Graphene films printable on flexible substrates for sensor applications

Fifteen-layered graphene films have been successfully deposited onto flexible substrates using a commercial ink consisting of graphene particles dispersed in an acrylic polymer binder. A value of 74.9 × 10 5 cm − 2 was obtained for the density of defects, primarily located at the flake edges, from the ratio of the D and G Raman peaks located at 1345 cm − 1 and 1575 cm − 1 respectively. 0.5 μ m thick drop-cast films on interdigitated silver electrodes exhibited Ohmic conduction with a small activation energy of 12 meV over the temperature range from 260 to 330 K . The photo-thermoelectric effect is believed to be responsible for photoconduction through graphene films under illumination intensity of 10 mW m−2 at 270 nm , corresponding to the UV absorption peak. The photo-transient decay at the bias of 1 V involves two relaxation processes when the illumination is switched off and values of 8.9 × 10 3 and 4.3 × 10 4 s are found for the relaxation time constant using the Kohlrauch stretched exponential function analysis.


Introduction
The mechanical, electrical, and optical properties of graphene are outstandingly promising for exploitation in the wide range of possible applications such as field effect transistors, sensors, for flexible and wearable electronics [1][2][3]. Successful mechanical exfoliation of highly oriented pyrolytic graphite sample to produce thin layers of graphene has been reported in 2004. A one atom thick single layer graphene consists of sp 2hybridised carbon atoms arranged in a hexagonal lattice. The overlap between the conduction and valence bands is found to occur in ambipolar, monocrystalline thin films showing a large carrier concentration of 10 17 cm −2 and high mobility up to 10 000 cm 2 V −1 s −1 [4]. Mechanical exfoliation and chemical vapour deposition (CVD) are commonly used to synthesise high quality graphene films, often in conjunction with high temperature annealing to reduce the number of defects [5]. Few-layer graphene was grown on a catalyst copper foil substrate by controlled thermal evaporation of polystyrene carbon source under atmospheric pressure [6]. The semimetallic properties of zero band-gap and zero local density of states at the Fermi level limit the significant applications of pristine graphene in nanoelectronics. Boron and nitrogen atoms are extensively used as doping agents for graphene because of their similarity with carbon atoms in atomic size. Nitrogen doping in high concentration contributes electrons to delocalised states of graphene, causing the shift of the Dirac energy [7]. The choice of doping strategies is very important to achieve stable performance of the device over time. Stable, high transparency has been achieved for CVD multilayer graphene doped with ferric chloride and tin (II) chloride [8]. Transverse electric fields are applied to the sample for tuning band gap without involving any chemical doping. For example, values of 250 meV and --1000 cm V s 2 1 1 were obtained for tuneable band gap and mobility for an exfoliated graphene bilayer used in a dual gate fieldeffect transistor configuration between gold top and platinum bottom electrodes [9]. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Printing is regarded being as an alternative, economically viable technique for producing micropatternable conductive graphene films at room temperature over large area on flexible substrates under ambient conditions without using vacuum or inert atmospheres. Graphene electrodes having low sheet resistance of W  0.3M and high optical transparency in the order of 86% has been successfully fabricated by photothermal reduction of inkjet printed graphene oxide sheets using the IR heat lamp in ambient environment for about 10 min [10]. The challenge with printing graphene is in formulating the ink since the various rheological properties such as density, surface tension and viscosity have a strong effect on the printing process. Low boiling point and environmentally-friendly solvents, such as ethanol and water in the volume ratio of 1:1 gives a surface tension of -31 m Nm 1 which meet the requirement of producing graphene-and few-layer graphene-based inks [11]. If the graphene loading is too low, below the percolation threshold of the system, then there will be no conductivity. A key advantage of graphene is its high aspect-ratio, flake-like, structure which will result in a much lower threshold than if it were low aspect-ratio morphology. Above the percolation threshold the conductivity will improve slightly as the loading is increased, but at the expense of the coating's mechanical properties, such as flexibility and extensibility. Poor mechanical stability can lead to cracks in the films and a loss of conductivity. The nature of the binder material is also critical in determining the flexibility of the deposits [12]. The liquid-phase exfoliation methods suffer from poor controllability in size and thickness, poor production efficiency and dispersion instability. The shear exfoliation technique is recently reported to be faster than liquid-phase exfoliation methods and therefore is regarded as being scalable for mass production using commercially available inkjet printers ranging from desktop to rollto-roll. The graphene ink (GI) is formulated with the graphene dispersion through tuning the viscosity by adding ethylene glycol using a mild sonication process. During the process, ethyl cellulose is added as stabiliser in order to keep the shear exfoliated graphene flakes stable inside the ink. The resulting few layer graphene films are highly conductive and stable over several months [13]. Graphene flake size should be fifty times smaller than the printing nozzle diameter. The flake size can be controlled by fluorination of multilayered graphene flakes in suspension of 3% aqueous solution of hydrofluoric acid [14]. The aerogel is fabricated by combining drop-on-demand 3D printing and freeze casting into continuous, boundary free microstructure for 3D architectures [15].
We have recently reported a positive value of --1.5 10 K 3 1 for the temperature coefficient of resistance for printed m 0.5 m thick graphene films on flexible plastic substrates, over the temperature range from 260 to 330 K [16]. In light of this preliminary observation, further work has been undertaken to investigate the scope of fabricating low cost graphene thermistors and UV photosensors, using this material. A number of earlier studies have focused on the use of graphene in hybrid architectures for photodetector devices [17]. This article presents the results of an investigation into the photocurrent transient relaxation behaviour, of simple graphene films, that can be printed over large area and on flexible substrates. Raman Spectroscopy, Scanning Electron Microscopy (SEM) and UV-Visible spectroscopy studies have been carried out on graphene films, in an effort to get a better understanding of the material's structure and morphology, and hence shed light on its interesting electrical properties.

