Effects of Connecting Polymer Structure and Morphology at Inter‐Tube Junctions on the Thermoelectric Properties of Conjugated Polymer/Carbon Nanotube Composites

Conjugated polymer (CP)/carbon nanotube (CNT) composites have been actively used for thermoelectrics for more than a decade. Thermoelectric performance of CP/CNT composites is greatly improved compared with that of the individual components; however, the underlying origin of the performance improvement remains vague, without clear explanations at the molecular scale. Moreover, the nature of the heterogeneous system limits quantitative analysis and restricts physical understanding of the thermoelectric effect in the composites. By combining experimental approaches with molecular dynamics simulations, the contribution of the CPs to the thermoelectric properties at inter‐tube junctions between adjacent CNTs is revealed. Indacenodithiophene‐co‐benzothiadiazole (IDTBT), which has a highly planar backbone and does not aggregate at CP/CNT interfaces, can better mediate effective intramolecular charge transport along backbone chains at inter‐tube junctions than poly[2,5‐bis(3‐tetradecylthiophene‐2‐yl)thieno[3,2‐b]thiophene] (PBTTT). The isotropic and continuous distribution of IDTBT backbone chains enables both holes and phonons to be transported effectively at inter‐tube junctions; this effect greatly increases electrical conductivity, but also increases thermal conductivity. Thus, to obtain a high thermoelectric figure of merit, the balance between the two effects must be optimized. These results may enable CP/CNT composites, whose development is currently stagnating, to be developed into commercially available thermoelectrics, complementing their conventional inorganic counterparts.


Introduction
The demand for eco-friendly and sustainable power generation technology is increasing. Among the available technologies, thermoelectrics are attracting attention because they can harvest waste heat energy and convert it to electricity without large facilities. [1,2] The energy-conversion efficiency of thermoelectric materials is expressed by the figure of merit ZT, which is defined as where σ (S cm −1 ) is the electrical conductivity, S (µV K −1 ) is the Seebeck coefficient, κ (W m −1 K −1 ) is the thermal conductivity, and T (K) is the absolute temperature. Also, σS 2 (µW m −1 K −2 ) in the numerator of ZT is defined as the thermoelectric power factor. Conjugated polymers (CPs) have a unique combination of properties, which include solution processability, low toxicity, and good mechanical flexibility. They are therefore being considered as complementary thermoelectric materials to their inorganic counterparts. [3][4][5][6] In terms of thermoelectric applications, CPs have high Seebeck coefficient and low thermal conductivity; thus, they can potentially increase the ZT. However, CPs cannot be used as a thermoelectric material by themselves because of their poor electrical conductivity. [7,8] Therefore, electrical conductivity of CPs should be improved to promote their use in thermoelectric applications. One easy approach to increasing the electrical conductivity of CPs is to blend them with carbon nanotubes (CNTs). [9][10][11][12] CP/CNT composites simply exploit the high Seebeck coefficient of CPs and the high electrical conductivity of CNTs simultaneously to attain a higher thermoelectric power factor than can be achieved with either individual materials. In addition, junction structures between CPs and CNTs introduce the possibility of suppressing the thermal conductivity while only slightly losing the electrical conductivity, and thereby ultimately increase the ZT of the resultant composites to a competitive level. [13] In early reports on CP/CNT composites for thermoelectric materials, a few studies focused on the phenomenon that CPs are aligned on the surface of CNTs with increased planarity by Conjugated polymer (CP)/carbon nanotube (CNT) composites have been actively used for thermoelectrics for more than a decade. Thermoelectric performance of CP/CNT composites is greatly improved compared with that of the individual components; however, the underlying origin of the performance improvement remains vague, without clear explanations at the molecular scale. Moreover, the nature of the heterogeneous system limits quantitative analysis and restricts physical understanding of the thermoelectric effect in the composites. By combining experimental approaches with molecular dynamics simulations, the contribution of the CPs to the thermoelectric properties at inter-tube junctions between adjacent CNTs is revealed.

Indacenodithiophene-co-benzothiadiazole (IDTBT), which has a highly planar backbone and does not aggregate at CP/CNT interfaces, can better mediate effective intramolecular charge transport along backbone chains at inter-tube junctions than poly[2,5-bis(3-tetradecylthiophene-2-yl)thieno[3,2-b]thiophene]
(PBTTT). The isotropic and continuous distribution of IDTBT backbone chains enables both holes and phonons to be transported effectively at intertube junctions; this effect greatly increases electrical conductivity, but also increases thermal conductivity. Thus, to obtain a high thermoelectric figure of merit, the balance between the two effects must be optimized. These results may enable CP/CNT composites, whose development is currently stagnating, to be developed into commercially available thermoelectrics, complementing their conventional inorganic counterparts.
