Graphene-Based Materials as Strain Sensors in Glass Fiber/Epoxy Model Composites.

The ability of graphene-based materials to act as strain sensors in glass fiber/epoxy model composites by using Raman spectroscopy has been investigated. The strain reporting performance of two types of graphene nanoplatelets (GNPs) was compared with that of graphene produced by chemical vapor deposition (CVD). The strain sensitivity of the thicker GNPs was impeded by their limited aspect ratio and weak interaction between flakes and fibers. The discontinuity of the GNP coating and inconsistency in properties among individual platelets led to scatter in the reported strains. In comparison, continuous and homogeneous CVD grown graphene was more accurate as a strain sensor and suitable for point-by-point strain reporting. The Raman mapping results of CVD graphene and its behavior under cyclic deformation show reversible and reliable strain sensing at low strain levels (up to 0.6% matrix strain), above which interfacial sliding of the CVD graphene layer was observed through an in situ Raman spectroscopic study.

Introduction

Fiber-reinforced composites continue to be used in engineering structures in areas such as automobiles, aircraft, and aerospace. However, conventional test methods for fiber composites usually destroy the integrity of specimens and cannot monitor the local behavior of a single fiber inside a composite structure. Herein, a real-time, in situ, and nondestructive technique has been developed to sense the fibers’ strain condition. Using Raman spectroscopy to monitor the deformation behavior of fibers sheds light on the understanding of the mechanism of fiber-reinforced composites,[1−3] and it is normally undertaken when well-defined Raman spectra can be obtained from fibers. For objects such as glass fibers that do not give Raman scattering, incorporation of Raman-active materials is a feasible option,[4−6] which, in comparison with other nondestructive method such as imprinting fibers with gratings, does not require a specific fiber composition and a core-cladding structure.[7]

Graphene, an allotrope of carbon consisting of a single layer of sp2 carbon atoms arranged in a hexagonal lattice,[8] has opened the door for exploring the world of two-dimensional materials. Its resonantly enhanced and strain-sensitive Raman bands make it an ideal candidate for working as a strain sensor. The phenomenon of graphene’s strain-induced peak shifts is well-understood,[9−11] based on which the performance of graphene in different composite systems was also investigated.[12,13] Compared with carbon nanotubes, graphene’s two-dimensional nature and the high intensity of the Raman 2D band are preferable as a Raman-active coating.[14,15]

In this study, we demonstrated the possibility of using three types of graphene-based materials: commercially available graphene nanoplatelets (GNPs), lab-made electrochemically exfoliated graphene (EG) with oxygen functionalities, and graphene grown by chemical vapor deposition (CVD) as strain sensors in a glass fiber/epoxy model composite system combined with Raman spectroscopy. The strain sensitivities of different types of graphene were compared, and point-by-point strain reporting was achieved along an embedded CVD graphene-coated glass fiber.

Experimental Section

Materials

The fibers employed in this study were SE 4220 (tex 2400) glass fibers from 3B (Belgium) and were used as received. 1-Methyl-2-pyrrolidinone (NMP) was purchased from Sigma-Aldrich Co., Ltd. (UK). The epoxy resin was a mixture of Araldite resin LY5052 and Aradur hardener HY5052 supplied by Mouldlife Ltd. (UK). The mixing ratio of the resin and hardener was 100:38 by weight.

Two types of graphene nanoplatelets were used. The first type is the commercially available PR0953 Elicarb materials grade multilayer graphene powder (Thomas Swan & Co., Ltd.). The other type is the lab-made electrochemically exfoliated graphene (EG graphene) flakes.[16] The exfoliation process was performed in a two-electrode system by using a strip of graphite foil as the working electrode (anode) with a platinum mesh as the counter electrode (cathode) and 0.15 M sulfuric acid as the electrolyte. A voltage of 10 V was applied for 10 min for the exfoliation of the graphite. The EG graphene liberated in the H2SO4 electrolyte was collected and then washed, vacuum filtrated, and dried before further usage.

The graphene produced by the chemical vapor deposition method (CVD) was synthesized on a 0.025 mm thick Cu foil substrate purchased from Alfa Aesar, UK. After the substrate was loaded into a quartz furnace, the specimens were heated to 1000 °C and then annealed for 20 min under a H2 atmosphere (50 sccm), which was followed by 10 min graphene growth under a CH4/H2 gas mixture (5 and 50 sccm, respectively) before the furnace was cooled to room temperature with no methane flow.

