Thermoelectric Paper: Graphite Pencil Traces on Paper to Fabricate a Thermoelectric Generator

Paper‐based thermoelectric generators are a promising and economical alternative to expensive organic conductors that are normally preferred for flexible generators. In the present work, graphite pencil traces on regular Xerox paper are successfully employed to constitute a thermoelectric generator. In conjunction with polyethylenimine polymer, the graphite traces act as both the p‐type and n‐type thermoelectric “legs,” of a graphite‐based thermoelectric generator. The fabrication method is facile and requires no conducting paste or silver paste to connect individual thermoelectric legs. A test module containing five pairs of p‐n legs is fabricated on paper to test its performance. The device produces a thermoelectric voltage of 9.2 mV, generating an output power of 1.75 nW at a temperature difference of ≈60 K. The present work demonstrates that ordinary pencil on paper may be used as the foundation for a cheap, flexible, easily disposable, and environmentally friendly thermoelectric generator.


Introduction
Thermoelectric generators can generate electricity under temperature gradients, or vice versa, and therefore they provide a pollution-free method to produce useful electric power from waste heat or solar radiation. [1] Employing thermoelectric generators to power electrolysis allows hydrogen to be produced from these renewable sources, thereby providing a means of storing the energy in order to circumvent intermittency of the available supply. [1b,2] At present, commercial thermoelectric generators are mainly based on elements such as Bi, Te, and Pb, but issues such as the high cost, scarcity, and toxicity of these elements have inhibited progress of the technology. [3] In order to overcome the problems associated with traditional thermoelectric materials, researchers have begun to

Thermoelectric Properties of the TD1 Sample
The room temperature Seebeck coefficient of an as-drawn TD1 sample was +16.1 µV K −1 , where the positive value is consistent with p-type electrical conductivity; as shown in Figure 3a, increasing the temperature up to 100 °C had a negligible effect on this value. Following PEI-treatment of the graphite trace, however, the room temperature Seebeck coefficient changed to a negative value of −21.5 µV K −1 , which is indicative of the successful conversion from p-to n-type behaviour. Furthermore, it is evident from Figure 3b that the Seebeck coefficient of the PEI-treated trace exhibited a small but significant variation with respect to temperature, increasing in magnitude to a value of −23.2 µV K −1 at 100 °C.
Alongside the Seebeck coefficient, the electrical conductivity and power factor of a thermoelectric material are critical to its overall performance as part of a device; these measurements are plotted from room temperature to 100 °C for the as-drawn and PEI-treated TD1 devices in Figure 3c,d, respectively. Despite the considerable increase in resistivity yielded by PEI-treatment, with the room temperature conductivity decreasing from 3.32 × 10 2 S m −1 for as-drawn graphite to 2.24 × 10 2 S m −1 , the power factor was correspondingly improved by the treatment process: the measured value of 85 nW m −1 K −2 in the case of untreated graphite was increased to 100 nW m −1 K −2 after modification by PEI. Moreover, increasing the temperature resulted in a still greater difference between the power factors of the two samples, as shown by Figure 3d.The strategy of changing conductivity type of a pencil trace was inspired by earlier reports where researchers have changed the behaviour of CNT's from p-type to n-type by using PEI and other polymers. [16] Reproducibility of the thermoelectric properties is also an important aspect of generator design. To test, measurements have been repeated 2-3 times on the pencil trace to ensure the reproducibility in the values and observed stable values over repeated measurements. It was also noted that there were small deviations in the Seebeck values from one sample to another but this change was found to be less than ± 3 µV K −1 .
The thermoelectric figure of merit (ZT) of the as-drawn and PEI-treated graphite traces have been estimated using the standard expression, ZT = S 2 σT/κ, where S, σ, κ, and T are the Seebeck coefficient, electrical conductivity, thermal conductivity, and temperature respectively. Thermal conductivity value (inplane, κ = 3.5 W m −1 K −1 at 100 °C) has been obtained from the literature on reduced graphene oxide films having similar electrical conductivity. [17] The estimated ZT values of the as-drawn and PEI-treated graphite trace at 100 °C are 1.01 × 10 −5 and 1.32 × 10 −5 respectively, which are very low in contrast to widespread thermoelectric materials but comparable to previous reports. [17,18] A key advantage of employing paper as a foundation for thermoelectric devices is its flexibility, as pictured in Figure 4. However, in order for an application to exploit this property over the long-term, it is necessary for the device to demonstrate sustainable performance through multiple cycles of distortion. To this end, changes in the Seebeck coefficient and electrical conductivity of a TD1 sample were measured over multiple bending cycles, as depicted in Figure 5a; separate but identical TD1 type samples were used for the Seebeck coefficient and electrical conductivity measurements, which are plotted in Figure 5b,c, respectively. The electrical conductivity decreased by ≈2.2% over 600 bending cycles, although there was concurrently negligible variation in the Seebeck coefficient within the experimental errors. The small decrease in electrical conductivity may be attributed to the formation of fractures between graphite structures during the cyclical bending process; nevertheless, the durability of the graphite-based thermoelectric generator is demonstrated by the low magnitude of this conductivity variation, despite the substantial and repeated distortion applied to the device. Further, the Seebeck coefficients of the TD1 type samples were measured under different bending conditions   (≈45° and ≈90°) and it was observed that the values were not much affected but a slight change of about 1 µV K −1 ( Figure S1, Supporting Information).

