Tunable Conductive Composite for Printed Sensors and Embedded Circuits

Multimaterial 3D printing is an attractive route for low‐cost fabrication of electronic systems having different types of embedded devices. Herein, tunable thermoplastic polyurethane (TPU)‐based conductive composite filaments are presented for development of either strain sensors or different circuit elements. The filaments are developed with two filler materials, namely, silver and multiwalled carbon nanotubes (MWCNT). The influences of filler aspect ratio (AR), concentration, functionalization, and morphology on the composites' mechanical, thermal, and electrical properties are studied. Printed tracks of the 10 wt% high‐AR MWCNT/TPU filament exhibit a maximum electrical conductivity of 0.92 S cm−1 and withstand powers >1 W and currents >100 mA. The filament shows negligible change in impedance over the frequency range 1 kHz–1 MHz and a change in the resistance of <5% with 90° bending. Conversely, printed tracks using filaments with 3 wt% low‐AR MWCNT exhibit a change in resistance of ≈30% with 90° bending, allowing a clear distinction between various bending angles, and thus could be used for embedded strain/bend sensors. These results suggest that, with the correct optimization, multimaterial additive manufacturing can be utilized with tunable conductive filaments to fabricate complex 3D electronic systems by constructing reliable circuit tracks, bendable interconnects, and sensors.

Multimaterial 3D printing is an attractive route for low-cost fabrication of electronic systems having different types of embedded devices.Herein, tunable thermoplastic polyurethane (TPU)-based conductive composite filaments are presented for development of either strain sensors or different circuit elements.The filaments are developed with two filler materials, namely, silver and multiwalled carbon nanotubes (MWCNT).The influences of filler aspect ratio (AR), concentration, functionalization, and morphology on the composites' mechanical, thermal, and electrical properties are studied.Printed tracks of the 10 wt% high-AR MWCNT/TPU filament exhibit a maximum electrical conductivity of 0.92 S cm À1 and withstand powers >1 W and currents >100 mA.The filament shows negligible change in impedance over the frequency range 1 kHz-1 MHz and a change in the resistance of <5% with 90°bending.Conversely, printed tracks using filaments with 3 wt% low-AR MWCNT exhibit a change in resistance of %30% with 90°bending, allowing a clear distinction between various bending angles, and thus could be used for embedded strain/bend sensors.These results suggest that, with the correct optimization, multimaterial additive manufacturing can be utilized with tunable conductive filaments to fabricate complex 3D electronic systems by constructing reliable circuit tracks, bendable interconnects, and sensors.low gauge factor and stiffness are limitations. [13]To overcome these limitations, conductive polymeric nanocomposites are widely investigated by researchers to develop piezoresistive devices. [14]FDM-printed strain sensors with combinations of base polymers (acrylonitrile butadiene styrene, polylactic acid, TPU, etc.) and conductive fillers (MWCNT, CB, graphene, silver nanoparticles, etc.) have been researched and demonstrated. [15]The use of CNTs has led to the development and demonstration of high-performance polymer/CNT-based strain sensors. [16]pecifically, CNT/TPU composites have been shown to have good properties for strain sensing in biomedical applications. [17]unctionalization of CNTs is quoted to enhance the homogenous dispersion of the conductive nanofiller in the nonconductive polymer.[16b]Functionalization of CNTs can be done using either covalent functionalization where the structure is altered by adding a connecting group, or non-covalent functionalization where functional groups are attached based on supramolecular complexation using various adsorption forces.Several approaches are based on non-covalent functionalization techniques; however, the interfacial interaction of non-covalent functionalized CNTs with polymer is weak compared to covalent functionalization CNTs. [18]Since CNT type, size, and structure greatly affect the electrical properties of the polymer/CNT nanocomposite, this article explores MWCNTs with different aspect ratios (ARs) and functionalization to determine the best performing strain sensor.
For bendable interconnects, TPU can be stretched and bent easily while returning to something very close to its original shape upon the removal of stress.TPU is also processable as a melt.This makes it a suitable base polymer for FDM 3D-printable bendable interconnects.TPU was chosen because of its high melting point suggesting that its conductive composites can operate under a wide range of currents and voltages providing versatility in circuit design.It was also chosen due to its superior bendability and durability.Silver flakes were chosen as they are highly conductive and do not oxidize in the way copper fillers do during printing. [19]Their stability allows them to retain their conductivity after printing.They were also preferred over other morphologies (e.g., nanowires [NWs]) of silver.This was because they exhibit better dispersion and less agglomeration in polymer matrices compared to NWs. [20] Despite Ag NWs offering a lower percolation threshold and higher conductivity values for the same weight percent (wt%), they are more prone to entanglement whereas the flakes are expected to flow across one another improving dispersion.MWCNTs were chosen due to their high electrical conductivity, excellent mechanical properties, and versatility with regards to processing conditions.Because of these factors, non-functionalized and covalently functionalized MWCNTs were also chosen as filler materials for conductive filament fabrication and testing.It may be noted that the conductivity values of the composites developed in this work are not high enough to be considered a practical interconnect materials at this stage.Nonetheless, the presented work has established a foundation for further optimization and development of composites with higher conductivities.The results obtained showcase the advantages and possibilities that arise from tuning the conductive composites for multi-material additive manufacturing of electronic systems.