Graphene Nanoplatelets Coating

The coating solution comprised NMP solvent and epoxy resin with a weight ratio of 8:2, in which 0.5 wt % Elicarb graphene or EG graphene was also added. First, Elicarb graphene or EG graphene was added to the NMP solvent, followed by 2 h sonication (37 Hz, 420 W) to disperse the flakes. The epoxy resin was then added, and the mixture was stirred for an hour. Subsequently, the whole system was stirred for another hour after the addition of the hardener. In the next step a single fiber filament was immersed into the coating solution for 10 min. Finally, the coated fiber was taken out and dried at 100 °C for 24 h under vacuum to remove NMP and cure the resin. The total thickness of the coating, determined from an SEM cross section (see Figure S1) of a coated fiber, was approximately 200–400 nm.

CVD Graphene Coating

First, the graphene/Cu foil was cut to approximately 0.5 cm × 2 cm before being spin-coated with PMMA (4% in anisole) at 4000 rpm. Next, the Cu was etched by 1 M aqueous FeCl3. After the copper was fully removed, samples were transferred to a glass Petri dish filled with deionized water and left in water for a few minutes. This step was repeated five times before a single fiber filament with a length of around 2 cm was immersed in the water and “fished” the floating graphene/PMMA stack out. The coated fiber was then left to dry overnight before the PMMA was removed by immersing the coated sample in acetone for 10 min. A schematic diagram illustrating the detailed coating procedure for both graphene nanoplatelets and CVD graphene is shown in Figure S2.

Single Fiber Deformation

The specimens were prepared by mounting a single fiber onto a piece of cardboard with a 20 mm gauge length window using cyanoacrylate adhesive (super glue). During the experiment, the card with the fiber was fixed on a single fiber deformation rig with super glue. After the glue had dried, both sides of the card were cut, and the rig was placed under a Raman spectrometer. The fiber was then strained incrementally, and Raman spectra were collected for each strain value. The strain was calculated though measuring the change in the fiber length divided by its original length.

Epoxy Model Composite Deformation

To produce a model composite specimen, first epoxy resin was mixed with hardener and degassed under vacuum for 30 min, after which the mixture was poured into a rectangular mold and left for half an hour to partially cure the resin. A coated fiber was then embedded in the central region of the mold, parallel to the axial direction, and the whole system was cold cured at room temperature for at least 24 h. Before testing, a strain gauge was attached to the sample surface with super glue, and two wires were soldered to the gauge. The specimen was deformed by using a four-point bending rig, and the strain was monitored by connecting the strain gauge to a voltmeter.

Characterization

The Raman spectrometer used was a Horiba LabRAM HR Evolution system (λ = 633 nm) with an 1800 lines/mm grating. The laser power was kept below 1 mW to avoid damaging the samples. The polarization of the incident light was always parallel to the deformation direction. A scanning electron microscope (SEM, Tescan Mira 3 FEGSEM) was employed to characterize the morphology of fibers coated with the graphene-based materials. All the specimens were gold-coated before imaging. Fourier-transform infrared (FT-IR) spectra were obtained in the transmission mode by using a Nicolet 5700 spectrometer (ThermoFisher Scientific Inc.). X-ray diffraction patterns (XRD) were obtained using a PANalytical X’Pert X-ray diffractometer (Philips) equipped with a Cu Kα radiation source (λ = 1.542 Å). An X-ray photoelectron spectroscope (XPS) equipped with a monoenergetic Al Kα X-ray source at 20 eV pass energy with a step size of 100 meV was used. The morphology of the graphene platelets was investigated using a NanoWizard atomic force microscope (AFM) from JPK Instruments (Germany). Before imaging, a small amount of graphene powder was dissolved in a mixture of isopropanol and deionized water with a volume fraction of 1:1, which was then sonicated for 2 h and drop-cast onto silicon wafers to be imaged.