Characterization of the Graphite Trace on Paper
In order to characterise the surface morphology of the graphite trace, SEM was employed; alongside an optical image of the underlying Xerox paper in Figure 6a, an SEM micrograph of the paper in Figure 6b is accompanied by corresponding images of as-drawn graphite trace in Figure 6c,d and a PEI-treated trace in Figure 6e,f. It is clear from the images that the porous arrangement of cellulose paper fibres was completely covered by each graphite trace, and moreover there are no significant visible differences between the layered graphite structures of as-drawn and PEI-treated samples. SEM images of higher-magnification that show thick graphitic flakes/layers can be seen in Figure S2 (Supporting Information). From the cross-sectional SEM image pictured in Figure S3 in the Supporting Information, an estimate of 3.20 (±0.05) µm is obtained for the thickness of an asdrawn graphite trace.
Adv. Mater. Technol. 2020, 5,2000227   The chemical surface compositions of as-drawn and PEItreated pencil traces were analyzed using XPS; survey spectra of these materials are shown in Figure 7, while higher resolution C1s spectra are displayed in Figure 8a,b, respectively, alongside the corresponding N1s spectrum of PEI-treated graphite in Figure 8c. The clear appearance of the N peak in the survey spectrum of PEI-treated graphite confirms the incorporation of PEI at the sample surface, while the presence of CN chemical environments is further evident from the large component at 286.3 eV in the C1s spectrum of PEI-treated trace. From the corresponding N1s spectrum depicted in Figure 8c it is apparent that the majority of the N atoms were contained within the amine groups of PEI and pyrrolic environments at the graphite surface, with relatively little substitution of C atoms for N within the surface graphite lattice. Indeed, deconvolution of the C1s spectra also indicates that the composition of the graphite lattice was not altered significantly by the PEI-treatment: despite the emergence of CN chemical environments, the intensity ratio of the component associated with sp 2 carbon atoms, at 284.4 eV, [19] and the other surface environments did not vary appreciably between the two samples. Alongside the CN groups identified in Figure 8b and the sp 2 environments of both spectra, a common component at 284.8 eV can be assigned to C atoms connected in an sp 3 configuration, [20] while peaks at 286.1 and 287.2 eV are characteristic of the C atoms in ether, alcohol or carbonyl groups. [20c,21] As shown in the figure, the N1s peak can be split into two peaks with a major peak centred at 400.0 eV and a minor peak at 401.4 eV. It is normally proposed that the three predominant bonding configurations are pyridinic N, pyrrolic N, and graphitic N when nitrogen is doped into graphene or graphite. [22] In the present case, the main peak at 400.0 eV can be assigned to pyrrolic N bonding and the small peak of 401.4 eV to the graphitic N bonding. [23] It is well studied that the conversion of p-to-n-type conductivity is due to the electron-donating capability of PEI polymer through the amine groups. [24] The n-type doping is more favourable in case of PEI polymer as it contains a high density of amine groups compared to other polymers. [24]