Experimental procedure
The commercially available ink from DZP technologies Ltd (product number G087) was formulated with few-layer graphene particles of lateral dimensions m <40 m, obtained by liquid phase exfoliation. The graphene particles were then dispersed in an acrylic polymer binder system to obtain the ink suitable for deposition on polyethylene terephthalate (PET) substrates (Melinex DuPont Teijin). Organic binding additives are not normally used in graphene films deposited for example by CVD processes. The ink used in this work is specifically formulated for industrial, large area printing and coating methods such as slotdie, gravure and flexography which are high-speed, volume manufacturing processes. However, the samples for electrical measurements consisted of drop cast GI films, m ( ) t 0.5 m in thickness , on predesigned interdigitated silver electrode structures, on PET substrates. The dropcast method was chosen here for the small scale, rapid deposition and the ink was mixed using high-speed mixer immediately before dropcasting using a micro-pipette with a view to ensuring a high degree of the film uniformity and reproducibility. This interdigitated electrode structure has been chosen in order to facilitate direct illumination on an enlarged photodetection area. A Keithley 617 electrometer in the microprocessor controlled measuring system was employed to record both dark current and photocurrent characteristics as a function of bias voltage V for the drop cast samples, which were held under vacuum of -10 mbar 5 in an Oxford Instrument liquid nitrogen constant bath cryostat. The mechanical stability of the graphene coating was investigated by measuring the room temperature resistance of drop cast films, on 0.1 mm thick PET substrates between two planar silver electrodes at the ends of the film GI under both extension and compression conditions of bending. The PET substrate was much longer than the GI film in order to apply uniform stress on the GI film.
A monochromator (HR320 Jobin-Yvon, HOR-IBA) was used as a light source in the spectral range of 300-700 nm for photoconduction experiments. UVvisible absorption spectra of the printed film on quartz substrates were recorded using a Perkin-Elmer LAMBDA 650 spectrophotometer between 190 and 900 nm scanning at the rate 654.8 nm min −1 . The surface microstructure of the GI drop-cast on an aluminium foil was investigated using an S3400N scanning electron microscope. The samples were characterised by means of x-ray diffraction using Bruker D8 advance scanning instrument. Cu K α radiation of wavelength 0.15406 nm was used with scattering angle varied between 10°and 90°at scanning rate of 0.02°per sec for crystallographic analysis of the samples. The Raman spectra were obtained by Horiba Jobin Yvon Lab RAM HR800 with incident laser beam of wavelength 514.5 nm (equivalent to the excitation laser energy = ) E 2.41 eV L of 2 μm spot size.

