Graphene-based surface heater for de-icing applications

Graphene-based de-icing composites are of great interest due to incredible thermal, electrical and mechanical properties of graphene. Moreover, current technologies possess a number of challenges such as expensive, high power consumption, limited life time and adding extra weight to the composites. Here, we report a scalable process of making highly conductive graphene-based glass fibre rovings for de-icing applications. We also use a scalable process of making graphene-based conductive ink by microfluidic exfoliation technique. The glass fibre roving is then coated with graphene-based conductive inks using a dip-dry-cure technique which could potentially be scaled up into an industrial manufacturing unit. The graphene-coated glass roving demonstrates lower electrical resistances (∼1.7 Ω cm−1) and can heat up rapidly to a required temperature. We integrate these graphene-coated glass rovings into a vacuum-infused epoxy–glass fabric composite and also demonstrate the potential use of as prepared graphene-based composites for de-icing applications.


Introduction
Recent years have seen signicant interest in developing smart de-icing systems due to problems associated with the ice formation in many industrial applications such as aircras, 1,2 wind turbines, 3 power transmission lines 4 and telecommunications. 5 Ice accretion can cause huge disruptions to day to day activities of human life 6 and performance degradation of the structure. 7 For example, ice accumulation on planes can change the aerodynamic performance and increase the weight; 8,9 thus produces signicant threats to aircra safety. Whereas ice accumulation on a wind turbine can cause mechanical and electrical failures; results in up to 50% power losses. 10 Several thermal, chemical and mechanical methods [8][9][10][11] have been developed for eliminating the surface ice such as chemical uid systems, 9,11 pneumatic boots, 12 electro-impulsive, 13 hot air system, 11 ultrasound system, 7,14 microwave technology 15 and electro-thermal de-icing system. 2 Among these, electro-thermal de-icing technique is considered to be the most effective and energy efficient due to its ability to control the temperature and heat dissipation by Joule heating. 16 Currently, bre reinforced polymer composites are increasingly popular in aerospace, automobile and civil engineering industries due to their higher strength and lower weight. 17 However, ice accumulation reduces the advantages that the composite brings to the structure. The electro-thermal system is identied as one of the most promising de-icing systems for polymer composites, as it does not cause delamination and damage to composite structure. 9,11 However, the application of the electro-thermal system within composites is limited by the poor thermal conductivity and high thermal sensitivity of polymeric materials. 18,19 Many studies [20][21][22] have reported uses of conductive polymers, metals, CNT and carbon black to make conductive polymer composites; however they still suffer from poor thermal and electrical conductivity, and higher energy consumption. Therefore, it is desirable to use a conductive material that can provide excellent electro-thermal properties as well as can achieve desired temperature without compromising existing mechanical and thermal properties of composites.
Graphene, a single atom thick two-dimensional closely packed honeycomb lattice of sp 2 carbon allotropes, has been focus of mass investigations in recent years due to its large surface area, 23 record thermal conductivity, 24 excellent mechanical strength, 25 and superior electronic mobility. 26 It has demonstrated much higher thermal conductivity $5000 W mK À1 than CNT ($3000 W mK À1 ) and amorphous carbon ($1 W mK À1 ). 27 Previous study has demonstrated the use of graphene-based inks to produce highly conductive electrical conductor with lower resistance. 28 Moreover, a small amount of graphene would be sufficient to form a percolative electric network due to higher aspect ratio and surface area; 29 thus reduce overall weight of the composite. Recent studies have highlighted using of Graphene Nanoribbon (GNR), 2,30 per-uorododecylated GNR 31 and Graphene Nanoplatelets (GNP) 32 for de-icing applications. However, there are number of problems still exist with these techniques such as complicated and expensive manufacturing route for GNR, not suitable for industrial applications and material waste. Moreover, strong van der Waals forces between graphene llers may cause restacking in the polymer matrix, which may result in defective composites. 33 Moreover, increased GNR concentration modies the resin viscosity and increases the brittleness of composites. 34 Here we report a scalable manufacturing route for next generation graphene-based de-icing composites. We utilise microuidic exfoliation technique to synthesize graphenebased ink in a scalable quantity and use a dip-dry-cure coating technique to make highly conductive graphenebased glass bre rovings. These graphene-coated glass rovings are characterised both electrically and thermally; then integrated into a composite structure. We demonstrate the use of as prepared composite for de-icing applications using Joule heating phenomena.