Materials
TPU pellets (RepRap World B.V., Netherlands) were purchased and used as the base polymer for the composites.The conductive fillers were purchased from different suppliers: higher AR MWCNTs (Cheap Tubes, USA), lower AR MWCNTs (Sigma Aldrich, UK), functionalized MWCNTs (COOH graphitized) (Cheap Tubes, USA), silver flakes with a diameter size distribution of 80% < 20 μm (Alfa Aesar, USA), and silver flakes with diameters between 4 and 8 μm (Alfa Aesar, USA).Table 1 shows a comparison of the ARs of the MWCNTs used in this work with those from other TPU/CNT filaments reported in literature.N-N-Dimethylformamide (DMF, Sigma Aldrich, UK) was used as the solvent for formulating the composites.NinjaFlex TPU 85 A (NinjaTek, USA) was used to print the nonconductive parts of the test structures.
The commercial and custom filaments were printed using an Ultimaker S5 3D printer (Ultimaker B.V., The Netherlands).The nonconductive TPU filament was extruded through a 0.4 mm nozzle while the conductive filaments were extruded through a 0.6 mm nozzle.The 0.6 mm nozzle was suitable for use with abrasive materials and composites as it had a ruby tip which offered durability and reliable printing of composites.

MWCNT/TPU Filaments
Using an ultrasonic bath (c575t, Camsonix, UK) at 550 W, 37 kHz, and 25 °C for 1.5 h, the MWCNTs were first dispersed well in the DMF solvent.TPU pellets were then added slowly to the DMF dispersion at 70 °C with magnetic stirring for 6 h until they fully dissolved (10 g of TPU per 100 mL of DMF).The pellets were not all added at once to prevent them from coagulating and not dissolving.The DMF/TPU/MWCNT solution was then poured in a glass tray and dried at 100 °C on a hot plate for 18 h in a fume hood.While drying, the solution was manually stirred every hour for the first 3 h to ensure homogenous dispersion.After drying (DMF evaporation), the remaining flexible sheet of the TPU/MWCNT composite was extracted from the tray and was cut into 5 mm Â 5 mm pieces.The wt% of the composites refers to that after solvent evaporation.The small pieces were dried in an oven at 100 °C for 30 min.Finally, the pieces were fed into a single-screw benchtop extruder (EX2, Filabot) at 192 °C and 26 rpm through a 2.85 mm diameter nozzle to produce the conductive filament.For each filament, 25 g of conductive TPU pieces were fed into the extruder to obtain a usable filament length for printing.Images of this process are shown in the Figure S1, Supporting Information.

Silver (Ag)/TPU Filaments
TPU pellets were first dissolved in DMF at 70 °C on a hot plate via magnetic stirring for 6 h (10 g of TPU per 100 mL DMF).The TPU pellets were slowly added and dissolved as in the process for preparing the MWCNT/TPU filaments.After dissolving the TPU, the silver flakes were added to the solution and stirred for 3 h to adequately disperse the flakes.The same drying, cutting, and extruding techniques were then performed on the mixture as performed with the MWCNT/TPU filaments.The same extrusion settings and mass of conductive TPU pieces were also used.

MWCNT/Ag/TPU Filaments
The same first two steps of MWCNT/TPU filament preparation were performed (MWCNT dispersion using ultrasonication and dissolving the TPU pellets in the DMF).After all the TPU pellets were slowly added and dissolved, the same silver flake addition, drying, cutting, and extruding techniques were performed as with the Ag/TPU filaments.The same extrusion settings and mass of conductive TPU pieces were also used.Figure 1 summarizes the filament preparation steps for all three conductive filament types.A total of 17 different conductive filaments were made with a TPU base polymer and conductive fillers of different materials, functionalization, morphology, AR, and concentration.The filaments made are summarized in Table 2.

Design and Printing of Test Structures
Three test structures, with an embedded conductive track in an insulating TPU package, were printed for each composite.First, two layers of the nonconductive NinjaFlex TPU were printed using the 0.4 mm nozzle at 235 °C (0.2 mm layer height).The conductive track was then printed on top of this substrate.The conductive tracks had a length of 20 mm, a width of 4 mm, and a thickness of 0.6 mm.The conductive filaments were printed using the 0.6 mm diameter nozzle with a layer height of  0.2 mm.Therefore, three conductive layers were printed for each track.After the first two layers of the conductive track were printed, the print was paused to allow for wire bonding.External wires were attached to the ends of the tracks using conductive silver paint (RS Pro 186-3600, RS Components, UK).The silver paint was dried for 1 h with the test structure on the print bed and the bed temperature set to 60 °C.Once the silver paint was dried, the third conductive layer was printed on top securely bonding the external wire to the printed conductive track.After the conductive layers were printed, they were embedded with two layers of NinjaFlex printed on top.
The fabrication process for printing the test structures is shown in Figure 2. The print temperature of each conductive filament was varied with its effect on printed track conductivity measured.