Results and discussions
Experimental results are analysed in order to derive physically meaningful information regarding lattice structure, defect type and density, and electrical and optical device parameters. Values of material parameters have been compared with relevant published data for identifying new knowledge. Figure 1 shows a scanning electron micrograph of the graphene deposit. It is apparent that the lateral dimensions of the flakes are in the micrometre range, although of course their thickness is three orders of magnitude smaller. Such materials pose challenges for this form of microscopy because the thickness of the flakes is massively smaller than the penetration depth of the electron beam used, and is even small compared to the range of secondary electrons used to form the image. As a result the three-dimensional structure of the whole near surface contributes to the image, rather than the normal situation when the image is dominated by the topography of the outermost surface. Additionally, the backscattered electron and secondary yields of graphene are both extremely low [18], and so ultimate resolution must be compromised in order to achieve sufficient signal to form an image. The XRD spectrum in figure 2 of the sample shows a strong and sharp [002] peak at 26.19°corresponding to an interlayer distance of = d 0.34 nm 002 determined using Bragg's law, where the subscript (002) refers to the diffraction plane. These results are in good agreement with those obtained for exfoliated graphite oxide precursor after electrochemically reduction at 1.5 V, and are close to those for pristine graphite [19]. A value of 5.57 nm is estimated for the average crystallite size, D , 002 from the Debye-Scherrer formula in the form:  The Raman spectra in figure 3 for the GI sample were recorded for the wavenumbers ranging from 1000 to 3000 cm −1 . The characteristics peaks of D and G have been observed at 1345 cm −1 and 1575 cm −1 respectively in figure 3(a). These peak positions are in agreement with those observed for the annealed graphene layer ink-jet printed on a Si/SiO 2 substrate [22]. The D band is due to the breathing modes of sp 2 rings of six-atom rings. The intense G band arises from inplane vibration of sp 2 carbon atoms present in the sample [23]. The intensity ratio of I D /I G is found to be 0.3, reflecting the defect concentration of carbonaceous materials [24]. A value of 20.6 nm is estimated for an average inter-defect distance L D in nm using the relation [25]:

Microstructural, crystallographic and compositional studies
is the laser excitation energy. Although the spot size is nominally 2 μm, the laser beam is expected to penetrate into the depth of  >50 nm encountering a large number of graphene flakes. Therefore, the estimated value of L D is considered to be reasonable in the context of the scattering within a large volume [26]. The value of -74.9 10 cm 5 2 is found for the defect density n D from the knowledge that These defects are believed to be located at the flake edges responsible for activation by double resonance giving rise to the D peak. Figure 3(b) shows that a broad 2D peak due to absorptions by second order zone-boundary phonons appears at 2690 cm −1 , representing a blue-shift by ∼30 cm −1 in relation to monolayer samples [27]. Also, a value of 80 cm −1 for the FWHM of the 2D peak is approximately three times larger than that reported from single layer graphene. These observations are consistent with earlier XRD studies that the sample is multi-layered. This 2D peak, usually referred to as a second-order overtone of the D peak, is not assigned to the defects for their activation since its origin is attributed to the momentum conservation by two phonons with opposite wave vectors. However, the positions of both D and 2D peak are dispersive dependent on the laser excitation energy. Figure 3(b) also illustrates an analysis of the 2D peak envelope, assuming a Lorentzian peak shape. Three components are found to fit the envelope with their peaks to be located at 2675, 2700 and 2725 cm −1 .

Steady state electrical measurements
As shown in figure 4(a), in-plane dc conductivity measurements were made using the interdigitated electrode systems with dimensions of = d 0.5 mm, = W 0.82 cm and = N 5, where d, W and N are the distance between the fingers, the overlapping width and the number of fingers in the electrode system, respectively. Figure 4  comparable value of -1100 S m 1 for the room temperature conductivity is reported for a ∼25 nm thick film spin-coated with dispersions of few layered graphene in N,N-dimethylformamide (0.7 mg ml −1 ) solvent [29]. However, the resistivity of thin printed samples of UV-driven water-based graphene oxide/acrylic poly(ethylene glycol) diacrylate nanocomposite is reported to vary between 10 4 and W 10 m 6 with the increase in the film thickness from 4 to 12 μm [30]. This shows the drop cast film of the present investigation is considerably more conducting than these nanocomposites.
Four further measurements were carried out at temperatures between 268 and 330 K and the dependence of the dark resistivity r D on T is shown in figure 5(a). This behaviour of r ( ) T D can be described by a simple well-known relation in the form: where r D0 and α denote the resistance at 0 K and the temperature coefficient of resistivity of the GI sample. The ratio of the slope to the intercept at = T 0 K gives ´--( ) 3.07 0.12 10 K 3 1 for the temperature coefficient α. This value is larger than one reported for graphene nanowalls by an order of magnitude [31]. It is, however, within the same order of magnitude as other highly conducting materials such as copper and silver [32]. The existence of positive temperature coefficient may be attributed to an increase in the interlayer distance with the rise of temperature, thereby reducing the tunnelling current between the layers [33]. The contributions of both optical phonons and intervalley scattering by transverse and longitudinal modes become increasingly significant with rising temperature [34]. The value of the activation energy ΔE is estimated to be 12 meV from the slope of Arrenhius plot in figure 5(b) of s ( ) ln D as the inverse of temperature -T . 1 D =  E 12 1.5 meV represents a non-zero band gap between the conduction and valence band and it is smaller than room temperature thermal energy. The small value of D = E 12 meV in the present work implies the possible break up of lattice symmetry at the Dirac points (K) of the graphene films due to the presence of defects [35]. This observation is consistent with our earlier observations from Raman and SEM studies. ΔE is generally sensitive to the measurement environment such as the presence of moisture, local humidity and the hydrophobicity of the binder resin. A value of D = E 29 meV which is almost equal to thermal energy is reported for CVD deposited graphene films from Van Der Pauw conductivity measurement in vacuum for the comparable temperature range [36]. Both temperature coefficient and activation energy have been measured for graphene flakeand solar exfoliated reduced graphene within a similar temperature range with a view to developing wearable temperature sensors. The temperature coefficient is negative with the value of the same order magnitude as obtained in the present investigation, implying the potential applications of the present GI in the development wearable temperature sensors. Also, values of 24.19 and 90.7 meV for activation energies of graphene flakeand solar exfoliated reduced graphene respectively are higher than one obtained in the present drop-cast graphene film, indicating nonmetallic characteristics of these graphene [37].