π-π interaction, and the authors attributed the improved performance to the enhancement of carrier mobility resulting from the alignment of CPs. [14][15][16][17] The alignment of CPs on the CNT surfaces increases electrical conductivity while retaining the Seebeck coefficient moderately, which resulted in a high power factor of >100 µW m −1 K −2 . [14][15][16] Further, with the alignments of CPs, as well as the addition of Au nanoparticles on the CNT surfaces, C.
J. An et al. reduced contact resistances between conductive CNTs and reported an extremely high power factor of >1000 µW m −1 K −2 in CP/CNT composites. [17] However, the alignment of CPs on the surface of CNTs cannot explain the improved charge transport of CP/CNT composite films alone because numerous charge transport pathways (i.e., inter-and-intra-CNTs, inter-and-intra-CPs, and inter-CNTs-and-CPs) and heterojunctions exist in the composites. Since CNTs exhibit much higher electrical conductivity than CP molecules in the composite films, the charge carriers are expected to move easily along intra-CNTs and to be predominantly transported through the CNT-CNT junctions (type I), which have the lowest electrical resistance among the available pathways. [18,19] Nevertheless, CP molecules blended in excess surround the CNTs and hence, the probability of charge transport at CNT-CP junctions (type II) cannot be ignored. In other words, the improvement of carrier mobility through the aligned CPs on the surface of the CNTs can locally promote charge transport along the axial direction of the CNTs, but the charge transport in the composite films is bottlenecked by the CNT-CP junctions (type II) at which the resistance is rapidly increased. Therefore, the thermoelectric performance of the composite films depends on the ability of the CPs to mediate charge transport between adjacent CNTs, and the CPs' structural properties that are advantageous to "bridge" charge carriers at the CNT-CP junctions (type II) need to be clarified at the molecular level. [9,11] In the present paper, we reveal the underlying role of the polymer structure at the inter-tube junctions (type II) in governing the thermoelectric properties at the molecular level by combining experimental results and molecular dynamics simulations. By comparing two CPs with different crystalline ordering in the bulk state, we found that indacenodithiophene-co-benzothiadiazole (IDTBT), a nearly amorphous CP with high backbone planarity, can increase electrical conductivity to a greater extent than poly[2,5-bis(3-tetradecylthiophene-2-yl)thieno [3,2-b]thiophene] (PBTTT) by facilitating intramolecular charge transport along the polymer backbone chains at the inter-tube junctions in CP/CNT composites. This effect also increased the thermal conductivity somewhat, but the ZT was higher in the IDTBT/CNT composites than in the PBTTT/CNT composites because of the maximized power factor. Our results show that the CP structure at the intertube junctions governs the thermoelectric performance of CP/ CNT composite systems and also provide guidelines for selecting CPs for use in composites as organic thermoelectric materials.

Inter-Tube Junctions and Crystalline Ordering of Conjugated Polymers in Composite Films
CP/CNT composite films include many charge transport pathways and heterojunctions between the two substances. The CNT-CNT junction (type I) has the lowest resistance among the available heterojunctions and is likely to be preferred as a conduction path. However, the proportion of the CNT-CP junction (type II) becomes dominant as a significant amount of CPs are incorporated into the composites (Figure 1b). The direct contact between the CP and the CNT means that the CPs at the inter-tube junctions determine transport properties and consequently, the thermoelectric performance of the composite films. In the present paper, using PBTTT and IDTBT, which are known to have completely different crystalline structures in the film state, we attempted to identify the structural properties of CP that are beneficial for improving the thermoelectric performance at the inter-tube junctions in CP/CNT composites (Figure 1a). Solubility of CPs that might have other effects on the formation of CP/CNT composites and on their solid-state structures needed to be excluded from our experiment to directly correlate junction structure and thermoelectric properties. [20] To solve it, we dissolved CPs in chloroform solvent with a dilute concentration (5 mg in 4 mL) and sonicated the mixture vigorously for an enough time (30 min) to fabricate a homogeneous solution which was experimentally observable in the main text later (UV-vis spectroscopy, SEM, AFM).
