Thermoelectric properties enhancement of p-type composite films using wood-based binder and mechanical pressing

Thermoelectric generators (TEGs) fabricated using additive manufacturing methods are attractive because they offer the advantages of scalability, lower cost, and potentially higher power density than conventional TEGs. Additive manufacturing of TEGs requires active thermoelectric particles to be dispersed in a polymer binder to synthesize printable slurries, and printed films to be subsequently subjected to a long and high temperature curing to enhance their thermoelectic properties. A large amount of polymer binder present in composite films results in a sizable loss in the electrical conductivity. In addition, a long and high-temperature film curing results is a slow and energy intensive fabrication process. In this work, we demonstrate the feasibility of using a small amount (≤10−3 wt ratio) of novel nanofiber cellulose (NFC) as a binder to provide sufficient adhesion strength to hold the TE particles together in the composite films. We also demonstrate a pressure induced densification process to enhance the thermoelectic properties of printed composite films. This novel approach has the potential to fundamentally transform the manufacting method for printing TEGs by eliminating the need of long-duration and high-temperature curing. A higher applied pressure leads to a compact packing and densification of films resulting in an improvement in the electrical conductivity. The highest power factor achieved for best performing p-type thermoelectric-NFC composite film subjected to pressure induced densification is 611 μW/m-K2.

www.nature.com/scientificreports www.nature.com/scientificreports/ in the rhombohedral Bi 0.5 Sb 1.5 Te 3 single phase 10 . Also, the addition of 8 wt% extra tellurium did not change the structure of Bi 0.46 Sb 1.34 Te 3.2 . Figure 2 shows the SEM image of MA p-type Bi 0.46 Sb 1.34 Te 3.2 particles only, the particles size is in the range of 0.1~5 μm suitable for printing purposes.
TE composite films were prepared using various weight ratios (1:100, 1:1,000, and 1:10,000) of the NFC hydrogel binder and commercially available Sb 2 Te 3 active particles on chitosan coated Kevlar substrate. Kevlar is a rough substrate therefore chitosan coating was used to achieve a smooth surface to deposit NFC hydrogel and TE particles composite films 23 . For 1:100 NFC hydrogel(binder) to Sb 2 Te 3 particles wt ratio composite film, electrical conductivity values was 0.5 S/cm, Seebeck coefficient was 110 μV/K, and power factor was 0.6 μWm −1 K −2 . Subsequently, different samples were subjected to mechanical pressure of 200, 400, and 600 MPa. TE characterization were performed at room temperature. Figure 3 shows the TE properties after the mechanical pressing of Sb 2 Te 3 -NFC hydrogel composite films for various mass loading. An increase in the electrical conductivity was observed with the increase in the applied pressure for the 1:100 ratio samples. There was significant increase in electrical conductivity by varying the pressure from 0 MPa at 600 MPa, as shown in Fig. 3(a). The highest electrical conductivity was obtained for 1:1000 wt ratio of binder to TE particles at 600 MPa. No significant changes in the TE properties were observed for the applied pressure beyond 600 MPa. The electrical conductivity also improved as a result of increasing the wt% of active particles. The electrical conductivity increased significantly when binder to particles ratio was increased from 1:100 to 1:1,000. This increase is due to the lower amount of insulating binder in the 1:1,000 sample as compared to 1:100 sample. The difference in the electrical conductivity of 1:1,000 and 1:10,000 samples was not significant. We discuss the pressure-induced film densification and improvement in electrical conductivity using imaging analysis towards the end of this section. Figure 3(b) shows the Seebeck coefficient value vs. applied pressure for various mass loading of hydrogel to Sb 2 Te 3 particles. The positive value of the Seebeck coefficient confirms these are p-type particles and the Seebeck coefficient reduces with increase in the applied pressure for a given weight ratio. This observation is in agreement with findings from previous research studies that document inverse relationship between electrical conductivity and the Seebeck coefficient. Also, for a given applied pressure, a slight increase in the Seebeck coefficient is observed with the increase in the weight ratios of particles. This increase is due to more TE active particles compactly packed in the films. The power factor captures the effect of changes in both the electrical conductivity and the Seebeck coefficient. The best reported sample (1:1,000 weight ratio) has the highest electrical conductivity 233 Scm −1 , Seebeck coefficient was 104 μV/K and power factor was 214 μWm −1 K −2 at 600 MPa applied pressure.
