GRAPHENE FUNCTIONAL COMPOSITES

Transcription

1 GRAPHENE FUNCTIONAL COMPOSITES A thesis submitted to The University of Manchester for the degree of Master of Philosophy in the Faculty of Engineering and Physical Sciences 2011 ARUN PRAKASH ARANGA RAJU School of Materials

2 TABLE OF CONTENTS List of tables... 5 List of figures... 6 List of abbreviation List of symbols Abstract Declaration Copyright statement Acknowledgement INTRODUCTION AND AIMS Introduction Aims and objectives LITERATURE REVIEW Structure and morphology of graphene Synthesis of graphene Mechanical exfoliation Micromechanical cleavage Solution phase exfoliation Epitaxial growth Chemical vapour deposition Thermal decomposition of SiC Other methods Un-zipping of Carbon nanotubes Properties of graphene

3 Mechanical properties Electronic properties Thermal properties Raman spectroscopy Raman spectroscopy of graphene Strain sensitivity of the Raman bands of graphene Strain sensitivity of the G band Strain sensitivity of the G band Existing strain-gauge technology Strain gauges Mechanical strain gauges Resistance-based strain gauges Optical strain gauges Other types METHODOLOGY Evaluation of sources of graphene Production of graphene Micromechanical cleavage Solvent exfoliation method Preparation of graphene composites and films Composites with micromechanically-cleaved graphene Composites from solvent-phase exfoliated graphene Composites from extrusion Raman Spectroscopy Experimental Setup Deformation under Raman Spectroscopy

4 4. RESULTS AND DISCUSSION Model single flake composite: Performance on a plastic substrate Characterization of graphene composite Cyclic deformation Accuracy measurements Deformational behaviour Residual stress and energy dissipation Strain hardening behaviour Model single flake composite: Performance on steel Multilayer graphene and its composite coatings: Performance on steel Bulk graphene composite Study on solution-phase exfoliated graphene Effect of sonication time PERSPECTIVES AND FUTURE WORK Perspectives Future works CONCLUSIONS REFERENCES Word count:

5 LIST OF TABLES Table 2-1: G' band shift rates observed by various groups Table 3-1: Different grades of xgnp Table 3-2: Different grades of Angstron materials Table 4-1: Residual stress and energy dissipation in the graphene composite during cyclic deformation Table 4-2: Calculated effective Young's modulus of graphene in composites during cyclic deformation

6 LIST OF FIGURES Figure 1-1:(a) Photograph (in normal white light) of a relatively large multilayer graphene flake (~3nm thick), (b) AFM image of single layer graphene on SiO 2 /Si substrate observed by Geim and his co-workers at University of Manchester in Figure 2-1: Graphene is a 2D building material for other allotropes of carbon. It can be stacked into 3D graphite, rolled into 1D carbon nanotubes or wrapped to 0D fullerenes. 23 Figure 2-2: a) Top view of real space unit cell of monolayer graphene Figure 2-3: 'Rippled' graphene from Monte Carlo simulation Figure 2-4: Mechanical exfoliation of graphene by the Scotch tape method Figure 2-5: TEM image of flakes produced from SDBS/water solution, a) monolayer, b) bilayer and c) trilayer Figure 2-6: a) Photograph of a centrifuge tube following the first iteration of density gradient ultracentrifugation (DGU). The suspension was diluted by a factor of 40 to ensure that graphene bands could be resolved in the photograph. b&c) Representative AFM images of images o of graphene deposited using fractions f4(b) and f16(c) on Si/SiO 2. d) Height profile of regions marked in blue (b) and in red (c) Figure 2.7: A) Experimental setup for the growth of graphene by CVD technique. B) a. Roach leg on top of Cu foil, b. Roach leg under vacuum, c. Residue from roach leg after annealing Figure 2-8: TEM image of partially un-zipped MWCNT structure opened by Li intercalation Figure 2-9: a) A SEM micrograph of a graphene sheet suspended above a trench etched in SiO 2 wafer b) SEM image of graphene monolayer deposited on the substrate with 1µm and 1.5µm circular arrays. c) Schematic representation of the nanoindentation on a suspended graphene membrane Figure 2-10: Electronic band structure of single-layer graphene (HOMO and LUMO meets at the Dirac point Figure 2-11: The energy-level diagram of Rayleigh, Stokes Raman and anti-stokes Raman scattering...35 Figure 2-12: a) Raman spectrum from monolayer graphene edge showing prominent features, b) Raman spectrum of monolayer graphene and bulk graphite

7 Figure 2-13: Raman spectrum of different layers of graphene a) in 514 nm b) in 633 nm c) Splitting of 4 components of G band of bilayer graphene...37 Figure 2-14: Change in G peak position with number of layers. Inset show the intensity difference between HOPG, bilayer and monolayer Figure 2-15: Shift of G band of mono- and tri-layer graphene on PET substrate on uniaxial strain... Error! Bookmark not defined. Figure 2-16: a) Position of G band with respect to strain (note the splitting) b) Shift rates of two components of G band Figure 2-17: Shift of G' band peak position as a function of strain Figure 2-12: Design of wire strain gauge Figure 2-19: The characteristic design of a strain gage with etched metal foil measuring grid Figure 2-20: Schematic representation of semiconductor strain gauges Figure 2-21: Schematic representation of a) Fibre Bragg grating structure b) Refractive index profile c) Spectral response and d) FBG s reflected power as a function of wavelength Figure 3-1: SEM image of xgnp M Figure 3-2: Raman spectra of commercial grades of graphene. Raman spectra of Sigma Aldrich graphite flakes are also shown for comparison purpose Figure 3-3: Schematic representation of solvent phase exfoliation of graphene Figure 3-4: (a) Schematic representation of the graphene composite (b) Cross-sectional view of the composite Figure 3-5: Schematic view of sandwich composites on steel substrates a) SU- 8/graphene/SU-8/steel b) SU-8/graphene/steel Figure 3-6: Schematic representation of graphene/pet strip...55 Figure 3-7: Schematic representation of solvent phase exfoliated graphene on steel substrate...55 Figure 3-8: Schematic representation of graphene/epoxy composite film on steel Figure 3-9: Picture of bulk composite of xgnp/pmma prepared from mini compounder56 Figure 3-10: Schematic diagram of Raman spectroscopy system Figure 3-11: Schematic representation of four-point bending test: (a) On loading, (b) On unloading

8 Figure 3-12: Actual setup of modified G-clamp for deformation of spring steel based substrate Figure 3-13: Cyclic deformational sequence of graphene composites Figure 4-1: Optical image of monolayer graphene from sandwich composite Figure 4-2: Raman spectra of different layers observed in prepared composite (spectra offset for clarity Figure 4-3: Position of the G' band with respect to deformation cycle. Two red lines corresponds to the data points taken from this position for calculating the accuracy of the strain measurements Figure 4-4: Variation of the G' band with respect to deformation cycle. Data obtained from reference Figure 4-5: Shift of the Raman G' band for cyclic deformation from % maximum loading strains Figure 4-6: Shift of Raman G' band for cyclic deformation for 0.4 and 0.5% maximum loading strains Figure 4-7: Derived stress-strain curve for graphene composites (3rd cycle) Figure 4-8: Residual stress and energy dissipated during cyclic deformation of graphene composite Figure 4-9: Plot between effective modulus of graphene vs deformation cycle Figure 4-10: Bright (a & c) and dark (b & d) field images of monolayer graphene under X50 and X100. The atomic ripples were not visible from the dark field images. The bright spots in the dark field images belong to the imperfections on the sample substrate Figure 4-11: Schematic view of sandwich composites on steel substrates a) SU- 8/graphene/SU-8/steel b) SU-8/graphene/steel Figure 4-12: Optical image of bilayer graphene identified in SU-8/bilayer graphene/su- 8/steel sandwich composite sample Figure 4-13: Raman spectra obtained from the identified bilayer graphene. Inset show the four Lorentzian peaks fitted in G band Figure 4-14: Shift of Raman G' band with strain in SU-8/bilayer graphene/su-8/steel sandwich composite Figure 4-15: Shift of Raman G band with strain in SU-8/bilayer graphene/su-8/steel sandwich composite Figure 4-16: Optical image of multilayer graphene identified in SU-8/graphene/steel sandwich composite sample

9 Figure 4-17: Raman spectra obtained from the identified multilayer graphene in SU- 8/graphite/steel sandwich composite sample Figure 4-18: Shift of Raman G' band with strain in SU-8/multilayer graphene/steel sandwich composit Figure 4-19: Shift of Raman G band with strain in SU-8/multilayer graphene/steel sandwich composite Figure 4-20: Raman spectra obtained from graphene/pet strip and PMMA/graphene/PET strip. Spectra offset for clarity Figure 4-21: Shift of Raman G' band (a) and G band (b) on deformation of the graphene/pet strip Figure 4-22: Shift of Raman G' band (a) and G band (b) on deformation the of PMMA/graphene/PET strip Figure 4-23: Raman spectra obtained from graphene/steel and SU-8/graphene/steel samples. Spectra offset for clarity Figure 4-24: Shift of Raman G' band (a) and G band (b) on deformation of graphene/steel sample Figure 4-25: Shift of Raman G' band (a) and G band (b) on deformation of SU- 8/graphene/steel sample Figure 4-26: Raman spectra obtained from graphene-epoxy thin coat on steel substrate. Spectra offset for clarity Figure 4-27: Shift of Raman G' band (a) and G band (b) on deformation in grapheneepoxy hot cured sample Figure 4-28: Shift of Raman G' band (a) and G band (b) on deformation in grapheneepoxy cold cured sample Figure 4-29: TEM images of graphene produced from sonication of XG graphite nanoplatelets Figure 4-30: Optical image of5 wt. % xgnp M25 graphite nanoplatelets/pmma bulk composite showing fairly homogenous dispersion of graphene platelets within the matrix. The flake chosen for the deformation experiment is highlighted Figure 4-31: Shift of Raman G' band of xgnp M25 graphite nanoplatelets/pmma bulk composite over deformation Figure 4-32: Photograph of various dispersions prepared from Branwell grade RFL 99.5 obtained after centrifugation which have been sonicated for the times indicated

10 Figure 4-33: Photograph of various dispersions prepared from Branwell grade 2369 obtained after centrifugation which have been sonicated for the times indicated Figure 4-34: Photograph of various dispersions prepared from Branwell grade 9842 obtained after centrifugation which have been sonicated for the times indicated Figure 4-35: UV-vis Absorption spectra of three different suspensions prepared from different grades of graphite. Sonication time: 3 hours Figure 4-36: Absorbance vs. sonication time for different grades of graphite (660 nm) Figure 4-37: Graphene concentration after centrifuge as a function of sonication time for different grades of graphite Figure 4-38: SEM images obtained from 90 min sonicated samples (a, b, c) and a 5 hr sonicated sample (d, e, f) at different magnifications Figure 4-39: Raman spectra obtained from random positions of the film from 90 min sonication sample. Inset is the vacuum filtered film on aluminium oxide filter paper Figure 5-1: Hand-held Mini Raman instruments from B&W TEK Inc

11 LIST OF ABBREVIATION CNT: SWCNT: MWCNT: Carbon Nanotubes Single-walled carbon nanotubes Multi-walled carbon nanotubes 1D: One-dimensional 2D: Two-dimensional 3D: Three-dimensional HOPG: Si/ SiO 2 : NMP: SDBS: DGU: FBG: PMMA: PET: AFM: SEM: TEM: CVD: UV-vis: He-Ne: HOMO: Highly oriented pyrolytic graphite Silicon/ Silicon dioxide N-methyl pyrrolidine Sodium dodecyl benzene sulphonate Density gradient ultracentrifugation Fibre Bragg grating Poly (methyl methacrylate) Poly (ethylene terephthalate) Atomic force microscope Scanning electron microscope Transmission electron microscope Chemical vapour deposition Ultra violet- Visible Helium-Neon Highest occupied molecular orbital 11

12 LUMO: C: Lowest unoccupied molecular orbital Degree centigrade N: Newton ev: MJ: GPa: TPa: µm: nm: pm: mg: ml: wt.%: rpm: Electron volts Mega joule Giga pascal Tera pascal Micrometre Nanometre Picometre Milligram Millilitre Weight percentage Rotations per minute 12

13 LIST OF SYMBOLS ν: Frequency ε: Strain GF: ΔR/R: ΔL/L: Gauge factor Fractional change in resistance Fractional change in length E: Young s modulus ω: Raman wavenumber I o : Δω: ŋ o : Maximum intensity Full width half maximum Krenchel orientation distribution factor 13

14 ABSTRACT Raman spectroscopy is a powerful technique for characterizing carbon materials and their derivatives. In this project, Raman spectroscopy was used to characterize graphene and its composites with the ultimate aim of developing graphene composites for use as wide area strain sensors in structural engineering applications. Model coating systems were made by depositing micromechanically-cleaved graphene onto PMMA substrates. These sensors could sense uniaxial strain to an accuracy of ± 720 microstrain over repeated cyclic deformation. This relatively low accuracy was due to both the strain hardening behaviour of graphene and the interfacial damage that occurred between the graphene and the matrix during the deformation. The interfacial damage led to increasing energy dissipation during each cycle and a build-up of residual stress. The strain sensitivity of micromechanically cleaved graphene deposited on engineering materials like spring steel was also demonstrated. Finally, in order to assess the scale-up potential of the graphene coatings, composites were prepared from solventphase exfoliated graphene and commercial graphene nanoplatelets, using either casting or compound mixing methods. In order to achieve large flakes for maximum stress transfer, the solvent phase exfoliation of graphene as a function of sonication time was investigated. The higher the sonication time, the higher concentration of graphene achieved. But presence of large graphite like flakes were evident from the Raman spectroscopy and SEM results, suggesting that low rotation speeds of 500 rpm in NMP was not sufficed enough to remove larger/thicker flakes. 14