Electrical properties under bending stress
The bending is commonly applied as strain for uniaxial tensile stress. As shown in figure 6(a), the strain was applied parallel to the length of the film so that the electrodes were not subjected to any strain. The strain is estimated in percentage from the knowledge of the substrate thickness and radius of curvature. Three radii of curvature of 2, 7 and 12 nm were used. Figure 6  is found to increase with increasing strain, implying an increase in the resistance (R BS ) on stretching. Similar behaviour was reported for CVD grown graphene layers [38]. Phenomenologically, the stretching causes the increase in length of the film with simultaneous decrease in its thickness and both these dimensional changes contribute to the increase in Ohmic resistance on stretching. The Dirac points in the strained graphene are believed to have undergone displacement from the K points, possibly opening up the energy gap. The GI film surface may be not completely flat because of imperfect adhesion to the PET substrate. The scattering of charges due to the non-flat nature of the strained GI surface may also contribute to the increase in resistance [39]. The measurements were repeated on the same film under compression over the same strain range and the results are also illustrated in in which R BC is the resistance of the compressed film.
As expected, the resistance is found to decrease with the increase in the strain. The opposite dimensional changes of the films occur on compression, reducing the Ohmic film resistances. In order to investigate the repeatability of the gauge performance, the resistance of the film (R Bw ) was measured after the withdrawal of bending stresses and the magnitude of recovery was estimated as the ratio ( ) and it is evident from figure 6(c), values of this ratio lie between 97% and 92% as the strain is reduced from 0.71% to 0.41% and the recovery is consistent between both stretched and compressed films. These results indicate a very good degree of the flexibility of the GI films, leading to possible application of these flexible GI films in resistance strain gauges for measuring strain within the range from 0.41% to 0.71%.