Materials
The natural ake graphite (average lateral size $ 50 mm) was kindly supplied by Graphexel Limited, UK. Sodium deoxycholate (SDC) powder was purchased from Sigma Aldrich, UK. The glass bre roving (Glass S, 758-AB-675) was purchased from AGY (USA). 290 gsm (grams per square meter) plain woven glass fabrics, EL2 Epoxy Laminating Resin and AT30 Epoxy Hardener were purchased from Easy Composites, UK.

Graphene exfoliation
We use microuidization technique to exfoliate graphene in a scalable quantity following previously reported methods. 28,35 Briey, 50 g graphite powder and 10 g SDC are added into a glass bottle and mixed with 500 ml deionised (DI) water. This mixture is sonicated for 30 minutes using an ultrasound bath to allow homogenous dispersion and added into an input reservoir of a Microuidizer (M-110P Microuidizer, Microuidics Corp, USA). The dispersion is slowly passed through 'Z-type' microuidic channels of $200 mm and $87 mm diameter with diamond construction at high pressure ($200 MPa). This allows the exfoliation of graphite to few-layer graphene (FLG) at 100 ml min À1 ow under high shear rate [$10 8 s À1 ] with a 100% exfoliation yield. The exfoliated dispersion is then passed through a cooling channel surrounded by cold water ($25 C) to prevent over-heating of the dispersion and collected. This process is repeated 20 times to produce FLG (MF akes) which is then used as a conductive ink for glass roving coating.

Glass roving coating
We use a simple dip-dry-cure coating technique to coat glass bre roving with graphene-based ink. The graphene-based ink is added into a 100 ml cylinder ask. Glass rovings are cut to 25 cm and dipped into graphene dispersion for 5 seconds. The coated glass bre rovings are then passed through eyelets to squeeze extra ink out of coated glass tows and to maintain coating evenness. We then use a Mathis Laboratory dryer (Mathis, Switzerland) to dry (at 100 C) and cure coated glass bres. We also use various curing temperatures (110 C to 240 C at 10 C interval), curing times (5 min to 30 min) and a number of coating cycles to optimise coating conditions.

Composite manufacturing
We use vacuum resin infusion process and room temperature thermoset EL2 epoxy resin to manufacture de-icing composites that contains graphene-based (coated) glass rovings. Briey, 6 layers of glass fabric (dimension: 10 Â 10 cm) are laid on a precleaned and pre-coated (with a release agent) metal plate. The graphene-based glass rovings are then inserted aer 3 layers of glass fabric and connected to wires for electrical and thermal characterisation. The sample is sealed by a plastic bag and vacuumed pressed using a pump. EL2 Epoxy Laminating Resin and AT30 Epoxy Hardener are degassed separately for 30 minutes and mixed together immediately before we use. The resin with hardener is then own over layered glass fabrics at a constant ow rate using a vacuum pump, which enables the impregnation of glass fabrics with resin. The resin infused preform is then cured at room temperature for 24 hours to make graphene-based glass composites for de-icing application.

Characterisation of MF akes and coated glass rovings
The graphene-based dispersion containing MF akes is diluted 1000 times and then drop-casted on Si/SiO 2 (290 nm oxide on plain silicon). The images and Raman spectra of MF akes are taken at 10 different locations on the drop-casted sample to assess the ake size, ake thickness and ake types. We use an optical microscope to measure the ake size and Philip XL30 Field Emission Gun (FEG) Scanning Electron Microscope (SEM) to assess the surface topography of the untreated and coated glass rovings. A Dimension Icon (Bruker) Atomic Force Microscopy (AFM) is used to determine the ake thickness. A Renishaw Raman System equipped with 633 nm laser is used to collect Raman spectra of MF akes, untreated and coated glass rovings. The surface of MF akes, untreated and coated glass rovings is characterised using a Kratos Axis X-ray Photoelectron Spectroscopy (XPS) system. The thermal decomposition of graphite, SDC and MF akes are evaluated using a TA instrument, TGA Q5000. The resistances of graphene-coated glass rovings per unit length are measured using a two probe multimeter (DL9309 Auto Ranging Multimeter, Di-Log, UK). The average resistance is calculated from 5 measurements at different positions along the length of graphene-based (coated) glass bre rovings.