Filament Diameter, Morphology, and Thermal Properties
For each filament, 10 measurements of its diameter were taken using vernier callipers.These measurements were used to determine the average diameter of each filament compared to the 2.85 mm target as well as the diameter consistency along the filament's length.The results for these measurements are shown in Figure S2, Supporting Information.The distribution of the filler particles in the composite was important in insuring homogeneity in the extruded filament and printed parts.An optical microscope (Eclipse LV100NP, Nikon, Japan) fitted with a camera (MC170 HD, Leica, Germany) was used to obtain images of the composites after casting, filament extrusion, and printing to determine filler distribution.These images are shown in Figure S3 and S4, Supporting Information.It was important to optimize the size and shape of the cut composite pieces after drying for uniform printing (Figure S5, Supporting Information).Scanning electron microscope (SEM) images were also obtained for determining the filler morphology and distribution post printing for the Ag/TPU composites with the images shown in Figure S6, Supporting Information.
X-ray diffraction (XRD) measurements were performed on some of the composites to identify composition across various fabrication stages.For the Ag-MWCNT/TPU composites, XRD was used to confirm the presence of Ag flakes and ensure there was no filler loss during fabrication.For the COOH-MWCNT/TPU composites, XRD was used to confirm the integrity of the functionalization after filament extrusion (Figure S7, Supporting Information).
Thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) were performed on composites to observe different filler and processing effects on the composites' thermal properties.These measurements provided insight to the thermal properties of the TPU-based conductive composites considering the effects of filler concentration and functionalization, filler morphologies and hybrid interactions, and the printing effect through a heated FDM nozzle.The MWCNT composites considered for thermal characterization were those with high AR MWCNTs.In the TGA measurement, the composites were tested at a ramp of 10 °C min À1 .In the DSC measurements, the composite samples were measured under nitrogen gas by default.But the 7 wt% MWCNT/TPU and the 5 wt% COOH-MWCNT/TPU composites were measured under argon gas.

Electrical Conductivity of the Filaments
To measure the conductivity of the filaments, five locations were specified along the length of each filament.At each location, silver paint (RS Pro 186-3600, RS Components, UK) was applied on the surface of the filament at two points a few centimeters apart.The silver paint was used to decrease the contact resistance between the filament surface and the probes.The silver paint was left to dry at room temperature for 1 h and was then dried in an oven at 70 °C for another 1 h.At each location, the distance between the two silver points and the diameter of the filament between them were measured to accurately calculate the resistivity (and conductivity) of the filament at that location.The resistance between two points at a given location was measured using an Agilent 34461 A digital multimeter (Keysight Technologies, Santa Rosa, CA, USA).The conductivity values at the five locations for each filament were averaged.For the printed embedded tracks, the two-wire resistance was also measured using the digital multimeter.

Electrical Performance of the Printed Structures
Current was applied to the printed conductive tracks and the voltage across them was measured using a precision source measurement unit (SMU) (Keysight B2912A, Keysight Technologies, Santa Rosa, CA, USA).The temperature of the embedded printed tracks during electrical performance tests was measured using an infrared (IR) thermometer which was used to scan the top surface of the samples from a fixed distance.For incremental current tests, the starting current applied to the tracks was 1 mA.This current was increased incrementally every 10 s until the limit of the SMU was reached.For the constant current stress test, the current from the SMU was set to values of 50, 60, 80, and 100 mA and applied for 60 min at each value.
The impedance of printed tracks with different filler concentrations and ARs was measured to observe the composites' electrical performance at different frequencies for potential application in a wide range of circuits.A frequency sweep was performed using an LCR meter (E4980A, Keysight Technologies, USA) with the measured samples placed in a Faraday cage to minimize the effect of surrounding noise on the measurements.

Bending Tests of the Printed Structures
The printed test structures were fastened onto rigid platforms that were attached to two stepper motors.The linear stepper motors were controlled using a computer running a custom LabVIEW program.As the motors pushed the platforms closer to each other, the test structure would bend upward.When they moved further away from each other, the sample would go back to its rest (non-bent) state.The electrical resistance of the test structures was measured during bending using an Agilent 34461 A digital multimeter (Keysight Technologies, Santa Rosa, CA, USA).

Morphology and Composition of 3D-Printable Filaments
The effect of filler morphology and concentration on the printability of the filaments is studied.It is to note that all composites apart from the Ag/TPU-based filaments were printable and studied as potential interconnects.For interconnects, we considered if the composite was printable and maintained its conductivity after printing.The printability of the Ag (30 wt% 80% < 20 μm)/ TPU filament was very poor as its extrusion through the nozzle could not be classified as continuous.The printed tracks from this material did not achieve high enough electrical conductivity to be used as a printed interconnect.In the performed tests, this was not printable and therefore, was not further explored for interconnect purpose in this work.The printed track was not conductive and as can be seen from Figure S6, Supporting Information, the nozzle was easily clogged, and very small amounts of silver flakes were observed in the printed structure from the SEM images.The SEM images also reveal why the printed track was not conductive as the eventual silver content left in the composite was very low.Despite the filament being printed using a 0.8 mm nozzle, the silver content was very high and caused complete clogging of the nozzle after only a few minutes of printing.The high silver content affected the viscoelastic flow of the TPU and sintered in the high-temperature nozzle.This led to the nozzle clogging.Due to these factors, the Ag/TPU was not pursued for further characterization (unlike the MWCNT/TPU composites) as they were not practical for the printing of interconnect and circuits.
Figure 3 shows a comparison of the XRD results between the filaments with different concentrations and between the casted, filament, and printed forms of the Ag-MWCNT/TPU composites.Grease was used to hold down the samples on the plate for measurements and so the peaks associated with it in the graph should be neglected.TPU was seen to have a broad peak at 21°which was also seen for the other composites around the same angle albeit shifted slightly.It is to note that the intensity of this broad peak at 21°is lower for the sample with 10 wt% compared with the other samples.The decrease in the XRD peak intensity in general could be due to many factors such as experimental issues because of the sample alignment (height alignment), amount/volume of the material under investigations, etc.In the present case, it could be because of decrease in amorphous TPU content due to higher filler concentration.Further investigation is needed to understand why the intensity of the 10 wt% composite is lower at the 21°peak.Nonetheless, the measurements of interest for this study are those of the filler particles.The three conductive composites with varying filler concentrations exhibited peaks at 38.73°, 44.97°, and 65.04°.These peaks are very similar to those reported in the academic literature for silver flakes and are associated with 111, 200, and 222 crystal planes, respectively. [21]Between these three filaments, on the Ag-MWCNT/TPU with 10 wt% filler exhibited the 002-crystal plane peak associated with MWCNTs and was present here at 26.12°. [22]his is likely due to the low concentration of MWCNTs in the 6 and 7 wt% filaments.The other peak associate with MWCNTs at around 43°(100) was masked for all measurements due to the large peak contributed by the grease adhesive material.
Comparing the presence of the fillers between the composite at different stages of the fabrication process and after printing, the characteristic peaks for the Ag flakes were present in all three forms of the 10 wt% composite.The 10 wt% samples were chosen for the fabrication steps because at this concentration, the filament had the highest conductivity compared to the 6 wt% and the 7 wt% samples (see Figure 5).Due to the intention of using this filament as a bendable interconnect, only the composite with the highest conductivity was considered for this study as it would be the only one of interest to fabricate given the low conductivity values of the other filaments in this configuration.The casted and filament forms exhibited larger 111, 200, and 222 peaks compared to the printed form of the same material.This is attributed to the fact that the measurement was taken from the filament's surface and not its cross section giving the impression that less silver flakes were present on its surface.All three forms exhibited the MWCNT 002 plane peak which meant that there was no severe loss of MWCNTs during fabrication and printing.Nonetheless, the peak for the casted form is higher than that for both the filament and printed forms signifying that there was some MWCNT loss during filament extrusion.
The XRD measurements performed on the filament and printed forms of the 10 wt% high AR MWCNT/TPU composite indicated a slight decrease in filler concentration post printing.Moreover, the measurements performed on the functionalized and non-functionalized MWCNT composites at 5 wt% indicated that the functionalization remained and was not destroyed by extruding the composite through the single-screw extruder (Figure S7, Supporting Information).