The above results demonstrate the successful use of NFC hydrogel as a binder for Sb 2 Te 3 based thermoelectric films. In an effort to further improve the thermoelectric performance and achieve best possible results with NFC hydrogel binder, we utilize Mechanical alloy (MA) process to prepare thermoelectric particles by adding different concentrations of tellurium powder to Bi 0.5 Sb 1.5 Te 3 to obtain Bi 0.48 Sb 1.4 Te 3.12 , and Bi 0.46 Sb 1.34 Te 3.2 . The next set of experiments were motivated by results from existing literature indicating an addition of extra tellurium to Bi 0.5 Sb 1.5 Te 3 can significantly improve its thermoelectric properties 6,10,21,37 . Subsequently, we used these MA active  www.nature.com/scientificreports www.nature.com/scientificreports/ particles together with NFC hydrogel binder to dropcast the films and their thermoelectric properties were characterized. Figure 4 shows thermoelectric properties for TE composite films with binder to active particle weight ratios of 1:100, 1:1,000 and 1:10,000. Bi 0.5 Sb 1.5 Te 3 , Bi 0.48 Sb 1.4 Te 3.12 , and Bi 0.46 Sb 1.34 Te 3.2 were used as active particles. The image of TE composite film on Kevlar substrate is shown in Fig. 4(a). Because 600 MPa applied pressure yielded best TE results for composite Sb 2 Te 3 films, the mechanical pressing for all variants of BiSbTe with NFC hydrogel was performed at this pressure. Figure 4a-d show the electrical conductivity, the Seebeck coefficient, power factor and, carrier concentration respectively. The results show the electrical conductivity increased with increase in binder to particle weight ratio for all the samples. The increase in the electrical conductivity was due to the reduced proportion of insulating binder in the film. The electrical conductivity value increased when binder to particle weight ratio changed from 1:100 to 1:1000 and 1:10,000 respectively. The sample with NFC hydrogel to Bi 0.46 Sb 1.34 Te 3.2 ratio of 1:10,000 produced the best electrical conductivity value of 141 Scm −1 .
There was no significant effect of added extra Te concentration on the electrical conductivity. The electrical conductivity for all the three composites were comparable for NFC hydrogel to particle weight ratios of 1:100, 1:1,000 and 1:10000. Mostly, an increase in the Seebeck coefficient was observed with an increase in the amount of added Te. Because Te is a high Seebeck coefficient metal, the Seebeck coefficient of the composite become higher when Te is added in higher concentrations. The best Seebeck coefficient (226 μVK −1 ) was observed for the sample with binder to Bi 0.46 Sb 1.34 Te 3.2 ratio of 1:10,000. These high Seebeck coefficient values with extra Te are similar to our previous work of MA Bi 0.5 Sb 1.5 Te 3 with addition of extra Te epoxy composite films 3,10,38 . The highest power factor of 611 μWm −1 K −2 was obtained for the Bi 0.46 Sb 1.34 Te 3.2 sample with binder to particle weight ratio of 1:10,000, with a corresponding electrical conductivity of 141 Scm −1 and a Seebeck coefficient of 226 μVK −1 . The 8% extra Te films (i.e. the Bi 0.46 Sb 1.34 Te 3.2 sample) resulted in the best power factor. Also, from Fig. 4d it can be observed that the films with higher amounts of Te had lower bulk carrier concentration values. Thus explaining the slight increase in Seebeck coefficient.