PBTTT can form highly crystalline regions in the 2D plane with interdigitated side chains, and this configuration corresponds to the classical view of charge transport in organic semiconductors, where intermolecular and intramolecular charge transport can occur effectively within the crystalline regions. [21][22][23] By contrast, IDTBT is a nearly amorphous CP that does not form large crystalline regions, but has highly planar backbone chains, along which intramolecular charge transport can occur. Intermolecular charge transport can occur in tiny aggregates. The discovery of these traits has dramatically changed the perspective on charge transport in organic semiconductors. [24] When these two CPs were spin-coated to form thin films, their grazing-incidence wide-angle X-ray scattering (GIWAXS) patterns indicated a completely different crystalline structure ( Figure 1c; Figure S1, Supporting Information). In the diffraction pattern of PBTTT, (h00) peaks associated with a lamellar structure were observed up to the fourth order in the out-of-plane direction and a (010) peak that corresponds to π-stacking was observed at q ≈ 1.7 Å −1 (d ≈ 3.8 Å) in the in-plane direction. These results indicate that PBTTT molecules are predominantly arranged in an edge-on manner with the polymer backbone chains perpendicular to the substrate. In the pattern of IDTBT, a broad (010) peak was observed at q ≈ 1.5 Å −1 (d ≈ 4.2 Å) in the out-of-plane direction; this peak corresponds to the π-stacking in tiny aggregates. Peaks for (00l) and alkyl chains were also observed in the in-plane direction; however, their weak intensity implies a nearly amorphous structure. Despite the difference in intrinsic crystallinity, the two CPs exhibited similar changes in crystallinity when blended with CNTs to form composite films (Figure 1c; Figure S2, Supporting Information). In the film with a CNT content ([CNT]) of 10 wt.%, the crystallinity of CP was somewhat weakened; however, the intensity of all peaks in the GIWAXS pattern was maintained. At [CNT] = 30 wt.%, the GIWAXS peaks broadened, indicating that the crystalline ordering of CP had begun to degrade. At [CNT] = 50 wt.%, all of the crystalline peaks in the patterns of both CPs eventually disappeared, leaving only a broad scattering at q ≈ 1.5 Å −1 in the out-of-plane direction, which originated from the CNT structure. Therefore, we concluded that, at [CNT] ≥ 50 wt.%, the CPs in the composites exist in an amorphous state in which all crystalline ordering is destroyed irrespective of the CPs' intrinsic crystallinity. The formation of crystalline regions was impossible because the large amount of CNTs interfered with interactions between CP molecules, thereby limited their ability to nucleate and form crystals. Consequently, the crystallinity of CPs could not be a decisive factor in improving charge transport in CP/CNT composite films, [25] and other structural features that affect charge transport are necessarily investigated.

Experimental Approaches
Imaging and spectroscopy techniques were used to observe the morphology of CP in the composite films. Transmission electron microscopy (TEM) images showed CNT bundles with a diameter of 10-20 nm, composed of several CNTs (Figure 2a,b). The attraction between CNTs was strong; thus, not all single

www.advelectronicmat.de
CNTs could be ideally dispersed. However, the occurrence of these relatively homogeneous bundles corresponds to the uniform dispersion of CNTs. In addition, the size distributions of the CNT bundles did not differ significantly between the composite films formed using the two CPs; thus, the bundle size could not be the cause of the difference in their electrical conductivity. [26] We solidified this point by confirming CNT structure of both composites using scanning electron microscopy (SEM), which preferentially observes the CNTs because their electron density is greater than that of CPs ( Figure S3, Supporting Information). Instead of structural differences between the CNTs, we observed salient differences in CP structures around the CNTs. PBTTT aggregated on the CNT surfaces to form a thick polymer shell ( Figure 2a). In particular, several CNT bundles were bonded and connected by PBTTT aggregates, and a few cavities formed. Similar structures have been reported in composites of polyaniline (PANI)/CNTs, poly(3,4-ethylen edioxythiophene):polystyrene sulfonate (PEDOT:PSS)/CNTs, and regioregular poly(3-hexylthiophene-2,5-diyl) (P3HT)/CNTs ( Figure S5, Supporting Information). [10][11][12] From these results, we speculate that typical crystalline CPs, which tend to form aggregates (or crystals) because of attraction between adjacent polymer chains, develop a polymer-shell layered morphology on the CNT surfaces via strong van der Waals forces. However, the crystalline ordering of CPs was completely destroyed by the blended CNTs, as previously discussed; thus, intermolecular or intramolecular charge transport within the crystalline region could