15 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. Arun Prakash Aranga Raju 15

16 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 mentioned) 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, trademarks and other intellectual property (the Intellectual Property ) and any reproductions of copyright 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 and/or Reproductions described in it may take place is available in the 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. 16

17 ACKNOWLEDGEMENT Firstly, I must thank The Almighty for giving me strength and patience to complete this study for my MPhil degree. I am truly indebted and thankful to my principal supervisor Dr Ian Kinloch for the continuous support of my project with his guidance, supervision, patience, motivation and immense knowledge. I owe sincere and earnest thankfulness to my co-supervisors Prof. Robert J Young and Prof. Konstantin Novoselov for their support and assistance in supporting my academic aspirations. I would like to extend my gratitude to UMIP for funding this project. Special thanks to Mr Andy Zadoronshnji for his assistance in Raman sessions, Dr Sarah Haigh for TEM measurements, Dr Christopher Wilkins for SEM images and Ms Polly Crook for general experimental setups. I am obliged to many of my colleagues namely, Dr Libo Deng, Mr Lei Gong, Mr Tamer Wafy and Dr Rahul Nair for their assistance in my experiments and comments on the results. Last but not the least; I would like to thank my family and friends for their continuous support. This thesis is dedicated to my parents and my sister for their unconditional love towards me 17

18 Chapter 1: INTRODUCTION AND AIMS 18

19 Chapter 1 Introduction and aims 1. INTRODUCTION AND AIMS 1.1. Introduction Carbon is the fourth most abundant element in the universe and one of the essential elemental building blocks of all life on Earth. Carbon can form remarkably complex molecules such as fatty acids, carbohydrates etc. due to its unusual four valence electrons. This flexibility in bonding also means that carbon allotropes vary significantly in their physical properties. The two well-known allotropes are graphite, which is soft and brittle, and diamond, which is hard and strong. Other well-known allotropes of carbon are: (i) amorphous carbon (e.g. coal, carbon black), (ii) fullerenes (bucky balls) comprising of hollow cages of sp 2 carbon, (iii) carbon nanotubes (bucky tubes) which take a cylindrical (one dimensional, 1D) nanostructure, (iv) glassy carbon and (v) carbon nanofoam. Graphene is a two dimensional (2D) allotrope and it is a single atomic layer of graphite. Whilst graphene had been modelled widely, it was not successfully isolated until 2004 when Geim and his co-workers at The University of Manchester produced graphitic sheets with thicknesses down to few atomic layers (including single- and few- layer graphene) (Figure 1.1) from highly oriented pyrolytic graphite (HOPG).[1] Graphene is a crystalline material which is one atom thick and has a Young s modulus of 1 TPa and intrinsic strength of 130 GPa, making it the strongest material ever measured.[2] Since its discovery, graphene with its unique and extraordinary electronic, electrical, mechanical and thermal properties, has become an exciting area for theoretical studies and experimental research. Figure 1-1:( a) Photograph (in normal white light) of a relatively large multilayer graphene flake (~3nm thick), (b) AFM image of single layer graphene on SiO 2 /Si substrate observed by Geim and his co-workers at University of Manchester in 2004 [1] 19

20 Chapter 1 Introduction and aims Raman spectroscopy is a well-established materials characterisation technique which measures the energy of the phonons in a sample. The fact that the energy of the phonons are dependent on the strain in material means that Raman spectroscopy can be used to study local deformation. This strain dependency is used widely in the field of composite materials to study local deformation patterns and interfacial stress transfer between the matrix and the reinforcement. This approach has been particularly successful with carbon- based reinforcements such as carbon fibres, carbon nanotubes etc.[3-6]. Graphene has a well-defined, intense Raman spectrum which is highly strain-sensitive due to graphene s high stiffness [7, 8]. These factors mean that graphene coatings probed by Raman spectroscopy have the potential to be an effective, precise and non-contact technique for strain measurements. In this thesis, graphene composite and coatings were evaluated as strain sensors as part of a University of Manchester Intellectual Property project. 20

21 Chapter 1 Introduction and aims 1.2. Aims and objectives The main aim of this project is to investigate the use of graphene composites as a strain sensor by using Raman spectroscopy to follow the strain sensitive Raman G band. In order to achieve a reliable and sensitive sensor, it is also necessary to understand the interfacial stress transfer that occurs during the deformation of the graphene composites. In order to investigate the strain sensor application of graphene composites the following objectives were identified: 1. Preparation of model flake graphene composites to evaluate the potential of graphene-composite strain sensors. 2. Cyclic deformation of the model composites to determine the accuracy in strain sensor applications. 3. Understand the stress transfer processes which limit the utility of the sensor, which also have significant implications in the more general area of graphene composites. 4. Preparation of model flake graphene composites on a steel substrate to demonstrate its sensor application on an engineering material. 5. Scale-up of the composite. a. Solvent exfoliated graphene-epoxy coating for evaluation as a functional paint. b. Suitably identified commercial graphene nanoplatelets extrusion compounded in PMMA for evaluation as strain sensor in a bulk composite. c. Study on the effect of sonication time on the concentration of graphene produced in the solvent exfoliation process in order to produce larger flakes for better stress transfer and hence sensor sensitivity. 21

22 Chapter 2: LITERATURE REVIEW 22

23 Chapter 2 Literature review 2. LITERATURE REVIEW The interest in the field of graphene goes back to the pioneering theoretical work of Wallace in 1947 which was used for many years as a model system for all sp 2 carbon. [9, 10] Since then several publications have reported on the study of graphitic thin layers of carbon and its synthesis, [11-14], but little emphasis was given to the field until 2004 when Andre Geim and Konstantin Novoselov prepared high quality monolayer graphene and explored its electronic properties. [1] The preparation of monolayer graphene not only proved the existence of infinite 2D materials (when held on a substrate) which for several decades were believed not to exist, but also the continuity and crystallinity in its structure.[15] 2.1. Structure and morphology of graphene Graphene is a two-dimensional carbon allotrope comprising of an one-atom-thick planar sheet of sp 2 bonded carbon atoms that are densely packed as a honeycomb crystal lattice. Graphene is considered to be the basic building block of other allotropic sp 2 forms; it can be stacked to form 3D graphite, rolled to form 1D nanotubes or wrapped around to form 0D fullerenes (Figure 2.1). Figure 2-1: Graphene is a 2D building material for other allotropes of carbon. It can be stacked into 3D graphite, rolled into 1D carbon nanotubes or wrapped to 0D fullerenes [16, 17] A monolayer of graphene comprises of a single layer of carbon atoms with hexagonal packing. Bi- and few- layer graphene has two and 3 to 10 such 2D layers of 23

24 Chapter 2 Literature review carbon atoms. The carbon-carbon (sp 2 ) bond length in graphene is approximately nm and the thickness of the sheet is 0.35 nm.[18-21] The unit cell of monolayer graphene has two atoms in it (Figure 2.2a) and in the case of bilayer has four atoms per unit cell. The stacking of these individual layers of graphene, in the direction perpendicular to the plane in an AB (or Bernal) stacking arrangement, is shown in Figure 2.2b. The carbon atoms in the graphene sheets are bonded together with in-plane σ bonds, and each carbon atom in the lattice has out-of-plane π bonds (Figure 2.2c). The σ bonds are the strongest and keep the hexagonal structure strong whilst the π bonds control the interaction between different graphene layers. Figure 2-2: a) Top view of real space unit cell of monolayer graphene [22] b) AB (Bernal) Stacking arrangement in graphene, c) Schematic of the in-plane σ bonds and the π bonds perpendicular to the plane of sheet [23] This π orbital leads to the delocalised network of electrons that are responsible for the extraordinary electronic properties of graphene. Intrinsic ripples have been observed in the graphene and been modelled by Monte Carlo simulations. [24] The scanning electron microscopic corrugations (Figure 2.3) were estimated to have lateral dimension of about 8 to 10 nm and a height displacement of about 0.7 to 1 nm. Apart from these ripples, graphene in real 3D space can have other defects including topological defects (e.g., pentagons, heptagons and their combinations), vacancies, adatoms, edges/cracks and adsorbed impurities.[25] Figure 2-3: 'Rippled' graphene from Monte Carlo simulation. Red arrows are ~8nm long [24, 25] 24

25 Chapter 2 Literature review 2.2. Synthesis of graphene The earliest reported synthesis of graphene was in 1975 when Lang et al. thermally decomposed carbon which resulted in the formation of mono- and multi-layered graphite on a single crystal Pt substrate. [26] Several methods for synthesising atomically thin layers of graphite have been reported since then. [11-14, 27, 28] In particular, Geim and his co-workers isolated graphene at The University of Manchester by the mechanical exfoliation technique using cellophane tape (Figure 1.1). [1] This process was the first repeatable synthesis of graphene through exfoliation. Some of the various methods reported are discussed below Mechanical exfoliation Mechanical exfoliation is used to produce thin samples from layered compounds and one such study was carried out by Ruoff s group on graphite in 1999 using an atomic force microscope (AFM) tip and plasma etching, which resulted in 200 nm (~600 layers) thick slabs of graphite. This approach was then adapted to produce very thin flakes of graphite (10 nm, ~30 layers) on Si/SiO 2 substrates. [29] Micromechanical cleavage Geim et al. isolated graphene from HOPG using a simple but most efficient technique called micro mechanical cleavage of graphite, [1] in which exfoliation is achieved by repeated peeling of layers from the bulk graphite. A thin piece of graphite (HOPG) flake is kept inbetween the Scotch tape and the graphite is repeatedly peeled so that the flake gets thinner and thinner to obtain an optically transparent flake (Figure 2.4). This tape is pressed down ultimately against a substrate (e.g. silicon/silicon dioxide Si/SiO 2 wafer). The flakes present on the tape are thicker than one layer. However, due to the van der Waals attraction between the substrate and the graphene sheet, a single graphene sheet is delaminated on removing or lifting the tape away. Then the single atomic sheet of graphene can then be identified under an optical microscope. [30] 25

26 Chapter 2 Literature review Figure 2-4: Mechanical exfoliation of graphene by the Scotch tape method. Inset is the Si/SiO 2 substrate [31] Recent advances have resulted in the production of large graphene flakes up to ~10 µm by controlled exfoliation [32], and mm-sized mono- to multi-layer graphene produced by bonding graphite to borosilicate glass followed by exfoliation. [33, 34] The advantage of this method is that it produces high quality graphene but the disadvantages are neither high-throughput nor high-yield Solution phase exfoliation The solution-based exfoliation of graphene was first demonstrated by Ruoff s group in 2006 in producing single-layer graphene. In this study Hummer s method was used to modify graphite into graphite oxide, which in turn exfoliate into graphene oxide sheets upon ultrasonication. (It should be noted that this method dates back to the 19 th Century so graphene oxide has probably been produced for over 150 years.) This exfoliation is due to the strength of interactions between water and the oxygen containing functionalities (epoxide and hydroxyl) which are introduced during oxidation. This hydrophilicity leads the water to intercalate between the sheets and individual graphene oxide sheets being dispersed. These sheets can then be reduced by thermal annealing or chemical agents (eg. hydrazine hydrate) to reduce the number of functionalities in the flakes. [34-36] The product obtained, though, is not fully reduced and is found to have some oxygen left in the structure which is evident from its Raman spectra (large D band at 1350 cm -1 ). Graphene obtained from this method is termed as partially-reduced graphene oxide sheets. 26

27 Chapter 2 Literature review Hernandez et al. demonstrated the production of graphene in solution phase without modifying the graphite by using N-methyl-pyrrolidone (NMP) as a solvent. The yield of monolayer was about ~1 wt. %. The energy required to exfoliate graphite into monolayer graphene was countered by the graphene-solvent interaction, with the solvent having a similar surface energy to that of graphene. [37] The process was further improved to give a monolayer yield of 4 wt. % by increasing the sonication time from 30 min up to 343 hours. The TEM analysis of their suspensions obtained from 100 hours sonication showed over 90% of the flakes had < 5 layers. The effect of sonication time was reflected both in the graphene concentration and the size of the flakes. As the sonication time increases, graphene concentration increases whereas the flake size (mean length and mean width) decreases. As centrifuge speed increases, concentration of graphene as well as flake size decreases. [38] Despite its advantage of high monolayer yield, the disadvantage of this process is the high cost of the solvent and the high boiling point of the solvent that makes the following graphene deposition difficult. A similar kind of process was used to produce single- to few-layer graphene sheets by dispersing graphite in surfactant-water solution. Loyta et al. used sodium dodecyl benzene sulfonate (SDBS) as the surfactant resulting in graphene concentration of ~ mg/ml (Figure 2.5). [39] Sodium cholate was also used as surfactant, resulting in monolayer yield of about 1.1 wt. % and a graphene concentration varying from ~ 0.2 mg/ml to ~ 0.03 mg/ml depending on the centrifuge speed used. [40] The graphene monolayers are stabilized against aggregation by a relatively large potential barrier caused by the Coulombic repulsion between the surfactant-coated sheets. The dispersions are reasonably stable with larger flakes precipitating out over more than 6 weeks. Figure 2-5: TEM image of flakes produced from SDBS/water solution, a) monolayer, b) bilayer and c) trilayer[40] 27

