A study on mechanical, electrical, and thermal properties of. graphene nanoplatelets reinforced epoxy composites

Transcription

1 A study on mechanical, electrical, and thermal properties of graphene nanoplatelets reinforced epoxy composites A thesis submitted to the University of Manchester for the degree of Master of Philosophy in the Faculty of Science and Engineering 2016 Wentao Zou School of Materials

2 Table of Contents List of figures... 5 List of tables List of abbreviation List of Symbols Abstract Declaration Copyright statement Acknowledgement Introduction and aim of the project Introduction Aim of the project Literature review Introduction to graphene, carbon nanotubes, and expanded graphite Introduction to graphene Introduction to carbon nanotubes and expanded graphite Synthesis of graphene Top-down Bottom-up Hot curing epoxy system Preparation methods of the nanocomposites Shear mixing Three-roll milling

3 2.4.3 Ultrasonication Quantification of dispersion state Properties of nanocomposites Mechanical properties of graphene-reinforced nanocomposites Electrical conductivity of the nanocomposites Thermal properties of the nanocomposites Methodology Materials Preparation of the samples Nomenclature of the samples Preparation process Sample geometry Experiment set-up Tensile strength and modulus test set-up Electrical conductivity test set-up Thermal conductivity test set-up Thermal diffusivity test set-up Results and discussion GNPs characterisations Optical microscope images of the fillers dispersion state Tensile strength and modulus properties Tensile strength and modulus results of EG/epoxy composites Tensile strength and modulus results of MWCNTs/epoxy composites

4 4.3.3 Tensile strength and modulus results for GNPs/epoxy composites Tensile strength and modulus results of GNPs (with different loadings)/epoxy composites Mechanical results discussion SEM images of tensile fracture surface Electrical Conductivity EG/Epoxy composites Electrical Conductivity MWCNTs/Epoxy composites electrical conductivity GNPs/Epoxy composites electrical conductivity Electrical results discussion GNPs reinforced epoxy thermal properties Conclusions and future work References Appendixes Word count:

5 List of figures Figure 1. The schematics of the honeycomb structure in monolayer graphene where the purple ball stand for carbon atom and the red rod stand for covalent bond between the carbon atoms [15] Figure 2. Different crystal structures of the carbon allotropes: diamond, graphite, graphene, carbon nanotube and Fullerene (from the left to right) [29] Figure 3. Schematics of the different structures and diameters between the SWCNT and the MWCNTs [31] Figure 4. Schematics of the different atoms arrangements of the SWCNT: armchair, zigzag and chiral (from the left to right) [33] Figure 5. Scanning electron microscope image of the structure of the expanded graphite showing the lamellar structure of the layers of graphene, arranged together in a worm like form [42] Figure 6. Schematic illustration of the top-down and bottom-up routes for graphene production [43] Figure 7. The chmeical structure of the graphite oxide and the functionalities exist in the graphite oxide [45] Figure 8. The chemical structures difference between the graphene oxide (a) and the reduced graphene oxide (b) and the functionalities exist in these materials [47] Figure 9. The reaction during the synthesis of the DGEBA [43] Figure 10. Scheme of reactions between the epoxy resin and the amine hardener during the curing process: (1) the reaction between primary amine and epoxide form a secondary amine; (2) the reaction between secondary amine and epoxide form the tertiary amine and hydroxyl; (3) the etherification reaction between hydroxyl and the epoxide to form the ether and hydroxyl

6 Figure 11. High-speed shear mixer and dispersion flow (in inset) which occurs during shear mixing [60] Figure 12. Exfoliation mechanisms occurring during the shear mixing process; the shear force, jet cavitation, and collisions are three primary mechanisms of the mixing [63] Figure 13. Schematic of three-roll milling working mechanism; the different rotation speeds and directions of the rollers exert shear force on the fillers, resulting in the exfoliation and dispersion of fillers [66] Figure 14. Schematic of exfoliation mechanism in sonication; the micro-jet work plays a primary role in exfoliation of fillers whiles the wedge and shear effect play a secondary role [68] Figure 15. Schematic of the nanofillers dispersion states; homogeneous distribution of fillers results in better mechanical properties, while inhomogeneous distribution leads to better electrical properties [43] Figure 16. Applications of electrical conducting composites (A) and conducting network of composites (B) [60] Figure 17. SilverSon L5MPA high-speed laboratory mixer (a) and Low-speed laboratory mixer; resin and hardener mixed together at 1000 rpm for 3 minutes (b) Figure 18. Heraeus Vacutherm degassing chamber (a) and Heraeus Thermoscientific Oven (b) (one curing cycle: 80 C for 2 hours and 140 C for 8 hours) Figure 19. Silicon rubber moulds (the left mould was used for the thermal sample and the right one, for the mechanical and electrical samples) Figure 20. The geometries of the samples used for mechanical, electrical and thermal test

7 Figure 21. SEM and optical microscopy samples (the silver samples on the top are the SEM samples and the glass slides at the bottom are the optical microscopy samples) Figure 22. Left image: Instron 5969 machine and the tensile sample under the test; right image: structure of the sample Figure 23. Left image: combination of the Impedance Analysis Interface (IAI), Phase Sensitive Multimeter (PSM) 1735, and Frequency Response Analyser; right image: electrical sample under two-probe test Figure 24. The relationship between sample conductivity and AC frequency, where log ω is the common logarithm of the current frequency and log δ is the common logarithm of the sample s conductivity [99] Figure 25. FOX-50 Thermal Conductivity Thermometer (the upper plate set as 30 C and the bottom plate set as 25 C) Figure 26. IR thermography machines and the measurement of thermal diffusivity of the sample (left to right: heat resources, samples, and IR thermography camera). 58 Figure 27. Position of the time for the sample to reach half-maximum temperature during the test Figure 28. SEM images of GNPs, EG, and MWCNTs (top to bottom) (a), (c) and (e) are the low magnification images while the (b), (d) and (f) are the high magnification images Figure 29. Raman spectroscopy result for the GNPs used in the project (laser used in the test is 633 nm, spectral range in the test was to (centre ) (Raman shift/cm -1 ) Figure 30. Raman spectroscopy results for graphene and graphite using nm laser [104] Figure 31. Optical microscope images of EG-reinforced epoxy composites: sample (a) EG, 0.1 wt%, 1000rpm, 10 min, sample (b) EG, 0.1 wt%, 5000 rpm, 10 min, 7

8 sample (c) EG, 0.1 wt%, 1000 rpm, 2 hours, sample (d) EG, 0.1 wt%, 5000 rpm, 2 hours Figure 32. Optical microscope images of MWCNTs-reinforced epoxy composites: sample (a) MWCNTs, 0.1 wt%, 1000rpm, 10 min, sample (b) MWCNTs, 0.1 wt%, 5000 rpm, 10 min, sample (c) MWCNTs, 0.1 wt%, 1000 rpm, 2 hours, sample (d) MWCNTs, 0.1 wt%, 5000 rpm, 2 hours Figure 33. Optical microscope images of GNP-reinforced epoxy composites: sample (a) GNPs, 0.1 wt%, 1000 rpm, 1 hour, sample (b) GNPs, 0.1 wt%, 5000 rpm, 1 hour, sample (c) GNPs, 0.1 wt%, 1000 rpm, 2 hours, sample (d) GNPs, 0.1 wt%, 5000 rpm, 2 hours, sample (e) GNPs, 0.1 wt%, 9000 rpm, 1 hour Figure 34. Optical microscope images of different loadings of GNP-reinforced epoxy composited (with/without acetone): sample (a) is GNPs, 0.1 wt%, 5000 rpm, 1 hour, without acetone, sample (b) is GNPs, 0.1 wt%, 5000 rpm, 1 hour, with acetone, sample (c) is GNPs, 5 wt%, 5000 rpm, 1 hour, without acetone, sample (d) is GNPs, 5 wt%, 5000 rpm, 1 hour, with acetone Figure 35. Tensile strength results for EG-reinforced epoxy composites prepared by different shear mixing speeds and times Figure 36. Modulus results for EG-reinforced epoxy composites prepared by different shear mixing speeds and times Figure 37. Tensile strength results for MWCNTs-reinforced epoxy composites prepared by different shear mixing speeds and times Figure 38. Modulus results for MWCNTs-reinforced epoxy composites prepared by different shear mixing speeds and times Figure 39. Tensile strength results of GNPs-reinforced epoxy composites prepared by different shear mixing speeds and times Figure 40. Modulus results of GNPs reinforced epoxy composites prepared by different shear mixing speeds and times

9 Figure 41. Tensile strength results for different loadings of GNPs-reinforced epoxy composites (without and with acetone) Figure 42. Modulus results for different loadings of GNPs-reinforced epoxy composites (without and with acetone) Figure 43. SEM images of the GNPs/Epoxy samples: sample (a) Pure epoxy, sample (b) GNPs, 0.1 wt%, 5000 rpm, 1 hour, sample (c) GNPs, 5 wt%, 5000 rpm, 1 hour, without acetone, sample (d) GNPs, 5 wt%, 5000 rpm, 1 hour, with acetone Figure 44. Electrical conductivity results for EG-reinforced epoxy composites prepared by 10 min shear mixing with speed range from 1000 to 5000 rpm Figure 45. Electrical conductivity results for EG-reinforced epoxy composites prepared by 1 hour shear mixing with speed range from 1000 to 5000 rpm Figure 46. Electrical conductivity results for EG-reinforced epoxy composites prepared by 2 hours of shear mixing with speed range from 1000 to 5000 rpm Figure 47. Electrical conductivity results for MWCNTs-reinforced epoxy composites prepared by 10 min shear mixing with speed range from 1000 to 5000 rpm Figure 48. Electrical conductivity results for MWCNTs-reinforced epoxy composites prepared with 1 hour of shear mixing with speed range from 1000 to 5000 rpm Figure 49. Electrical conductivity results for MWCNTs-reinforced epoxy composites prepared by 2 hours of shear mixing with speed range from 1000 to 5000 rpm Figure 50. Electrical conductivity results for GNPs-reinforced epoxy composites prepared by 1 hour of shear mixing with speed from 1000 to 5000 rpm Figure 51. Electrical conductivity results for GNPs-reinforced epoxy composites prepared by 2 hours of shear mixing with speed range from 1000 to 5000 rpm Figure 52. Electrical conductivity results for GNPs-reinforced epoxy composites prepared by 1 hour of shear mixing at a speed of 5000 rpm and loading range from 0.1 wt% to 5 wt%

10 Figure 53. Thermal diffusivity results for the different loadings of GNPs-reinforced epoxy composites prepared by 1 hour of shear mixing at a speed of 5000 rpm without and with acetone Figure 54. Thermal conductivity results for the different loadings of GNPs-reinforced epoxy composites prepared by 1 hour shear mixing at a speed of 5000 rpm without and with acetone Figure 55. Thermal diffusivity mapping images: (a) GNPs, 1 wt%, 5000 rpm, 1 hour, without acetone, sample (b) GNPs, 1 wt%, 5000 rpm, 1 hour, with acetone, sample (c) GNPs, 2 wt%, 5000 rpm, 1 hour, without acetone, sample (d) GNPs, 2 wt%, 5000 rpm, 1 hour, with acetone, sample (e) GNPs, 5 wt%, 5000 rpm, 1 hour, without acetone, sample (f) GNPs, 5 wt%, 5000 rpm, 1 hour, with acetone