not be effectively mediated in the aggregated polymer shell layer. [23] By contrast, IDTBT formed a uniform polymer matrix without severe aggregation on the surface of the CNTs ( Figure 2b). The IDTBT/CNT composite films showed CNTs embedded in a continuous polymer matrix; they did not form the polymer shell layer or the cavities observed in the PBTTT/ CNT composite films. IDTBT, which has a highly planar backbone chain, minimized aggregation at the CNT surfaces and maintained the original amorphous morphology despite strong van der Waals forces; this behavior is distinct from that of conventional crystalline CPs, such as PBTTT. PBTTT, which attempts to increase the coplanarity of the backbone chains via intermolecular interactions, preferentially adsorbs to the CNTs surface. The PBTTT molecules approaching to the innermost adsorbed layer form an aggregated polymer shell because of the increased curvature and structural disorder by the CNT surface; this certainly could not lead to a high crystalline ordering. By contrast, individual IDTBT molecules possess intrinsically high planarity of backbone chains and they do not require intermolecular interactions to increase their coplanarity. Accordingly, IDTBT molecules similarly adsorbed on the CNT surface, but IDTBT molecules approaching the innermost adsorbed layer rarely aggregated themselves around CNT interfaces and formed a uniformly distributed polymer matrix. [21,24] Despite the differences in morphology, similar CNT bundle diameters were observed in PBTTT/CNT and IDTBT/CNT and it means that there is no significant difference in the dispersing ability of the polymers to CNTs ( Figure S4, Supporting Information).
We used UV-vis absorption spectroscopy to further analyze the differences in microstructure between the PBTTT and IDTBT in composite films. The absorption spectrum of the pure PBTTT film shows a clear vibronic signature: a 0-0 peak at 586 nm and a 0-1 peak at 547 nm ( Figure 2c). When PBTTT www.advelectronicmat.de and CNT were blended in a mass ratio of 1:1, the absorption spectrum broadened and these peaks were red-shifted to 602 and 559 nm, respectively. By contrast, the spectrum of the pure IDTBT film showed a camel-back-like spectral feature, with an intramolecular electron transition peak at 677 nm and a shoulder peak at 626 nm ( Figure 2d). [27,28] As observed in the spectra of the PBTTT films, the peaks in the spectra of the IDTBT films broadened by the blended CNTs, but the positions of the peaks changed only slightly. The mutual tendency of broadening in the absorption spectra after the addition of CNTs is attributed to an increase in structural disorder of the CP molecules. The blended CNTs destroyed both large and small CP crystals; thus, CP molecules could not easily maintain intermolecular order. Despite these common increase in structural disorder for both CPs by CNTs, the greater red-shift of the absorption peaks in the spectrum of the PBTTT/CNT composite than in the spectrum of the IDTBT/CNT composite might be attributed to the difference in arrangement of the polymer chains on the CNT surfaces. As evident in the corresponding TEM images, PBTTT was intensively aggregated and arranged on the surface of the CNTs, which caused a substantial change in molecular arrangement from that in a pure PBTTT film. In particular, the PBTTT backbone chains could arrange with greater planarity on the CNT surface, resulting in a remarkable red-shift of the absorption peaks. [14] Despite the red-shift effect caused by molecular arrangement, the relative intensity of the 0-0 absorption peak in the high-wavelength region decreased; this change indicates a reduction in the proportion of highly ordered chromophores. By contrast, although the blended CNTs destroyed the tiny crystals of IDTBT, IDTBT molecules in the composite film could maintain a high backbone planarity with an amorphous morphology similar to that of pure film; thus, the absorption peaks of the IDTBT/CNT composite film did not substantially shift. [29]

Molecular Dynamics Simulations
We conducted molecular dynamics (MD) simulations to predict the structures of CP molecules at the inter-tube junctions. [30,31] This approach can directly predict the structure of the CPs at the inter-tube junctions at the molecular scale, overcoming the limitations of previous research of only providing qualitative explanations for the performance improvement of CP/ CNT composites. In this simulation, the (12, 12) chirality was chosen for the CNT to investigate the alignment of CP at the inter-tube junctions by increasing the π-π contact area with CP. We attempted to predict the polymer structure at the intertube junctions by assuming a composite system in which CNTs and a CP are blended in a mass ratio of 1:1 and in which seven CNTs with a diameter of 1.6 nm form a closed-packing structure ( Figure S6, Supporting Information). We set up the simulation system by taking into account the density of each composite film and