28 Chapter 2 Literature review Green and Hersam produced graphene suspensions by using sodium cholate as the surfactant, and they to isolated the resultant graphene sheets of controlled thickness using density gradient ultracentrifugation (DGU). The exfoliation of graphite yielded a dispersion of monolayer graphene and graphite with few layers of graphene which have different buoyant density. DGU separation of such mixtures produces graphene sheets with mean thicknesses that increase as a function of their buoyant density (Figure 2.6). [31, 41] Figure 2-6: a) Photograph of a centrifuge tube following the first iteration of density gradient ultracentrifugation (DGU). The suspension was diluted by a factor of 40 to ensure that graphene bands could be resolved in the photograph. b&c) Representative AFM images of images o of graphene deposited using fractions f4(b) and f16(c) on Si/SiO 2. d) Height profile of regions marked in blue (b) and in red (c)[41] Low boiling solvents like isopropanol and chloroform have also been used to produce graphene from graphite using ultrasonication. Graphene concentration obtained from isopropanol was about ~ 0.5 mg/ml and from chloroform was ~ 0.2 mg/ml. But the yield of monolayers was much less when compared to NMP and surfactant/water mixtures. Most of the flakes obtained had 7 layers or less. [42] A solvent exchange method was used to produce stable graphene suspensions in low boiling solvents like ethanol. Initially graphene suspension was produced from NMP and then filtered. This filtered cake was redispersed in ethanol by sonication and centrifugation until a homogenous graphene suspension in ethanol was achieved. The concentration of graphene obtained by this method was ~0.04 mg/ml. [43] 28

29 Chapter 2 Literature review Production of very high concentration of graphene up to 5.33 mg/ml was reported when using ionic liquids like 1-hexyl-3-methylimidazolium hexafluorophosphate (HMIM) as the solvent for the ultrasonication process. [44] The process of liquid phase exfoliation shows good promise to synthesis large scale graphene and some recent modifications in the process has helped even in producing large size. However, liquid phase exfoliation has its own disadvantages like removing the solvent, high boiling points of the solvents and oxidation and reduction of the basal plane of graphene can generate defects Epitaxial growth Chemical vapour deposition (CVD) This method depends on the carbon-saturation of a transition metal upon exposure to a hydrocarbon gas at a high temperature. The first reported CVD synthesis of few layer graphene was in 2006 from camphor on a nickel (Ni) substrate at 850 o C in an argon atmosphere. [45] In another approach, 1-2 nm thick graphene sheets were grown on a Ni substrate using H 2 and CH 4 (92:8 ratio) as the precursor gas. [46] Later, Yu et al. reported three to four layers of graphene on Ni foils using CH 4, H 2, and argon (Ar) (0.15:1:2 ratio) at 1000 o C for 20 min. [47] Attempts have been made to produce graphene in large scale. A new method of substrate-free few-layered graphene involved using CH 4 and Ar (1:4 volume ratio) and MgO-supported Co as the catalyst. 50 mg of graphene could be produced from 500 mg of catalyst powder. [48] In a recent development, highly crystalline few-layer graphene was grown directly on a 1 cm 2 area of polycrystalline Ni substrate by carefully controlling the gas ratio, growth time and temperature. [49] In a further development of this process, graphene was synthesized on a 1 cm 2 Copper (Cu) foil by a thermal CVD process. The graphene thus obtained was of high quality and uniformity. However, it was found that graphene growth on Cu substrates was self-limiting, probably due to the low solubility of C in Cu. [50] Very recently research from Rice University has produced pristine high-quality monolayer graphene by CVD growth on Ni foils using girl scout cookies, chocolate, grass, dog waste, PMMA and roach (cockroach leg) as the carbon source (Figure 2.7). Graphene 29

30 Chapter 2 Literature review produced from this method was high quality with very few defects and was ~97% transparent, confirmed by Raman spectroscopy and UV-vis spectroscopy. Figure 2-7: A) Experimental setup for the growth of graphene by CVD technique. B) a. Roach leg on top of Cu foil, b. Roach leg under vacuum, c. Residue from roach leg after annealing[51] The CVD growth was performed at 1050 o C for 15 min under low pressure with H 2 /Ar gas flow. With this technique, different kind of solid materials that contain carbon can potentially be used without purification as the feed stock to produce high quality graphene. Furthermore, through this method low-valued food and negative-valued solid waste are successfully transformed into high-value graphene which brings new solutions for recycling of carbon from impure sources. [51] Thermal decomposition of SiC The process of thermal decomposition of silicon carbide (SiC) to graphene on the surface plane of single crystal of 6H-SiC is highly popular. In this method the silicon carbide is reduced at very high temperatures to form graphene sheets. Silicon generally desorbs around 1000 o C in ultrahigh vacuum, which leaves behind small islands of graphitic carbon. This method was proposed by de Heer and his group. The graphene that grows epitaxially on the surface typically had one to three graphene layers with the number of layers dependent upon the decomposition temperature. [34] 30

31 Chapter 2 Literature review Graphene prepared by this method shows different physical properties when compared to the mechanically cleaved graphene. This is because in epitaxially grown graphene the interfacial effects are highly dependent on the both the silicon carbide substrate and other growth parameters. [30, 52] Other methods Un-zipping of Carbon nanotubes (CNT) A very recent technique of graphene synthesis was multiwalled carbon nanotubes (MWCNT) as the starting material. This process is known as un-zipping of CNTs. MWCNT was opened up longitudinally using the intercalation of Li and ammonia, followed by exfoliation in acid and abrupt heating (Figure 2.8). [53] The resulting product, among the nanoribbons and partially-opened MWCNT, contained graphene sheets. In another approach, graphene nanoribbons were produced by plasma etching of MWCNTs, partially embedded in polymer film. [54] The etching process basically opened up the MWCNT to form graphene. In a recent method, MWCNTs were un-zipped by a multi-step chemical treatment, including exfoliation by concentrated H 2 SO 4, KMnO 4 and H 2 O 2, stepwise oxidation using KMnO 4 and finally reduction in NH 4 OH and hydrazine monohydrate. [55] Figure 2-8: TEM image of partially un-zipped MWCNT structure opened by Li intercalation [53] These new process routes of un-zipping MWCNTs to produce graphene create the possibility of synthesizing graphene in a substrate-free manner. [30] 31

32 Chapter 2 Literature review 2.3. Properties of graphene Mechanical properties Only after three years of isolation of graphene from graphite did research start exploring the mechanical properties of graphene. One of the first works reported was in 2007 by Frank et al. who measuring the spring constant of multilayer (<5 layers) graphene from force-displacement measurements of an AFM tip on a strip of graphene suspended over trenches (Figure 2.9a). The Young s modulus of the graphene was measured at 0.5 TPa and the effective spring constant to range from 1-5 N/m and was found to scale with the dimensions of the suspended region and the layer thickness (from 5 to 30 layers). [56] Circular membranes of few-layer graphene were also characterized by force-volume measurements using AFM by Poot et al. [57] The most significant results were reported by Lee et al. in mid The elastic properties and intrinsic breaking strength of a free-standing monolayer graphene was measured by nanoindentation in AFM. In this method, the graphite flakes were mechanically deposited on the silicon substrate which had circular cavities. The graphene sheets were held together by the van der Waals force. The AFM tip was placed on the centre of free standing graphene sheets (Figure 2.9b &c). The experimental results showed that the breaking strength was 42 Nm -1 which represents the intrinsic strength of a defect free monolayer sheet. The Young s modulus was calculated to be around 1 ± 0.1 TPa, the third-order elastic stiffness -2.0 TPa and the intrinsic strength 130 GPa. These values make the graphene the strongest material ever measured. [2, 30] Figure 2-9: a) A SEM micrograph of a graphene sheet suspended above a trench etched in SiO 2 wafer [56] b) SEM image of graphene monolayer deposited on the substrate with 1µm and 1.5µm circular arrays.[2, 30] c) Schematic representation of the nanoindentation on a suspended graphene membrane[2] 32

33 Chapter 2 Literature review Electronic properties It is primarily because of graphene s unusual and extraordinary electronic properties that it has become an exciting field of research after its isolation in Monolayer graphene is a zero band gap semiconductor or semi-metal in which the highest occupied molecular orbital (HOMO) touches the lowest unoccupied molecular orbital (LUMO) at a single Dirac point (Figure 2.10). Figure 2-10: Electronic band structure of single-layer graphene (HOMO and LUMO meets at the Dirac point [15] Graphene is a high electron mobility material at room temperature, a reported value of 15,000 cm 2 /Vs. [15] The mobilities of holes and electrons are found experimentally to be the same and are independent of temperature from 10 K to 100 K which implies that the dominant scattering mechanism is defect scattering. [58] Scattering by the acoustic phonons of graphene places intrinsic limits on the room temperature mobility to cm 2 /Vs at a carrier density of cm 2. [59] Fuhrer et al. [60] reports the room temperature resistivity of graphene sheet as around 10 6 Ω cm, which is 35% less than the resistivity of silver. The charge carriers in the graphene sheet behave as massless Dirac fermions. [30] Some of the recent experiments have demonstrated the influence of chemical dopants on the carrier mobility in graphene. Schedin et al. doped graphene with various gaseous species, and found the initial undoped state of a graphene structure can be recovered by gently heating the graphene in a vacuum. [61] No change in the carrier mobility was observed even with a dopant concentration in excess of cm -2. Chen et al. doped graphene with potassium in ultrahigh vacuum at low temperature and they found that potassium ions act as expected for charged impurities in graphene and can reduce the 33

34 Chapter 2 Literature review mobility by about 20-fold. This mobility reduction is reversible on heating the graphene to remove the potassium. [30, 59] Recently Zhang et al. demonstrated that the band gap of graphene can be tuned from 0 ev to 0.25 ev by applying voltage at room temperature to a dual-gate bilayer graphene field-effect transistor.[30, 62] Thermal properties The first work on the thermal properties of graphene was reported in 2008 by Balandin et al. using the non-contact technique of Raman spectroscopy. The Raman laser was focused on monolayer graphene suspended on a Si/SiO 2 wafer. The thermal conductivity of graphene was calculated using the dependence of the Raman G peak frequency to the exciting laser power and found to be in the range of ~4.84±0.44 x 10 3 to 5.30±0.48 x 10 3 W/mK. [63] This value of thermal conductivity indicates superior performance to carbon nanotubes in terms of heat conduction Raman spectroscopy Raman spectroscopy works on the principle of Raman effect or Raman scattering. The Raman effect was discovered by Sir. C. V. Raman is 1928 in liquids. [64] When monochromatic light is scattered from a molecule, most of the scattered photons have the same energy (frequency) and wavelength as the incident photons, which is called elastic scattering. But very few photons (1 in 10 million) have a different frequency from the incident photons, which is inelastic scattering. This effect is called Raman effect. These inelastic scatterings are characteristic of the bonds in the molecule allowing identification of samples. There are two types of Raman scattering; Stokes and anti-stokes. Interaction of a photon with a molecule may give rise to three types of scattering (Figure 2.11): The incident and the emitted photons have the same energy Rayleigh (elastic) scattering The molecule absorbs energy and thus the emitted photon has a lower energy than the incident photon Stokes scattering 34

35 Chapter 2 Literature review The molecule loses energy and the emitted photon has a higher energy than the incident photon anti-stokes scattering Figure 2-11: The energy-level diagram of Rayleigh, Stokes Raman and anti-stokes Raman scattering Raman spectroscopy is a form of vibrational spectroscopy, and these energy differences arise from the molecular vibrations. Moreover, the intensity of Raman bands depends on the number of molecules present in different vibrational states. According to the Boltzmann distribution (which gives number of molecules present at different states of energy level) more molecules will occupy the lowest energy state. Thus, Stokes scattering is more intense than anti-stokes scattering, with the latter increasing with increasing temperature. The vibrational information is specific to the chemical bonds and symmetry of molecules, which act as a fingerprint of that particular molecule. The Raman spectrum is typically plotted as the intensity of the scattered radiation as a function of the wavenumber difference from the incident radiation. Raman spectroscopy is widely used to characterise all sp 2 carbon forms from three to zero dimensions, including 3D graphite, 2D graphene, 1D carbon nanotubes and 0D fullerenes. 35