11 List of tables Table 1. Data sheet of as-received MWCNTs and GNPs used in the project [93, 94] Table 2. Data sheet of the acetone used in the project [96]

12 List of abbreviation NPCs: Nanoparticles reinforced polymeric composites CNTs: SWCNT: MWCNT: GNPs: Carbon Nanotubes Single-walled carbon nanotube Multi-walled carbon nanotube Graphene nanoplatelets EG: Expanded graphite GO: Graphite/Graphene oxide 0D: Zero-dimensional 1D: One-dimensional 2D: Two-dimensional 3D: Three-dimensional HOPG: Highly oriented pyrolytic graphite SEM: Scanning electron microscope TEM: Transmission electron microscope CVD: Chemical vapour deposition N: Newton GPa: Giga Pascal TPa: Tera Pascal 12

13 µm: Micrometre nm: Nanometre wt%: Weight percentage AC: Alternating current

14 List of Symbols Symbol Quantity Unit E Elastic modulus GPa δ Electrical conductivity S/m ω Frequency of AC H q Heat flux W/m 2 λ Thermal conductivity W/mK 14

15 Abstract Graphene is a material that has superior mechanical, electrical, and thermal properties. It has drawn the attention of many the scientific researchers for this purpose. In this project, three different types of fillers, graphene nanoplatelets (GNPs)-, multi-walled carbon nanotubes (MWCNTs)-, and expanded graphite (EG)- reinforced epoxy nanocomposites were prepared by shear mixing. Different shear mixing speeds and shear mixing times were used during the preparation of the nanocomposites with 0.1 wt% loading of the fillers. The effects of various types of fillers and different shear mixing speeds and durations on mechanical and electrical properties of the final composites were examined. The GNPs-reinforced epoxy nanocomposite was the only one that showed a 13% improvement in elastic modulus as compared to pure epoxy when the shear mixing conditions were 3000 rpm for 2 hours. The project also studied the effects of different loadings of GNPs and the addition of acetone as a solvent on the final mechanical, electrical and thermal properties of the composites (with the fixed shear mixing speed and time). The tensile strength of the composites reduced drastically when the loading of GNPs increased while the elastic modulus show some increase with the growth in GNP loading. The GNPsreinforced composites did not show the percolation threshold even with 5 wt% (with the ratio to the weight of epoxy) loading of the GNPs. The GNPs-reinforced epoxy composites showed an 116% improvement in the thermal conductivity as compared to the pure epoxy samples when the GNPs loading was 5 wt%. The results also showed that the samples prepared with the addition of acetone had higher thermal diffusivity than the samples prepared without acetone. 15

16 Declaration No portion of the work referred to in the thesis has been submitted in support of an application for another degree or qualification of this or any other university or other institute of learning Wentao Zou 16

17 Copyright statement The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the Copyright ) and s/he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made. The ownership of certain Copyright, patents, designs, trade marks and other intellectual property (the Intellectual Property ) and any reproductions of copyright works in the thesis, for example graphs and tables ( Reproductions ), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property University IP Policy (see in any relevant Thesis restriction declarations deposited in the University Library, The University Library s regulations (see and in The University s policy on Presentation of Theses 17

18 Acknowledgement Firstly, I wish to thank Edward Pullicino for helping me to complete my MPhil degree. Without him, I would not have completed this project. I am sincerely thankful to my supervisor Dr Aravind Vijayaraghavan for his continuous support throughout my project and his guidance, supervision, patience, and immense knowledge. I also want to thank my Co-supervisor Prof Constantinos Soutis and Dr Matthieu Gresil for their kindness, advice, and help in my project. Special thanks to Quentin Poutrel for the Raman session and Zixin Wang for the continuous support throughout the project. Last but not the least, I want to thank my family for their continuous support. 18

19 1. Introduction and aim of the project 1.1 Introduction The physical properties of polymer-based nanocomposites are affected by the state of dispersion of the nanofillers [1]. Composites are multi-phase materials that have superior engineering properties; these properties are achieved by the combination of their constituent phases. Nanoparticle-reinforced polymeric composites (NPCs) are new types of composites which have attracted a lot of attention recently. The NPCs are materials which contain toughening fillers at the nanoscale. The mechanical or chemical properties of the final composites could be improved as a result of the larger specific surface area (SSA) of the nanoscale fillers. Apart from the large SSA, the natural properties of the nanofillers also play a significant role in the final properties of NPCs. Therefore, various kinds of fillers such as carbon nanotubes (CNTs), graphene, and other carbon derivatives, which have superior properties and nanoscale dimensions, have been used to synthesise the NPCs [2 4]. Jancar and his co-workers reported how the selection of a dispersion strategy is becoming important. Dispersion techniques like high-speed shear mixing (HSM) and three-roll milling (TRM) have frequently been applied in the dispersion of nanofillers [5 7]. Polymer composites reinforced by CNTs have shown many improvements in their electrical, mechanical, and thermal properties [8 10]. Graphene is a relatively new material which has drawn the attention of researchers. It has been reported to possess superior properties such as extremely high modulus, high electrical and thermal conductivity, high tensile strength [11, 12]. It is a 2-dimentional (2D) material with a large SSA. Also, the entanglement issue which occurs in CNTs becomes a minor issue for graphene-reinforced polymeric 19

20 composites. These advantages make graphene an ideal material to strengthen polymer composites [13, 14]. 1.2 Aim of the project The aims of this project are as follows: (1) To use different shear mixing speeds and shear mixing times to disperse graphene, CNTs, or expanded graphite (EG) into an epoxy matrix and synthesise graphene/cnts/eg-reinforced epoxy composites. (2) To examine the effect of different loadings of graphene at constant mixing speed and time to determine the effect of filler loading on the mechanical, electrical, and thermal properties of graphene/epoxy composites (3) To examine the effect of acetone on the mechanical, electrical, and thermal properties of graphene/epoxy composites The thesis layout is as follows. There are five chapters in total. The first chapter presents an introduction to the background and aims of the project. Chapter 2 presents the literature review. It includes background information on the nanofillers used in the project. The methods used for synthesising graphene have been outlined. Then, background information on epoxy resin is presented, followed by dispersion methods and the properties of graphene-reinforced composites. The third chapter provides information about the experiments conducted in this project. The details of the materials, the set-ups of characterisation tests, and the geometries and dimensions of the sample have all been illustrated in this chapter. Chapter 4 presents the results of all the tests and the related discussions. The effects of different shear mixing speeds and times have been discussed. The effects of different loading of the nanofillers and the application of the solvent have also been discussed. The final chapter presents the conclusion of the main findings of the project and also recommendations for future works in this area. 20

21 2. Literature review 2.1 Introduction to graphene, carbon nanotubes, and expanded graphite Introduction to graphene Graphene, which is a single-layer carbon-atom sheet, is a carbon allotrope. Figure 1 shows the carbon arrangement of the atoms in the monolayer graphene; the honeycomb structure can also be observed [15]. The dimensions and structure of graphene have been evaluated and reported in many studies. The length of the carbon-carbon (sp 2 ) bond in graphene is around nm, the thickness of the monolayer layer sheet is around 0.35 nm, and the distance between two neighbouring graphene layers in the graphite is around nm [16, 17]. A. Fasolino et al. proved that monolayer graphene has ripples in its structure and not a flat 2D structure [18]. The attempts made to discover graphene date back to 1947; Wallace had predicted Figure 1. The schematics of the honeycomb structure in monolayer graphene where the purple ball stand for carbon atom and the red rod stand for covalent bond between the carbon atoms [15]. 21

22 that the building cell of graphite is graphene ; he also predicted that these isolated monolayers could have both extraordinary electrical and mechanical properties [19]. Then, Boehm isolated graphene, as a monolayer sheet of carbon atoms, from graphite in 1986 [20]. Interestingly, the term graphene was officially used in However, the graphene-like material was already synthesised in the 1960s from graphite oxide [21]. The monolayer graphene was assumed to be unstable in the ambient environment and incapable of existing; however, Geim and Novoselov from the University of Manchester successfully isolated graphene (nanoscale) from graphite (microscale) at room temperature [22 24]. They were also awarded the Nobel Prize in Physics in 2010 for this discovery. Graphene has many outstanding characteristics compared to other materials. Concerning mechanical properties, graphene has a Young s modulus of around 1 TPa, and it also has an intrinsic tensile strength of 130 GPa which makes it the strongest material to date [25]. Its thermal conductivity is around W/(mK) (Watt per Kelvin-metre) and electrical conductivity, S/m (Siemens per meter) [26, 27]. As mentioned before, the thickness of monolayer graphene is around 0.35 nm, meaning it is the thinnest 2D material to date [16]. Concerning the optical property of graphene, Kuzmenko reported a value of 2.3% absorption of the light, which indicates low opacity [28]. As mentioned before, graphene is the building block of graphite. However, graphene could also be used to build other carbon allotropes such as CNTs and fullerenes. Figure 2 shows the different crystal structures: diamond, graphite, Figure 2. Different crystal structures of the carbon allotropes: diamond, graphite, graphene, carbon nanotube and Fullerene (from the left to right) [29]. 22

23 graphene, carbon nanotubes, and fullerenes [29]. As it shows, the stack of graphene can form the graphite; it could also be rolled or wrapped into the carbon nanotubes or the fullerenes Introduction to carbon nanotubes and expanded graphite CNTs are another kind of carbon allotrope with many superior properties; the excellent properties make them the preferred materials for toughening polymer composites. Sumio Iijima was the first person to discover CNTs in 1991 using the arc-discharge method [30]. CNTs have two typical structures: (1) Multi-walled carbon nanotubes (MWCNTs) and (2) single-walled carbon nanotubes (SWCNTs). MWCNTs consist of at least two or more rolled concentric cylindrical layers of graphene sheets. Meanwhile, SWCNTs only consist of one rolled-up tubular layer of the graphene sheet. The diameter of SWCNTs could ranges from 0.5 to 6 nm, while the diameter of MWCNTs is larger than 100 nm; the length of SWCNTs ranges from 100 nm to several micrometres, while that of MWCNTs ranges from 0.1 to 50 microns [31,32]. Figure 3 shows the schematics of these two materials. There are three kinds of atoms arrangements in a graphene sheet: armchair, zigzag, and chiral; these structures could lead to three different structures of SWCNTs, as shown in Figure 4. Figure 3. Schematics of the different structures and diameters between the SWCNT and the MWCNTs [31]. 23