diameter of CNT bundles presented in the TEM images (Figure 2a,b). Consistent with the experimental results, at the molecular scale, PBTTT molecules were aggregated and arranged parallel to the CNT axis at the CNT surfaces; this aggregation could maximize the overlap of p orbitals between the two materials at the interfaces, which is a structure that promotes charge transfer from the CNTs to CP and consistent with earlier reports on PANI (Figure 3a left). [14] This anisotropic arrangement of CP backbone chains on the CNT surface was confirmed by inspection of the angular distribution between the axial vector of CNT (υ 1 ) and the normal vector of PBTTT backbone plane (υ 2 ). For PBTTT/CNT composites, a perpendicular relationship (90°) was most frequent between the two vectors and the distribution density decreased sharply with an increasing deviation of the angle from 90° (Figure 3b upper). Surprisingly, between the inner layer and the outer layer of PBTTT chains aligned on the CNT surface, a node with a sparse density of PBTTT backbone chains at the level ≈3 Å formed, which is smaller than the π-stacking distance of PBTTT (≈3.8 Å) (Figure 3c left). This observation indicates that PBTTT did not form crystalline regions on the CNT surfaces, but aggregated in an anisotropic way with an amorphous state; consequently, the intermolecular charge transport path could not function effectively and the transport of charge carriers to adjacent CNTs requires an intramolecular charge transport path along the polymer backbone chains. Because the electrical resistance of CNTs is much lower than that of CPs, charge transport along the aligned CPs on the CNT surface actually could not contribute to the improvement of electrical conductivity of CP/CNT composites; thus, intramolecular transport of CP should occur effectively in the CP regions between adjacent CNTs (i.e., in a radial direction to the CNT axis). Unfortunately, the anisotropic arrangement of PBTTT chains along the CNT axial direction restricts the possible intramolecular charge transport direction within the PBTTT region (type II junctions in Figure 1b). Furthermore, the torsional angles of the PBTTT backbone chains in the amorphous state, as calculated by density functional theory (DFT), were so large that the wavefunction could not be extended along the backbone chains; [32] this result implies that the structure of PBTTT is unfavorable to charge transport at the inter-tube junctions. By contrast, IDTBT did not aggregate on the CNT surfaces at the molecular level (Figure 3a right); thus, charge transfer from CNTs to the CP may have lagged somewhat compared to PBTTT due to the reduction in the overlap of p orbitals. However, the distribution of IDTBT backbone chains was isotropic in all directions at a relatively continuous density from the CNT surfaces (Figure 3b lower and Figure 3c right). Intramolecular charge transport along its backbone chains could be facilitated in the IDTBT region between adjacent CNT bundles (type II junctions in Figure 1b). In particular, potential barriers to rotation in the backbone chain were far greater in IDTBT than in PBTTT, which lowered the torsional angles of IDTBT backbone chains and thereby provided an advantageous structure for charge delocalization. [32][33][34][35] The estimated passage time for charge carriers to migrate across the type II junction through the polymeric backbone revealed by the kinetic Monte Carlo simulation also indicated that IDTBT can be more efficient in charge carrier transport ( Figure S8, Supporting Information). In summary, intermolecular transport pathway of CPs in the CP/CNT composite was largely disrupted as evidenced in GIWAXS results. Even in that situation, IDTBT could retain charge delocalization along planar backbone chains and provide transport pathways for charge carriers at type II junctions which is an actual bottleneck for electrical conduction in CP/ CNT composites; these structural characteristics of IDTBT www.advelectronicmat.de offered an opportunity to carry out the "bridge" function better than PBTTT at the inter-tube junctions.

Thermoelectric Properties
To analyze the relationship between the polymer structure at the inter-tube junctions and the thermoelectric properties of the composites, we measured the electrical conductivity (σ) of each composite (Figure 4a). In both composites, the electrical conductivity increased as [CNT] increased; however, the IDTBT/ CNT composite exhibited a much higher electrical conductivity than the PBTTT/CNT composite over the entire CNT content range. At [CNT] ≥ 50 wt.%, IDTBT/CNT exhibited an electrical conductivity >800 S cm −1 , which is more than five times higher than the conductivity of the corresponding PBTTT/CNT composite. Although the electrical conductivity of both composite films was lower than the conductivity of the pure CNT film, IDTBT resulted in a smaller reduction in electrical conductivity compared with PBTTT (Table S2, Supporting Information).