36 Chapter 2 Literature review 2.5. Raman spectroscopy of graphene The Raman spectrum of carbon-based materials has a set of characteristic common features in the region between 800 and 3000 cm -1. The three major peaks of graphene in its Raman spectrum (Figure 2.12) are found around 1580 cm -1 (G band), 1350 cm -1 (D band) and 2700 cm -1 (G band). The G band is associated with the doubly degenerated E 2g phonon at the Brillouin zone centre. The D band is a disorder-induced band and is a finger print peak of graphite material which has a significant concentration of defects. The G band is the second order of the D band and is caused by double resonant Raman scattering with two phonon emissions. [65, 66] Figure 2-12: a) Raman spectrum from monolayer graphene edge showing prominent features [67] b) Raman spectrum of monolayer graphene and bulk graphite [65] The shape of the Raman G band of graphene is sensitive to the number of layers and it is free from defect and has no D peak. Ferrari et al. in 2006 first reported the sensitivity of the G band to the number of layers in the flake. As the number of layers increase, the shape, width and the position of the G band varies due to the corresponding change in the electron bands. A characteristic feature of monolayer graphene in the Raman spectrum is that the G band is approximately 3-4 times intense than the G band, and vice versa as the number of layers increases. This feature is often used to identify monolayer graphene. Monolayer graphene have a single sharp G band and bilayer graphene has a much broader band. When fitted with a Lorentzian function, the bilayer curve splits into four components (Figure 2.13). [65] 36

37 Chapter 2 Literature review Figure 2-13: Raman spectrum of different layers of graphene a) in 514 nm b) in 633 nm c) Splitting of 4 components of G band of bilayer graphene [65] Figure 2.13 show the sensitivity of the G band to number of layers. The peak position changes slightly with the laser used. The four components of bilayer graphene are 2D 1B, 2D 1A, 2D 2A and 2D 2B of which 2D 1A and 2D 2A have relatively higher intensities than the other two. As number of layers increases, the relative intensities of the 2D 1 peaks decrease significantly resulting in peaks similar to bulk graphite. Raman spectra of more than 5 layers become hardly distinguishable from bulk graphite. [65] Grap et al. showed that the G band is also sensitive to the number of layers. As the number of layers increases, the intensity of the G band increases. In addition, the peak position of the G band shifts to lower wavenumber as the number of layers increases (Figure 2.14). [68] Figure 2-14: Change in G peak position with number of layers. Inset show the intensity difference between HOPG (upper peak), bilayer (middle peak) and monolayer (lower peak)[68] 37

38 Chapter 2 Literature review 2.6. Strain sensitivity of the Raman bands of graphene It is well known that the G and G band of carbon based materials like carbon fibres [3, 69-71] and carbon nanotubes [5, 72] are sensitive to strain. On application of uniaxial tension, the bands shift to a lower wavenumber. These shift rates can be used to measure the stress transfer efficiency between a matrix and a carbon-based reinforcement. This sensitivity of Raman bands could be used to measure strains, and therefore act as a sensitive strain sensor. Not surprisingly, even the G and G band of graphene are sensitive to an applied strain and exhibit considerable shifts Strain sensitivity of the G band There are two types of strain that could be introduced to graphene through a substrate: uniaxial strain and biaxial strain. Zhen et al. reported early studies on the behaviour of the Raman scattering from graphene under uniaxial strain. The graphene prepared by micromechanical cleavage, was deposited on the top of a thin PET substrate. The strain was induced in graphene by stretching the PET film in one direction. [73] The strain in the graphene was assumed to be the same as the change in length of the substrate. A slope of ± 0.7 and ± 0.6 cm -1 /% strain was observed for monolayer and trilayer graphene respectively (Figure 2.15). Figure 2-15: Shift of G band of mono- and tri-layer graphene on PET substrate on uniaxial strain [73] 38

39 Chapter 2 Literature review Mohiuddin et al. in 2009 reported a detailed study on the sensitivity of the G and G bands of graphene to straining. Uniaxial strain was applied in very small steps of 0.05% up to ~1.3% to mechanically-exfoliated monolayer graphene on a thin PET film which was coated with a photoresist. The G band (doubly degenerated E 2g optic mode) splits into two components (G + and G - ) at a sufficiently high uniaxial strain (0.5 %). One of the components polarized along the strain and the other polarized perpendicular to the strain. Both components shift to a lower wavenumber and the splitting increases as strain increases (Figure 2.16). The shift rates of the two components were ~ cm -1 /% strain (G+) and ~ cm -1 /% strain. [74] Figure 2-16: a) Position of G band with respect to strain (note the splitting) b) Shift rates of two components of G band [74] It has to be noted that there are two different conclusions drawn from studies by Zhen et al.[73] and Mohiuddin et al.[74]. The first concerns the splitting of the G band, which does not occur in the study of Zhen et al., although the strain reached 0.8% (far higher than the 0.5% of Mohiuddin et al.). The next concerns the value of the shift rate, ± 0.7 cm -1 /% strain for Zhen et al. whereas for Mohiuddin et al. it is ~ cm -1 /% and ~ cm -1 /% strain. These significant differences cannot be attributed to the inaccuracy of the apparatus. Moreover, the two sets of results are in accordance with the theoretical calculations. Mohiuddin et al. performed the calculations with a more accurate Grüneisen parameter (γ, represents the rate of phonon mode softening {hardening} during tensile {compressive} strain and determines the thermo mechanical properties) (1.99), 39

40 Chapter 2 Literature review whereas Zhen et al. used γ (1.24) which is for CNT. Also G band splitting is a common phenomenon observed for carbon nanomaterials under strain, such as nanotubes. [75, 76] The reasons which led to the different results for Zhen et al. may be: The graphene is not a monolayer but a multilayer, because strain is more effectively applied to a thinner graphene sheet. The actual strain in graphene is less than the calculated value. Probably the strain was calculated directly from the PET substrate instead of directly from graphene. If the strain of the substrate is not the same as the the graphene, the resulting shift is smaller than the real value. These may also explain why the G band did not split even at higher strain levels of 0.78% for Zhen et al. [66] Strain sensitivity of the G band The Strain sensitivity of the G band of monolayer graphene was studied by Zhen et al. [73] and Ting et al. [77] on mechanically-cleaved graphene on PET film. Strain was applied to the samples by stretching the PET film by Zhen et al. and by bending the PET film by Ting et al. In both the cases the G band shifted to a lower wavenumber. This shift of the G band could be attributed to the elongation of the carbon-carbon bonds, which weakens the bonds and therefore lowers their vibrational frequency. Shift rates recorded by different groups are given in Table 2.1. The possible reasons for such significant differences in the observed shift rates were discussed in the previous section. Table 2-1: G' band shift rates observed by various groups Sample Shift rate (cm -1 /% strain) Reference Monolayer on PET film ± 0.8 [73] Trilayer on PET film ± 1.1 Monolayer on PET film ~ -8 [77] Monolayer on PET thin film ~ -64 [74] Georgia et al. in 2009 reported initial work on the peak shift of graphene embedded within a polymer (composites). Mechanically-cleaved graphene was placed inbetween a 40

41 Chapter 2 Literature review sandwich structure of PMMA (5 mm)/ SU-8 (~200 nm)/ graphene/ PMMA (~100nm). A high shift rate of cm -1 /% strain was observed, which indicates higher strainsensitivity of graphene and efficient stress transfer between the polymer matrix and the reinforcement. A value cm -1 /% strain was also observed for sample in which graphene was deposited on top of PMMA without any top coat over the graphene. [78] Gong et al. did a tensile deformational study on similar samples by using a four-point bending technique and observed a shift rate of ~-60 cm -1 /% strain (Figure 2.17). [7] Figure 2-17: Shift of G' band peak position as a function of strain [7] This study [7] also showed that in graphene composites the interfacial stresstransfer will only occurs through van der Waals bonding across an atomically smooth surface. It was concluded that larger graphene flakes 0f >3 µm, which was the critical length, with respect to shear-lag theory, were needed to obtain efficient reinforcement in the composites. All the above studies show the high sensitivity of the graphene bands with strain, in which the G band is much more sensitive than the G band. This high strain-sensitivity of graphene makes graphene-polymer composites a potential candidate for mechanical strain sensors, a concept with is explored in this thesis Existing strain-gauge technology Strain gauges Strain gauges are sensors that are used to measure tensile or compressive strain in a particular part or over an area results from an applied load. The measurement of strain 41

42 Chapter 2 Literature review using strain gauges always assumes that the strain on the object under investigation is transferred equally to the gauge. There are different types of strain gauges that are available and can be used to measure strains either at a single point or over an area Mechanical strain gauges One of the earliest strain gauges used to measure strains were mechanical strain gauges, also called extensometers, which can be used to measure static or gradually varying load conditions. These extensometers have two knife edges which are clamped firmly to the test component by means of clamping spring at specific gauge length. When the specimen is strained the knife edges undergoes displacement, this displacement is amplified by the mechanical linkages and the strain is displaced on the calibrated scale. The Berry strain gauge, Huggenbeger extensometer, and Johansson extensometer are typical examples of mechanical strain gauges. [79] Resistance-based strain gauges Resistance based strain gauges are the most widely used strain gauges. They work on the principle that when a conductor is stretched its length will increase and the area of cross section will decrease and this will result in a change in measured resistance. This change in resistance is then converted to an absolute voltage by a Wheatstone bridge. The change in resistance per unit strain is defined as the Gauge factor. Gauge factor indicates the sensitivity of the strain gauge. GF = Equation 2.1 Where, GF is the Gauge factor, ΔR/R the Fractional change in resistance, ΔL/L the Fractional change in length and ε is strain. Equation 2.1 is valid for low strain levels. Commercially available foil strain gauges can measure up to 5% maximum strain level.[80] Different types of resistance-based strain gauges are explained below: Wire strain gauge The wire strain gauge operation is dependent on the well-known fact that the resistance of a wire increases as the wire is stretched. The first wire-type strain gauge was developed in 1938 by Arthur Claude Ruge Father of strain gauge. [81] This type of strain gauge has a 42

43 Chapter 2 Literature review grid of wire (resistor) bonded directly to the strained surface by a thin layer of epoxy resin. The wires used to make strain gauges vary in diameter from 12 to 25 µm. The most common alloys for preparing the strain gauge wire are copper-nickel alloys and nickelchromium alloys. [82] A simplified model of a wire strain gauge is shown in Figure Figure 2-18: Design of wire strain gauge [81] Foil strain gauge Advancements in the field of printed circuits by Paul Eisler, lead to the development of foil strain gauges. Foil strain gauges are currently by far the most common strain gauges. A film of copper-nickel alloy several micrometres thick is laminated onto a thin, flexible substrate (eg, polyimide), which act as a carrier material. The foil is then etched to produce a grid of the desired pattern (Figure 2.19). Figure 2-19: The characteristic design of a strain gage with etched metal foil measuring grid [81] This gauge is bonded to the test specimen with a thin insulating layer of epoxy resin or cyanoacrylate adhesive. Variations in the electrical resistance of the grid are measured as an indication of strain. Overall resistance of a typical foil strain gauge may range from around 30 to as much as 3000 ohms. [83] Gauges with a resistance of few hundred ohms are more commonly used. The advantages of the foil strain gauges are: Can be produced with uniform characteristics Light weight and durable Multiple strain gauges can be made on same substrate Inexpensive Measured strain value has an accuracy better than 0.001% strain 43

44 Chapter 2 Literature review Small physical size and low mass Foil strain gauges are also only moderately affected by temperature changes. The gauge factor of foil based strain gauges is typically between 2 to 5. The bonded resistance strain gauges can be used to measure both static and dynamic strains. [30, 81-83] Semiconductor strain gauges A semiconductor-based resistance strain gauge works on the principle of the semiconductor piezoresistive effect, discovered by C.S. Smith in 1954 in germanium and silicon. Semiconductor strain gauges are similar to foil strain gauges in construction. The measuring element consist of a strip of a few tenth of a millimetre wide and a few hundredth of a millimetre thick which is fixed to an insulating foil carrier and is provided with connecting leads (Figure 2.20). Diode effects are prevented by using a thin gold wire as the connection between the semiconductor element and the connecting strips. [81] Figure 2-20: Schematic representation of semiconductor strain gauges [81] Semiconductor strain gauges have gauge factors of more than times greater and sensitivities more than 100 times greater than wire or foil based strain gauges, and are used in transducer manufacture. Despite their high sensitivity, this type of strain gauge is not used widely in stress analysis because: Non-linear characteristics in the resistance-to-strain relationship More expensive than wire- foil- based strain gauges Temperature effects on the sensitivity Handling is difficult due to the brittle nature of the semiconductor [81] Optical strain gauges Fibre Bragg grating (FBG) Fibre optic sensors are a group of sensors that use optical fibres as the sensing element ( intrinsic sensors ). This type of sensor can be used to measure strain, 44

45 Chapter 2 Literature review temperature or pressure by modifying the fibre. Intensity, phase, polarization, wavelength and transit time of the light in the fibre are the sensitive factors that can be followed for measurements. Fibre optic sensors have been developed to measure temperature and strain simultaneously with high accuracy by using Fibre Bragg gratings (FBG). [84] A FBG is a distributed Bragg reflector in an optical fibre. In an FBG, the refractive index of the core change radically along a short length in the optical fibre. When a light with many wavelengths is passed through a FBG, one particular wavelength is in-phase with the grating period (Ʌ) and this wavelength is reflected back to the input end of the fibre leaving all other wavelengths to transmit to the other end. Figure 2-21: Schematic representation of a) Fibre Bragg grating structure b) Refractive index profile c) Spectral response and d) FBG s reflected power as a function of wavelength [84] The spectral response of a FBG is shown in Figure 2.21c. The response of the reflected spectrum is sensitive to strain and the temperature with the wavelength-shift of ~10 pm/ o K and ~1 pm/microstrain. [85] When a FBG is subjected to tensile strain the spectra shifts to higher wavelength and to lower wavelength on compressive strain. Advantages of FBG s are: They can be directly integrated into composites, or can be fixed directly or as patches on to the surface of the test specimen for measuring strains Can measure very high strains (>10000 µm/m) and can be used in highly stressed composite constructions Long term stability is very high Special grades can be used at very high temperatures (>700 o C) They can be used for simultaneous measurements, such as strain and temperature, by integrating two FBG s in to the optical fibre 45