Results and Discussion

Fibers Coated with Graphene-Based Nanoplatelets

Characterization for the two types of graphene nanoplatelets, Elicarb and EG graphene, is shown in Figure . Their graphitic structures were confirmed by XRD patterns which exhibit a dominant (002) Bragg peak at ∼26.7°, corresponding to an interlayer spacing of ∼0.34 nm. This can be further proved by their Raman spectra showing characteristic D, G, and 2D bands.[17] Compared with that of Elicarb, the intense D band and appearance of D′ and D+D′ in the Raman spectrum of EG graphene were due to defects and functionalities caused by the exfoliation and oxidation process,[18] which may also cause the broadening of its (002) diffraction peak in the XRD pattern due to the decreased crystallite size according to Scherrer’s equation[19] and the appearance of the asymmetrical (100) reflection at ∼42.5° due to the turbostratic disorder in layer stacking.[20] The differences in the chemical composition of Elicarb and EG graphene were further characterized by FT-IR and XPS analyses. As shown in Figure c, the FT-IR spectrum of EG graphene exhibits characteristic bands of C–O groups located at ∼1064, 1120, and 1220 cm–1 compared with that of Elicarb graphene. The band at 1620 cm–1 is due to the vibration of the adsorbed water molecules.[18] The existence of oxygen functionalities for EG graphene was also confirmed by the XPS analysis as shown in Figure d. The C 1s spectrum can be fitted into five components, located at 284.8, 286.4, 287.1, 288.2, and 290.2 eV, corresponding to C–C, C–O, C=O, and O–C=O groups and the π–π* shakeup satellite structure, respectively.[21] The deconvolution of the C 1s peak suggests that its atomic C/O ratio is around 13.8, which is similar to the oxygen content of the weakly oxidized graphene.[22]

Figure 1

Comparison of physical and chemical characteristics of Elicarb and EG graphene nanoplatelets: (a) representative Raman spectra, (b) XRD patterns, (c) FT-IR, and (d) XPS C 1s spectra.

The morphology of Elicarb and EG graphene coated glass fiber was observed by SEM as shown in Figures b and 2c, respectively. Compared with the as-received glass fibers (Figure a), it can be seen that the epoxy coating layer is almost invisible apart from the evenly distributed graphene platelets. The lateral size of the two types of graphene flakes is normally below 5 μm; however, it can be observed that Elicarb graphene is thicker than EG graphene. A thorough study of the morphology for both types of graphene flakes was performed using AFM. Figure S3 presents typical AFM images of Elicarb and EG graphene platelets, in which a majority of EG appear to be more exfoliated, e.g., flatter and thinner. A statistical analysis on their lateral size and thickness is summarized in Figure S4. Because of the variation in shape and uneven surface of individual flakes, the longest diameter and the maximum height were measured to represent the lateral size and thickness, respectively. As can be seen, although both graphene flakes are polydimensional, the EG graphene has a generally higher aspect ratio (the ratio of the lateral size over thickness) than the Elicarb graphene.

Figure 2

SEM images of (a) as-received, (b) Elicarb graphene, and (c) EG graphene coated glass fibers.

Figures a and 3b show the changes in the Raman 2D band position of Elicarb and EG graphene as a function of the tensile strain applied to coated glass fibers, respectively. As can be seen, for both graphene types, the 2D peaks shift linearly to lower wavenumber under tensile deformation, indicating stress can be transferred from the fibers to the coating layer. In comparison with Elicarb, EG graphene showed higher average strain shift rate (−7.5 ± 0.9 cm–1/% vs −3.5 ± 0.7 cm–1/%), indicating better stress transfer for EG graphene attributed to the relatively higher aspect ratio and the improved interaction between the flakes, fiber, and epoxy coating as a result of functionalities on the graphene plane which may react with the epoxy resin and the fibers’ surfaces. Therefore, the EG nanoplatelets were chosen for strain mapping.

Figure 3

Representative Raman 2D band position of (a) Elicarb and (b) EG graphene coated onto glass fiber surfaces with respect to applied strain before being embedded in the epoxy resin matrix. (c) Raman 2D position shift with strain for EG coated glass fiber after embedment.