We also investigated the dependence of the electrical conductivity on temperature to clearly elucidate the effect of CP molecules at inter-tube junctions on charge transport (Figure 4b).
The electrical conductivity in the composite films varied with temperature, whereas the conductivity in a pure CNT film did not. [36] This result indicates that charge transport through the polymer at the inter-tube junctions strongly influenced the electrical properties of the composite films (effects of type II junctions exceed those of type I junctions in the CP/CNT composites in which excess CPs are incorporated). Specifically, PBTTT/CNT showed hopping transport behavior (dσ/dT > 0) near room temperature (i.e., typical transport behavior of a CP). The transition from a temperature-independent region to a thermally activated region suggests that, when the elevated temperature supplied sufficient thermal energy for charge carriers to hop between localized states, electrical conduction occurred primarily via hopping transport through PBTTT at inter-tube junctions. [37] Because the activation energy for hopping (E a ≈ 36 meV) extracted from the slope of the Arrhenius plot is similar to the thermal energy at room temperature (≈26 meV), this transition upon an increase in temperature is plausible. By contrast, IDTBT/CNT showed a weak metallic transport behavior (dσ/dT < 0), which is an extremely unusual transport behavior in a CP and is attributed to the low energetic disorder of IDTBT. [38,39] IDTBT exhibits the highest resilience against energetic disorder ever reported for a CP, and this www.advelectronicmat.de resilience might minimize the density of states (DOS) broadening caused by the amorphous phase in composites, thereby enabling charge delocalization in the backbone chains. [32,40] Rather, the electrical resistance increased because of thermal fluctuations of the IDTBT chains with increasing temperature; thus, the electrical conductivity decreased. These contrasting results between the two CP/CNT composites confirmed that structures of CP molecules at the inter-tube junctions strongly affect electrical conduction in the composites.
The differences in electrical conductivity between the PBTTT/CNT and IDTBT/CNT composites depend on how fast each CP mediates charge transport between adjacent CNTs. Therefore, we performed Hall effect measurement to quantify the carrier mobility in each composite film. The carrier densities (n H ) of the PBTTT/CNT and IDTBT/CNT composites containing 50 wt.% CNTs were 6.2 × 10 22 and 7.7 × 10 22 cm −3 , respectively (Figure 4c). Given that the carrier density of a pure CNT film was 9.1 × 10 22 cm −3 , the carrier densities of the composites were somewhat reduced by the blending of CPs with CNTs and the difference between the carrier densities of the two composites was small because of the samples' identical compositions (CP:CNT = 1:1 by weight). The validity of Hall effect measurements of organic materials is controversial. Early studies involving Hall effect measurements of CNTs reported carrier densities of ≈10 18 to 10 19 cm −3 . [41,42] Deviations from coherent transport reduce the Hall voltage, and this effect causes an overestimation of the carrier density. Despite these possibilities, first-principles calculations for aligned metallic CNTs showed that maximum carrier density could theoretically be as high as ≈10 22 to 10 23 cm −3 . [43] The CNTs used in our experiment contained a substantial metallic component (≈33%) and formed a uniform film with a much higher CNT bundle density than the films in the aforementioned early reports ( Figure S9, Supporting Information). These factors resulted in a carrier density close to the theoretical limit, and we confirmed the reliability of our measurements by reproducing a comparable carrier density with ac Hall effect measurements, which ensures reliable measurements in materials with low carrier mobility www.advelectronicmat.de (Table S1, Supporting Information). [44] Before discussing the role of the CP on charge transport, we found that the carrier mobility of the pure CNT film was quite low ≈0.1 cm 2 V −1 s −1 . Considering the mobility of an individual CNT could exceed 79000 cm 2 V −1 s −1 , this low mobility value indicated that the transport-limiting step at inter-tube junctions was very dominant and frequent as we hypothesized. [18,19,45] Although the PBTTT/CNT and IDTBT/CNT composites had similar carrier densities, they showed drastically different carrier mobilities (µ H ): 1.3 × 10 −2 cm 2 V −1 s −1 in the PBTTT/CNT composite and 5.0 × 10 −2 cm 2 V −1 s −1 in the IDTBT/CNT composite (Figure 4c). These values are substantially lower than the 1.0 × 10 −1 cm 2 V −1 s −1 for a pure CNT film. Blended CPs increased the ratio of type II junctions and reduced carrier mobilities of CP/CNT composite films compared to that of pure CNT film. In the CP/CNT composites, IDTBT exhibited a carrier mobility approximately four times greater than that of PBTTT because IDTBT can mediate fast intramolecular charge transport along the backbone chain direction at the inter-tube junctions. We can confirm that the difference in electrical conductivity is primarily caused by the difference in carrier mobility by introducing the Drude relation: where e is the elementary charge (e = 1.602 × 10 −19 C). In the optimized configuration of