. Reeves, Function minimization by conjugate gradients, Comput J. 7 (1964) 149–154. doi:10.1093/comjnl/7.2.149. [131] I.T. Todorov, W. Smith, K. Trachenko, M.T. Dove, DL_POLY, J. Mater. Chem. 16 (2006) 1911–1918. [132] S.L. Mayo, B.D. Olafson, W. a G. Iii, E. Eb, E. a E.T. El, DREIDING: A Generic Force Field for Molecular Simulations, J. Phys. Chem. 101 (1990) 8897–8909. doi:10.1021/j100389a010. 171 [133] K. Binder, Monte Carlo and Molecular Dynamics Simulations in Polymer Science, 1995. doi:10.1111/j.1471-8286.2004.00618.x. [134] H.J.C. Berendsen, J.P.M. Postma, W.F. van Gunsteren, a DiNola, J.R. Haak, Molecular dynamics with coupling to an external bath, J. Chem. Phys. 81 (1984) 3684–3690. doi:10.1063/1.448118. [135] B. Wilfred, V. Gunsteren, C. Berendsen, Computer Simulation of Molecular Dynamics: Methodology, Applications, and Perspectives in Chemistry, Anger,. Chem. Lnr. Ed. Engl. 29 (1990) 992–1023. doi:10.1002/anie.199009921. [136] K. Kholmurodov, W. Smith, K. Yasuoka, T. Ebisuzaki, Highly vectorized `link-cell’ FORTRAN code for the DL_POLY molecular dynamics simulation package, Comput. Phys. Commun. 125 (2000) 167–192. doi:10.1016/S0010-4655(99)004853. [137] A. Adnan, C.T. Sun, H. Mahfuz, A molecular dynamics simulation study to investigate the effect of filler size on elastic properties of polymer nanocomposites, Compos. Sci. Technol. 67 (2007) 348–356. doi:10.1016/j.compscitech.2006.09.015. [138] Q.H. Zeng, a. B. Yu, G.Q. Lu, Multiscale modeling and simulation of polymer nanocomposites, Prog. Polym. Sci. 33 (2008) 191–269. doi:10.1016/j.progpolymsci.2007.09.002. [139] W.M. Lai, D. Rubin, E. Krempl, Introduction to Continuum Mechanics, 2010. doi:10.1016/B978-0-7506-8560-3.00006-2. [140] M. Bohlén, K. Bolton, Molecular dynamics studies of the influence of single wall carbon nanotubes on the mechanical properties of Poly(vinylidene fluoride), Comput. Mater. Sci. 68 (2013) 73–80. doi:10.1016/j.commatsci.2012.10.010. 172 [141] M. Grujicic, Y.P. Sun, K.L. Koudela, The effect of covalent functionalization of carbon nanotube reinforcements on the atomic-level mechanical properties of polyvinyl-ester-epoxy, Appl. Surf. doi:10.1016/j.apsusc.2006.06.050. 173 Sci. 253 (2007) 3009–3021.