Thermal analysis
The electro-thermal behaviour of coated glass bre rovings is measured using a thermal camera (TIM 160, Micro-Epsilon, Germany) and a power supply unit (Stabilised Power Supply L 30D, Farnell Instruments LTD, UK) which supplies direct current at different voltages from 0-30 V. The graphene-based glass bre rovings are clamped on a wooden plate using two metals clamps at 1, 5, 15 and 20 cm distances. A multi-meter is also used to measure the current. The thermal camera is installed on a tripod and connected to the computer. The thermal images obtained by the camera is shown on the computer through a soware (TIM Connect, Rel.2.12.2202.0). The change in temperature of the coated glass is observed and recorded at various volts and currents. Thermal images of coated glass at various volts and currents are also captured.

Graphene ink characterisation
We use microuidization technique to exfoliate few layers graphene (MF akes) from graphite into a water-based dispersion. Microuidizer can pass uids through micro-channels (diameter, d < 100 mm) at high pressure (up to 209 MPa), 35 which generates liquid velocities of 400 m s À1 and several order of magnitudes higher shear rates (>10 8 s À1 ) 36 than conventional rotor-based or other homogenisers. It is used primarily for particle size reduction, 37 nano-emulsion of immiscible liquids, 38 for disrupting or lysing cells 39-41 and de-agglomeration and dispersion of carbon nanotubes and graphene nanoplatelets (GNP) into polymers. 41,42 Recent studies have highlighted using microuidization technique to produce graphene, 28,35 graphene quantum dots 36 and two-dimensional (2D) boron nitride nano-sheets. 43 It is a simple and environmental friendly technique that can produce graphene-based ink with a 100% exfoliation yield. 28 Moreover, this process could be scaled up to produce hundreds of kilos graphene-based inks, which would suit perfectly for industrial applications such as manufacturing of graphene-based multi-functional composites.
We use Graphexel 2736 grade graphite (average ake size < 50 mm) supplied by Graphexel, UK. Fig. 2a shows the average lateral size of exfoliated akes aer 20 cycles (MF akes) is $1.45 mm. Flake thickness analysis by AFM shows $20% akes are <10 nm (ESI, Fig. S1 †). Fig. 2b shows Raman spectra of exfoliated MF akes aer 20 cycles, which is a typical for liquidphase exfoliated graphene, with characteristics D peak at $1350 cm À1 , G peak at $1582 cm À1 and an asymmetric 2Dband at $2730 cm À1 . 28,35 For XPS analysis, the exfoliated MF akes without any surfactant is drop-casted onto a PEL paper and attached onto a carbon tape. Similarly, starting ake graphite is attached onto a carbon tape directly. The wide scan XPS spectra reveal only C1s and O1s peaks for both starting graphite and MF akes, Fig. 2c. C/O ratio of starting graphite materials and MF akes (without surfactant) are 24.91 and 24.84, respectively. High resolution C1s spectra also reveals similar peaks for both starting graphite and MF akes, which is dominated by C-C/C]C bond in aromatic rings ($284.6 eV), (ESI, Fig. S2a and b †). During microuidization, further oxidation or chemical functionalisation is not evident from C1s spectra, which in agreement with the previous study. 28 TGA analysis shows slight decomposition ($2 wt%) of starting graphite above 750 C; whereas MF akes start to decompose at lower temperature ($5.5 wt% up to 750 C), may be due to the lower thermal stability of smaller MF akes. 44 As expected, SDC suffers signicant decomposition around 400 C.