field-effect transistors (FETs), PBTTT and IDTBT in the film state showed field-effect mobilities similar to that of amorphous silicon (≈1 cm 2 V −1 s −1 ). [46,47] However, when they were used in CP/CNT composites, in which the crystalline ordering of CPs is destroyed, the structure and charge transport properties of PBTTT were no longer ideal for charge transport at inter-tube junctions. By contrast, the structure and charge transport properties of the IDTBT at the inter-tube junctions were ideal for the improvement of electrical conductivity by facilitating intramolecular charge transport; thus, IDTBT is suitable for use in composite thermoelectric materials. The Seebeck coefficient (S) did not substantially differ between the PBTTT/CNT and IDTBT/CNT composites in the [CNT] range measured in our experiment. It gradually decreased with increasing [CNT] and ultimately converged to the value of a pure CNT film ≈37 µV K −1 (Figure 4d; Table S2, Supporting Information). This result was predictable because the Seebeck coefficient is known to be less affected by structure and morphology than by the other parameters that contribute to the ZT. [48][49][50] Because of the high electrical conductivity of the IDTBT/CNT composite, its power factor (σS 2 ) was substantially greater than that of the PBTTT/CNT composite. As a result, at [CNT] = 50 wt.%, the IDTBT/CNT composite exhibited an unprecedentedly high power factor of 697 µW m −1 K −2 , whereas the power factor of the PBTTT/CNT composite was 130 µW m −1 K −2 (Figure 4e). To the best of our knowledge, the power factor of the IDTBT/CNT composite is the highest ever reported for a CP/CNT composite not subjected to a post-treatment, including doping ( Figure S10, Supporting Information). Our results suggest that the structural properties of the CPs are fundamental factors that governs the thermoelectric and charge transport properties in composite systems.
Thermal conductivity (κ) of organic-based thermoelectric materials should be reduced to make their commercialization feasible. Thus far, organic thermoelectric materials have been considered to have low thermal conductivity; consequently, thermal conductivity of these materials have not been fully investigated or their importance has been underestimated. [51][52][53] However, in our CP/CNT composite systems, the CNTs mixed in an excess amount and substantially different CP structure around them could result in a noticeable difference in thermal conductivity. We therefore compared the thermal conductivity of the two composites ( .78 W m −1 K −1 , respectively. By applying the Wiedemann-Franz law, we find that the contribution of charge carriers to the total thermal conductivity, κ electron , was 2%-6% in the PBTTT/CNT composites, but 9%-13% in the IDTBT/CNT composites. Because of the high electrical conductivity of the IDTBT/CNT composites, the contribution of charge carriers to the total thermal conductivity increased compared with that in the PBTTT/CNT composites and represented a non-negligible proportion of the total thermal conductivity. At the same time, the contribution of the lattice to the total thermal conductivity, κ lattice , was much lower in the PBTTT/CNT composites than in the IDTBT/CNT composites with the same CNT content (Table S3, Supporting Information). The CP structure at the inter-tube junctions may have determined the contribution of the lattice to the thermal conductivity. PBTTT that aggregated on the surface of CNT and whose crystallinity was destroyed www.advelectronicmat.de created a high scattering probability for phonons at the intertube junctions; in addition, a node between the inner layer and outer layer near the interface inhibited propagation of phonons (Figure 3c). By contrast, IDTBT maintained an amorphous morphology similar to that of its bulk state without forming aggregates on the surface of CNTs; in this case, the relatively continuous density of the polymer chains at the interface was not suitable to inhibit phonon propagation (Figure 3c). Because the backbone chain of IDTBT has an extended chain conformation compared to the PBTTT aggregated in an amorphous phase at the inter-tube junctions, heat transport along the 1D chain direction which is the main pathway of heat conduction in CPs could be facilitated in IDTBT. [54,55] Consequently, IDTBT exhibited structural and charge transport properties that lowered both the electrical and the thermal resistance at the inter-tube junctions, thereby offering effective transport paths for electrons and phonons simultaneously. Nevertheless, the increase in the power factor was larger than the increase in the thermal conductivity; thus, at [CNT] = 50 wt.%, the IDTBT/CNT composite exhibited a higher ZT (ZT = 0.043) than the PBTTT/ CNT composite (ZT = 0.019) ( Table 1). The optimal ZT value of the IDTBT/CNT composite (ZT = 0.043) rivals that of tailored semiconducting single-walled carbon nanotube (s-SWCNT) networks with sophisticated chemical doping. [56] Although our system might have a slightly lower ZT compared to the CP/ CNT system with chemical doping or metal nanoparticles, the effect of the CPs at the inter-tube junctions on each thermoelectric parameter (σ, S, κ) was clearly presented. [17] Therefore, our results pave the way for the CP/CNT composite systems to become promising thermoelectric materials.