46 Chapter 2 Literature review However, a low value of gauge factor ( , differs with fibre types) and high cost are some of the disadvantages of FBG. [84, 86] Other types Piezoelectric paint sensors Wire strain, foil strain and FBG strain gauges can be used in measuring strain at one-point. Whereas, paint/coating-based sensors can be used for wide-area sensing. The strain obtained from paint-based sensors averages the strain over the sensor area. This type of sensor works on the principle of piezoelectricity, which is the accumulation of electrical charges in solid materials in response to an applied mechanical stress/strain. The sensitivity of the sensor can be estimated as the electrical displacement generated by the sensor relative to the strain experienced by the substrate surface, under the hypothesis that the electrical charge generated is proportional to the sum of the principal strains.[87] Piezoelectric ceramic-polymer composites (Lead zirconate titanate-acrylic polymer) have shown their sensing ability under uniaxial strain over a wide area (steel and plastic substrates) in dynamic range of microstrains [88] and with biaxial strain in the range of ± 200 microstrains [87] Luminescent strain-sensitive coating This patented strain-sensitive coating, is a mixture of light-emitting dyes and epoxy resin developed by researchers from University of Florida, USA. The Coating can be applied on any shape substrate by spray painting resulting in a glossy reddish paint. On exposure to ultraviolet light, the light transmits through the epoxy resin and its polarization changes in direct proportion to the amount of strain the part is experiencing. The lightemitting chemicals then transmit these polarization differences to a computer or digital camera resulting in a 3D graphic map of the stress/strain levels sustained over the entire coating. [89, 90] 46

47 Chapter 3: METHODOLOGY 47

48 Chapter 3 Methodology 3. METHODOLOGY 3.1. Evaluation of sources of graphene The graphene used in this project was either prepared in-house or used from the asreceived materials bought from commercial suppliers. Time restraints meant that only commercial source of graphite could be used to make composite samples. (Future work would compare all the sources) Therefore, the grades and specifications of the different commercial products were evaluated using both the supplier s data and Raman spectroscopy and SEM measurements conducted by the author. Note that whilst some of the materials are sold as graphene, the thickness of the flakes mean that there are really nanosized graphite particles. XG Sciences xgnp Graphene nanoplatelets are nanoparticles consisting of short stacks of graphene sheets having platelet shape. These nanoplatelets were produced by ex-situ exfoliation of graphite (graphite is intercalated with alkalis, acids, salts etc. and expanded) followed by pulverisation resulting in graphene nanoplatelets. The different grades of xgnp are given in the Table 3.1 Table 3-1: Different grades of xgnp Grade X-Y Dimensions (µ) Price Remarks xgnp M $ 379/kg xgnp M $ 379/kg xgnp M 5 5 $ 399/kg Average thickness 6-8 nm Typical surface area m 2 /g Different grades were classified according to the average size of the flakes as shown in the table. Average thickness of 6-8 nm corresponds to layers of graphene. It has > 99.5% carbon and the rest are impurities like oxygen groups, hydroxyl groups etc. which are present at the edges of the particles. The individual particles are of high quality and have a clean surface consisting of sp 2 carbon, which is reflected in its Raman spectrum 48

49 Chapter 3 Methodology (Figure 3.2). Other miscellaneous impurities are less than 0.5 wt%. SEM images of xgnp shows the morphology of the platelets (Figure 3.1). [91] Figure 3-1: SEM image of xgnp M25 Angstron materials Two commercial grades of Angstron materials were purchased. Both grades comprised of a very light and fluffy powder with density 2.2 g/cm 3 but the platelets in the two grades varied with thickness and lateral dimensions. The graphene was produced by a reduction method which is evident from its considerable impurity content and its Raman spectra. The specification of the products is given in Table 3.2. Table 3-2: Different grades of Angstron materials Grade X-Y dimension (µ) Thickness (nm) Price Remarks N002-PDR < 5 < 1 $ 200/g N P < 14 < 10 $ 1/g ~ 3 layers Surface area m 2 /g ~ 30 layers Surface area 110 m 2 /g The oxygen content is estimated to be approximately 2.1 % in N002-PDR and 1.5 % in N P. Hydrogen, nitrogen and ash content in both grades are around 1%, 0.5% and 1.5%, respectively. This is evident from its Raman spectra, showing a larger and more intense D band compared to the G band (Figure 3.2). [92] 49

50 Chapter 3 Methodology Nanocs Nanocs produce single layer graphene flakes via reduction. The flakes range in size from 0.5 to 5 µm. Over 80% of the flakes are single atomic layers. A Raman spectrum of this grade of graphene is shown in Figure 2.2. The spectrum has a large, broad D and G band without any G band. It costs around $180/50 mg. Durham Graphene Sciences Ltd. Durham graphene is from Durham Graphene Sciences Ltd., UK. It is produced from epitaxial method of preparing graphene (CVD process). It costs 500/g. The Raman spectrum of this graphene show a very broad D and G band with no G band, which indicates the presence of a large amount of impurities from the production process (Figure 3.2) D G G Figure 3-2: Raman spectra of commercial grades of graphene. Raman spectra of Sigma Aldrich graphite flakes are also shown for comparison purpose 50

51 Chapter 3 Methodology Selection of material for composite manufacture XG sciences xgnp was chosen for composite preparation, as it has a well-defined G band which is more sensitive to strain than the G band. Branwell graphite Steps were taken in this project to study the effect of sonication time on the concentration of graphene prepared from a solvent exfoliation method. For this study graphite was obtained from Branwell Graphite Ltd, UK. Different grades of graphite are: Graphite RFL 99.5, Graphite 2369 and Graphite Production of graphene Micromechanical cleavage In this method, a small piece of HOPG was repeatedly peeled inbetween Scotch tape resulting in peeling of graphene layers from the stacks. The tape was then pressed against a Si/SiO 2 substrate. Acetone was used to dissolve the tape, leaving the graphene attached to the substrate. The substrate was then heated on a hot plate to remove any residual solvent. One final peel was done with a fresh tape to remove any large graphite flakes, leaving a population of graphene layers (monolayer to multilayer) on top of the substrate. [1] Solvent exfoliation method Hernandez et al. demonstrated the production of pristine graphene directly from graphite powders, unlike earlier works that produced graphene from reduction of graphene oxide. [37] The particular technique was followed in this project using NMP as a solvent. The raw material used to prepare graphene was either XG Sciences graphene nanoplatelets or Branwell graphite powders. A schematic representation of the whole process is shown in Figure 3.3. Sonication experiments were carried out initially at the School of Physics and Astronomy, University of Manchester, until new instruments were purchased at the School of Materials. XG sciences graphene (0.1 mg/ml) was sonicated in NMP in a Dawe Soniclean (Automatic 300/150 W) ultrasonic bath for different sonication times and then centrifuged at different speeds and time in a Sigma Centrifuge (max rpm). Branwell 51

52 Chapter 3 Methodology graphite (5 mg/ml) was sonicated in NMP in a new instrument at the School of Materials (Elmasonic P, max power-800w). The power level used was 30% with sweep mode ON. The actual power output measured from the increase in temperature of a known mass of water was 32 W at 30% power. The resulting solution was left overnight at room temperature to allow the deposition of any unstable aggregates of graphite powder. The solution was then transferred to a centrifugal tube and centrifuged in a Thermos Scientific Sorvall LEGEND RT+ centrifuge (carbon fibre sample holders) for 45 min at 500 rpm. Experimental conditions were adopted from reference [38]. Figure 3-3: Schematic representation of solvent phase exfoliation of graphene The decanted suspension obtained was mixed with epoxy resin to prepare composite films, which are explained in the next section. The samples for Raman spectroscopy were prepared by placing drops of the suspension on a Si/SiO 2 wafer and then heating it up to 100 o C for five min to remove the NMP. 52

53 Chapter 3 Methodology Preparation of graphene composites and films Composites with micromechanically-cleaved graphene Graphene used in this process was produced by the Scotch tape method as described in section The composite specimen used was a sandwich layer of PMMA/SU-8/graphene/PMMA. The base of the sandwich is a 5 mm thick PMMA beam. A 300 nm thick layer of SU-8 photoresist was spin-coated on top of this PMMA beam and crosslinked. This SU-8 photoresist ensures optimal visible contrast for graphene identification. The mechanically exfoliated graphene was then carefully deposited on the surface of the SU-8 resin. A thin (50 nm) layer of PMMA was then spin-coated on the top of the beam so that the graphene remains visible inbetween these sandwich layers of two polymers. The schematic representation of the specimen is shown in the Figure 3.4. [30] Figure 3-4: (a) Schematic representation of the graphene composite (b) Cross-sectional view of the composite [30] This particular sample was prepared by Prof. Konstantin Novoselov from The School of Physics and Astronomy, University of Manchester. [1, 30] Mechanically cleaved graphene was also used to prepare sandwich composites samples on a steel substrate. These samples were prepared to investigate the strain sensitivity of graphene on engineering materials like spring steel. Spring steel was chosen because of its high yield strain (0.3%) compared to other steels. The results are explained in the following chapter. Two different sandwich samples were prepared: SU- 8/graphene/SU-8/steel and SU-8/graphene/steel (Figure 3.5). 53

54 Chapter 3 Methodology Figure 3-5: Schematic view of sandwich composites on steel substrates a) SU-8/graphene/SU-8/steel b) SU-8/graphene/steel The spring steel (0.45 mm thick) was cleaned in a sonic bath with acetone followed by isopropanol followed by distilled water. This step was repeated several times to remove impurities like oil from the steel surface. A thin layer of SU-8 was spin coated (2500 rpm) on top of the steel and cross-linked by exposing it to a UV lamp for 10 min. Mechanically cleaved graphene from the Scotch tape was pressed against the top surface of the SU-8. The graphene adheres well to the SU-8/steel substrate. One final peel was done using a fresh tape. Over this, a thin layer of SU-8 was coated and cross-linked to ensure the graphene is visible and fixed in the position without any slippage. The sample shown in Figure 3.5b did not have any SU-8 on top of the steel. Instead, graphene was directly deposited on top of the steel to observe any differences in stress transfer between the sandwich composite specimens Composites from solvent-phase exfoliated graphene Several model composites were prepared using solvent phase exfoliated graphene in order to investigate their stress transfer ability under Raman deformation. Graphene produced from the solvent-phase exfoliation method was spray coated on top of a 100 µm thick PET sheet. The initial concentration of graphite was 0.1 mg/ml, which was sonicated in NMP for 24 hours followed by 10 min centrifuge at rpm. The NMP was removed by heating the PET sheet on a hot plate to 100 o C for five min. This sample was prepared by Prof. Novoselov s group at the School of Physics and Astronomy. A thin layer of PMMA was coated on top of the strip to observe any difference in the shift rate and interfacial stress transfer (Figure 3.6). 54

55 Chapter 3 Methodology Figure 3-6: Schematic representation of graphene/pet strip For another set of samples, a few drops of graphene suspension produced from XG graphite (grade M25, 24 hours sonication, centrifuge rpm) were deposited on a spring steel substrate and heated (100 o C for 5 min) to remove the NMP. In one specimen, SU-8 was coated on top of the suspension to study any difference in the Raman shift behaviour on deformation (Figure 3.7). Figure 3-7: Schematic representation of solvent phase exfoliated graphene on steel substrate Solvent-phase exfoliated graphene was also used to produce thin composite films with epoxy resin as the matrix g of epoxy resin (Araldite LY5052 matrix) was added to 1.16 g of graphene-nmp suspension (xgnp M25, 24 hours sonication, centrifuged for 10 min at rpm. The mixture was sonicated in a Dawe Soniclean (300/150 W) for 15 min to facilitate homogenous mixing. This mixture was left overnight at 100 o C under vacuum to remove NMP. Hardener (Araldite HY5052) was added to this mixture in a ratio of 100:38 by weight (0.4 g), and the mixture was once again sonicated to 15 min for achieve a homogenous dispersion. A thin coat of this mixture was applied on the middle of two spring steel substrates (0.45 mm thick), with the aid of a razor blade. To study the effect of residual stresses on deformation, one sample was cured at room temperature for seven days (cold cure), whereas the other sample was cured at 100 o C for two hours in an oven (hot cure) (Figure 3.8). 55

