Enhancement of p-type thermoelectric power factor by low-temperature calcination in carbon nanotube thermoelectric films containing cyclodextrin polymer and Pd

Thermoelectric conversion is an important technology that recovers energy from heat and converts it into electricity through a simple mechanism that converts temperature gradients in semiconductors and metals into electrical potential.^1–3 Although inorganic materials have long been preferred as thermoelectric materials,^4,5 our preliminary results reported in 1999 demonstrated organic polyaniline materials as candidate thermoelectric materials.⁶ Based on the steady progress made since then,^7,8 organic thermoelectric materials have attracted much attention owing to their potential to recover energy from commonly unused heat sources below 150 °C.⁹ Organic materials such as poly(3,4ethylenedioxythiophene)-poly(styrenesulfonate) (PEDOT-PSS),¹⁰ carbon nanotubes (CNTs),¹¹ and their composites,¹² which have recently been developed with sufficient material performance, are being widely studied for thermoelectric conversion.^13,14 Moreover, these materials have unique advantages over inorganic materials, such as sheet flexibility, light weight, nontoxicity, material abundance, and scalability of production processes including large-area solution processing, paving the way to a series of new thermoelectric applications, including biomedical implants, power generation for the Internet of Things, and wearable heating and cooling devices.^15–17 Because the technology to accurately measure the in-plane thermal conductivity of CNTs and other organic thermoelectric films is currently lacking,¹⁸ their conversion efficiency is estimated using the thermal power factor (PF) as follows:

where S and σ are the Seebeck coefficient and electrical conductivity, respectively. Materials with a higher S and higher σ are more advantageous. Both p-type and n-type materials with excellent PF values are required to realize a π-type (two-leg type) CNT thermoelectric generator with a high conversion efficiency. However, there are no n-type CNT materials that are stable in air,¹⁹ and with some exceptions such as crown ether complexes²⁰ and diphenylhydrazine/polymer composite systems,²¹ there are no reports on materials that can be used in the temperature range of 80–150 °C. Therefore, to develop an effective unileg-type thermoelectric generator without n-type materials, design guidelines to enhance the functionality of p-type CNTs are expected to become highly desired.

CNTs fabricated using floating chemical vapor deposition in a catalytic atmosphere, which are directly prepared by fiber/yarn forming, have been reported to exhibit a large σ and a high PF owing to the increased longitudinal carrier mobility derived from their highly aligned structure.^22–24 This fabrication approach is reasonable because the overall resistance of CNT structures with dimensionality is determined by the appearance of contacts between tubes. On the other hand, CNT-based films, which have no dimensionality, must be dispersed once in a solubilizer solution such as a polymer of insulating nature or a surfactant to loosen the nanotube bundles of the film that are densely bound by van der Waals forces. Except in the case of highly specialized removal techniques using special polymers, as reported by Norton-Baker et al.,²⁵ it has been difficult to achieve a high σ in film structures because the insulating organic material interferes with the electrical contact at the bundle–bundle interface. Therefore, no simple and general method exists to significantly enhance the σ of CNT-based films coated with various polymers. Alternative strategies that can overcome this drawback will further expand the potential applications of film-type flexible/wearable power conversion devices.

Herein, we report a suitable approach to efficiently enhance the σ of CNT composite films by using soluble γ-cyclodextrin polymer (PγCyD), an excellent CNT solubilizer,²⁶ and heat treatment under atmospheric conditions. A simple method was developed to remove the organic insulating layer from the CNT surface through the pyrolysis of the PγCyD matrix on the nanotubes. In addition, we investigated the addition of palladium(II) acetate as a Pd catalyst^27,28 and its effects on promoting oxidation in the thermal decomposition of PγCyD. This process is expected to be an excellent route for thermal decomposition that can be activated under significantly milder conditions than the heat treatment of conventional CNT thermoelectric conversion films at 400–800 °C. We fabricated and evaluated CNT films with optimized PF values to determine their suitability for typical unileg thermoelectric generators. These achievements are expected to lead to future industrial applications of high-PF CNT thermoelectric films.