Thermal Characterization of the Filaments
From TGA, information about the polymer composite's thermal degradation can be extracted.A summary of the key parameters from the TGA measurements is presented in Table 3. From the table as well as Figure S8, Supporting Information, a comparison can be made with respect to increasing filler concentration for the MWCNT/TPU and Ag-MWCNT/TPU composites in filament form.For both sets of filaments, increasing the filler concentration decreased the onset degradation temperature.The increase in conductive filler concentration contributed to the destruction of the polymer's network structure therefore, allowing it to start degrading at lower temperatures.Increasing the filler concentration also reduced the temperatures at which maximum rate of decomposition occurred (first and second peaks).As expected, the percentage weight loss was higher for higher concentration filaments.However, in all cases, the remaining weight percentage after the experiment was higher than the filler wt% signifying the potential presence of inorganic residues in the polymer or char which would have formed during the decomposition of the material.The first peak did not go back down to the baseline before the second peak arose for the filaments, indicating that maximum decomposition rates were part of the same decomposition step.
The thermal degradation of composites after printing and functionalization of fillers is analyzed and shown in Figure S9 and S10, Supporting Information.For the 10 wt% MWCNT/ TPU and 5 wt% COOH-MWCNT/TPU composites, the printed structures exhibited a lower onset degradation temperature as well as a slightly higher temperature for the first maximum degradation rate peak compared to their filament forms.The former exhibited a slight increase in the second peak temperature whereas the latter exhibited a slight decrease in it.The 10 wt% Ag-MWCNT/TPU and 7 wt% MWCNT/TPU composites showed opposite results.In this case, the printed parts of the composites exhibited an increase in the onset degradation temperature and a decrease in temperature for the first maximum degradation rate peak.The silver-based composite had a higher second peak post printing whereas the 7 wt% MWCNT composite had a lower second peak temperature after printing.In all cases, there was a slight decrease in weight loss percentage of the measured samples post printing.It is clear from the results that printing affects the thermal properties of the conductive composites; however, since the printing parameters were different for each filament, it is not possible to draw up a concrete conclusion or find a clear trend as to what the effect is.For this study, further experiments are needed to produce large set of samples with similar printing parameters and conditions.
The effect of having a hybrid Ag-MWCNT composite with two different filler materials and morphologies compared to a single MWCNT-type filler was analyzed.From Table 3 and Figure S11, Supporting Information, the results of the TGA measurements for the 10 wt% MWCNT/TPU and 10 wt% Ag-MWCNT/TPU composites can be seen for both filament and printed forms.In the filament form, the hybrid filler (Ag-MWCNT) compositesbased filaments exhibited a slightly lower onset degradation but in the printed form, they exhibited a higher onset degradation temperature.In both forms, the silver-based composite had a lower first peak temperature and a higher weight loss percentage.For the second peak, the temperature was the same for both composites in the filament form but was slightly higher in the printed forms for the silver-based composite.The thermal properties of the TPU composites clearly differ when considering fillers of different morphologies and materials.
Finally, covalent functionalization of the MWCNTs appeared to improve the thermal stability of the composites compared to non-functionalized MWCNTs for the same filler concentration.The onset, first peak, and third peak temperatures for the 5 wt% COOH-MWCNT/TPU filament were all higher than those for the 5 wt% non-functionalized MWCNT/TPU filament.The bonds between the functionalized MWCNTs and the TPU polymer matrix improved the thermal stability of the composite and added a third maximum rate of decomposition peak at 427.3 °C.The weight loss percentage for the functionalized filler composite was less than that for the non-functionalized filler composite.This reduced weight loss signifies the presence of a larger amount of residue when running the TGA test on the functionalized filler composite.This is likely because of the enhanced filler-polymer bonding in the functionalized filler composites.
From DSC (Figure 4 and S12, Supporting Information), information about a composite's glass-transition temperature (Tg) and other potential endothermic or exothermic peaks can be extracted.In all cases, the typical endothermic hook was observed as a drop in the heat flow at the start of each measurement.A difference in baseline of either being around À1.5 or around À0.5 was observed for the measured composites.This was due to using two different instruments which offer different sensitivities.However, the location of the baseline is irrelevant to the thermal characterization as the important information lies in at what temperature the thermal events occur and by how much they deviate from the baseline.
For the MWCNT/TPU filaments, increasing the filler concentration did not alter the temperature at which the main endothermic events occurred significantly (Figure 4).The glass-transition midpoint temperature was measured as %93.4 °C for the 3 wt% filament and as %94.7 °C for the 10 wt% filament.This small increase in Tg with higher filler concentration could be due to reduced mobility of molecules in TPU or due to measurement errors.The main endothermic event shifted slightly to a lower temperature with increased filler concentration.The maximum peak temperatures were %207.7,%202.2, and %182.9 °C for the 3, 7, and 10 wt% filaments, respectively.The thermal event for the 7 wt% filament is more pronounced.This could be because this filament was run on a different DSC instrument than the other two filaments which potentially had greater sensitivity.For the Ag-MWCNT/TPU filaments, increasing the filler concentration also marginally affected the main endothermic events by causing a potential shift to occur at lower temperatures.There is a shift of the glass-transition midpoint from 97.3 to 92.9 °C and then to %92.1 when increasing the concentration from 6 to 7 wt% and to 10 wt%.We have also observed a small shift in the main endothermic event occurring at %209, %205.9, and %205.6 °C for the 6, 7, and 10 wt% filaments, respectively.The main endothermic peak for the 10 wt% MWCNT filament appeared to be relatively shallow compared to the 3 wt% MWCNT filament indicating a lower amount of heat absorbed by the polymer at the higher loading to facilitate a phase change.Similarly, for the 7 wt% Ag-MWCNT and 10 wt% Ag-MWCNT composites, the latter exhibited a shallower peak at the main endothermic event.In both cases, the polymer chains potentially had weaker bonds between them at higher concentrations.The shift in the Tg value signifies a change in the mobility of the polymer molecules.It appears that for the MWCNT-based filaments, increasing the filler wt% have made it more difficult for the polymer molecules to move and thus, there was slight increase in Tg was observed with higher filler loading.In contrast, opposite trend, i.e., decrease in Tg was observed for the Ag-MWCNT based filaments with higher filler loading.
Comparing the filament and printed forms of the 10 wt% MWCNT/TPU and 10 wt% Ag-MWCNT/TPU composites, differences in the thermal profiles can be seen in Figure S9, Supporting Information.It appears that for both composites, the printing of the materials shifts the Tg midpoint to a higher temperature and the main endothermic peak to a lower temperature.There are also differences in the thermal profiles when comparing the use of hybrid fillers.There is influence from each filler's individual content in the composite, the difference in material properties, and the different filler morphologies on the thermal profiles.