56 Chapter 3 Methodology Figure 3-8: Schematic representation of graphene/epoxy composite film on steel Composites from extrusion In this project, steps were taken to prepare bulk composites from commercial graphene nanoplatelets. A masterbatch was produced using a Thermo Scientific MiniLab II Haake Rheomix CTW5, with co-rotating screws. The speed of the screws was 50 rpm and the temperature was set at 200 o C. xgnp M 25 grade was chosen for this particular experiment because of its larger dimensions (~25 µm flakes). 5 wt% of xgnp was mixed with PMMA pellets in a plastic bag before feeding into the compounder. Small shots of 3 g were fed into the mini compounder at once. It was allowed to mix homogenously by cycling it inside the compounder for about 5 minutes. Then the resultant compounded mixture was collected and cut into small pieces. This process was repeated until approximately 25 g of material was collected. In order to achieve a homogenous mix within different shots, the collected pellets were further compounded two times. Figure 3-9: Picture of bulk composite of xgnp/pmma prepared from mini compounder 25 g of compounded pellets was then compression moulded in a rectangular mould at 200 o C for 10 min to produce a 3 mm thick bulk graphene/pmma composite (Figure 3.9). This composite was prepared to investigate the peak shift rates during deformation using Raman spectroscopy. 56

57 Chapter 3 Methodology 3.3. Raman Spectroscopy Raman spectroscopy was the primary characterization technique used in this project. The experimental setup for the experiments conducted using Raman spectroscopy is explained below Experimental Setup Raman spectra were collected using a Reinshaw 2000 spectrometer connected to an optical microscope (Figure 3.10). The laser used was a low-power He-Ne laser with an excitation energy of 1.96 ev and corresponding wavelength of 633 nm. Figure 3-10: Schematic diagram of Raman spectroscopy system [30] The light beam passes through the spatial filters, which contains a pair of lenses and a 10 µm pinhole. The light then passes through holographic notch filters which reflect the light towards the microscope. During the deformation tests, the polarisation of the laser beam was always parallel to the tensile axis. The sample was placed on the adjustable microscope stage and the laser beam was focused on the sample surface with a 50x microscope objective lens. The diameter of the laser spot was about 2 µm. The laser power on the sample was < 1 mw/cm 2. The scattered light from the sample was collected in the backscattering geometry and passed through the holographic notch filters, which filters out the Rayleigh scattered light. The diffraction grating and triangular mirror separates the scattered Raman lines spatially as function of wavelength. After this process, a charge coupled device (CCD) camera collects the light. 57

58 Chapter 3 Methodology Charges were generated inside the CCD chip due to the scattered photons, which were collected with electrodes. The measured current was proportional to the number of photons detected and in turn to the intensity of scattered light. The intensity of the light was collected as a function of its wavelength (its difference with the incident light wavelength gives the Raman wave number in cm -1 ), processed and stored on the computer. The absolute resolution of the spectrometer was 1 cm -1 and relative resolution (e.g. peak shift with respect to a spectrum collected on the same machine) was 0.1 cm -1. The peak positions in the Raman spectra collected were curve-fitted using a Lorentzian equation: Equation 3.1 where ω is the Raman wave number (cm -1 ), ω c at the centre, I o is the maximum intensity and Δω the full width at half maximum of the peak. [30, 93, 94] The Raman system was calibrated using a standard silicon sample which has a characteristic peak at 520 cm -1. Any change in peak position is corrected before the start of an experiment Deformation under Raman Spectroscopy Composites with PMMA as the base substrate were subjected to a tensile deformation in a four-point bending rig, whereas for sandwich composites with spring steel as the base substrate the deformation was imposed in a two-point bending (G-clamp) rig. A schematic representation of four-point bending rig is shown in Figure Actual setup of G-clamp is shown in Figure Figure 3-11: Schematic representation of four-point bending test: (a) On loading, (b) On unloading[30] 58

59 Chapter 3 Methodology Figure 3-12: Actual setup of modified G-clamp for deformation of spring steel based substrate In the four-point bending rig, the force was applied to the sample by loading the bar stepwise. In tensile deformation, the two loading bars were under the specimen to support the specimen and push it upwards in order to generate the tension on the top surface. The fixed bars were located at the top of the sample in order to fix its position. Whereas with the G-clamp, the ends of the sample were fixed in slits made in the faces of the clamp (Figure 3.12). Deformation was induced by rotating the screw stepwise allowing the sample to bend between the two fixed points. An electrical resistance strain gauge was attached to the sample s surface using cyanoacrylate adhesive in order to monitor the deformation. (Figure 3.12) In order to investigate the interfacial stress transfer between the matrix and the reinforcement and the stability of the shift rate, the model PMMA/graphene/SU-8/PMMA sandwich composite was subjected to cyclic deformation. The loading and unloading of the sample was carried out up to an increasing maximum strain level of 0.1, 0.2, 0.3, 0.4, and 0.5 % (Figure 3.13). When the maximum strain was achieved for each cycle, the sample was relaxed to 0 % strain. For the model composites on spring steel, a single cycle of loading and unloading was carried out. The maximum strain in most of the spring-steel samples was restricted to 0.3 % because of the limitations of the bending rig (G-clamp). The deformation was gradually increased in intervals of 0.02 ± 0.005% strain, until the maximum strain was achieved. 59

60 Chapter 3 Methodology Figure 3-13: Cyclic deformational sequence of graphene composites All the spectra were obtained using 25 % laser power (He-Ne laser, 633 nm) with an exposure time of 20 seconds and 5 accumulations in extended mode. It should be pointed out that all the spectra were collected from the same area of the sample in each test to detect the in situ deformation behaviour of the composites. The presented values are an average of three readings at the same point. As mentioned in the literature review, this project focuses primarily on the G band shift, rather than the G band shift, because of its higher strain sensitivity. 60

61 Chapter 4: RESULTS AND DISCUSSION 61

62 Chapter 4 Results and discussion 4. RESULTS AND DISCUSSION 4.1. Model single flake composite: Performance on a plastic substrate Characterization of graphene composite It is well known that graphene can be characterized by optical microscopy and Raman spectroscopy. The sandwich composite prepared was carefully scanned under an optical microscope in order to identify the transparent layers which could be potentially monolayer graphene. Raman spectroscopy was then used to confirm the identified flakes as being monolayer or not. Figure 4.1 show the optical image of large identified monolayer flake on which a deformation experiment was carried out. Figure 4-1: Optical image of monolayer graphene from sandwich composite For all the images shown throughout this thesis, the Raman spectra were collected from the position shown by the crosshairs in the image. It should be noted that (not shown) as number of layers increases, the transparency of layers decreases under white light. Raman spectroscopy can be used to identify the number of layers by observing the shape of the G band in a spectrum. [65] Spectra for different layers of graphene in the composite are shown in Figure 4.2. One significant feature of monolayer graphene is that the intensity of the G band is much higher (nearly 4 times) compared to the G band. Also the shape of the G band is narrow for a monolayer. The spectrum shown at the bottom of 62

63 Chapter 4 Results and discussion figure 4.2 belongs to a monolayer. As the number of layers increases the shape of the G band is seen to broaden. For bilayer, the G band splits into four components when fitted with a Lorentzian function as discussed in literature [65] and shown by the spectrum in red in Figure 4.2 Figure 4-2: Raman spectra of different layers observed in prepared composite (spectra offset for clarity) The absence of a D band at 1350 cm -1 indicates that the identified graphene is high quality without any defects. The other two peaks at ~ 1700 cm -1 and ~1450 cm -1 belongs to the PMMA and SU-8 in the composite Cyclic deformation Cyclic deformation gives an indication of the long term stability of graphene as a strain sensor in applications such as monitoring a bridge. Data from the author s previous work [30] is used in this work for comparison purposes. In the earlier experiments, cyclic deformation of graphene was done up to the same maximum strain level, whereas in this project, cyclic deformation was carried out with increased maximum strain intervals. 63

64 Chapter 4 Results and discussion Accuracy measurements The main aim of the project is to investigate the use of graphene as a strain sensor, in which the accuracy of the strain sensor plays a major role in determining its selection over other sensors. In this section, the accuracy of graphene in measuring strain during a cyclic deformation is discussed. As mentioned earlier, the model single-flake graphene composite was subjected to a cyclic deformation with increasing maximum strain intervals. The peak position of the G band with respect to the deformation cycle is shown in Figure 4.3. Figure 4-3: Position of the G' band with respect to deformation cycle. Two red lines corresponds to the data points taken from this position for calculating the accuracy of the strain measurements From the Figure 4.3, it can be clearly seen that the graphene G band position has followed the deformation of the PMMA beam. As the strain increases the position of G band shifts to a lower wave number which is attributed to the elongation of carbon-carbon bonds, which weakens the bonds and therefore lowers their vibrational frequency. It is very important to know the accuracy of the system in measuring strain during a cyclic deformation. The accuracy of this system can be calculated by considering the difference in the G band peak position compared to its reference value. This can best be explained with an example, the G peak position at 0.1% strain (reference strain) during the first deformation cycle is cm -1 (reference position). During the next deformation cycle, the position of the G band at 0.1% strain is cm -1 (new position). The data that can be extracted is the difference in the wavenumber (new position reference position). A set of values at 64

65 Chapter 4 Results and discussion all the deformation cycles can be collected and the accuracy of the system can be calculated from the standard deviation of these values divided by the shift rate obtained. The main assumption made only for the accuracy calculation was that all the deformation cycles show a standard shift rate of -50 cm -1 /%strain which is equivalent to having an effective modulus of 1TPa. In this method, the G band peak position at 0.1 % and 0.3 % strain during the first deformation cycle were chosen as the reference points, which were cm -1 and cm -1, respectively. It was hoped that in further deformation cycles the G band peak position at the chosen strain level to be the same compared to the first deformation cycle. So that even an unknown strain could be calculated just with the information of the G band peak position. But in this case, at each deformation cycle, even during loading and unloading, the G band peak position at the chosen strain level was not constant. Instead, a standard deviation of ± 3.5 cm -1 for 0.1 % strain and ± 3.6 cm -1 for 0.3 % strain was observed with respect to its reference G band peak position at the first deformation cycle. This calculates to an accuracy of strain measurement of ~ ± 720 microstrains (± % strain) at both the strain levels. As mentioned before, data from reference [30] was used for comparison purposes, in which the graphene composite was subjected to the same maximum strain level at each deformation cycle. The peak position of the G band with respect to the deformation cycle is shown in Figure 4.4. Figure 4-4: Variation of the G' band with respect to deformation cycle. Data obtained from reference [30] 65

66 Chapter 4 Results and discussion The accuracy calculation was applied to the set of data obtained from Figure 4.4. The accuracy of the system in measuring strain was ~ ± 230 microstrains (± % strain), which is better than the result (± 720 microstrains) obtained from this study, wherein the composite was deformed with increased maximum strain intervals. Reference calculations were done at 0.3 % strain (red line in Figure 4.4). The accuracy of graphene as a strain sensor measured from these cyclicdeformational experiments has values ± 720 microstrains and ± 230 microstrains, which is very much lower when compared to resistance-based strain gauges (accuracy of < ± 10 microstrains) and wide area strain sensors (accuracy of ~ ± 100 microstrains) [82] The accuracy obtained in this study is less when measuring low strain levels (< 1 %). The interface between the matrix and the graphene and the strength of the interface becomes a critical factor for measuring higher strains. Interfacial stress transfer is discussed in the following section. The reason for this low value of accuracy may be due to compression of graphene which it undergoes at the end of each cycle and also strain hardening behaviour which is discussed in section Deformational behaviour In this section, emphasis is given to the deformational and interfacial behaviour of the graphene composites. The variation of the G band position with respect to applied strain, or in other words the response of the graphene when the composite was repeatedly deformed, is shown in Figures 4.5 and

67 Chapter 4 Results and discussion Figure 4-5: Shift of the Raman G' band for cyclic deformation from % maximum loading strains Figure 4-6: Shift of Raman G' band for cyclic deformation for 0.4 and 0.5% maximum loading strains 67

68 Chapter 4 Results and discussion Figures 4.5 and 4.6 shows the variation of the G band position during cyclic deformation ranging from 0.1 to 0.5 % strain. It can be noticed that for the initial two cycles, 0.1 and 0.2 % strain, the G band shifts to lower wavenumber approximately linearly with the strain and the loading curves almost overlap with the unloading curve suggesting that full stress transfer has occurred between the matrix and graphene. Therefore, the graphene is exhibiting roughly elastic deformation in the initial cycle of deformation. Similar results were observed with the SWCNT composites subjected to cyclic deformation with increased maximum strain intervals. [6] In which, SWCNT exhibited elastic deformation up to 0.4 % strain. [6] Graphene exhibited elastic deformation up to 0.3 % strain level in the first deformation cycle when subjected to cyclic deformation with same maximum strain intervals. As the maximum loading strain increased, the strain-shift relationship of the G band becomes non-linear for the loading and the unloading cycle. This non-linearity between the loading and the unloading cycle leads to the development of a hysteresis loop in further deformation cycles. The hysteresis area was found to increase in size with the level of maximum loading strain. This could be attributed to permanent interfacial damage occurring between the matrix and the graphene. The interface was damaged during the loading cycle, which in turn changes the stress transfer efficiency of the unloading cycle leading to the hysteresis loop. From Figure 4.6, it could be noted a slight upshift of the hysteresis loop, which indicates that the interface has been subjected to irreversible damage in the composite during the loading cycles and therefore two curves do not follow each other. Similar behaviour was observed in cyclic deformational studies of SWCNT composites and graphene composites. [6, 30] The interfacial damage that occurred during the deformation cycles resulted in a change of the G band peak position at the same strain levels in different cycles, resulting in a low value of accuracy in the strain measurements. This suggests that the interface plays a major role when using graphene composites as strain sensor. Based on shear-lag theory, it has been shown that the stress transfer between the matrix and reinforcement happens through the shear stress at the matrix-graphene interface. The maximum shear stress calculated for similar model composite was around ~2.3 MPa at 0.4 % strain which is relatively low value compared to carbon fibre composites with shear stresses in the range of ~20-40 MPa. [7] This suggest that good stress transfer ability would occur with strong 68