24 Figure 4. Schematics of the different atoms arrangements of the SWCNT: armchair, zigzag and chiral (from the left to right) [33]. The electrical conductivity of the SWCNTs could vary from metallic to semiconducting owing to these differences in structure [33]. However, the electrical conductivity of MWCNTs is more complicated to predict as each MWCNT contains several layers of graphene sheets and the atoms arrangement could vary from layer to layer. There are many reports about the tensile strength and the elastic modulus of the CNTs. Yu et al. have found the tensile strength of MWCNTs ranges from GPa and they also concluded that only the outer layer of an MWCNT is capable of taking higher loading of strength owing to the weak load transfer ability of the inner walls [34]. Furthermore, Yu et al. also found that 30 GPa was the average strength of breaking a SWCNT [35]. Concerning the elastic modulus properties, Lourie and Wanger reported that the elastic modulus of SWCNTs ranges from 2.8TPa to 3.6 TPa and that of MWCNTs, from 1.7 TPa to 2.4 TPa [36]. Apart from the mechanical properties, the CNTs also have superior electrical and thermal properties. The electrical conductivity of CNTs highly depends on their structure and diameter. The 24

25 electrical conductivity of CNTs could either be metallic or semiconducting [37, 38]. Ebbesen et al. measured the electrical conductivity of MWCNTs and found that it ranged from 10 7 S/m to 10 8 S/m [39]. Dai et al. studied the effect of defects in CNTs and found that electrical resistance increases with the existence of the defects [40]. CNTs also showed superior thermal conductivity; a value of 6000 W/mK was reported by Berber et al. when they measured the values for isolated CNTs [41]. EG is another material which has been widely used to reinforce polymer matrix. The preparation process of EG can be briefly divided into two steps. The first step is that of using concentrated H 2 SO 4 and HNO 3 acids to oxidise natural graphite and extract EG; next, the extract is heated rapidly at 600 C and, finally, the EG is obtained [42]. Figure 5 shows the scanning electron micrograph (SEM) of EG. According to the image, EG shows a relatively loose and worm-like structure. Figure 5. Scanning electron microscope image of the structure of the expanded graphite showing the lamellar structure of the layers of graphene, arranged together in a worm like form [42]. 25

26 2.2 Synthesis of graphene Graphene is a monolayer of carbon atoms with a honeycomb structure. Pristine graphene has many superior properties such as extremely high Young s modulus, desirable electrical and thermal conductivity, etc. Therefore, finding the methods for large-scale production of pristine, defect-free graphene has become a favourite project in this area. Many methods have been proposed in the last few years. All these methods could be divided into two categories: (1) top down (2) bottom up. Figure 6 shows the general idea about these two types Top-down The top-down methods involve the synthesis of graphene or graphene-like materials from graphite or carbon-related materials such as graphite oxide (GO). Exfoliation is the most common mechanism used in this category. Two typical up-to-bottom methods are explained below: In 2004, Gem et al. successfully produced monolayer graphene by using a scotch tape [24]. Highly ordered pyrolytic graphite (HOPG) was used as the source material to prepare graphene; Figure 6. Schematic illustration of the top-down and bottom-up routes for graphene production [43]. 26

27 1-mm-thick HOGP was cleaved mechanically by using the scotch tape. First of all, the scotch tape was used to stick and peel the graphite rapidly several times; then, the Si wafer was used to finish the final transfer of the graphite. After that, the tape which contained a few layers of graphite was stuck on the Si wafer and peeled gently to obtain fine cleavage of graphite. A new scotch tape was then used to peel the graphite from the Si wafer rapidly. After this step, there were some monolayers and multilayers (2 to few layers) of graphene left in the Si wafer. This method is an example of mechanical cleavage. The other popular method involves producing graphene or graphene-like materials through the reduction of graphite/graphene oxide (GO). Before the reduction of GO, it is necessary to process the GO first. Several methods can be used to synthesise GO. The method of synthesising GO was first reported by Brodie and then modified by Hummer who made this experiment much safer and quicker [44]. Hummer s method is still widely used; the details are given below: Natural graphite flakes are first added into the mixture consisting of concentrated H 2 SO 4, NaNO 2, and KMnO 4 by gentle stirring for around 2 hours; the temperature is not allowed to reach over 45 C, which means an ice bath is needed. The ice bath is removed after 2 hours, and the mixture is warmed back to room temperature. After the oxidisation of graphite, H 2 O 2 is used to wash the residual KMnO 4. Finally, distilled water is added to the mixture many times; finally, the brown colour liquid containing GO is obtained. GO usually includes functionalities such as epoxide, hydroxyl, and carboxyl, shown in Figure 7 [45]. The presence of these functionalities could help the GO react with polymer-like epoxy, resulting in better interactions between the fillers and matrix. However, oxidisation of graphite has some disadvantages. Pristine graphene has high electrical conductivity owing to the long-range order of conjugated network in 27

28 Figure 7. The chemical structure of the graphite oxide and the functionalities exist in the graphite oxide [45]. the graphene lattices. The oxidisation brings the functionalities to the carbon atoms and breaks this order, and then causes defects in the lattices. Finally, it decreases the electrical conductivity of the graphene. Therefore, the reduction of GO is necessary when electrical conductivity is important for the composites [46]. After the preparation of GO, the materials are dissolved in water. Before the reduction of GO, the freeze drying process is carried out to collect the GO powders. Then, these materials are reduced thermally or chemically, and respectively, named as thermally reduced graphene oxide (TRGO) or chemically reduced graphene oxide (CRGO). Reduced GO is a graphene-like material. The chemical structure of the reduced GO is shown in Figure 8 [47]. Concerning the thermal reduction of GO, the materials are heated rapidly in an argon atmosphere of over 2000 C. In this situation, the GO powders generate CO or CO 2 gases which exert pressure to exfoliate the GO. The pressure, ranging from 40 to 130 MPa, depends on the environmental temperature; these pressures are higher than the theoretical prediction of pressure needed to exfoliate the GO, which is 2.5 MPa [48]. However, thermal treatments lead to the formation of wrinkled structures in TRGO sheets. 28

29 (a) (b) Figure 8. The chemical structures difference between the graphene oxide (a) and the reduced graphene oxide (b) and the functionalities exist in these materials [47]. Concerning the chemical reduction of GO, many different methods have been reported. Hydrazine is the most commonly used reagent in the chemical reduction of GO. However, the toxicity of hydrazine is seen as a major drawback of its application. Therefore, hydride salts have been used to replace hydrazine. Salt-like NaBH 4 (sodium borohydride) has been reported to reduce GO [49]. Apart from hydrazine salts, other reagents have been reported to be used in the chemical reduction of GO, such as ascorbic acid [50] and alcohol [51] Bottom-up With the bottom-up methods, the carbon atoms are converted to graphene or its derivatives. Many different methods have been reported in the past. Chemical vapour deposition (CVD) is a representative method. Wang et al. stated that this method can be used for the large-scale production of graphene [52]. Chae et al. also reported similar findings [53]. Arc-discharge is another representative bottom-to-up method for the synthesis graphene. Nan Li et al. reported that multi-layer graphene can be produced using 29

30 the arc-discharge method [54]. Further, Sprinkle et al. reported the method of using epitaxial growth to synthesise graphene [55]. Apart from the top-down and bottom-up methods, unzipping of CNTs is another interesting approach worth being noted. The method of using plasma etching to unzip multiwall CNTs which are embedded in polymer film has also been reported [56]. Each method has its advantages and disadvantages; the selection of a method depends on the final application of the materials. The graphene sheets produced by using CVD and epitaxial growth methods provide larger-sized and more defect-free graphene than mechanical cleavage and GO reduction. However, bottom-up methods also have their limitations. Both CVD and epitaxial growth require high temperatures during the production compare to mechanical cleavage and GO reduction. Cost of production is also a consideration, in terms of feed stock, energy and substrates used. 2.3 Hot curing epoxy system Polymer matrices play many important roles in composites. In terms of the mechanical property of the composites, they act as the load-transfer medium in the composites which transfer the load to the toughening components. They also maintain the shape of the composites. They function as the adhesions between the toughening components. The thermal stability of the matrixes also decides the physical behaviour of composites in a high-temperature environment. In addition, the polymer matrixes also protect composites from degradation due to environmental attacks. Epoxy resin is a typical thermoset polymer. Thermoset polymers provide better thermal stability, mechanical properties such as high modulus and tensile strength, and hardness than thermoplastic polymers [57]. Concerning the epoxy resin system, 30

31 it provides superior properties such as high modulus, high-temperature resistance, low moisture absorption, etc. [58]. Apart from these, epoxy resins also show high resistivity to corrosion and excellent adhesiveness to most substrates [59]. Given the presence of these outstanding properties, the epoxy resin system has been widely used in the industry. However, the epoxy system has a drawback which limits its applications. The highly cross-linked structures give the cured epoxy materials low fracture toughness. The epoxy resin system usually contains two or three components depending on the types of epoxy resins and the final application of the polymers. These components are resin, hardener, and accelerator. The accelerator is mostly used for speeding up the curing process. There are different kinds of epoxy resins. The most commonly used one is diglycidyl ether of bisphenol A (DGEBA). It is synthesised using bisphenol A and epichlorohydrin; the synthesis process of DGEBA is shown in Figure 9 [43]. Concerning the curing process, epoxy resin reacts with the hardener to form 3D branched structures and then form solid materials. There are many different kinds of curing agents suitable for curing epoxy resins. The most commonly used hardeners are amines. However, other types of hardeners could also be used depending on the final application of the composites. The curing reaction is strongly dependant on the temperature. The curing process usually consists of two steps: Partial formation of cross-links and post-curing. It is well known that the glass transition temperature (Tg) of polymers increases as the degree of cross-linking increases. When the Tg is up to the curing temperature, the curing reaction speed will slow down. Therefore, the material needs to be cured at a temperature which is higher than the highest Tg to obtain the final product with the desirable mechanical properties. The reactions occurring in the curing of epoxy resins and a primary amine hardener 31

32 Figure 9. The reaction during the synthesis of the DGEBA [43]. are shown in Figure 10. As the figure shows, both amine and hydroxyl react with epoxide and generate more hydroxyl which can further react with the epoxide and finally form a highly cross-linked structure. This point indicates the complexity of the curing process as these three reactions could compete with each other. Apart from primary amines, secondary and tertiary amines could also be used as a hardener. Figure 10. Scheme of reactions between the epoxy resin and the amine hardener during the curing process: (1) the reaction between primary amine and epoxide form a secondary amine; (2) the reaction between secondary amine and epoxide form the tertiary amine and hydroxyl; (3) the etherification reaction between hydroxyl and the epoxide to form the ether and hydroxyl. 32

33 2.4 Preparation methods of the nanocomposites The manufacture of the nanocomposites is a relatively long process; it consists of the selection of the nanofillers and matrixes, dispersion of nanofillers, solidifying of composites, etc. Therefore, numerous factors could affect the final properties of nanocomposites. However, the most important part of the process is the dispersion of nanofillers which can directly decide the final properties of the composites. Therefore, the following parts will focus on the methods for dispersing the nanofillers into the matrices. The dispersion of nanoparticles is a major factor which can decide the final properties of the composites. The aim of this process is to achieve a homogeneous dispersion of nanofillers in the composites. It is believed that the homogenous dispersion of nanofillers could result in beneficial properties for the final composites, especially the mechanical and thermal properties. During the preparation of nanocomposites, the dispersion technique has two primary functions: (1) the dispersion of nanofillers and (2) exfoliation of the nanofillers. The sizes of raw materials are randomly distributed from the macroscale to the microscale. Therefore, the exfoliation of the raw materials is necessary for preparing the nanocomposites. However, the selection of the dispersion techniques depends on many factors such as the physical properties of the raw materials and matrixes and the final application of the composites. Currently, there are many dispersing techniques available such as shear mixing, three-roll milling, sonication, ball milling, etc. The most widely used methods are (1) shear mixing, (2) three-roll milling, and (3) sonication Shear mixing A high-speed shear mixer and schematic of shear-mixing flow during dispersion have been shown in Figure 11 [60]. 33