Conclusion
Despite the versatile properties of CP/CNT composites, analyses of their interfaces and structure-performance relationship for thermoelectric applications have not been previously reported because of their heterogeneous nature. In particular, there were only findings on the alignment effect of the CP on the CNT surface, and enough considerations were not given to the effect of the CP at the inter-tube junction, which can act as an actual transport-limiting step. Related research has therefore long been stagnant. We used two CPs with completely different crystalline structures and charge transport properties to investigate the function and importance of the CPs at the inter-tube junctions in composite films. PBTTT was aggregated at the CNT surface in a direction parallel to the CNT axis; however, IDTBT formed a uniform amorphous matrix without aggregation of the polymer chains; these structural differences resulted from how they realized coplanarization of their backbone chains. The PBTTT resulted in low electrical conductivity because the discontinuous and anisotropic arrangement of its backbone chains at inter-tube junctions was not favorable for intramolecular charge transport to neighboring CNTs. By contrast, IDTBT had a continuous and isotropic backbone chain arrangement with high chain planarity, which effectively mediated the intramolecular charge transport at the inter-tube junctions, resulting in high electrical conductivity. The structure of IDTBT is not favorable to the suppression of phonon propagation at inter-tube junctions; IDTBT therefore somewhat increased the thermal conductivity of the composites compared to PBTTT did. Nonetheless, the ZT of the IDTBT/CNT composites was higher than that of the PBTTT/CNT composites because of the maximized power factor. Our research is the first attempt to systematically reveal the relationship between the molecularlevel structure of a CP at inter-tube junctions and the thermoelectric properties in CP/CNT composite systems by applying experimental and computational methods complementarily. These results suggest a new research direction that approaches understanding by starting from fundamental aspects and may lead to breakthroughs in CP/CNT composite systems for thermoelectric materials. Additionally, the transport of carriers at inter-tube junctions could also be affected by many different factors, such as the barrier height created by the CP between CNTs, the junction distance, and the energy-dependent distribution of charge carriers in CNTs. For a step forward from our results, the physical analysis on such topics will bring further insights on CP/CNT composite systems.
Fabrication of Composite Films: CPs with CNTs were blended in chloroform, then tip-sonicated (VCX-500, Sonics) them for 30 min while they were immersed in an ice bath to dissipate heat. The amount of solute and solvent (5 mg in 4 mL) was fixed and changed the ratio of CNT in composites: 10, 20, 30, 40, 50, or 60 wt.%. At [CNT] ≥ 70 wt.%, CNTs did not disperse uniformly, but formed aggregates. Then, the solution was spin-coated onto glass substrates or silicon wafers and conducted various measurements. Sonication and spin-coating were performed in ambient environment. When metal electrodes were needed for measurement, 50-nm-thick gold (Au) electrodes were evaporated onto the spin-coated composite films.
Structural Analysis: Grazing-incidence wide-angle X-ray scattering (GIWAXS) to analyze the crystalline structure of pure CPs and composite films was performed at the 3C beamline of the Pohang Accelerator Laboratory (PAL). To minimize scattering from substrate, measurements used a silicon wafer on which the SiO 2 layer was <1 nm thick. Atomic force microscopy (AFM) was conducted using a MultiMode 8-HR