Electrical Conductivity of the Filaments
Figure 5 shows the electrical conductivity values of the filaments made from composites with different filler materials.Schematic representing how the filament conductivities were measured along its length are shown in Figure S13, Supporting Information.The Ag/TPU filaments showed very high conductivities after the percolation threshold was passed (around 25 wt%).For Ag/TPU filaments (with silver flake size distribution of 80% < 20 μm) at 30 wt%, the average conductivity was 87.5 S cm À1 .This was higher than that for the Ag/TPU filament with silver flakes of dimensions 4-8 μm.This shows that the morphology of the filler particles greatly affects the conductivity at a constant concentration value, specifically that higher surface area Ag flakes provide more overlapping surface for contacts to be made in the matrix.Although the silver-based composites showed higher conductivity values, but because the TPU filament at this loading was not printable, it was not explored further.
The 10 wt% high AR MWCNT/TPU filament showed the highest average conductivity of the other filaments with a value of 1.65 S cm À1 .The functionalized COOH/MWCNT/TPU composite-based filaments showed very low conductivities for different weight percent of filler loading.Functionalization was performed to achieve a better dispersion of MWCNTs in the composite.But during the chemical functionalization process, covalent bonds between the MWCNTs and the functionalization reagent are formed which modifies the MWCNT structure.This could adversely impact the intrinsic MWCNT properties such as electrical and mechanical.In the present case, the diminished electrical conductivity of COOH-functionalized MWCNTs-based filaments is could be due to structural damage and/or generation of surface defects of MWCNTs. [23]Due to these low conductivity values, the functionalized MWCNT filaments were not considered for further characterization.The hybrid composite filaments exhibited an expected increase in conductivity with increased filler concentration.However, there seems to be a negligible effect from the MWCNT:Ag ratio on the average conductivity values along the filament.The 10 wt% Ag-MWCNT/TPU filament had an average conductivity value of 0.31 S cm À1 which is not high enough to be considered for printed interconnects.Despite the addition of silver flakes to enhance conductivity of the hybrid filament (due to silver's higher conductivity compared to MWCNTs), it appears that their contribution at low concentrations is not as high as expected.The 10 wt% hybrid filament Despite the achieved filament conductivity of 1.65 S cm À1 being lower than the conductivities of printed interconnects using other additive manufacturing methods such as direct ink writing, it is on the higher end of the spectrum for printable filaments using FDM, which is the fabrication method of interest in this study (Figure S18, Supporting Information).3b,10,20,24] In some of them, functional circuits were demonstrated to justify their use as interconnects.However, they each have drawbacks whether with the base polymer prohibiting their use at higher currents due to its thermal properties, or the need to use a much higher filler concentration than that used in this article.This in turn affects cost of manufacturability.Nonetheless, the conductivity achieved by 10 wt% high AR MWCNT/TPU filament in this work remained a limiting factor to their use as circuit interconnects.However, utilizing the trends observed in these results, it can be imagined that a further iteration of this material or new material can have a muchimproved conductivity.
The coefficient of variation (CV) for the filaments' conductivity values was calculated to gauge the extent of variability in relation to the average value.This was done by dividing the standard deviation over the average conductivity for a given filament.The CV values are shown in Figure 6.There is a general trend of a decreasing CV value as the filler concentration was increased for all the filaments.For the highest conductivity filaments (high AR MWCNT/TPU and 10 wt.% Ag-MWCNT/TPU), the CV values were only 4% and 6%, respectively.As the concentration increased, a complete filler network was established in the polymer matrix with the addition of conductive branches.With the matrix filled with more conductive pathways in its volume, there was less variation in electrical resistance along it.Conversely, when the concentration was low, the average separation distance between the filler particles was larger and varied a lot which meant that the variation in resistance depending on the filler distribution and conductive pathways varied greatly (up to 50% in some cases).
Figure 7 shows the conductivity values of the printed tracks and comparison with the conductivity values in their filament form.Figure 7a shows the electrical conductivity values obtained for the printed structures of filaments with varying AR of MWCNT fillers only whereas Figure 7b shows the conductivity values with varying ratio and weight percent of hybrid MWCNT/ Ag fillers.The conductivity values of the low AR MWCNT/TPU filaments were slightly higher than those for the high AR ones at the same filler loading in the filament form.In contrast, the high AR MWCNT/TPU composite-based printed structures showed higher conductivity compared to filament form.This is because of the higher AR MWCNTs' ability to form more robust conductive pathways and networks which could withstand the generated stress while printing (discussed in the following paragraphs).The highest conductivity of 0.92 S cm À1 was obtained with the high AR 10 wt% MWCNT/TPU-printed track (Figure 7a).
Regarding the CV for the printed tracks, a trend of decreasing CV with increasing filler concentration was observed (Figure 7c).This is because of the same explanation outlined for the observed trend in the measured conductivity of filaments.At a particular wt%, the low AR MWCNT/TPU tracks provided a smaller CV than that from the high AR MWCNT/TPU.Similarly, the high AR MWCNT/TPU-printed tracks provided a smaller CV than that from the Ag-MWCNT/TPU tracks.The influence of the filler AR and hybrid interactions may have caused this variation by altering the flow of the filaments through the nozzle and therefore influenced the bonding of the printed rasters and the conductivity.With regards to the application presented here of printed embedded interconnects, a low CV value is required to give predictable resistance values of tracks for designing 3Dprinted circuits.The high ARs 10 wt% MWCNT/TPU tracks offer a 5% CV which is acceptable for this application.
17a,25] This is in part due to the extrusion tunnel abruptly narrowing and the melt  composites suffering from great internal stresses potentially causing the fillers to bend and tear. [10]In Figure 7d, it can be observed that the low AR MWCNT/TPU composites follow this hypothesis.However, the high AR MWCNT/TPU and Ag-MWCNT/TPU composites exhibit an opposite behavior, but only for lower concentrations (<10 wt%).These composites exhibited a very large increase in conductivity after printing at lower filler concentrations.This could potentially be due to the MWCNTs aligning during printing in the well-fused rasters.As the concentration increased, there was a trend toward a decrease in conductivity of the printed part relative to the filament.At higher weight percent, the polymer matrix is more saturated with fillers, and so, it could be more difficult to align them, and the mechanical stresses influenced their morphology in the printed tracks.More studies are needed to confirm this hypothesis.
The conductivity of the printed tracks depends heavily on the printing conditions.This is due to their different thermal properties.Particularly, the printing temperature of the nozzle plays a big role in the flow of the polymer composite melt through the nozzle, the fusing of the rasters on the print structure, and the dimensional accuracy of the printed track.All these factors influence the resistance of the printed track making it very important that the optimal printing temperature is used for each filament.The effect of printing temperature on resistance of the tracks is presented in Figure S14, Supporting Information.As the filler concentration increased, so did the nozzle temperature at which the resistance of the printed tracks was optimized.Moreover, for a given filament, there was a general trend of decreased track resistance with increased nozzle temperature.This is because with an increase in filler content, the melt viscosity of the polymer composite increases.It is to note that melt in the nozzle liquefier requires a higher temperature allowing it to flow through the nozzle and extrude continuous rasters for the infill.