34 Figure 11. High-speed shear mixer and dispersion flow (in inset) which occurs during shear mixing [60]. The working mechanism of a shear mixer is as follows: Firstly, the blades start rotating and generate a vortex in the vessel. As the rotating speed increases, a strong suction is generated that sucks the suspensions which contain the matrix and primary fillers into the centre of the vessel. Then, the suspensions are mixed and pushed to the gap between the blade edge and the stator which are the places for exfoliating the fillers. After that, the suspensions are pushed back to the matrix with smaller-sized fillers. This is counted as one cycle of shear mixing. After one cycle is finished, the mixed suspensions are sucked into the rotator with the new suspensions and then be mixed again. The mixing stops at the pre-set time decided by the users. Finally, the suspensions form a relatively homogeneous dispersion of the micro/nanoscale fillers. Graphite nanoplatelets (GNPs) are the raw materials 34

35 used in this method. These materials consist of many layers of graphene which are stacked together by van der Waals forces. Therefore, the interaction between each layer is not strong. Mechanisms like shear force, collision, shock wave, etc., which could provide enough energy to delaminate graphite, have been used to exfoliate the GNPs [61], [62]. Lei Liu et al. explained the exfoliation mechanism of graphite during high-speed shear mixing [63]. Figure 12 shows the exfoliation mechanisms of graphite occurring during high-speed shear mixing Lei Liu reported that shear force is easily induced during the mixing when the speed is high, and this high shear force could peel the GNPs with relatively high efficiency. Apart from the shear force, jet cavitation is another mechanism used to exfoliate the GNPs. Jet cavitation is generated by the sudden changes in the velocity and geometry of the flow [64]. The flow which is pushed out from the stator could apply the jet cavitation on the GNPs flakes owing to the speed and geometry changes. At last, the collisions of the particles can also exfoliate the GNPs. The collisions can be divided into two types: random collision and edge collision. Random collisions occur in any part of the suspensions during the mixing, while edge collisions periodically occur in the stator hole edges [65]. Figure 12. Exfoliation mechanisms occurring during the shear mixing process; the shear force, jet cavitation, and collisions are three primary mechanisms of the mixing [63]. 35

36 2.4.2 Three-roll milling Three-roll milling is another technique that has been widely used in dispersing nanofillers into matrices. The mechanism of three-roll milling is described below: As the Figure 13 (a) shows, a TRM machine has three adjacent cylindrical rollers. The left roller is the feed roller, and the right one is the apron roller. The left roller and the right roller rotate in the same direction, and the roller in the centre rotates in the opposite direction to the other two. As the Figure 13 (b) shows, the suspensions are fed into the narrow gap between the feed roller and the centre roller. The two rollers rotate in different angular velocities. Therefore, a high shear force is generated [66] (ω 1 <ω 2,ω 1 pertains to the feed roller and ω 2 pertains to the centre roller). The same mechanism occurs between the centre roller and apron roller as well. The suspension flow covering the surface of rollers is collected by the scraper blade which is in contact with the apron roller. In this way, one cycle of TRM is completed. The mixing cycle could be repeated several times until the desired dispersion state is achieved. There is a unique advantage of three-roll milling; the gap between the rollers is adjustable, so the dispersion state of the fillers can be controlled. Apart from this, three-roll milling has been found to be a suitable technique for dispersing CNTs into polymers [67]. However, there are some concerns about the three-roll milling. The first concern is the gap width between the rollers; if the width is in the microscale (generally from ~1-5 µm), large agglomerates can easily be broken into micro-scale agglomerates while the further exfoliation of the agglomerates is limited [67]. Second, three-roll milling can only be applied to viscous materials, thus limiting the types of polymers that can be used. In addition, the cleaning of the machine after the mixing could be another issue. 36

37 Figure 13. Schematic of three-roll milling working mechanism; the different rotation speeds and directions of the rollers exert shear force on the fillers, resulting in the exfoliation and dispersion of fillers [66] Ultrasonication Ultrasonication exerts energy by ultrasound to disperse and exfoliate the nanofillers in the matrices. During ultrasonication, ultrasound causes liquid cavitation and highspeed liquid jet which results in intense dispersion of nanofillers. A schematic of the exfoliation mechanism via sonication is shown in Figure 14 [68]. Figure 14. Schematic of exfoliation mechanism in sonication; the micro-jet work plays a primary role in exfoliation of fillers whiles the wedge and shear effect play a secondary role [68]. 37

38 The cavitation generates many bubbles. The collapse of these bubbles creates micro-jets and the shock waves apply compressive stress on the fillers; then, the tensile stress waves are generated owing to these compressive stress waves, resulting in the exfoliation of the fillers [68]. Owing to intense bubble collapses, the ultrasonication tends to exert a strong peeling effect on the fillers in the suspensions. However, ultrasonication has many drawbacks. First of all, ultrasonication works efficiently in low-viscosity liquids which may need the addition of a solvent during the process; therefore, an extra step of removing the solvent becomes necessary, resulting in a longer process time. Secondly, there is a rapid heating effect during the process owing to the intense energy exerted by the sonicator; therefore, an ice bath is necessary for cooling down the environment temperature. Most importantly, ultrasonication could damage the nanofillers during the dispersion process. Bracamonte et al. found the relationship between the sonication time and the defects in graphene; the short time leads to defects to the edge while longer sonication time leads to defects in the basal plane of graphene [69] Quantification of dispersion state The dispersion state of nanofillers is crucial to the nanocomposites. Chandrasekaran et al. reported that the homogenous dispersion of nanofillers is an important factor affecting the mechanical properties of the composites (Figure 15); in the meantime, improved electrical conductivity is not strongly related to the homogenous dispersion of nanofillers [43]. Several instruments such as transmission electron microscope (TEM), SEM and optical microscope, etc. can be used to analyse the dispersion state of the nanofillers in the composites [70, 71]. However, the TEM is a complex and expensive instrument for this task while a combination of the SEM and the optical microscope can be more suitable for the project. 38

39 Figure 15. Schematic of the nanofillers dispersion states; homogeneous distribution of fillers results in better mechanical properties, while inhomogeneous distribution leads to better electrical properties [43]. 2.5 Properties of nanocomposites As mentioned in the previous section, graphene is a material with many superior properties. Therefore, it could be an ideal material to reinforce the polymer matrix. Numerous studies have reported how to use graphene or graphene-related materials to improve the properties of the composites Mechanical properties of graphene-reinforced nanocomposites Pristine graphene has been reported as the stiffest material to date; it has the modulus ~1 TPa and extremely high tensile strength of about 130 GPa [25]. These high values make an ideal material to improve the engineering properties of composites. Materials like CNTs and nanoclays which have high stiffness and nanoscale sizes could improve the modulus of the composites; the high stiffness of nanofillers could reduce the mobility of the polymer chains by nesting in the polymer matrix and resulting in the increase of modulus [72]. The same mechanism is suitable for graphene as well. It has been reported that thermally reduced graphene (TRG)- reinforced polymethyl methacrylate (PMMA) composites have a 33% higher elastic 39

40 modulus with 0.01 wt% addition of the TRG via high-speed shear mixing [73]. There is another 31% improvement in modulus by the addition of 0.1 wt% of TRG in the epoxy matrix with the help of sonication and stirring [74]. These improvements are achieved by the wrinkled surface of the TRG and oxygen functionalities in the TRG. The wrinkled surface provides a stronger mechanical interlocking effect between the fillers and matrixes and the oxygen functionalities promote the hydrogen-bonding interaction between fillers and matrixes, consequently reducing the mobility of chains and increasing the modulus of the composites. However, not only the wrinkled surface but also the functionalities of the graphene improve the mechanical properties of the composites. It has been reported that different lateral sizes of graphene fillers affect the mechanical properties as well. As compared to pure epoxy, an 82% improvement in fracture toughness can be obtained by adding 2 wt% of 25 µm of GNPs and a 60% increase in fracture toughness can be obtained by the addition of 5 µm of GNPs [75]. It has been reported that the larger lateral size of the GNPs flakes promotes the effect of crack deflection and crack bridging which finally results in the improvement of fracture toughness. Apart from the modulus and fracture toughness, 130% improvement of tensile strength has been reported by adding 2 wt% of GNPs into polyvinyl chloride [76]. All in all, various factors decide the final mechanical properties of nanocomposites such as the intrinsic mechanical properties of the fillers and matrixes, the loading of the fillers, the adhesions between the fillers and the matrix, the sizes of the fillers, the dispersion of the fillers in the matrixes, etc. These factors could all affect the mechanical properties of the nanocomposites Electrical conductivity of the nanocomposites Most polymers behave similar to insulators. Therefore, the addition of conducting 40

41 fillers could transform these insulators into conductors. The conducting fillers like carbon derivatives (GNPs, CNTs, EG, etc.) could be used for improving the electrical conductivity of polymer composites. The percolation theory could be used to explain the electrical conduction mechanism of the composites. At the beginning, the loading of the fillers is low; therefore, the polymer matrixes dominate the electrical conductivity of the composites, and the composites behave like insulators. When the loading increases to a certain point, the electrical conductivity of the composites steeply increases by several orders of magnitude, and this loading is called as the percolation threshold. After this loading, continuous electron paths are formed in the composites, and the insulators are transferred to conductors [60]. A figure demonstrating the applications of electrical conducting composites and the percolation threshold and conducting network is given in Figure 16. However, direct contact between the conducting fillers is not necessary owing to the tunnelling effect; this is also the reason why composites have low percolation thresholds. Both Munson-McGee and Balberg found that nanofillers orientation plays a major role in forming the percolation threshold network; they both reported that increased filler alignment leads to an increase in the percolation threshold [77, 78]. Stankovich et al. reported a low percolation threshold of about 0.1 vol% for graphenepolystyrene composites and found the electrical conductivity of the composites had increased from 0.1 to 1 S/m when the graphene volume fraction rose from 1 to 2.5 vol% [79]. They believed that the high aspect ratio of graphene and homogenous dispersion states contributed to such a low percolation threshold and high electrical conductivity of the composites. It appears that, the lowest percolation threshold of about 0.07 vol% was reported in graphene-reinforced polymer composites [80]. Concerning CNT-reinforced polymer composites, a percolation threshold about 0.15 vol% of the MWCNTs/HDPE composite has been reported [81]. A high electrical 41