Photoresponse studies
The UV-vis absorption spectrum in figure 7(a) for the graphene film is due to for these interband transitions showing a pronounced and asymmetric peak at 270 nm [40]. This peak position is in close agreement with that observed for films deposited from n-methyl-2-pyrrolidone and ethanol/water based dispersions [11]. The peak is due to a van Hove singularity in the density of states, which occurs close to the hopping energy [41,42]. The considerable red-shift with respect to the absorption peak of graphene oxide at 226 nm implies reduced disruption of the π-conjugation in the multi-layered graphene film [43]. The flat absorption band over the visible region results from the linear dispersion of Dirac electrons in graphene. The weak broad absorption band (shoulder) around 353 nm corresponds to residual graphitic sp 2 domains in the samples. However, other structural parameters like the number of graphitic layers, curvature of aromatic layers, crystallite size and so on, also affect the shifting of (π-π * ) bond position. The action spectrum in figure 7(b) shows the dependence of photocurrent on the incident wavelength, at a fixed bias of 50 mV. As expected, it shows close similarities to the UV-vis absorption spectrum, with a sharp peak occurring at 270 nm. The minor differences may be attributed to scattering and recombination of charges in the GI film.
where R il is the resistance for the GI sample under illumination. This decrease in s ph by a factor of nearly 3 with an increase in temperature of 62 K from 268 to 330 K may be interpreted by the photo-thermoelectric effect due to a temperature gradient that is possibly produced between the graphene layers under illumination. The photocurrent in ordinary semiconductors is primarily due to separation of excited electron-hole pairs by a built-in electric field. A photoexcited e-h pair leads to ultrafast heating of the carriers in graphene because of strong e-e interactions [44,45].
The photoresponsivity, S, is defined as the ratio of the photocurrent to the incident optical power P i . Therefore S can explicitly be written for the interdigitated electrode system in the form: The value of photosensitivity S at 268 K is estimated to be -161 A W 1 from equation (6). As the temperature is raised to 330 K, S is found to decrease to -56 A W . 1 This value is of the same order of magnitude as one obtained for the hybrid films containing graphene quantum dots in the reduced graphene oxide matrix [46]. The specific detectivity D, can be estimated from the knowledge of S using the expression is the absolute value of electronic charge. Using the expression above, values of4.9 10 12 Jones and1.9 10 12 Jones were estimated from equation (7) at 268 K and 330 K, respectively. These values of the specific detectivity represent the capability of the devices to detect a small photo signal.
Values of the short circuit current I sc at zero bias, V=0, are found to be m 2.0 A and m 0.4 A at 268 K and 300 K, respectively, indicating that the number of photo-generated charges are smaller at higher temperatures for a given illumination strength [47]. Similarly, values of 6.7 mV and 1.6 mV are estimated for open circuit voltage V oc using I=0 for 268 K and 300 K, respectively. Figure 8(b) shows a typical current versus time cycle for a fixed bias of 1 V as the light was turned on and off. Two conducting states, high and low, of the GI films are found to exist under illumination and the recovery to nearly low conducting occurs when the illumination is switched off. The response time t R is estimated to be 42 s from fitting the expo- to experimental rise curve in figure 8(b), where i 0 is the saturation photocurrent for = t 0 at which instant light illumination was switched on at 150 s. This value is found to be comparable with those obtained for the UV sensors based upon reduced graphene oxide decorated ZnO nanostructures [48].
where i 0 is the saturation photocurrent for = t 0 at which instant light illumination is switched off. The stretching parameter g normally varies between values of 0 and 1. For g = 1, the function in equation (6) approaches classical single-exponential behaviour and differences in energy transfer processes become indistinguishable under this condition [49]. It is obvious that the decay curve in the present work is characterised by two regimes corresponding to (i) fast and (ii) slow relaxation processes. Values of the relaxation time constant, t , D were found to bé 8.9 10 s 3 and4.3 10 s 4 for the time scale regime (i) from 371 s to 463 s and (ii) from 463 s to 482 s respectively by fitting equation (8) to the experimental data. The present g = 0.82 may be ascribed to the combined effect of both defects and potential barriers between the edges of the single sheets within the overall assemblies [50].

Concluding remarks
A low cost, easy fabrication method has been developed for depositing graphene films on flexible substrates and their optoelectronic properties have been investigated for practical applications. The films studied in this article could be deposited using for example rotary screen-printing, gravure or even offset lithography which are high-speed, volume manufacturing processes. The graphene film in the present investigation is composed of multiple, randomly distributed graphene nano-sheets, in contrast to CVD produced graphene which is typically continuous layers grown directly onto a substrate. It should also be noted that a polymer binder is added to the ink. A value of´--3.07 10 K 3 1 for the positive temperature coefficient of resistivity of the graphene film is comparable to that of common metals like copper and silver and these graphene films may therefore be employed in the manufacture of thermistors for their uses as electronic components in resettable control circuits. The graphene films show a high specific detectivity, of the order of 10 12 at its maximum absorbance wavelength of 270 nm under a constant light intensity of 10 m W m −2 . Two distinct high and low resistive states have been observed. The observation that the properties of our films are similar to 'real' graphene is important because these results may lead to potential exploitation for low cost, large-scale industrially development of highly efficient, large area UV sensors using printed graphene films. Future work will include optimisation of the deposition (printing) procedure to obtain uniform films and investigation into the effect of different polymer binders with a view to improving the mechanical and chemical stability of the film. The graphene content relative to that of polymer binder and the nature of the binder will be interesting features of further investigations into design and fabrication of resistance strain gauges. It may then be possible to improve the recovery performance by using a binder with better elasticity for strain gauge transducers.