Electrical Performance of the Printed Structures
For incremental current tests, the starting current applied was 1 mA and it was increased incrementally every 10 s until the current limit of the SMU was reached.The results are presented in Figure 8 for three samples of the embedded high AR 10 wt% MWCNT/TPU printed tracks.The tracks did not fail but the experiment was halted when the SMU limit was reached.This large operating range is signified by the power and temperatureprinted tracks could be withstood.3b] The tracks maintained their composition, form, and resistance values at a surface temperature up to 70 °C, which is again a considerable improvement.The resistance values were maintained or even slightly decreased throughout the measurements and maintained the new value after the measurements.Moreover, the printed tracks were functional above 11 V, which is enough to drive most of the commercially available integrated circuits (ICs) for readout and digital communication.It is to note that the three printed samples under the same condition exhibited different resistances.There could be many possible reasons for the observed variation in resistance: The new multi-material printing approach requires several process parameters to be optimized, including the custom-made composites.In our opinion, the prominent reason could be inhomogeneity of the filler particle distribution in the composite filaments and nonuniform thickness of printed layers.Further optimization during fabrication of filaments and printing parameters is required to achieve uniformly printed layers.
As can be seen from Figure 9, the tracks were stable even at the highest current applied continuously during measurements.Despite a small drop in the resistance value at the start for each measured sample, which increased in size with increased applied current, it maintained its stability.Furthermore, there was a large increase in surface temperature of the tracks due to Joule heating during the first 5 min; however, the temperature did not increase by much after that.Due to the base polymer's high melting temperature, the increase in temperature did not affect the tracks' electrical operation.
It was observed that with decreasing wt% and AR, a more capacitive behavior was exhibited by the composite (Figure 10).This is due to relatively larger distance between the filler particles, increasing the capacitive element of the series resistor-capacitor model of the composite.At high wt% and with high AR MWCNTs, the printed tracks exhibited an almost completely resistive behavior due to the complete filler network that is present in the polymer matrix.This resistive behavior is desired for circuit tracks.All these results indicate that the custom-printed composite can withstand the operating conditions of embedded circuits for reliable readout and data-transfer applications.