69 Chapter 4 Results and discussion interfaces with a relatively low possibility of permanent interfacial damage at low levels of strain, which in turn would increase the accuracy of the strain measurements. Furthermore, it should be noted that the G band peak position at 0 % strain in 2 nd deformation cycle is at a higher wavenumber than the 1 st deformation cycle, and similar behaviour was observed for the following cycles (small rectangular boxes shown in Figures 4.5 and 4.6) This behaviour can be related to the slippage of graphene during tensile deformation and on unloading it is subjected to in-plane compression, resulting in a shift to higher wavenumber. This behaviour is similar to the observations during deformation experiments by Gong et al. and Raju et al. [7, 30] Residual stress and energy dissipation The hysteresis loops observed in Figure 4.5 and 4.6, represents the energy dissipated in the composite system during the deformation process. The amount of energy dissipated per unit volume can be quantified by estimating the area of the loop. As an example, the derived stress vs applied strain for the deformation cycle with 0.3 % as the maximum strain is plotted in the Figure 4.7. These were then plotted for all deformation cycles and the values obtained for residual stress and energy dissipated are given in Table 4.1 and Figure 4.8. The observed G band shift can be converted into a stress by using the universal stress shift rate of -5 cm -1 /GPa of the G band for carbon fibres. [4, 6] Residual stresses developed at the end of each cycle can be calculated by dividing the difference between the band position at the end of cycle with the band position of the initial cycle by -5 cm -1 /GPa. Figure 4.8 shows the residual stress at end of each cycle along with the energy dissipated at each cycle. 69

70 Chapter 4 Results and discussion Figure 4-7: Derived stress-strain curve for graphene composites (3rd cycle) As mentioned earlier the initial two cycles of deformation with 0.1 and 0.2 % as the maximum strain exhibited elastic behaviour, therefore the loading data were fitted with linear functions. For the other deformation cycles, the loading and unloading data were fitted with polynomial curves. The loading and unloading energies were calculated by integrating the equations obtained within their maximum and minimum strain limits, and the energy dissipated in each cycle is just the difference between unloading and loading energies and are tabulated in Table 4.1. Table 4-1: Residual stress and energy dissipation in the graphene composite during cyclic deformation Cycle Loading energy (MJ m -3 ) Unloading energy (MJ m -3 ) Energy dissipated (MJ m -3 ) Residual stress developed at end of cycle (GPa)

71 Chapter 4 Results and discussion Figure 4-8: Residual stress and energy dissipated during cyclic deformation of graphene composite It can be observed from Figure 4.8 that residual stress and the energy dissipated in the graphene composite starts increasing as the maximum strain level deformation cycle continues. This is because at the end of each deformation cycle (after unloading) the graphene is undergoing compression in the composite resulting in the development of residual stress. The residual stress increases at the end of each deformation cycle as the maximum strain % increases. It can be seen that no significant energy is dissipated during the initial two cycles with 0.1% and 0.2% as maximum strain, which supports the earlier discussion of complete stress transfer at these strain levels. This is due to the good bonding between the graphene and the matrix. Interface damage starts occurring from the 3 rd cycle of deformation which is indicated by the increase in energy dissipation. Energy dissipation is high for the deformation cycle with 0.5 % as the maximum strain, indicating that most of the damage has occurred by this deformation cycle. When the maximum strain was increased from 0.5 % to 0.55 %, interfacial failure was observed (Figure 4.3) indicated by a sudden uplift of the G band peak position. 71

72 Chapter 4 Results and discussion Strain hardening behaviour The shift rates obtained from the plots of Raman G band wavenumber shift vs strain % can be converted into graphene s effective modulus. It is assumed for all the micromechanically-cleaved graphene composites that the graphene is oriented in the direction of load and so they are in-plane with the load. The G band undergoes a shift in excess of -50 cm -1 /% strain which is consistent with it having a Young s modulus of over 1 TPa. [4] The shift rate can be obtained from the slope of the curve in the plot and this calibration of -50 cm -1 /% strain = 1 TPa was used to calculate an effective Young s modulus during all the cycles of deformation. The calculated effective modulus values are given in the Table 4.2. Table 4-2: Calculated effective Young's modulus of graphene in composites during cyclic deformation Strain (%) Cycle Shift rate (cm -1 /%strain) Effective modulus (TPa) Loading Unloading Loading Unloading Loading Unloading Loading Unloading Loading Unloading Loading Unloading

73 Chapter 4 Results and discussion It can be observed from Table 4.2 that the effective modulus values for graphene in the composites are in the range of > 1 TPa after the first cycle of deformation, which makes graphene a very good reinforcement in the composite. It should also be noted that as the maximum strain % in each deformation cycle increases the effective modulus increases, which is shown in Figure 4.9. Figure 4-9: Plot between effective modulus of graphene vs deformation cycle The effective modulus increases up to a very high value of 1.69 TPa during the final loading cycle (0.5 %). This behaviour could be related to strain hardening of the graphene during loading, i.e., the carbon-carbon bonds in graphene exhibit higher resistance for deformation leading in each cycle to this behaviour. This strain hardening behaviour, as mentioned earlier, affects the accuracy of the strain measurements by showing a slight shift in peak position compared to its reference position. Surprisingly, strain hardening was found even during unloading by showing higher effective modulus than the previous cycle. Recently, dynamic strain hardening in CNT/PDMS (poly dimethyl siloxane) nanocomposites was observed by Brent et al. when the composite was subjected 73

74 Chapter 4 Results and discussion to cyclic compressive loading. [95] It was shown that there was a permanent increase in the stiffness of the composite that continued until the dynamic stress applied was removed and which resumed when it was reapplied. This study of CNT nanocomposites supports the results obtained in Figure 4.9. The effective modulus obtained in SWCNT composites in reference [30] during cyclic deformation showed an average value of 0.35 TPa with a shift rate of ~ -18cm -1 /% strain. But graphene shows an average value of ~1 TPa (~ -50 cm -1 /% strain). On observing these shift rates, it could be said that the shift rate that is approximately 3 times higher for the graphene compared to SWCNT composites, meaning that a graphene-based strain sensor is 3 times more sensitive than a nanotube-based one. The increase in the modulus over the cyclic deformation can be used in some applications where it is desired to achieve good reinforcement. The reason behind the observed strain hardening requires further detailed study and this is discussed in the further work section. One possible explanation is that whilst applying a tensile deformation to graphene, the ripples in the graphene could straighten out, meaning that more of the bonds in the graphene flake are in the direction of the load, hence increasing its effective modulus. (This is analogous to the difference in reinforcement obtained from straight versus wavy fibres) However, it was found that it was not trivial to check this theory, as the ripples are very small (~1 nm amplitude), [24] and were not visible even in the dark field image of the graphene surface (Figure 4.10). Thus the reason behind this strain hardening behaviour is not yet understood. 74

75 Chapter 4 Results and discussion Figure 4-10: Bright (a & c) and dark (b & d) field images of monolayer graphene under X50 and X100. The atomic ripples were not visible from the dark field images. The bright spots in the dark field images belong to the imperfections on the sample substrate 4.2. Model single flake composite: Performance on steel The strain sensitivity of graphene on an engineering material, spring steel, was investigated to assess its use in applications such as functional paint coatings. The mechanically-cleaved graphene was deposited on the steel substrate inbetween thin layers of SU-8 to form a sandwich composite. A second sample was also prepared (Figure 4.11), where the graphene was deposited directly on the steel substrate without any bottom layer of SU-8 to establish the importance of the epoxy to form a good interface. Optical microscopy and Raman spectroscopy were used to characterize the flakes. 75

76 Chapter 4 Results and discussion Figure 4-11: Schematic view of sandwich composites on steel substrates a) SU-8/graphene/SU-8/steel b) SU-8/graphene/steel In the sample SU-8/graphene/SU-8/steel, because of the contrast from the SU-8 and the steel, it was difficult to find a large monolayer flake for strain measurements. Nevertheless, strain measurements were obtained from a relatively large (~15 x 10 µm) bilayer graphene flake identified by optical microscopy (Figure 4.12) and Raman spectroscopy (Figure 4.13). The sample was deformed with a loading cycle up to a maximum strain of 0.3% and unloading cycle back to 0% strain in the two-point bending rig shown in Figure Figure 4-12: Optical image of bilayer graphene identified in SU-8/bilayer graphene/su-8/steel sandwich composite sample 76

77 Chapter 4 Results and discussion Figure 4-13: Raman spectra obtained from the identified bilayer graphene. Inset show the four Lorentzian peaks fitted in G band Figure 4.14 and 4.15 show the response of the G and G bands during the deformation cycles. A G band shift rate of ~ cm -1 /% strain and ~ cm -1 /% strain was observed during loading and unloading, respectively, for this sample. This shift in the Raman G band confirms that bilayer graphene can be used to measure strain on an engineering material. (It should be noted that the shift rate for a bilayer flake in a pure epoxy composite has just been established as ~ - 53 cm -1 /% strain, highlighting the poor interface in this composite [Gong et al., University of Manchester, private communication]. 77

78 Chapter 4 Results and discussion Figure 4-14: Shift of Raman G' band with strain in SU-8/bilayer graphene/su-8/steel sandwich composite Figure 4-15: Shift of Raman G band with strain in SU-8/bilayer graphene/su-8/steel sandwich composite 78

79 Chapter 4 Results and discussion A G band shift rate of ~ cm -1 /% strain and ~ cm -1 /% strain was observed in the bilayer graphene during loading and unloading, respectively. The shift rate in the G band is much lower than the G band shift rate, supporting the focus on the G band shift for the active strain measurements. From the composite point of view, the shift in the G band can be converted to its effective modulus by the universal calibration of ~ -50 cm -1 /% strain = 1 TPa. The obtained G band shift corresponds to a graphene effective modulus of ~ 280 GPa during loading and ~ 360 GPa during unloading. It should also be noted in Figure 4.14, the slope of the unloading curve is higher than the loading curve which is similar to observations in the literature. [7, 74, 96] However, surprisingly the position of the G band after unloading is approximately the same as before loading, indicating that graphene did not undergo any slippage in the composite during loading and no in-plane compression occurred during unloading. In the second sample, where the graphene was deposited directly over the steel without any bottom layer of SU-8, it was possible to identify only a multilayer in the composite sample. The absence of mono-, bi- or few-layer graphene in the sample indicates poor adhesion of graphene on to steel substrates without any surface modification. The optical image of the identified multilayer (~ 8 x 10 µm) is shown in Figure 4.16 and the Raman spectra in Figure Figure 4-16: Optical image of multilayer graphene identified in SU-8/graphene/steel sandwich composite sample 79

80 Chapter 4 Results and discussion Figure 4-17: Raman spectra obtained from the identified multilayer graphene in SU-8/graphite/steel sandwich composite sample Figure 4.17 shows the Raman spectrum of the obtained multilayer from the SU- 8/graphite/steel composite. A peak identified at ~1600 cm -1, belongs to the SU-8 spin coated over the graphene. The absence of a D peak at ~1350 cm -1 (not shown) indicates that the graphene has no defects and that the obtained spectrum was not from any edge of the graphene flake. Shift rates obtained from the G band and G band on deformation are shown in Figures 4.18 and 4.19, respectively. 80

81 Chapter 4 Results and discussion Figure 4-18: Shift of Raman G' band with strain in SU-8/multilayer graphene/steel sandwich composit Figure 4-19: Shift of Raman G band with strain in SU-8/multilayer graphene/steel sandwich composite 81

82 Chapter 4 Results and discussion A G band shift rate of ~ cm -1 /% strain on loading and of ~ cm -1 /% strain during unloading was observed for the multilayer graphene, which corresponds to ~100 GPa and ~120 GPa effective moduli. Very low G band shift rates of ~ cm -1 /% strain and ~ cm -1 /% strain were observed for the loading and unloading cycles, respectively Multilayer graphene and its composite coatings: Performance on steel Graphene prepared by the solvent exfoliation method in the form of a suspension, was used to make composite coatings for the investigation of strain sensitivity on steel substrates. A variety of samples were prepared from the graphene suspension and deformation was carried out for each sample: Graphene/PET strip PMMA(thin coat)/graphene/pet strip Graphene/steel SU-8/graphene/steel Graphene-epoxy coating (hot cure)/steel Graphene-epoxy coating (cold cure)/steel Figure 4.20 shows Raman spectra obtained from the graphene/pet strip and PMMA/graphene/PET strip. The shape of the G band is very broad and doesn t have a shoulder, indicating that the spectra are obtained from different populations of graphene in the sample. A large and broad D band, apart from a small overlapping PET peak, indicates that spectra were obtained from the edges of the graphene. These particular strips were deformed using a fibre deformation rig, where a thin strip was glued at both the end to the rig and strain was imposed by pulling apart both edges of the rig. Strain was manually calculated from the change in length of the strip. 82