42 conductivity of about 27 S/m has been reported by Xue Wang and his group; they achieved this high value by the method of in-situ polymerisation. They assumed that high electrical conductivity can be achieved by a higher level of polypyrrole doping [82]. Figure 16. Applications of electrical conducting composites (A) and conducting network of composites (B) [60]. 42

43 All in all, the electrical conductivity of nanocomposites can be decided by many factors such as the tunnelling effect, contact resistance between the fillers, and the original conductivity of nanofillers Thermal properties of the nanocomposites Thermal conduction can be achieved by phonons and electrons [60]. However, the epoxy polymer is an insulator which means the electrons will not play an important role in thermal conduction. There is no percolation threshold for the thermal conductivity of the GNPs- or CNTs-reinforced nanocomposites; this could be explained by relative small thermal conductivity difference between the fillers and the matrix compare to their electrical conductivity difference [83 85]. In addition, the electrical conductivity of GNPs- or CNTs-reinforced composites is mainly contributed by the conducting fillers while both the matrix and fillers contribute to thermal conductivity, so there is no thermal percolation threshold for composites. Khan M.F. Shahil has investigated the thermal conductivity of graphene/multilayer graphene-reinforced epoxy composites and found that graphene materials have a better thermal interface than CNTs, metal nanoparticles, etc. owing to their geometry and lower Kapitza resistance (interfacial thermal resistance) [86]. An 18- fold improvement in the thermal conductivity of GNPs/silicone composites has been reported with the 25 wt% loading of the GNPs by three-roll milling [87]. The author also found a 29% improvement in the thermal conductivity as compared to the equivalent sample prepared by shear mixing. However, this result could not prove that three-roll milling is always better than shear mixing. As mentioned before, the selection of the processing technique depends on the types of the filler and matrix and the final application of the composites. Furthermore, a 230% increase in the thermal conductivity of expanded graphite epoxy composite by using sonication has been reported by Carola Esposito Corcione [88]. An 800% improvement in thermal conductivity has been reported with 4 wt% loading of non-covalent functionalised 43

44 graphene [89]. They assume such high growth is caused by the functionalization of the graphene which leads to lower interfacial thermal resistance and better interactions between the fillers and the host polymer. Tengfei Luo and John R. Lloyd also studied the factors that could affect the thermal conductivity of the interface between graphene and polymer. They found that longwavelength phonons could promote the spectra coupling between graphene and polymer, resulting in the increase in thermal conductivity and stronger interactions between the fillers; changed in polymer density also result in the improvements in thermal conductivity [90]. However, different geometries of the fillers also affect the thermal conductivity of the composites. Several researchers have reported that graphene is more efficient in improving the thermal conductivity of the composites than materials like CNT (1D structures) [85, 91, 92]. 44

45 3. Methodology 3.1 Materials GNPs, MWCNTs, and EG are the three materials used in this project. All materials have been analysed using a SEM to investigate the geometry of the raw materials (presented in the later section). The information about as-received materials from the suppliers is shown in Table 1. The resin system used in the project was Araldite LY564* (resin) and Aradur 2954* (hardener) from Huntsman Inc., Switzerland. The resin was a bisphenol-a epoxy resin which has low viscosity; the hardener was a cycloaliphatic Table 1. Data sheet of as-received MWCNTs and GNPs used in the project [93, 94]. MWCNTs GNPs Supplier Cheap Tubes Inc., USA XG Sciences, Inc., USA Appearance Dry Black Powder Dry Black Powder Purity >95 wt% >99.5 wt% Lateral Diameter (um) Typical Length um ~25 um Average Thickness (nm) Outside Diameter nm nm Average Layer N/A ~18 layers Specific Surface Area (m 2 /g) 233 m 2 /g m 2 /g Electrical Conductivity (S/m) >100 S/cm Parallel to surface: 10 7 S/m; Perpendicular to surface: 10 2 S/m Thermal Conductivity (W/mK) N/A Parallel to surface: 3,000 W/mK; Perpendicular to surface: 6 W/mK 45

46 Table 2. Data sheet of the acetone used in the project [96]. Supplier Product Molecular Density at Assay Boiling Name Formula 20 C (g/ml 3 ) Point Fisher Scientific Ltd, UK 2.5LT Acetone, Certified AR for analysis C 3 H 6 O > 99.8% 56 C polyamine. The mixing ratio of resin to hardener was 100:35, parts by weight (epoxy : hardener). The curing cycle of this epoxy system used in the project was 2 hours under 80 C and 8 hours under 140 C. According to the data sheet, the samples which have been produced under these conditions would have a tensile strength in the range of MPa and their Young s modulus would be in the range of GPa [95]. The details of the acetone used in the project are shown in Table Preparation of the samples Achieving a homogeneous dispersion of nanoparticles in the matrix is always a challenge in nanocomposite science. Currently, many techniques are applied in this area such as high-speed shear mixing, balling milling, sonication, three-roll milling, etc. These techniques could be used either individually or by randomly combining any two techniques together to achieve the homogenous dispersion state of the nanoparticles in the matrix. Most dispersing methods induce high shear force to exfoliate the nanoparticles during the dispersion process. However, each technique has its advantages and disadvantages. For example, high-speed shear mixing could be a relatively easy and time-saving process as compared to other techniques, while three-roll milling may be beneficial for large-scale production. However, there 46

47 is no absolutely flawless dispersion method for preparing all types of nanocomposites. The suitability of a method will depend on the properties of the fillers and matrixes and the final application of the nanocomposites. High-speed shear mixing was chosen as the technique used for the whole project. A SilverSon L5MPA laboratory mixer shown in Figure 17 (a) was used to disperse the raw materials into the matrix during the experiments. (a) (b) Figure 17. SilverSon L5MPA high-speed laboratory mixer (a) and Low-speed laboratory mixer; resin and hardener mixed together at 1000 rpm for 3 minutes (b). 47

48 3.2.1 Nomenclature of the samples The samples prepared for the project can be divided into two groups. It is worth noting that all the weight percentages (wt%) of the samples are the material s weight with the ratio to the weight of resin before the shear mixing process. The first group contains three different sets of samples. All three sets of samples were prepared by using a high-speed shear mixer to disperse the nanoparticles directly into epoxy with different shear mixing times and speeds. Further, 0.1 wt% of EG and MWCNTs each was dispersed into the epoxy; the speed of mixing ranged from 1,000 rpm to 5,000 rpm; three different shear mixing times were used to prepare the samples: 10 min, 1 hour, 2 hours. The GNPs (0.1 wt%) samples were prepared in the same way, except for the samples prepared by 10-min shear mixing. This set of samples had an extra sample prepared with 1 hour and 9,000 rpm shear mixing. All the samples have been named using the following nomenclature: Particles, X wt%, Y rpm, mixing time (e.g. EG, 0.1 wt%, 1000 rpm, 10 min). The second group had two different sets of samples; shear mixing was used for the preparation of these samples as well. In the second group, the GNPs were nanoparticles. Different loadings of GNPs ranging from 0.1 wt% to 5 wt% were dispersed into the epoxy by fixed the mixing time and speed. The shear mixing speed was 5000 rpm for all the samples in the second group and the shear mixing time was 1 hour. One set of the samples was prepared with 5000 rpm, 1 hour mixing without the addition of acetone and the other set of samples was prepared with the same mixing time and speed with the addition of acetone during the mixing process. The samples were named using the following nomenclature: Particles, X wt%, Y rpm, mixing time, without/with acetone (e.g. GNPs, 0.1 wt%, 5000 rpm, 1 hour, without/with acetone). 48

49 3.2.2 Preparation process The process of preparing the samples is given below: (1) Measuring the weight of empty beaker (m 1 ) and the weight of the resin (m 2 ). (2) Measuring the weight of nanoparticles with the ratio to the weight of the resin and mixing the powder with the resin manually. Usually, 150 grammes of the epoxy resin are needed. (3) Setting the speed and time for shear mixing and preparing the ice bath and start shear mixing. (4) Measuring the weight after shear mixing (m 3 ) and calculating the weight of the lost resin; then, the weight of resin could be calculated (m 4 ). (5) Adding the hardener to the epoxy in at a ratio of 35:100 by using m 4 (hardener mixes with resin in the ratio by weight). (6) Mixing the suspensions at the speed of 1000 rpm for 3 minutes to make the resin and hardener mix well with a low-speed laboratory mixer (Figure 17 (b)). (7) Degasification of the suspensions in a Heraeus Vacutherm degassing chamber (Figure 18 (a)) until all the bubbles have been removed. (8) Preparation of optical microscope slides; pouring the suspensions into prepared silicon rubber moulds which show in Figure 19. (9) Curing samples for 2 hours at 80 C and post-curing at 140 C for 8 hours in the Heraeus Thermoscientific oven (Figure 18 (b)). After the preparation process, all the samples were ground and polished into the standard sizes. However, an additional step was carried out for the samples which were shear mixed with acetone. Before adding the hardener, the acetone in the suspensions needed to be removed by heating the suspensions in the oven at 80 C for 16 hours. After removing the acetone from the suspensions, the following steps remained the same as the other groups. 49

50 (a) (b) Figure 18. Heraeus Vacutherm degassing chamber (a) and Heraeus Thermoscientific Oven (b) (one curing cycle: 80 C for 2 hours and 140 C for 8 hours) Sample geometry The project focused on examining the mechanical, electrical, and thermal properties of the nanoparticle-reinforced epoxy composites. In order to evaluate these properties, the samples were prepared in specific geometries and dimensions according to the relevant testing standards. The geometries of these samples are shown in Figure 20. For the tensile test, the samples were shaped in a dog-bone structure according to the ASTM D638 standard [97]. Figure 19. Silicon rubber moulds (the left mould was used for the thermal sample and the right one, for the mechanical and electrical samples). 50

51 Electrical conductivity test sample Thermal conductivity test samples Tensile test sample Thermal diffusivity test samples Figure 20. The geometries of the samples used for mechanical, electrical and thermal test. The Type IV shape was chosen as the geometry for tensile testing samples. In addition, the thickness of the sample was maintained at 3.2 ± 0.4 mm. Each set of the samples contained seven individual tensile test specimens. For the electrical conductivity test, the samples were given cuboid shape and a length of 20 ± 0.5 mm; the width was polished to 10 ± 0.5 mm, and the height was polished to 5 ± 0.5mm. Each set of samples contained 5 individual electrical conductivity test specimens. For the thermal conductivity and diffusivity tests, the sets included 3 samples measuring ± 0.1 mm (length width thickness). For the thermal diffusivity test, 6 additional cylindrical samples were prepared; 3 of them measured 51 5 ± 0.1 mm (Diameter Height) and the other 3 measured ± 0.1 mm (diameter height). 51