Glass roving coating, optimisation and characterisation
For electro-thermal de-icing application, it is desirable to produce highly conductive and uniform glass-bre roving using a scalable process. We use a very simple dip-dry-cure technique that could potentially be scaled-up for industrial applications. As illustrated in Fig. 1, glass bre roving is dipped into graphene-based ink for few seconds and passed through an eyelet of specic diameter to squeeze out additional materials from the surface; thus produce uniformly coated glass-bre roving. The graphene-based glass roving is subsequently dried and cured at elevated temperature. These graphene-based (coated) glass-bre rovings could be integrated into a composite by weaving, knitting or braiding and heated up to a desired temperature for de-icing composite applications.   3a shows the effect of curing temperature and time on the resistance of coated glass-bre roving. As expected the resistance per unit length of graphene-based glass-bre roving decreases with the increase of temperature due to the thermal annealing of graphene akes and the removal of residual solvents and surfactants. 45 However, aer a certain temperature, further increase in temperature results in slight increase in the electrical resistance of graphene-based glass bre roving. This optimum temperature point for lowest resistance varies with the curing time, as higher curing time provides lowest resistance at lower optimum curing temp and vice versa. We use 220 C for 5 minutes as optimum curing condition to enable rapid drying and curing in a continuous manufacturing process. Fig. 3b shows the change of resistance of graphenebased glass rovings with temperature. The electrical conductivity improves with the increase in number of coating cycles due to the deposition of higher amount of conductive MF akes on the bre surface and the formation of a continuous percolated network. 2,46 Moreover, the contact resistance between graphene akes are reduced with the increase of coating cycles. The lowest resistance of coated glass-bre roving with graphene-based ink is obtained as $1.7 U cm À1 with 15 coating cycles.
Raman spectra of control glass roving shows smooth featureless background of uncoated bre, which is masked by strong uorescence, Fig. 3c. Whereas, graphene-based glass bre roving shows characteristics graphene peaks at $1350 cm À1 , $1582 cm À1 and $2700 cm À1 , corresponding to D, G and 2D peaks, respectively. The D peak is caused by the defects and disorders in the hexagonal graphitic layers, while G peak attributed to an E 2g mode of graphite due to the vibration of sp 2bonded carbon atoms in a two dimensional hexagonal lattice. 47 These Raman spectra are similar to the spectra of a typical liquidphase exfoliated graphene or graphitic materials. 28,35,48 The wide scan XPS spectra of control glass-bre reveal the presence of C1s, O1s, Si2s and Si2p and provide the evidence of some silane coating on the bre surface, Fig. 3d. However, aer coating with graphene-based ink, C/O ratio signicantly increased to $26.03 from $5.16 for control glass bre.
The high resolution C1s spectrum of control glass-bre can be tted into two main components: C-C/C]C (284.6 eV) and C-O (286.4 eV), Fig. 3e. The coating with graphene-based ink enables diminishing of oxygen containing functional groups from the coated surface, Fig. 3f. For graphene-based glass rovings, C1s is mainly dominated by C-C/C]C, which is similar to graphene or graphite. 49 SEM images provide further evidence of the presence of graphene akes, covering the glass bre, Fig. 4b and c. SEM images of uncoated glass show smooth glass bre, Fig. 4a. Aer coating with graphene-based ink, signicant deposition of MF akes on individual bre surfaces is observed, Fig. 4b. The individual bres are wrapped with MF akes, Fig. 4c, and create continuous conductive track, which enables current to ow through the structure. The diameter of the bres increases with the increase of number of coating cycles due to the deposition of more MF akes on the surface.

Joule heating of graphene-coated roving
We use four different lengths (5, 10, 15 and 20 cm) of graphenebased (coated) glass bre rovings for electro-thermal characterisation and ve samples for each length. We apply various voltages across the entire sample, and measure the current generated and the change in the temperature of graphene-based rovings (due to Joule heating) with voltages and the power consumed per unit length, Fig. 5a-c. I-V curves for various length of graphene-based glass roving show the linear relationship between the applied voltages and current passes through conductive rovings, Fig. 5a. However, a non-linear relationship is observed for the change of temperature with voltages, Fig. 5b. The temperature of coated roving increases quadratically with voltage, as indeed expected. When the obtained temperature is plotted against the power per unit length (p ¼ IV/L), the curves for the samples of different length collapse on the same straight line indicating that the major dissipation mechanism is the thermal radiation and convection rather than thermal conductivity through the contacts. Moreover, thermal images show uniform heat distribution along the length of conductive graphene-based glass rovings (ESI, Fig. S3 †). Unlike Peltier effect, the heat is distributed over whole length of the conductor during Joule heating, which is also evident from the thermal images.