83 Chapter 4 Results and discussion Figure 4-20: Raman spectra obtained from graphene/pet strip and PMMA/graphene/PET strip. Spectra offset for clarity Both the samples were strained up to a maximum strain level of 0.5 % and relaxed to 0 % strain. The response of the G and G bands for the deformation of samples are shown in the Figure 4.21 and Surprisingly, the G band and G band in both samples showed no considerable downshifts on deformation. One possible reason for this behaviour is explained at the end of this section. Figure 4-21: Shift of Raman G' band (a) and G band (b) on deformation of the graphene/pet strip 83

84 Chapter 4 Results and discussion Figure 4-22: Shift of Raman G' band (a) and G band (b) on deformation of the PMMA/graphene/PET strip The graphene suspension was deposited directly on the steel substrate, which was then heated to 100 o C for 5 min to remove the solvent NMP. A thin layer of SU-8 was deposited on top of graphene on one specimen to observe any change in shift rate. Figure 4.23 shows the Raman spectra obtained from these two samples. The presence of residual NMP, even after annealing was observed from the Raman spectra. The response of the G and G bands for these two samples during deformation in two-point bending is shown in Figures 4.24 and Even in these samples no considerable downshifts were obtained, rather a scattered shift was observed within a 2 cm -1 wavenumber difference. Figure 4-23: Raman spectra obtained from graphene/steel and SU-8/graphene/steel samples. Spectra offset for clarity 84

85 Chapter 4 Results and discussion Figure 4-24: Shift of Raman G' band (a) and G band (b) on deformation of graphene/steel sample Figure 4-25: Shift of Raman G' band (a) and G band (b) on deformation of SU-8/graphene/steel sample The graphene suspension was mixed with epoxy resin as explained in the experimental section, and a very thin layer of this composite mixture was coated on top of the steel substrate. Two sets of such samples were prepared, one cold cured and the other hot cured, to study the effect of residual stress on the shift rates. Figure 4.26 shows the Raman spectra obtained from both the samples. The deformational response of the Raman G and G bands is shown in Figures 4.27 and Even in these samples, no considerable downshift was obtained during deformation. Instead a scattered shift was obtained. 85

86 Chapter 4 Results and discussion Figure 4-26: Raman spectra obtained from graphene-epoxy thin coat on steel substrate. Spectra offset for clarity Figure 4-27: Shift of Raman G' band (a) and G band (b) on deformation in graphene-epoxy hot cured sample Figure 4-28: Shift of Raman G' band (a) and G band (b) on deformation in graphene-epoxy cold cured sample 86

87 Chapter 4 Results and discussion The reason for this behaviour in these samples is because the sizes of the produced graphene flakes were very small. The conditions used to prepare these samples were 24 hours of sonication and centrifugation at rpm for 10 min. The graphene flakes produced under these conditions were in the order of nm in their lateral dimensions or even smaller. TEM characterization of XG graphite nanoplatelets sonicated at these conditions was carried out. Few drops of prepared sample were dropped on a 400 mesh size holey carbon grid (Agar Scientific) for TEM imaging. Figure 4.29 show some TEM images of XG graphite nanoplatelets. Figure 4-29: TEM images of graphene produced from sonication of XG graphite nanoplatelets It can be clearly seen from Figure 4.29 that the graphene flakes are of the order of < 0.5 µm in their lateral dimensions. Figure 4.29d is an expanded image of the area highlighted in Figure29b which shows the flake morphology. The XG graphite nanoplatelets are produced by expanding graphite, and the sonication of these nanoplatelets has not resulted in complete exfoliation to individual flakes, rather irregular exfoliation in the solvent phase was observed (Figure 4.29d). 87

88 Chapter 4 Results and discussion Gong et al. has demonstrated using the shear-lag theory that in order to achieve reinforcement, which in turn gives the shift in the Raman G band, a graphene flake has be longer than 3 µm. [7] From the TEM measurements, it is clear that the flake sizes in the produced sample are much lower than the critical length of 3 µm and therefore gives little reinforcement and in turn only small shifts in the Raman bands Bulk graphene composite XG graphite nanoplatelets (grade M25) were mixed with PMMA in a mini extruder to form bulk composites, as explained in section The composite was characterized by optical microscopy and Raman spectroscopy. Optical microscopy of the bulk composites in different areas showed a fairly homogenous dispersion of XG graphite nanoplatelets in the PMMA matrix (Figure 4.30). Raman spectra from different regions of the sample show different G band peak positions for the graphite, suggesting that various degree of exfoliation happened during extrusion. A relatively large flake (~ 10 x 8 µm) at the composite surface was chosen for the deformation experiment. The response of the G band over the loading and unloading deformation cycle is shown in Figure Figure 4-30: Optical image of5 wt. % xgnp M25 graphite nanoplatelets/pmma bulk composite showing fairly homogenous dispersion of graphene platelets within the matrix. The flake chosen for the deformation experiment is highlighted 88

89 Chapter 4 Results and discussion Figure 4-31: Shift of Raman G' band of xgnp M25 graphite nanoplatelets/pmma bulk composite over deformation The G band shows a downward shift of cm -1 /%strain during loading and cm -1 /% strain during unloading. The universal calibration of ~ -50 cm -1 /%strain = 1 TPa could be applied to calculate the effective modulus in the composites by using a Krenchel orientation distribution factor, ŋ o. [97, 98] A range of modulus values could be obtained for loading and unloading cycles. For loading, ~ 81.6 GPa (unidirectional orientation parallel to applied force, ŋ o -1), ~ 40.8 GPa (biaxial orientation, ŋ o -1/2), ~ 30.6 GPa (random orientation, ŋ o -3/8). Similarly for unloading, the effective modulus ranges from ~ 61.8 GPa (ŋ o -1) to ~ 30.9 GPa (ŋ o -1/2) to ~23.2 GPa (ŋ o -3/8). The shift observed in the bulk composite is very low when compared to the shift obtained from the flake graphene in the sandwich composites. However, it is comparable to the shift obtained from multilayer graphene on the steel substrate (-4.75 cm -1 /% strain). This initial example shows that graphite is sensitive to strain in the extruded composites and could be used as a strain sensor as well as a reinforcement for the composites. 89

90 Chapter 4 Results and discussion 4.5. Study on solution-phase exfoliated graphene Effect of sonication time Poor reinforcement and low peak shift rates were found in the composites prepared from solvent phase exfoliated graphene. It was concluded in the previous section that this poor performance was due to: (i) (ii) The single and bilayer graphene flakes were smaller than the critical length. The larger flakes were multilayer and hence had a lower intrinsic modulus due to shear between the planes in the graphite. Thus for maximum accuracy for strain sensor applications, single-layer flakes should be used which are at least as long as the critical length of graphene (~ 3 µm). Thus a preliminary study was carried out to study the effect of sonication time on the dimensions of the graphene flakes. Three different grades of graphite were chosen: Branwell grades RFL 99.5, 2369 and The initial graphite concentration was 5 mg/ml and the solvent was NMP. Different sonication times were considered: 30 min, 60 min, 90 min, 2 hr, 3 hr, 4 hr and 5 hr. However, the centrifuging conditions of rpm were kept constant in all cases. Khan et al. has reported that these centrifuging conditions remove all visible large aggregates from the suspension. [38] Figures show the dispersions prepared from different grades of starting graphite at different sonication times in 30 ml bottles. Figure 4-32: Photograph of various dispersions prepared from Branwell grade RFL 99.5 obtained after centrifugation which have been sonicated for the times indicated 90

91 Chapter 4 Results and discussion Figure 4-33: Photograph of various dispersions prepared from Branwell grade 2369 obtained after centrifugation which have been sonicated for the times indicated Figure 4-34: Photograph of various dispersions prepared from Branwell grade 9842 obtained after centrifugation which have been sonicated for the times indicated The difference in the colour of the suspension depicts the difference in the concentration (Beer s law). From these images, it can be seen that grade 2369 and 9842 give good concentration of graphene when compared to RFL The concentration of the graphene was calculated from vacuum filtration of these suspensions on aluminium oxide (0.1µm pore size) and PTFE filter papers (0.2µm pore size). The difference in the mass of the filter paper divided by the volume of suspension filtered gives the concentration of graphene in the suspension. These dispersions were characterized by UV-vis absorption spectroscopy. The UVvis absorption spectra were measured for all the suspensions prepared, and it appeared flat and featureless (Figure 4.35). All suspensions were diluted by a factor of four for absorption measurements. 91

92 Chapter 4 Results and discussion Figure 4-35: UV-vis Absorption spectra of three different suspensions prepared from different grades of graphite. Sonication time: 3 hours Figures 4.36 and 4.37 show the change in absorbance (660 nm) and concentration of graphene obtained after centrifugation as a function of sonication time for the three different grades of starting graphite. It is clear that grade 2369 gives a better yield of graphene under the same conditions compared to other two grades. As the sonication time increases the graphene concentration obtained after centrifuagtion increases. Maximum yield of graphene obtained was ~ 0.21 mg/ml at 5 hr sonication time for grade Whereas the maximum concentration obtained for grade RFL 99.5 and 9842 are 0.02 mg/ml and 0.14 mg/ml, respectively. 92

93 Chapter 4 Results and discussion Figure 4-36: Absorbance vs. sonication time for different grades of graphite (660 nm) Figure 4-37: Graphene concentration after centrifuge as a function of sonication time for different grades of graphite SEM and Raman spectroscopy were used to characterize the graphene flakes obtained. For characterization samples obtained from grade 2369 were used because of its high graphene yield. Small strips of filter paper on which graphene was filtered were cut and observed in the SEM. (Figure 4.38) 93

94 Chapter 4 Results and discussion Figure 4-38: SEM images obtained from 90 min sonicated samples (a, b, c) and a 5 hr sonicated sample (d, e, f) at different magnifications From figure 4.38, it is clear that the flake morphology is completely different from the XG graphite nanoplatelets (see Figure 3.1 and 4.29). Individual flakes were easily distinguishable from the above images, whereas the expanded morphology was clearly visible in Figure 3.1 for XG graphite nanoplatelets. On comparing Figures 4.38a and 4.38d, it can be seen that the sample with a 5 hr sonication time had relatively smaller flakes than the sample with a 90 min sonication time. Red arrows are marked in the sample showing the presence of larger flakes. However, large chunks of graphite (red arrows pointing in Figure 4.38b and 4.38c) was observed clearly in the 90 min sonication sample, whereas these were absent in the 5 hr sonication sample (Figure 4.38e and 4.38f), indicating that the 94

95 Chapter 4 Results and discussion lower time of sonication did not completely exfoliate the starting graphite into graphene. This might be the reason for the low yield of graphene. Raman spectroscopy identified many graphite-like flakes in the material produced by a 90 min sonication (Figure 4.39). Figure 4-39: Raman spectra obtained from random positions of the film from 90 min sonication sample. Inset is the vacuum filtered film on aluminium oxide filter paper The top spectrum in Figure 4.39 is similar to the starting graphite powder (not shown). The two other spectra belong to multi- and few-layer graphene on the thin film. The presence of the D band indicates that spectra were obtained from the edges of small flakes of graphene on the film, which act as a defect. Nevertheless, many spectra showed characteristics of relatively thin graphene rather than graphite-like flakes. In contrast, the presence of large graphite flakes was not prevalent in the 5 hr sonication sample. This again supports the previous discussion that the presence of large graphite flakes in the samples is mainly because the rotation speed chosen for centrifuging was not sufficient to remove the large graphite flakes from the suspension. Selecting higher rotation rates would potentially remove the larger/thicker flakes from the suspension. But among those large flakes, large graphene flakes would also be included, which are what is needed for obtaining better reinforcement and shift rates. Therefore, there should be a compromise developed between the rotation speed and sonication time in order to produce relatively large graphene flakes in the suspension. 95

96 Chapter 5: PERSPECTIVES AND FUTURE WORKS 96

97 Chapter 5 Perspectives and future works 5. PERSPECTIVES AND FUTURE WORK 5.1. Perspectives The experiments in this study have shown that graphene composites and coatings can be used as strain sensors. Demonstration of the sensitivity of a Raman spectrum of a graphene coating on steel substrates to strain suggests that such sensors could be used in engineering applications. For example, strain sensitive graphene composite and coatings could be used to monitor strain over a wide area in engineering structures such as a bridge. The monitoring of strain can be achieved using commercial hand-held fibre optical Raman spectroscopy instruments as shown in Figure 5.1. Figure 5-1: Hand-held Mini Raman instruments from B&W TEK Inc. [99] The maximum achievable resolution of graphene-based Raman strain sensors can be calculated from the maximum shift rate obtained from the ideal graphene samples and the resolution of the spectrometer. The absolute resolution of the spectrometer with the peak fitting software is 0.1 cm -1. [30, 100] The maximum G band shift rate obtained from the cyclic deformation of graphene composites was ~ -85 cm -1 /%strain. So the maximum achievable resolution of the graphene strain sensor is (0.1 cm -1 )/(85 cm -1 /% strain), which is % strain (12 microstrain). Thus if this composite was used as strain sensor, it could measure strain from 0.0 % up to interfacial failure (~ 0.55 % strain) between the graphene and matrix to an accuracy of %. The resolution obtained (12 microstrain) is worse than commercial single point strain gauges (1 microstrain-arc weldable strain 97