The EG coated glass fibers showing the higher 2D band shift rate were embedded into the epoxy resin matrix to simulate a model composite containing one single fiber filament and Raman spectroscopy was used to map the strain distribution along the fiber. When the specimen was strained after embedment, an increase in the shift rate from −7.5 ± 0.9 cm–1/% to −10.3 ± 4.1 cm–1/% can be observed (Figure c). This is consistent with the previous study by Sureeyatanapas et al.,[23] suggesting that the shift rates of single-walled carbon nanotubes (SWNT) can be further increased with an extra epoxy layer outside the sizing layer containing SWNTs. Despite the relative high level of scatter in data points, the strain distribution along the fiber reported by EG graphene in Figure is still in good agreement with the shear-lag model presented by Cox[24] where for a fiber with a certain length l the strain builds up from each fiber end to a plateau in the middle approaching the applied matrix strain, provided the fiber and matrix are well bonded and both of them deform elastically. A general equation describing Cox’s theory is given as[24]for 0 < x < l whereandwhere εf and εm are the strain of the fiber and matrix, respectively; Em, Gm, and ν are the Young’s modulus, shear modulus, and Poisson ratio of the matrix, respectively; r is the fiber radius; and R is the distance between neighboring fibers. For a single fiber model composite, R may be represented by a cylinder of resin around the fiber into which the stress will decay radially.[1]

Figure 4

Strain distribution at the fiber–matrix interface with distance along the length of an EG coated glass fiber in a model composite at 0.4% strain.

The interfacial shear stress (IFSS), τ, along the fiber length can also be determined using the equation given by[25]where Ef is the Young’s modulus of the fibers.

The limitation of using graphene nanoplatelets as strain sensors can be clearly seen in Figure . On one hand, its strain sensitivity can be limited by the relatively small aspect ratio of the platelets compared with previous research focused on SWNTs with high aspect ratios,[23,26] although this may be partially compensated by further improving the interaction between fiber surfaces and graphene through the usage of a more appropriate coating solution and functionalization of graphene. On the other hand, since each graphene flake acted like an individual strain sensor and the variation in their dimensions as well as chemical composition for EG graphene resulted in different strain sensitivities for each flake, the strain reported by graphene flakes will be scattered as shown in Figure . The inconsistency in graphene platelets, apart from Figure S4, can also be indicated by the scatter plot in Figure S5, in which the variation in the intensity ratio of the D and G band of graphene nanoplatelets shows different extents of defects and functionalization.[18]

Fibers Coated with CVD Graphene

Because the sensing ability of graphene nanoplatelets may be hampered by their limited aspect ratio and inconsistency in properties, graphene produced by CVD was also employed to coat fibers and its strain reporting performance investigated. A representative Raman spectrum and a Raman line mapping of the intensity ratio of 2D and G band (I2D/IG) of CVD graphene coated on the glass fiber surface are shown in Figures a and 5b, where the absence of the Raman D peak shows that the CVD graphene was relatively defect free and I2D/IG ∼ 2–3 indicates that the coating remained mostly monolayer after being transferred onto fibers.[17] The SEM image in Figure c shows the morphology of a CVD graphene coated glass fiber, in which wrinkles and corrugation can be observed for the continuous graphene layer. This is likely to be caused by the relatively large graphene sheet used during the transfer process compared with the diameter of glass fibers (∼17 μm).

Figure 5

(a) Representative Raman spectrum of CVD graphene coated on a glass fiber. (b) Raman line mapping of the intensity ratio of the 2D and G peak along the CVD graphene coated glass fiber. (c) SEM image of the glass fiber coated with CVD graphene.

As before, the ability of CVD graphene to act as a strain sensor was investigated. Before embedment, only small Raman band shifts were detected for the coated fiber and slippage of the CVD coating was observed at low strain. An example of the shift of its 2D peak with strain after embedding a single coated glass fiber into the epoxy matrix is shown in Figure a,b. Compared with the result of glass fibers coated with Elicarb graphene, a dramatic increase in the strain sensitivity from −3.5 ± 0.7 cm –1/% to −14.7 ± 3.6 cm–1/% can be achieved through employment of CVD graphene, although its shift rate is still lower than the value (−60 cm–1/%) reported for the exfoliated monolayer graphene[11] possibly due to the polycrystalline nature of the CVD graphene used in this study,[27] corrugated coating structure, and inefficient stress transfer at the fiber/graphene interface.[28] An SEM image of the CVD grown graphene domains with a size of ≈3 μm grown on a copper foil prior to coalescence is shown in Figure S6.

Figure 6

(a) Raman 2D peak position before and after deformation, (b) 2D band shift with respect to strain, and (c) five cyclic deformations to around 0.4% strain and the response of 2D band position for the CVD graphene coated glass fiber.