Also, the best sample among various MA samples was the 8% extra Te (Bi 0.46 Sb 1.34 Te 3.2 ) sample with the binder to particle ratio of 1:10,000 as shown in Table 1. This sample achieved the power factor of about 611 μWm −1 K −2 . Overall, the power factor values increased with increase in the mass loading of particles and by addition of Te particles.
To examine the effect of addition of hydrogel on particles, XRD analysis was performed for the 1:10,000 hydro-gel_Bi 0.46 Sb 1.34 Te 3.2 sample. This sample was chosen because it has the best power factor. The pressed sample was ground into powder using a mortor and pestel and then mounted into the sample holder for XRD. As shown in Fig. 5, the XRD analysis of the 1:10,000 hydrogel_Bi 0.46 Sb 1.34 Te 3.2 sample shows evidence of amorphorization. This amorphization can be attributed to the mild sample oxidation and the presence of hydrogel in the composite system 11,39 . The reduction in the intensity of peaks can be attributed to the presence of hydrogel in the composite system. www.nature.com/scientificreports www.nature.com/scientificreports/ In addition, we examined the potential degradation of film properties due to possible oxidation of film due to its exposure to ambient atmosphere. The data is reported in the Table 2 below. The electrical conductivity slightly increased with time and then become stable which implies that there is only slight oxidation due to extended exposure of the sample to the ambient atmosphere. (The data reported in the paper was collected immediately after sample preparation).
Hydrogel is used in a small amount as a binder material for active thermoelectric particles. Because the thermoelectric properties of this composite are similar to those of other polymer-thermoelectric composites (such as epoxy-Bi 2 Te 3 composite), we use equipment setup that are routinely used for polymer-thermoelectric   www.nature.com/scientificreports www.nature.com/scientificreports/ composite materials. The role of the binder in the composite is to provide sufficient adhesion to hold the TE particles together. Because the composite films were uniform, they did not have visible cracks or flaws, and were not powdery (as shown the Fig. 6 in the manuscript) we concluded the NFC hydrogel provided sufficient adhesion to hold the TE particles together.
To study the mechanism behind the improved results seen in thermoelectric characterizations, high-magnification optical microscope and SEM images of various films were captured. The images were analyzed for the information about the distribution and the packing of the TE active particles in the binder. Figure 6(a,b) show optical microscopic images of films with binder to MA Bi 0.46 Sb 1.34 Te 3.2 weight ratio of 1:100 and 1:1,000, respectively. No mechanical pressure was applied for these two samples. These images show how the TE active particles are dispersed in hydrogel matrix. Yellow arrows show the NFC hydrogel binder and the red arrows indicate voids present in the TE composite films. A comparison of images shown in Fig. 7(a,b) reveals the films with binder to particle weight ratio of 1:1,000 has smaller amounts of NFC hydrogel binder phase than the sample with the weight ratio of 1:100. Also, the film, with binder to particle weight ratio of 1:100, has larger voids and pores than the film with the binder to particle ratio of 1:1,000. The presence of a large number of pores and voids limits the good contact between active TE particles. The electrical conductivity of TE hydrogel composite films depends on the electrical contacts between active TE particles. The electrical contact depends on the mass loading and the density of active particles. It is evident from the optical images that active particles are well connected and densely packed in samples with higher mass loading of particles. A higher density results in contact and less void space between active particles. Figure 7(c,d) show optical microscopic images of same samples as in 7(a) and (b) under 600 MPa applied mechanical pressure. As the applied pressure increased, active particles came close to the neighboring particles and started aggregating in hydrogel matrix. Additionally, particles and hydrogel interface is compressed and pores are filled with active particles resulting in better contact between particles. The outcome is a compact and dense film facilitating better electrical contact between grains of active particles. The contact between particles results in the formation of continuous path for charged carriers, which increases the electrical conductivity of the film 25,[40][41][42][43] . It is clear from these optical microscope images that uniaxial cold pressing provides a way to reduce the pores and voids and improvise better electrical contact and aggregation between the particles.