Bending Tests of the Printed Structures
It has been shown in the literature that the strength of adhesion between a conductor and the substrate affects the stretchability of a bendable/stretchable interconnect. [26]It is noted here that the conductive composite and TPU substrate are fused together because of high temperature during FDM printing.It has also been shown in the academic literature that the low stiffness of a substrate compared to a metal interconnect provided smaller plastic strain values in a perpendicular uniaxial test. [27]In the case of the printed structures in this work, the thickness and material (therefore the stiffness) of the substrate were kept constant for all the conductive composites.
Figure 11 presents the results of the cyclic bending tests of the printed tracks at a 90°bending angle.Further, Figure S15-S17, The change in resistance of the samples with bending is presented.The obtained values are negative because resistance of the sensors decreases under bending relative to resistance values under flat (relaxed) state.When the composite was bent, the MWCNTs dispersed in the matrix came into closer contact with each other, decreasing the resistance of the printed structure.Similarly, when the composites were relaxed, the distance separating the fillers increased which increases the resistance of the printed structures.In both bend and relaxed states, there is a downward drift to lower resistance.The drifting resistance phenomenon has been previously observed in MWCNT-based composites [28] and is likely because of stresses on the fillers from the elasticity of the polymer matrix which maintained a close contact of the filler particles even when the structure was in relaxed state.
The low AR MWCNTs showed higher change in resistance under bending in comparison with the high AR MWCNTs at the same concentration.This leads to the conclusion that higher AR nanotubes in this case overlap more over each other at the same concentration creating a more robust network and thus does not break easily under bending.With the lower AR nanotubes, they have a lower total overlapping area and the average distances between them are bigger.This causes a larger change in resistance when mechanical stress is applied.Under bending, the low AR MWCNTs come closer to each other and greatly decreases the overall resistance of the printed track.Additionally, for a given AR, increasing the filler concentration decreased the change in resistance with bending at the same angle.This was due to the same reason of filler particle interactions and conductive branches being less disturbed when the concentration is higher and more of the polymer matrix is filled with them.As shown in Figure 11c, for 7 and 10 wt% loading of high AR MWCNTs, the change in resistance under bending is negligible (%4% change at 90°).Further, both samples showed similar ΔR/R values which suggests that above 7 wt% MWCNT network has reached to saturation.
It was observed that MWCNT size and loading greatly affects the change in electrical resistance of a TPU/MWCNT nanocomposite with bending.Lower AR MWCNTs at a concentration near the percolation threshold provided the greatest resistance change for the composite at a set bending angle.This composite configuration also demonstrated distinct and repeatable resistance change values at particular bending angles (below 70°) where the other tested composites did not.Therefore, exploiting the  Comparing the stepwise change in resistance with bending between the MWCNT and the hybrid Ag-MWCNT 7 wt% printed tracks, little difference was observed.This comparison is shown in Figure 12.The data shows that the resistance change is largely influenced by the MWCNT content as opposed to the Ag flakes.The 7 wt% Ag-MWCNT/TPU composite was fabricated with 5 wt% MWCNTs and 2 wt% Ag fakes and it does not differ greatly in its performance from either the 5 wt% or the 7 wt% MWCNT/TPU.