52 SEM samples Optical microscope sample Figure 21. SEM and optical microscopy samples (the silver samples on the top are the SEM samples and the glass slides at the bottom are the optical microscopy samples). In order to analyse the dispersion of nanoparticles in the matrix, the optical microscopy samples and SEM samples were prepared as well. Before each set of samples was cured, three drops of suspensions from the samples were placed on the microscope slides. The SEM samples were prepared from the broken tensile test samples to observe the tensile fracture surface. Figure 21 shows the optical microscope sample and the SEM samples coated with platinum. 3.3 Experiment set-up Tensile strength and modulus test set-up The standard used in the tensile test was the ASTM D 638. It is the standard method for testing the tensile properties of plastics. An Instron 5969 machine was used to evaluate the tensile strength and modulus of the samples; the load cell in the test was set to 100 kn. One 2034 MTS Clip-On extensometer was used to measure the gauge length changes. The test speed used in the experiment was set at 1 mm/min according to the standard (Figure 22). 52

53 The tensile strength of each sample was calculated by using the maximum load in the test (in Newton) divided by the minimum cross-sectional area of each sample s original size (in square meters). The results were expressed in Pascals. The formula used is as follows: Tensile Strength (Pa) = Maximum load (N) Original minimum cross sectional area (m 2 ) (1) The other important value is the elastic modulus of the sample. For calculating each sample s elastic modulus, the load-extension curve needed to be converted to the stress-strain curve of the sample. The linear portion of the stress-strain curve was extended. Then, the modulus was calculated by using the stress difference in any section of this straight line divided by the strain difference of the corresponding section. The equation is shown below: Elastic Modulus (Pa) = Stress Strain = Load (N) Average cross sectional area (m 2 ) Length changes (m) Orginal length (m) (2) Figure 22. Left image: Instron 5969 machine and the tensile sample under the test; right image: structure of the sample. 53

54 3.3.2 Electrical conductivity test set-up Alternating current (AC) was applied to measure the electrical conductivity of the samples. A combination of the Impedance Analysis Interface (IAI), the Phase Sensitive Multimeter (PSM) 1735 and the Frequency Response Analyser (Newtons4th Ltd, UK) was used to measure the AC conductivity of the samples (Figure 23). The instruments could measure the resistance at a range of 1 mω to 500 MΩ; the frequency ranged from 10 µhz to 35 MHz, and the accuracy of the results was ~0.1 % when the frequency was lower than 1 khz [98]. A frequency range from 10 μhz to 1 MHz was applied in the conductivity test owing to the temperature rise caused by the effect of dielectric losses in the materials with high impendence. During the test, the resistance of the sample was calculated by the PSM with the help of a discrete Fourier Transformation. The results were computed under the model of parallel capacitance to improve the accuracy of results [98]. The two-probe Figure 23. Left image: combination of the Impedance Analysis Interface (IAI), Phase Sensitive Multimeter (PSM) 1735, and Frequency Response Analyser; right image: electrical sample under two-probe test. 54

55 test method was chosen because of the relatively low contact resistance between the wires and samples as compared to the sample impendence. After the resistance results had been collected, the electrical conductivity results were calculated using the equation below: Electrical conductivity (S m) = 1 = 1 = L (S m ) R (Ω m) r (Ω) A (m2 ) ra (3) L (m) Where R refers to the resistivity of the sample (Ω m), r refers to the resistance of the sample (Ω), A refers to the cross-section area of the sample (m 2 ), L refers to the length of the sample (m) and the results were expressed in Siemens per meter (S/m). Figure 24 shows the relationship between the conductivity of the semiconductor and Figure 24. The relationship between sample conductivity and AC frequency, where log ω is the common logarithm of the current frequency and log δ is the common logarithm of the sample s conductivity [99]. 55

56 the frequency of the AC [99]. The reason for using AC to measure the sample s resistance instead of direct current (DC) is that the resistance of the sample exceeded the DC range of the machine. The equation used for demonstrating the relationship between the AC conductivity and the frequency of current is given below [99]; σ AC = σ DC + Aω n (3) Where δ AC refers to AC conductivity; the δ DC refers to direct current (DC) conductivity; the A refers to pre-exponential factor; the ω refers to the frequency, and the n refers to the exponent of frequency which generally <1. As to the semiconductors, a relatively plain line could be observed in the low-frequency zone, and the value could be assumed as the DC conductivity of the sample [99,100]. When frequency increases, the effect of dielectric losses could generate heats and result in the increase of the sample conductivity Thermal conductivity test set-up The FOX-50 from Laser Comp Inc., USA, was used as the thermal conductivity meter. The FOX-50 measured the thermal conductivity of the sample based on the ASTM C518 and the ISO 8301 standards. Figure 25 shows the FOX-50. The machine could measure the thermal conductivity at a range of 0.1 to 10 W/(mK) with an accuracy of ±3% in the 2-thickness measurement method [101]. During the test, each sample was placed between isothermal plates to maintain the different temperatures. Therefore, the heat could flow through the sample. The calculation of thermal conductivity was based on Fourier s law as shown below [102]: q = λ(dt dx) 56

57 Figure 25. FOX-50 Thermal Conductivity Thermometer (the upper plate set as 30 C and the bottom plate set as 25 C). Where q refers to heat flux (W/m 2 ), λ refers to thermal conductivity (W/mK), and dt/dx refers to the temperature gradient of the sample s isotherm flat surface (K/m). The dt/dx was calculated by using the temperature difference between two plates (set at 25 C and 35 C to simulate the normal temperature) divided by the thickness of the sample. During the test, the two-thickness measurement method was applied during the measurement to exclude the thermal resistance of the contact surfaces between the sample and plates [102] Thermal diffusivity test set-up The infrared (IR) thermography technique was applied to measure the thermal diffusivity of the samples. Thermal diffusivity indicates how fast the heat conduction is through a given material. Figure 26 shows the IR thermography machines. 57

58 Heat resources Sample IR camera Figure 26. IR thermography machines and the measurement of thermal diffusivity of the sample (left to right: heat resources, samples, and IR thermography camera). During the test, the IR thermography camera was set to the transmission mode. After the sample was placed in the test area, the heat source started heating the front surface of the sample, and the IR camera recorded the thermal radiation on the rear surface. In each test, 1000 images which indicate the temperature changes of the sample during the flash were collected; the frame rate for computational calculating was 95 Hz. The equation used for calculating the thermal diffusivity is as follows [103]: α = In(0.25)L 2 (π 2 t 0.5 ) (3) Where α refers to thermal diffusivity (m²/s), L refers to sample thickness, and t 0.5 refers to the time to reach the half-maximum temperature. Figure 27 shows the position of the time to reach the half-maximum temperature. It was automatically measured the by computer and shown on the screen. The MatLab code used in the measurement is shown in Appendix 1. The code could be used to calculate the average thermal diffusivity of the sample and present the thermal diffusivity mapping. A 350 x 350-pixel image was collected for each sample, 58

59 and the thermal diffusivity of each pixel was calculated. In addition, the mapping image also provided information about particle dispersion in macroscale. This IR thermography technique could be used to effectively analyse dispersion quality; it can provide the dispersion quality information for large-scale samples without destructing the sample, and it is a relatively low-cost method as well. Figure 27. Position of the time for the sample to reach half-maximum temperature during the test. 59

60 4. Results and discussion 4.1 GNPs characterisations As-received materials were observed under the SEM; the images are shown in Figure 28.The SEM images showed the clear flake structures of the GNPs. Concerning the EG, the expanded layers and relative loose structures could be seen from the SEM images. Concerning the MWCNTs, intensive bundles of CNTs (a) (b) 20 µm 2 µm (c) (d) 30 µm 10 µm (e) (f) 20 µm 2 µm Figure 28. SEM images of GNPs, EG, and MWCNTs (top to bottom) (a), (c) and (e) are the low magnification images while the (b), (d) and (f) are the high magnification images. 60

61 could be seen, and the variations in bundle size could be observed as well. Raman spectroscopy was used to detect the number of layers of the pure GNPs used in the project. The Raman result for the GNPs is shown in Figure 29. The GNPs powders were examined using 633-nm lasers. The Raman shifts of the GNPs used in the project were at the D-peak at cm -1, G-peak at cm -1 and 2D-band at cm -1 ; the intensities of the D-peak, G-peak, and 2Dpeak are arbitrary units, arbitrary units, and arbitrary units, respectively. As Figure 30 shows, typical Raman spectroscopy of graphene had a lower G-peak than 2G-peak and typical Raman spectroscopy of graphite had a higher G-peak than 2G-peak; in addition, it was difficult to distinguish between graphene (>5 layers) and graphite by the Raman spectrum [104]. Figure 29. Raman spectroscopy result for the GNPs used in the project (laser used in the test is 633 nm, spectral range in the test was to cm

62 Figure 30. Raman spectroscopy results for graphene and graphite using nm laser [104]. The results showed that the GNPs used in the project had at least 5 layers. 4.2 Optical microscope images of the fillers dispersion state The optical microscopy sample for each group was prepared before the curing process. Then, all the samples were cured along with their group under the same curing process. However, these images did not represent the exact dispersion state of the nanofillers in the final composites owing to the curing process and Brownian motion. However, the images indicated a rough dispersion state of the nanofillers in the composites. In addition, the optical microscope provided a larger view of the nanofillers dispersion state than the SEM. All the samples were observed in the same condition at a magnification of 500, so the magnifications of all the images were the same. The differences in the lengths of the scale-bars can be attributed to the different operators. Optical microscope images of EG-reinforced epoxy composites prepared by different shear mixing speeds and times are shown in Figure

63 Sample (a) shows numerous large-sized EG agglomerates. The sizes of these agglomerates varied from dot to flake. The distances between these agglomerates were relatively smaller than in other samples. However, the distance between the flakes was relatively long. An uneven dispersion state of EG was observed in the sample (a) as well owing to the short mixing time; in this case, the rotators did not have enough time to repeat the cycles of the suction and projection of the suspensions. On comparing the images (a) and (b), it can be concluded that a higher shear mixing speed could improve the dispersion even with a short mixing time. This could be explained by the larger energy input due to the higher mixing speed. When comparing image (a) and (c), the longer mixing time without the increase in speed achieved a better dispersion of the agglomerates. (a) 100 µm 100 µm (b) (c) 100 µm 100 µm (d) Figure 31. Optical microscope images of EG-reinforced epoxy composites: sample (a) EG, 0.1 wt%, 1000rpm, 10 min, sample (b) EG, 0.1 wt%, 5000 rpm, 10 min, sample (c) EG, 0.1 wt%, 1000 rpm, 2 hours, sample (d) EG, 0.1 wt%, 5000 rpm, 2 hours. 63

64 Optical microscope images of MWCNTs-reinforced epoxy composites prepared with different shear mixing speeds and times (Figure 32). The MWCNTs images show that the trend of dispersion was similar that of the EG samples. An increase in the mixing time or mixing speed could lead to a better dispersion state of the MWCNTs. In the meantime, a higher shear mixing speed generated higher energy to break the agglomerates into smaller sizes. The variations in the agglomerates sizes and the relatively uneven dispersion state could be observed in the image (a). However, there are some differences between EG samples and MWCNTs samples. Sample (a) of EG shows that the sizes of agglomerates varied from dot to flake, while the agglomerates in the sample (a) of MWCNTs did not show such variation in size. 100 µm (a) 100 µm (b) (c) 100 µm 100 µm (d) Figure 32. Optical microscope images of MWCNTs-reinforced epoxy composites: sample (a) MWCNTs, 0.1 wt%, 1000rpm, 10 min, sample (b) MWCNTs, 0.1 wt%, 5000 rpm, 10 min, sample (c) MWCNTs, 0.1 wt%, 1000 rpm, 2 hours, sample (d) MWCNTs, 0.1 wt%, 5000 rpm, 2 hours. 64