To further confirm densification and better electrical contact phenomenon, SEM was performed on pressed films. Figure 8 shows the cross-sectional SEM images of pressed Bi 0.46 Sb 1.34 Te 3.2 films at various mass loading. We picked Bi 0.46 Sb 1.34 Te 3.2 films for the SEM analysis because it attained the best electrical conductivity and power factor values. Figure 8(a-c)   www.nature.com/scientificreports www.nature.com/scientificreports/ 1:10,000, respectively, at 1 micron resolution. Figure 8(d-f) are SEM images of the same samples at 10 micron resolution. The high-resolution SEM images shown in Fig. 8(a,c) confirm that particles become tightly packed and well connected with each other when mass loading of particles was increased from 1:100 to 1:10,000. Also, the higher applied pressure reduced voids and caused neighboring particles to aggregate with each other resulting in better electrical contacts. SEM images confirm a higher mass loading and a higher applied pressure led to aggregation, better packing, and densification of films. As a result, better electrical contacts were formed and electrical conductivity improved.
To understand the effect of high uniaxial pressure application on TE composite films deposited on chitosan coated Kevlar substrates, we performed high-resolution optical image microscopy as shown in Fig. 9. We chose a 1:10000 NFC hydrogel to Bi0.46Sb1.34Te3.2 TE composite films cured at room temperature and subjected  www.nature.com/scientificreports www.nature.com/scientificreports/ them to an applied pressure of 600 MPa (for 5 minutes). The optical images of these TE composite films show no evidence of embedding of TE films in the Kevlar substrate even after 600 MPa applied pressure. Our hypothesis is that the chitosan coating on the Kevlar substrate provides a smooth surface and prevents the embedding and infiltration of TE films in Kevlar substrate. As a result, the thickness of the pressed TE composite films is reasonably uniform 23 .

Conclusion
In summary, the NFC hydrogel was successfully used as a binder to drop-cast thermoelectric films. Unlike the need of high amount of binders in previous research for good adhesion of particles, a very small amount of novel nanofiber cellulose (NFC) (≤10 −3 wt ratio) was sufficient to provide the needed adhesion strength to hold the TE particles together in the composite film. Unlike most other TE films with thermoset polymers and removable binders that require high curing temperatures and long curing time, NFC hydrogel TE composite films are cured at the room temperature in a short time to achieve good properties. Also, all samples were prepared in ambient atmosphere and dried at room temperature for 5 minutes and measured in ambient air. Therefore, the entire process of synthesizing inks, drop-casting and drying films, and pressure application can be completed within a couple of hours in ambient air. This novel manufacturing method eliminates the need of long durations for curing at high temperatures as performed by various researchers for either removing the binder or sintering the TE particles to improve TE properties. The best sample obtained using commercially available TE particles was the composite film in which the hydrogel to Sb 2 Te 3 ratio was 1:1000 and a uniaxial pressure of 600 Mpa. This sample achieved the power factor of 214 μWm −1 K −2 . The best sample among various MA samples was the 8% extra Te (Bi 0.46 Sb 1.34 Te 3.2 ) sample with the binder to particle ratio of 1:10,000. This sample achieved the electrical conductivity of 141 S/cm and the Seebeck coefficient of 226 μV/K which lead to the power factor of about 611 μWm −1 K −2 . Overall, as a result of adding extra Te to the MA Bi 0.5 Sb 1.5 Te 3 , using a high mass loading of TE particles in the binder, and the application of uniaxial pressure, a large increase in the power factor was achieved for TE composite films.