Joule heating of graphene-based glass-bre composites for de-icing application
In order to demonstrate a potential de-icing application of graphene-based glass rovings, we manufacture a glass-epoxy composite using a vacuum resin infusion process, where ve graphene-based glass rovings are inserted in the middle of 6 glass fabrics layers (ESI, Fig. S4a and b †). Aer making the composite, the graphene-based glass rovings are connected with external electrical wires and a power supply unit. At rst, we investigate the time dependent temperature prole of the composite by applying voltages (5 V, 7.5 V and 10 V) across graphene coated glass rovings, Fig. 6a. The surface temperature of the composite increases with the increase of voltages, due to increase in the power. As shown in Fig. 6a, the temperature increases from room temperature ($24.1 C) to $36 C aer applying 5 V. Whereas for 7.5 V, the temperature increases rapidly from 30.3 C to 51.8 C within 30 seconds and becomes almost at at 60 C aer 90 seconds. The temperature increases at much faster rate for 10 V, up to 71.6 C within 30 seconds and continues to increase rapidly up to 100.8 C aer 180 seconds. Aer that slight increase in temperature is observed for 10 V. Similar time dependent temperature prole is observed in a previous study based on graphene nanoribbon (GNR); 2 however at much higher voltages (20-40 V). Moreover synthesis of GNR is based on complex, time consuming and expensive process of unzipping CNTs. 50 Furthermore, GNR was mixed with epoxy resin which may alter the rheology and uniformity of  epoxy/hardener mixture and limit the concentration of conductive materials. 51 In contrast, we use a scalable process for manufacturing graphene-based glass roving. Moreover, the coating of individual glass roving with graphene-based ink would enable insertion of coated warp or we glass roving into the fabric or composite structure as required. This would potentially reduce the material waste and provide much better control and exibility to remove various amount of ice from different parts of the structure.
Finally, we demonstrate the de-icing capability of our graphene-based glass/epoxy composites by dipping it into an ice-bucket, Fig. 6c. We then place this next to another bucket which contains only ice. We apply 10 V in order to see the deicing effect of graphene-based composite. Fig. 6b shows rapid increase in the temperature and the melting of ice in the bucket containing graphene-based composites by Joule heating. The temperature increases from À0.1 C to 27.3 C within 5 minutes, whereas the temperature of the bucket containing only ice remains almost same ($À1 C) even aer 30 minutes. Fig. 6d(ii) shows efficient removal of ice from the surface of graphene-based composite. Thermal image of both ice buckets shows similar temperature before applying heat, Fig. 6e(i and  ii). The thermal image of ice only bucket shows almost similar temperature (<À1 C) for 30 minutes. However, the temperature Fig. 6 (a) Heating profile of graphene-based glass composite at various voltages (5 V, 7.5 V and 10 V); (b) the comparative change in temperature of only ice bucket and the ice bucket containing graphene-based de-icing composites. (c) Before heating: (i) ice bucket and (ii) graphene-based de-icing composite dipped into an ice bucket; (d) after heating at 10 V: (i) ice still in the bucket and (ii) removal of ice due to joule heating of deicing composite; (e) before heating thermal images of (i) ice bucket and (ii) graphene-based de-icing composite dipped into ice showing similar temperature; (f) after heating thermal images: (i) ice still in the bucket and (ii) de-icing composite dipped into ice bucket demonstrates heating of the composites and removal of ice due to joule heating. increases to 42.3 C for the ice bucket containing graphenebased composite over that duration. This demonstrates the good perspective of our graphene-based composite for next generation de-icing applications.

Conclusions
We report a scalable process of manufacturing graphene-based surface heating for de-icing applications. The graphene-based glass roving demonstrates low resistance, $1.7 U cm À1 and efficient heating to a desired temperature at lower power consumption. We also demonstrate the use of this graphenebased glass rovings for manufacturing glass/epoxy composite for de-icing applications. The de-icing test shows efficient removal ice from graphene-based composite. We believe, the obtained results from this study would be an important step towards realising graphene-based next generation lighter, stronger and cost-effective smart de-icing composite for aerospace and other applications.

Conflicts of interest
There are no conicts to declare.