The coating was subjected to cyclic deformation consisting of five loading/unloading cycles to the maximum strain of ∼0.4% for each cycle (Figure c). As can be seen, the deformation remained elastic (Figure c and Figure S7a) with no obvious indication of damage at the interface even after five cycles to 0.4% strain. However, at the end of each unloading cycle there was a slight increase in the 2D band position at 0% strain and therefore shift rate (since the peak position at 0.4% strain was constant) than in the previous cycle, which stabilized after the first three cycles. This phenomenon is consistent with previous research by Raju et al.[15] in which the authors suggested that the induced compression during unloading and the flattening of ripples present in the graphene could be plausible explanations. After five cycles to 0.4%, the same specimen was deformed further to 0.6% strain and the deformation behavior of graphene is shown in Figure S7b, in which some hysteresis started to be observable.

Point-by-point mapping of local strain along a CVD graphene coated glass fiber was performed for 0%, 0.3%, 0.6%, and 1% matrix strain in sequence after the specimen was predeformed three times. After that the strain was completely released back to zero and reloaded to 0.3%. All the mapping results are displayed in Figure . It can be seen that at matrix strain of up to 0.6%, the strain reporting performance of CVD graphene was much more accurate, stable, and of higher resolution compared with that of EG graphene platelets (Figure ) and even SWNTs.[29] This can be attributed to CVD graphene’s high aspect ratio and homogeneous properties. This is illustrated more clearly by Figure showing the different morphology between a glass fiber coated with graphene flakes and CVD graphene. At 1% matrix strain, a clear discrepancy between the experimental data and theoretical analysis can be observed (Figure d), indicating the breakdown of the interface when the matrix strain increased from 0.6% to 1%. This is not a surprise though since a number of studies have mentioned the easy sliding of graphene deposited on substrates.[11,15,30] Apart from the interfacial sliding, since the strain reported by CVD graphene did not fall to zero in the middle region, both the graphene and the glass fiber inside appear to remain intact during the deformation.[23] The Raman mapping of the CVD graphene that was taken after the imposed strain was fully relaxed (Figure e) shows there was a slight residual compression on the coating layer. This could again be caused by slippage of graphene at high strain levels. During the initial straining, all components in the model composites deformed elastically since they were well-bonded with each other. At high strain, however, the graphene started sliding, and the relative movement is suspected to be more critical on the fiber–graphene interface since a test on a sandwiched model composed of epoxy resin, CVD graphene, and epoxy resin in sequence indicates that graphene in this type of structure can withstand higher strain (Figure S8) during loading without slippage. With ongoing loading, the model composite continued to be strained, but the graphene remained at a constant strain level. Once the load was released, the resin matrix along with the “clamped” graphene contracted to a relaxed state. Consequently, the graphene layer ended up being loaded in compression. The strain reporting performance of CVD graphene was, however, partially reversible even after exceeding the upper limit of its strain range as shown in Figure f.

Figure 7

Strain distribution along the length of a CVD graphene coated glass fiber in a model composite at different strain levels: (a) 0%, (b) 0.3%, (c) 0.6%, (d) 1%, (e) reversed 0%, and (f) reloaded 0.3%, along with dashed lines showing the theoretical shear-lag analysis.

Figure 8

Schematic illustration (not to scale) of glass fibers coated with graphene flakes (top) and CVD graphene (bottom).

Conclusions

This study has demonstrated the use of both discontinuous graphene nanoplatelets and continuous CVD grown graphene as strain sensors in a glass fiber/epoxy model composite using Raman spectroscopy, and their relative sensing performance was compared. It has been shown that the strain reporting performance of graphene nanoplatelets can be improved through functionalization, which improves matrix and fiber interaction, although the inconsistency in both dimensions and chemical properties among individual graphene nanoplatelets causes scatter in the measured data. Compared with graphene nanoplatelets, the sensing ability of CVD graphene was more sensitive and reliable due to its high aspect ratio and homogeneity of its properties. The strain distribution along a fiber reported by both graphene platelets and CVD graphene was in good agreement with classical shear-lag analysis. The strain reporting of CVD graphene remained reversible at low strain levels (<0.6% strain), above which sliding at the interface was observed. Overall, the best strain sensing behavior was obtained with the CVD graphene monolayer material.