Methods
Mechanical alloying. Sb 2 Te 3 (p-type) was purchased from the Santech Inc. and was used as TE active particles for ink preparation. Composite films were printed on flexible substrates using TE inks consisting of TE particles, a suitable binder, and a solvent. Previous research results indicate that with the addition of extra tellurium to Bi 0.5 Sb 1.5 Te 3 , the TE properties of printed composite films have improved 6,10,21 . Therefore, in this work we prepared three kinds of mechanically alloyed samples: Bi 0. 5  Ink synthesis. Empirical testing suggests the particle size of the active materials needs to be in the range from 1-10 μm for printing of good quality films. To produce the average particle size of 10 μm, wet grinding was performed on MA particles. A weight ratio of 1:1:10 of the resultant MA TE particle, isopropyl alcohol, and grinding balls were placed in stainless steel jars and ball milled at 30 hz for 150 minutes. The resulting fine powder was subsequently mixed with NFC hydrogel. The NFC hydrogel was obtained from Dr. Liangbing Hu's research group. Inks of proper viscosity were prepared by mixing the fine MA TE particles with NFC hydrogel and appropriate amounts of distilled water. Various weight ratios of binder to particles are evaluated to attain the best possible TE www.nature.com/scientificreports www.nature.com/scientificreports/ properties of drop-casted TE films. The inks were thoroughly mixed using a vortex mixer (MX-S, Scilogex) for 5 minutes and were sonicated for 30 minutes using ultrasonic bath (CPS2800H, Emerson). Film fabrication. The inks were then drop-casted on chitosan coated Kevlar substrates (Kevlar substrate, Fibre Glast developments Corporation) and dried at room temperature for 5 minutes. Drop casted films were prepared by using various inks in different NFC hydrogel to particle (e.g., Sb 2 Te 3 , Bi 0.5 Sb 1.5 Te 3 , Bi 0.48 Sb 1.4 Te 3.12 , and Bi 0.46 Sb 1.34 Te 3.2 ) weight ratios of 1:100, 1:1,000, and 1:10,000. The drop casted composite films were densified by applying mechanical pressure in the range of 10 MPa to 600 MPa for 5 minutes.
Characterizations. Seebeck coefficient and electrical conductivity measurements were performed in ambient air using a custom-built setup. The electrical conductivity was measured using the van der Pauw method. The van der Pauw method measures the vertical and the horizontal resistance and calculates the sheet resistance for arbitrary and irregular shape sample 44 . To measure the vertical and the horizontal resistance, a current of 1000 µA was forced through the films and corresponding voltage drop were measured using the Keithley 2400 source meter. In order to reduce measurement errors, multiple readings of resistance were taken for each direction. The thickness of the pressed sample was measured using a micrometer (Mitutoyo). Substrate thickness (222 μm) was subtracted from the total thickness (421 μm) to find the film thickness (218 μm). In the table below, we show measurements for Kevlar substrate thickness, total thickness, film thickness, and area of one film. Sample thickness was measured at 4 different locations on the film and the film thickness variation is relatively small (<10 μm). Area of the pressed TE composite film is 35 mm 2 . The Kevlar substrate is coated with chitosan to make the rough Kevlar substrate smooth. This chitosan coating also helps in reducing the potential film embedding into the substrate. The electrical conductivity of films was obtained by calculating the reciprocal of the product of sheet resistance and thickness of each film. The Seebeck coefficient measurements were performed using a custom-built test setup to measure the voltage difference, ΔV, for a given temperature difference, ΔT. To reduce measurement errors, the system is enclosed in a thermally insulated case and protected from viable fluctuations of the ambient atmosphere. For measuring the Seebeck coefficient, a custom-built setup was utilized in which two peltiers were used to create a temperature difference across the sample by forcing voltage through the peltiers. Metal electrodes were made on top of the sample using Au paste for Seebeck and electrical conductivity measurements. The voltage and temperature probes on each of the ends were positioned close together to minimize the difference between actual and measured values. In order to further reduce the possibility of errors, multiple readings of ΔV were taken for each ΔT. The voltage difference was measured at six different values of ΔT (0 K, 1.6 K, 3. where σ is the electrical conductivity and α is the Seebeck coefficient. The carrier concentration measurements were performed using ecopia-3000.