65 This could be explained by the relatively loose structure of the EG and the heavily entangled structure of MWCNTs. Therefore, under low energy input, the EG has smaller sizes of agglomerates than the MWCNTs. In addition, when the speed increased, the MWCNTs samples showed better dispersion than the EG samples as the EG could start restacking in the high-speed shear mixing condition. Optical microscope images of GNP-reinforced epoxy composites prepared by different shear mixing speeds and times are shown in Figure 33. The GNPs samples did not show an obvious improvement in dispersion state when the mixing time increased. It might because the shear mixing time difference was 1 hour to 2 hours instead of 10 min to 2 hours. As the shear mixing speed increased, an improvement in the dispersion state could be seen. Interestingly, some mediumsized agglomerates always existed in all the GNPs samples owing to the standard diameter of the GNPs used in the project being 25 µm, so a larger amount of energy may be required to exfoliate the GNPs or due to the restacking of the fillers. Optical microscope images of GNP-reinforced (in different loadings) epoxy composites prepared without/with acetone are shown in Figure 34. The images above show the dispersion state of GNPs when the loading increase. The left column shows the samples prepared without acetone and the right column shows the samples prepared with the addition of acetone. The network structure of the GNPs can be seen when the loading increase. The images (a) (b) and (c) (d) indicate the dispersion difference between the samples due to the addition of acetone. The viscosity of the suspensions dropped drastically owing to the addition of acetone. Therefore, the friction between the rotator and the suspensions decreased and the energy loss caused by friction may decrease. Then, the energy input may be larger in the samples prepared with acetone owing to less friction between the rotator and suspensions. Hence, image (b) shows fewer agglomerates than image (a). It is clear that the primary and secondary agglomerates were formed 65

66 (a) 100 µm 100 µm (b) (c) 100 µm 100 µm (d) 100 µm (e) Figure 33. Optical microscope images of GNP-reinforced epoxy composites: sample (a) GNPs, 0.1 wt%, 1000 rpm, 1 hour, sample (b) GNPs, 0.1 wt%, 5000 rpm, 1 hour, sample (c) GNPs, 0.1 wt%, 1000 rpm, 2 hours, sample (d) GNPs, 0.1 wt%, 5000 rpm, 2 hours, sample (e) GNPs, 0.1 wt%, 9000 rpm, 1 hour. when the loading of GNPs increased. It might be the reason that the less exfoliation of agglomerates during the mixing when the loadings of the fillers have increased. In addition, image (d) shows fewer secondary agglomerates than image (c) probably because of agglomerate restacking. This could be explained by the large energy input and high mobility of the agglomerates in the sample prepared with acetone. 66

67 100 µm (a) 100 µm (b) (c) 100 µm 100 µm (d) Figure 34. Optical microscope images of different loadings of GNP-reinforced epoxy composited (with/without acetone): sample (a) is GNPs, 0.1 wt%, 5000 rpm, 1 hour, without acetone, sample (b) is GNPs, 0.1 wt%, 5000 rpm, 1 hour, with acetone, sample (c) is GNPs, 5 wt%, 5000 rpm, 1 hour, without acetone, sample (d) is GNPs, 5 wt%, 5000 rpm, 1 hour, with acetone. 4.3 Tensile strength and modulus properties A set of pure epoxy samples was prepared with the same curing process as the other samples. The pure epoxy samples were prepared for the mechanical properties, electrical conductivity and thermal conductivity measurements. The pure epoxy samples have ± 3.27 MPa in tensile strength and 2.67 ± 0.19 GPa in elastic modulus Tensile strength and modulus results of EG/epoxy composites Figures 35 and 36 show the tensile strength results and modulus results for the EG- 67

68 Figure 35. Tensile strength results for EG-reinforced epoxy composites prepared by different shear mixing speeds and times. reinforced epoxy composites. The tensile strengths of the samples show a decrease, regardless of the shear mixing speed and time compare to pure epoxy. Figure 36. Modulus results for EG-reinforced epoxy composites prepared by different shear mixing speeds and times. 68

69 However, on comparing 1-hour and 2-hour samples with the 10-min samples, a longer mixing time led to a relatively lower decrease in the tensile strength. Apart from this, the 1-hour and 2-hour samples had almost the same values. There was a slight increase only when the shear mixing speed was up to 4000 rpm for all three samples. Concerning the modulus results, all three sets of the samples showed no obvious increase or decrease compare to pure epoxy, with the consideration of the error bars. The changes in the shear mixing speed and time did not exert an obvious influence on the modulus. In all, with the 0.1 wt% loading of EG, the tensile strengths of the composites showed a decrease as compared to the pure epoxy. There is no clear influence of speed changes on the tensile strength of the composites until the speed of up to 4000 rpm while a longer mixing time led to a relatively smaller decrease in tensile strength. The modulus values for the samples are quite similar to those for the pure epoxy, regardless of the changes in mixing speed and time Tensile strength and modulus results of MWCNTs/epoxy composites Figures 37 and 38 show the tensile strength results and modulus results for the MWCNTs-reinforced epoxy composites. No improvements were observed in the tensile strength results. However, all the average values of tensile strength for MWCNTs samples were higher than 60 MPa; this indicates the better tensile strength of the MWCNTs samples than the EG samples. Interestingly, the tensile strengths of the 2-hour samples show a relatively clear trend; the strength increases with the increase in mixing speed and reaches the maximum at 3000 rpm and then decreases again. The other two groups do not show a clear trend with the speed. Also, no obvious increase or decrease in the modulus was observed in the samples as compared to the pure epoxy samples. The shear mixing speed and time did not 69

70 Figure 37. Tensile strength results for MWCNTs-reinforced epoxy composites prepared by different shear mixing speeds and times. show an apparent influence on the MWCNTs samples either. In all, the tensile Figure 38. Modulus results for MWCNTs-reinforced epoxy composites prepared by different shear mixing speeds and times. 70

71 strengths of the MWCNTs samples showed no improvements as compared to the pure epoxy sample. Apart from this, a relatively clear trend was observed for the MWCNTs 2-hour samples, indicating that 3000 rpm has the best effect on the final tensile strength. However, the average strength of the MWCNTs samples was better than that of the EG samples. The modulus of the MWCNTs samples showed slight changes as compared to the pure epoxy samples, regardless of the mixing speed and time Tensile strength and modulus results for GNPs/epoxy composites Figures 39 and 40 show the tensile strength results and the modulus results for the GNPs-reinforced epoxy composites. The tensile strengths of the samples show no noticeable improvement as compared to the pure epoxy samples. A higher mixing speed of 9000 rpm did not increase the tensile strength of the composites either. However, the GNPs 2-hour samples showed a similar trend as the MWCNTs 2-hour samples. An increase in the strength was noted until the mixing speed of up to 3000 Figure 39. Tensile strength results of GNPs-reinforced epoxy composites prepared by different shear mixing speeds and times. 71

72 rpm and then the strength started decreasing again. Interestingly, a clear improvement was found in the modulus results. The modulus results of the GNPs 2- hour samples showed an increase with the rise in speed to up to 3000 rpm and started dropping after 3000 rpm. The maximum value was 3.02 ± 0.09 GPa which indicates a 13% increase as compared to the pure epoxy sample which has 2.67 ± 0.19 GPa in elastic modulus. However, the 1-hour samples showed no obvious increase or decrease in the modulus as compared to the pure epoxy samples. In all, the tensile strengths of the GNPs samples showed no improvements compared to the pure epoxy samples while the modulus showed a 13% increase. A relatively clear trend of the tensile strength could be observed in the GNPs samples prepared by 2 hours of mixing as well. The tensile strength increased with the increase in speed and reached the maximum at 3000 rpm. As to the samples prepared by 1 hour mixing, the tensile strengths of the samples vary a lot, while the elastic modulus results are relatively close to the pure epoxy sample. Figure 40. Modulus results of GNPs reinforced epoxy composites prepared by different shear mixing speeds and times. 72

73 4.3.4 Tensile strength and modulus results of GNPs (with different loadings)/epoxy composites The GNPs samples prepared by the different loadings and the addition of the acetone were tested. Results show in Figure 41 and 42. The mixing speed and time were fixed to see the effect of different loadings of GNPs and the addition of the acetone as a solvent. The tensile strength of the samples decreased when the loading increased, regardless of the addition of acetone. The lowest tensile strength was ~45.1 MPa with the 5 wt% sample prepared without acetone, which indicates a decrease of ~38% as compared to the pure epoxy sample. In terms of the modulus results, higher modulus results were observed with the samples that were prepared with the addition of the acetone and the loading of GNPs was higher than 1 wt%. It is interesting to find that the samples with 1 wt% and 5 wt% loading and prepared with acetone had a higher modulus than pure epoxy samples, while their tensile strengths decreased drastically as compared to the pure epoxy samples. Figure 41. Tensile strength results for different loadings of GNPs-reinforced epoxy composites (without and with acetone). 73

74 Figure 42. Modulus results for different loadings of GNPs-reinforced epoxy composites (without and with acetone). The 1 wt% sample with acetone had a modulus of ~2.92 GPa and the 5 wt% sample with acetone had a modulus of ~3.11 GPa; these results indicate improvements of about 9% and 16%, respectively Mechanical results discussion According to the tensile strength results, there was no improvement in any of the samples. Most results showed a decrease in the tensile strength as compared to the pure epoxy results, especially for the samples prepared with the EG. Some of the MWCNTs and GNPs samples showed slight improvements (e.g. samples prepared by 3000 rpm, 2 hours of shear mixing with 0.1 wt% loading of the fillers). It is clear that the tensile strength decreased drastically with an increase in the filler loading. However, the elastic moduli of the sample showed a different trend. The EG (0.1 wt%) and the MWCNT-reinforced (0.1 wt%) epoxy composites showed elastic modulus values similar to the pure epoxy sample values, regardless to the shear mixing speed and time. Interestingly, the GNPs samples like GNPs, 1 wt%,