Conclusions
In this work, polymer composites-based conductive filaments were developed and then printed using FDM 3D printing.Their mechanical, thermal, and electronic properties were studied for different filler materials namely silver (Ag) and MWCNTs, filler material's AR, and functionalization and morphology of filler materials.With the increased filler concentrations in the fabricated MWCNT/TPU and MWCNT-Ag/TPU composites-based filaments, the onset thermal decomposition temperature of the filaments decreased.But, after FDM printing of these filaments, temperature studies showed no clear trend for change in thermal properties.The AR of the filler materials is critical in determining the percolation threshold and adjusting the electrical conductivity of composites.The maximum attainable electrical conductivity was 0.92 S cm À1 achieved by printing the filament with 10 wt% high AR MWCNT/TPU composite.The printed 10 wt% high AR MWCNT/TPU embedded printed tracks withstood power values >1 W and current values >100 mA (cross-sectional area of 0.24 mm 2 and length of 20 mm).This custom-made filament exhibited %10 times better electrical performance in terms of withstanding high current values without failing in comparison with most of the commercially available conductive filaments.One of the reasons for withstanding such a high current is the base polymer's (TPU) high glass transition and melting temperatures.The custom filament also showed no significant change in impedance over a frequency range from 1 kHz to 1 MHz.Moreover, various applications of the printed conductive composites in flexible electronics and sensors were demonstrated by altering the filler concentrations and morphologies.The 10 wt% high AR MWCNT/TPU-printed tracks exhibited a change in resistance of <5% with bending of 90°.This suggests that the material can be used for embedded printable bendable interconnects, given its conductivity is improved.Conversely, the 3 wt% low AR MWCNT/TPU-printed tracks exhibited a change in resistance of %30% with bending of 90°and clear distinctions could be made of various bending angle from the response.This suggests that the material could be used for embedded printable strain and bend sensors.Therefore, the tunable conductive filaments could be used to expand the use of multi-material additive manufacturing to fabricate embedded electronic systems with functionalities such as reliable 3D circuit tracks, bendable interconnects, and integrated sensors.

Figure 1 .
Figure 1.Details of the processes used to fabricate the filaments composed of a) Ag flakes thermoplastic polyurethane (TPU), b) multi-walled carbon nanotube (MWCNT) TPU, and c) MWCNT/Ag flakes TPU.

Figure 2 .
Figure 2. a) The printing steps for the embedded conductive composite test structures showing images and the sequence of i) the first two conductive layers printed, ii) silver paste bonding external wires, iii) top conductive layer printed, and iv) the embedded track in an insulating thermoplastic.b) A diagram showing the printing process and materials used.All conductive tracks had a thickness of 0.6 mm and the layer height used was 0.2 mm.

Figure 3 .
Figure 3. XRD data for the Ag-MWCNT/TPU filaments comparing a) different loading concentrations, and c) different forms of the composite.b) Shows a zoomed-in graph of the circled area in (a), and d) shows a zoomed-in graph of the circled area in (c).

Figure 4 .
Figure 4. DSC graphs to see the effect of increasing filler concentration on the a) MWCNT-TPU filaments, and b) Ag-MWCNT-TPU filaments.

Figure 5 .
Figure 5. Electrical conductivity values of the filaments.a) MWCNT fillers, b) MWCNT fillers-zoomed in graph of (a); c) Ag flakes fillers; and d) hybrid fillers.For each filament, five locations were measured and averaged.The error bars represent one standard deviation.The green arrow in (a) shows that the data points inside green colored circle are presented in (b).

Figure 6 .
Figure 6.Coefficient of variation (CV) for the electrical conductivity values of the filaments averaged over five locations.

Figure 7 .
Figure 7. Electrical conductivity values obtained for the printed structures of filaments with a) varying aspect ratio (AR) of MWCNT fillers.The inset graph displaying a zoomed-in version for electrical conductivity values for printed structures of filaments with lower filler concentrations.b) Varying ratio and weight percent of hybrid MWCNT/Ag fillers.c) The CV for the conductivity measurements obtained from different filler materials.d) The change in conductivity of the composites after printing compared to their filament form.For each filament, five samples were measured and averaged.The error bars represent one standard deviation.

Figure 8 .
Figure 8. Incremental current test to show the electrical performance of three samples of the printed MWCNT (AR:2500) 10 wt% composite.With increasing current, the graphs show the a) calculated power, b) measured surface temperature, c) measured voltage, and d) calculated resistance of the three samples.Error bars represent one standard deviation.

Figure 9 .
Figure 9. Constant current test to show the electrical performance of three samples of the printed MWCNT (AR:2500) 10 wt% composite.The graphs show the a) calculated power, b) measured surface temperature, c) measured voltage, and d) calculated resistance of the three samples at different constant current values over a 60 min period.Error bars represent one standard deviation.

Figure 10 .
Figure 10.Impedance measurements of the composites over a frequency sweep.a) Impedance, and b) phase comparing different loading concentrations of high-AR MWCNT-TPU composites.c) Impedance, and d) phase comparing different loading concentrations of high-AR MWCNT-TPU composites with low-AR MWCNT-TPU and Ag-MWCNT-TPU.
tunability of TPU composites by modifying their composition can drastically change the material application.Based on the results presented here, the developed 3 wt% MWCNT (AR: 93)/TPU conductive filament exhibited excellent strain sensing behavior.

Figure 11 .
Figure 11.a) The temporal evolution of change in resistance (ΔR/R) for varying AR and weight percentage loading of MWCNTs when bent at 90°.The inset shows optical images of the filaments under bending and relaxed state.b) The temporal evolution of change in resistance for varying silver (Ag)/MWCNT filler loading when bent at 90°, and c) temporal evolution of change in resistance data showing comparison between MWCNT as filler (7 and 10 wt%) with AR of 2500 and hybrid Ag/MWCNT filler composites.For each material, three samples were measured and averaged.

Table 1 .
Comparison of MWCNTs used in this work and in reported TPU/MWCNT filaments in literature.

Table 2 .
Details of the variations of conductive fillers and concentrations used to develop the various TPU-based custom filaments.

Table 3 .
Summary of the information from the TGA data for the measured composite showing onset degradation temperature, maximum rate of degradation peak temperatures, and weight loss.