Performance of the Thermoelectric Generator
Through use of the experimental arrangement pictured in Figure 9, voltage, current and power outputs were measured as a function of temperature gradient for a TD2 configuration of p-and n-type legs, as plotted in Figure 10. Figure 10a shows that the thermoelectric generator produced a linearly varying potential of 0.15 mV K −1 for a temperature gradient ΔT of up to 60 K, indicating that the Seebeck coefficient remained approximately constant over this ΔT range. Correspondingly, the estimated power outputs of the generator along with respective voltage outputs are plotted in Figure 10b. The variation of resistance at different temperature gradients with respective power outputs can be seen in Figure S4   The power produced by the generator is comparable to the power output of alternative flexible devices reported previously in the literature, but offers particular advantages with respect to the overall device size and ease of fabrication. For example, the power output is consistent with values obtained from a reduced graphene oxide (rGO) device, which yielded 1.36 nW for a 50 K temperature gradient, but one should note that this arrangement required 24 pairs of p-n legs, in contrast to the five pairs utilised in the present work. [25] The generated power is also of similar magnitude to alternative systems of far greater cost and complexity; for instance, one pair of p-n legs composed of chemically exfoliated WS 2 (n-type) and NbSe 2 (p-type) nanosheets produced 0.63 nW from a temperature gradient of 60 K, which extrapolates to 3.15 nW in the case of a five-leg module. [26] Similarly, a configuration containing 4.5 Bi 2 Te 3 /Bi 0.5 Sb 1.5 Te 3 p-n pairs produced a power of ≈6.5 nW for a ΔT value of 60 K, [27]    device yielded 16 nW under application of the same temperature gradient; [27] these reported values are included in Table 1.
While these values are superior to the power output achieved in the present case, the materials and fabrication procedures utilised in these studies are notably more expensive than those employed herein.

Conclusion
The present study demonstrates a facile and novel approach to produce thermoelectric generators using ordinary pencil traces on Xerox paper, wherein it has been shown that graphitic p-n junctions may be reliably achieved via a rudimentary treatment of the graphite trace with PEI polymer. Although nature of the protocol is simple and low-cost, the power output generated by the thermoelectric generator module comprised of five graphitic p-n pairs is low. Further improvements are possible by tuning the electronic properties of the starting graphite material either by alloying or doping with other elements or compounds.

Experimental Section
Thermoelectric Generator Fabrication: Commonly available HB grade graphite pencil and office Xerox paper (80 g m −2 ) were obtained from a local stationery supplier, while polyethyleneimine (PEI, branched M.W. 600) and ethanol were purchased from Alfa Aesar; all of these materials were used without any further purification. As a foundation for the thermoelectric generators, graphite traces were drawn by hand onto Xerox paper, with multiple layers applied in order to achieve high connectivity between the graphitic structures. To alter the majority charge carriers in the graphite traces and thereby transform the material from p-type to n-type, they were treated with a solution of PEI (2 mL) in ethanol (20 mL); this volumetric ratio was selected following an optimisation procedure outlined in the Supporting Information, wherein Figure S5 in the Supporting Information compares various electrical  Trace Design 2 (TD2): Optimisation of thermoelectric devices containing multiple p-n junctions was achieved using a track in the form of a "square wave," as illustrated in Figure 1 alongside a schematic of the PEI immersion step. The n-type section of this design was composed of five rectangular graphite traces, or "legs," of approximate width 2 mm and length 15 mm, with a separation of ≈6 mm between legs. In order to ensure homogeneity within the design, the resistance along the length of each leg was measured and approximately equal resistance was achieved in all cases. After immersing this design in ethanolic PEI solution to achieve n-type characteristics, the five legs were connected by five parallel and uniformly spaced p-type graphite legs of identical length and width. The completed form of TD2 is shown in Figure 2, which displays photographs of the completed thermoelectric module; notably, it is evident from the image in Figure 2c that the p-type and n-type legs could be distinguished from a marked difference in their reflective properties, with the n-type legs appearing distinctively duller than the adjoining p-type traces.
Thermoelectric Testing: The as-drawn TD1 sample was cut from the paper and the Seebeck coefficient was measured by applying a temperature gradient along its length using a home-built apparatus. [10e] The electrical conductivity of the trace was measured using the standard four-probe method.
Current and potential measurements were carried out using HMC 8012 DMM digital multimeters and regulated power supply. The thermoelectric performance of the TD2 generator was tested by applying different temperature gradients by mounting it on a hot plate.
Material Characterization: The as-drawn and PEI-treated graphite traces were examined by scanning electron microscopy (SEM) using Hitachi S4800 FE-SEM system and X-ray photoelectron spectroscopy (XPS) using Kratos Axis Supra system having a monochromatic Al-Kα X-ray source. For XPS scanning, samples of pencil trace on paper (5 mm x 5 mm) were mounted on copper tape. To achieve electrical contact between the pencil trace and the stage, copper clips were screwed in; the XPS scanning's were performed over an area of 700 × 300 µm. Survey scanning was performed across energy range of 0-1200 eV at a step size 1 eV using pass energy 160 eV and dwell time 100 ms. Core level spectra were recorded using pass energy 20 eV, dwell time 1000 ms, with a step size 50 meV. Charge compensation was achieved with a charge neutralizer working at a potential of 3.3 V with filament bias of 1.0 V and filament current of 0.4 A. All the measured data of the core level scanning were fitted with Gaussian-Lorentzian components.