75 rpm, 1 Hour with acetone and GNPs, 5 wt%, 5000 rpm, 1 Hour, with acetone showed improvements in the elastic modulus as compared to the pure epoxy samples. The tensile strength of a nanocomposite depends on the weakest part of the material. Therefore, many factors can contribute to the decrease in tensile strength such as weak interactions between filler and matrix, the dispersion state of the nanofillers, the size of the agglomerate, mechanical properties of the nanofillers, etc. In this project, it was assumed that the suitable shear mixing speed or mixing time could contribute to an improved distension state and smaller size of the agglomerate, resulting in improved tensile strength. However, the tensile strengths of samples had not improved with high shear mixing speed or long mixing time. It is worth noting that different materials may have different criteria for shear mixing. Yan Yan Huang suggests that the energy induced by shear mixing should be larger than the binding energy of the agglomerates and smaller than the energy that could break the nanotubes [105], while exfoliation of graphite can only occur when the shear rate is above the specific shear rate [106]. Therefore, different materials have their suitable shear mixing speed and time. According to optical microscope images, the agglomerates have been exfoliated into smaller sizes when the mixing speed or time increased. However, the sizes of the agglomerates remained in the micrometre. It is well known that these agglomerates entangled together owing to the Van der Waals forces between the nanofillers. During the tensile test, the loading tension was transferred from the matrix to the agglomerates and the agglomerates started breaking due to weak Van der Waals forces. Then, the cracks propagated through the sample, consequently breaking the sample. Apart from this reason, the poor interface between the fillers and matrix could also contribute to the decrease in tensile strength. The SEM images in the next section show the gaps and voids between the fillers and matrix. The voids between the fillers and matrix 75

76 may be caused by insufficient degasification during the process. Apart from the above reasons, the geometry of the fillers could also lead to a decrease in tensile strength. The EG SEM image showed that it had the loosest structure among the three fillers. This could explain why the EG samples had the lowest tensile strength. Concerning the increase in the elastic modulus, it could be explained by the nesting of these micro/nanofillers in the matrix; the nesting of the fillers in the polymer matrix resulted in a reduction in the polymer chains mobility, increasing the elastic modulus of the composites [6]. However, improvements in modulus were found in the samples with high loadings prepared with acetone. According to the SEM images present in the next section, better interfaces between the filler and matrix were achieved with the addition of acetone. Therefore, both high loading and improved interface contributed to the improvements in the modulus SEM images of tensile fracture surface The SEM images (Figure 43) show the difference in tensile fracture surface among the pure epoxy, GNPs 0.1 wt%, 5000 rpm, 1 hour, and GNPs, 1 wt%, 5000 rpm, without/with acetone samples. The gaps and voids are easily observed in images (b) and (c). Image (d) shows a better interface between the fillers and the matrix could be explained by the addition of acetone. The image (d) also indicates a better interface may result in a better nesting effect of the fillers and finally results in the improvement of the elastic modulus. 76

77 (a) (b) 100 µm 20 µm (c) (d) 20 µm 10 µm Figure 43. SEM images of the GNPs/Epoxy samples: sample (a) Pure epoxy, sample (b) GNPs, 0.1 wt%, 5000 rpm, 1 hour, sample (c) GNPs, 5 wt%, 5000 rpm, 1 hour, without acetone, sample (d) GNPs, 5 wt%, 5000 rpm, 1 hour, with acetone. 4.4 Electrical Conductivity EG/Epoxy composites Electrical Conductivity According to the electrical conductivity results for the EG samples (Figure 44, 45, 46), only three samples showed the plane line results in the low-frequency zone (1-100) which could indicate the samples behave as the conductor and the value show in 1 Hz indicate their conductivity. The rest of the samples could be considered insulators. The three samples which acted as conductors are as follows: (a) EG, 0.1 wt%, 5000 rpm, 10 min, (b) EG, 0.1 wt%, 2000 rpm, 2 hours, and (c) EG, 0.1 wt%, 4000 rpm, 2 hours. The sample (a) which was prepared by 5000 rpm with 10 min mixing has an electrical conductivity slightly higher than 1E-6 S/m. The sample (b) which was 77

78 prepared at 2000 rpm with 2 hours of mixing had similar values to sample (a). Sample (c) which was prepared at 4000 rpm with 2 hours of mixing had an electrical conductivity of ~1E-5 S/m, indicating a 10-fold improvement in the electrical conductivity. In all, the three samples had become conductors with the addition of 0.1 wt% of EG. The rest of the samples remained as insulators. All the electrical conductivity results varied from 1E-6 to 1E-5 S/m. Figure 44. Electrical conductivity results for EG-reinforced epoxy composites prepared by 10 min shear mixing with speed range from 1000 to 5000 rpm. 78

79 Figure 45. Electrical conductivity results for EG-reinforced epoxy composites prepared by 1 hour shear mixing with speed range from 1000 to 5000 rpm. Figure 46. Electrical conductivity results for EG-reinforced epoxy composites prepared by 2 hours of shear mixing with speed range from 1000 to 5000 rpm. 79

80 4.4.2 MWCNTs/Epoxy composites electrical conductivity Concerning the MWCNTs sample results (Figure 47, 48, 49), more samples had become conductors than the EG samples. Among the samples prepared by 10 min of shear mixing, two became conductors: One was prepared by 1000 rpm mixing and the other, by 2000 rpm mixing. Both samples had an electrical conductivity of around 1E-6 S/m. With regard to the samples prepared by 1 hour of shear mixing, all had become conductors. Two of them had an electrical conductivity higher than 1E-5 S/m (the samples prepared by 2000 rpm and 5000 rpm shear mixing), two of them had conductivity values near 1E-5 S/m (the samples prepared by 1000 rpm and 4000 rpm), and one sample prepared by 3000 rpm shear mixing had conductivity values lower than 1E-5 S/m. Most MWCNTs samples prepared by 2 hours of mixing had turned into conductors (except from the sample prepared by 2 hours, 1000 rpm mixing). The one prepared Figure 47. Electrical conductivity results for MWCNTs-reinforced epoxy composites prepared by 10 min shear mixing with speed range from 1000 to 5000 rpm. 80

81 Figure 48. Electrical conductivity results for MWCNTs-reinforced epoxy composites prepared with 1 hour of shear mixing with speed range from 1000 to 5000 rpm. by 5000 rpm show the lowest conductivity while the others had conductivity values of around 1E-5 S/m. Figure 49. Electrical conductivity results for MWCNTs-reinforced epoxy composites prepared by 2 hours of shear mixing with speed range from 1000 to 5000 rpm. 81

82 4.4.3 GNPs/Epoxy composites electrical conductivity The results of the GNPs samples show that none of them were transformed to conductors (Figure 50, 51, 52). In the samples loaded with 0.1 wt% GNPs, the changes in the shear mixing or the mixing time did not lead to conversion into the conductor state. This finding is different from those for the samples loaded with EG or MWCNTs. As the loadings of the GNPs increased, the AC machine could not measure the conductivity of the samples in the low-frequency zone, indicating that the percolation threshold of the GNPs/Epoxy composites may be higher than 5 wt%, or the samples were not prepared properly. Figure 50. Electrical conductivity results for GNPs-reinforced epoxy composites prepared by 1 hour of shear mixing with speed from 1000 to 5000 rpm. 82

83 Figure 51. Electrical conductivity results for GNPs-reinforced epoxy composites prepared by 2 hours of shear mixing with speed range from 1000 to 5000 rpm. Figure 52. Electrical conductivity results for GNPs-reinforced epoxy composites prepared by 1 hour of shear mixing at a speed of 5000 rpm and loading range from 0.1 wt% to 5 wt%. 83

84 4.4.4 Electrical results discussion Very few EG and MWCNTs samples showed electrical conductivity. The GNPs samples showed no electrical conductivity in the low-frequency zone regardless of the shear mixing speed, time, and the loading of the fillers. This indicates that 0.1 wt% loading of EG and MWCNTs could be the percolation threshold for the EG- and MWCNTs-reinforced epoxy composites and the percolation threshold for GNPsreinforced epoxy composites may be higher than 5 wt%. According to the results for EG and MWCNTs samples, the changes in electrical conductivity did not show a clear trend with the increase in mixing speed. However, the length of the mixing time had an effect on the final electrical conductivity. The results showed that the EG samples prepared by 2 hours of mixing had 10 times higher electrical conductivity than the EG sample prepared by 10 min of mixing; the MWCNTs samples show the best electrical conductivity among the samples prepared by 5000 rpm and 1 hour of mixing. The rest of the MWCNTs samples showed electrical conductivities 1E-5 S/m. A longer mixing time leads to more suction cycles during the shear mixing which could result in the formation of smaller agglomerates and a better dispersion state. The factors like filler s geometry, the aspect ratio, and dispersion state could all decide the percolation threshold for the composites [ ]. As the loading of fillers in the EG and MWCNTs sample was 0.1 wt%, it was difficult to ensure direct contact between the fillers after shear mixing. Therefore, the tunnelling effect dominated the conduction mechanism in this case. Then, the homogeneous dispersion state and the small distance among the fillers promoted the tunnelling effect, finally increasing the electrical conductivity. However, the MWCNTs sample prepared by 5000 rpm and 2 hours of mixing showed a decrease in the electrical conductivity that could be explained by the intense energy input that led to the breaking of the nanotubes. The GNPs samples did not show the percolation 84

85 threshold even in the 5 wt% loading, explained by the high difference in the conductivity between the parallel and perpendicular direction of graphene [94]. 4.5 GNPs reinforced epoxy thermal properties The Figure 53 and 54 show the thermal diffusivity and the thermal conductivity of the GNPs/epoxy composites prepared with different loadings with/without acetone. The thermal diffusivity results of the pure epoxy samples are not available in this measurement; the flash directly goes through the pure epoxy sample and the machine could not generate the image and the result of the sample. The samples prepared with the addition of acetone showed higher diffusivity than the samples prepared without the addition of acetone. The largest improvement was found at 5 wt% loading, in the sample prepared with acetone whose diffusivity was ~0.35 mm 2 /s while the sample without acetone had a diffusivity of ~0.25 mm 2 /s which indicated an increase of about 25%. Figure 53. Thermal diffusivity results for the different loadings of GNPs-reinforced epoxy composites prepared by 1 hour of shear mixing at a speed of 5000 rpm without and with acetone. 85

86 Figure 54. Thermal conductivity results for the different loadings of GNPs-reinforced epoxy composites prepared by 1 hour shear mixing at a speed of 5000 rpm without and with acetone. All samples reinforced with the GNPs had higher thermal conductivity than the pure epoxy samples. The maximum thermal conductivity was ~0.41 W/mK, achieved by the sample with 5 wt% loading without acetone and the minimum thermal conductivity was ~0.24 W/mK, indicating 116% and 26%, respectively, improvements as compared to the pure epoxy. These improvements are much higher than those reported by Chandrasekaran, who showed only a 14% increase with the loading of 2 wt% [42]. However, it is worth noting that the samples prepared with 5 wt% GNPs without acetone showed high viscosity, resulting in processing problems and a bigger error bar. Thermal diffusivity mappings of the GNPs-reinforced epoxy composites prepared by 1 hour of shear mixing at a speed of 5000 rpm without and with the acetone are shown in Figure

87 (a) (b) (c) (d) (e) (f) Figure 55. Thermal diffusivity mapping images: (a) GNPs, 1 wt%, 5000 rpm, 1 hour, without acetone, sample (b) GNPs, 1 wt%, 5000 rpm, 1 hour, with acetone, sample (c) GNPs, 2 wt%, 5000 rpm, 1 hour, without acetone, sample (d) GNPs, 2 wt%, 5000 rpm, 1 hour, with acetone, sample (e) GNPs, 5 wt%, 5000 rpm, 1 hour, without acetone, sample (f) GNPs, 5 wt%, 5000 rpm, 1 hour, with acetone. 87