PγCyD/CNT films were fabricated using the procedure described in the supplementary material. The addition of the dispersant PγCyD resulted in good colloidal dispersion in water, resulting in CNT films with stable morphology, as reported in our previous study.²⁹ The film was subjected to air flow in a calcining oven in the 240–380 °C range to remove the wrapping polymers on the CNT surface. A series of samples were photographed externally, and their surface morphologies were confirmed by scanning electron microscopy (SEM, Fig. 1). In polymer-free prepared films, rope-shaped nanotube bundles bound by van der Waals forces are observed (Fig. S1). It was confirmed that the morphology of these films consisted of a continuous high-density network of nanotubes that were randomly oriented. Further, the CNT bundles were confirmed to be encased in PγCyD and the width of the calcined CNT bundle in the unfired state was 23.8 ± 9.7 nm. Moreover, the CNT bundle diameters of the films calcinated at 260, 300, 340, and 380 °C were 21.6 ± 9.7, 19.7 ± 7.8, 18.0 ± 7.4, and 16.2 ± 7.4 nm, respectively, indicating a trend of decrease with gradual increase in the air calcining temperature. Macroscopically, the PγCyD/CNT films did not lose their flexibility or shape after calcination, but the apparent bundle diameter of each nanotube varied with the calcining temperature. According to the thermogravimetric analysis (TGA) results (Fig. S2), PγCyD thermally decomposed in the temperature range of 275–372 °C in air. The first-step reduction behavior of the PγCyD/CNT composite film was presumed to originate from its organic polymer, and the loss rate was 52.3 wt. %. Therefore, the decrease in the bundle diameter was attributed to the pyrolysis of the polymer on the CNT bundle surface. In fact, the SEM images of the samples calcined at 380 °C showed no polymer coating, and fibrous CNT bundles were visible, indicating a film structure with numerous cavities and voids.

To accurately observe the PγCyD pyrolysis on the nanotubes, isotherms for N₂ adsorption (−196 °C) and water vapor (25 °C) were measured for each sample, and their respective specific surface areas (S_BET) were estimated through Brunauer–Emmett–Teller (BET) analysis (Table S1). The N₂S_BET and H₂OS_BET of the dispersant-free pure CNT films prepared as reference samples were 445 and 13.6 m² g⁻¹, respectively, and the adsorption area occupied by N₂ was larger than that occupied by water.³⁰ On the other hand, the N₂S_BET and H₂OS_BET of the uncalcined PγCyD/CNT film were 1.0 and 235 m² g⁻¹, respectively, and the adsorption area occupied by N₂ was smaller than that occupied by H₂O owing to the PγCyD coating. This indicates the presence of a hydrophilic polymer matrix on the surface of the hydrophobic nanotubes in the uncalcined PγCyD/CNT film, as evidenced by the SEM images. For the calcined samples, N₂S_BET increased with increasing heat-treatment temperature, while H₂OS_BET decreased, confirming the contribution of the gradual pyrolysis of the polymer coating the CNT bundles. The N₂S_BET/H₂OS_BET ratio correlates with the presence or absence of the polymer on the surface, as it induces the film surface to switch from hydrophilic to hydrophobic in the temperature range of 300–340 °C. Furthermore, the N₂S_BET/H₂OS_BET ratio of the films calcined at 380 °C with many cavities and gaps was almost the same as that of the CNT film without the dispersant, suggesting that the polymer is generally lost by pyrolysis.

Figure 2 summarizes the Seebeck coefficient S, conductivity σ, and power factor PF (= S²σ) for PγCyD/CNT films calcined at various temperatures. All the films had positive S values and, thus, were p-type conducting materials. The S value of the uncalcined PγCyD/CNT film was approximately 69.4 μV K⁻¹ and decreased with increasing calcination temperature, reaching 41.2 μV K⁻¹ at 340 °C. Correspondingly, the σ value of the unburned PγCyD/CNT film was approximately 487 S cm⁻¹ and increased with increasing calcination temperature up to 3168 S cm⁻¹ at 340 °C. These results demonstrate that the positive S value decreases with increase in the σ value of the sintered film, confirming the general trend of a trade-off between these properties.³¹ Furthermore, the film heat-treated at 340 °C showed the highest PF value of 536 μW m⁻¹ K⁻². According to the thermogravimetry–differential thermal analysis (TG–DTA) measurements, SEM observations, and hydrophobic–hydrophilic evaluation of the films, the PF values of the calcined CNT films may correspond to the loss of the insulating polymer matrix on the nanotubes by heat treatment. That is, the PF values of the calcined films are greater than those of the polymer/CNT composite films because of improved electrical contact between the CNT bundles due to the removal of the dispersant polymer by the calcination.

Although several attempts to improve the electrical conductivity of CNTs by removing the insulating potential barrier between CNTs through sintering have been reported previously,³² the advantages of this method in terms of organic thermoelectric output remain unclear. Figure S3 shows a wide range of polymers that were used to verify the versatility of the proposed method. The method can be applied to both PγCyD and CNT thermoelectric films coated with various polymers, which indicates its potential to improve the thermoelectric performance. In addition, surfactants that act as insulating potential barriers between CNTs were similarly examined for the effectiveness of this approach. The thermoelectric properties of sodium dodecylbenzenesulfonate (SDBS)/CNT film and cetyltrimethylammonium bromide (CTAB)/CNT film before and after calcining are summarized in Fig. S4. The calcining did not have a significant impact on the S and σ values of SDBS/CNT, and thus, its p-type PF value was kept constant. On the other hand, calcination induced the negative S value of CTAB/CNTs to become positive and at the same time increased the σ value, thus resulting in a 4.2-fold increase in the PF value (note that some CNTs with molecularly adsorbed cationic compounds are known to exhibit n-type semiconductors due to electrostatic effects^20,33). According to these results, the increase in σ value of CNTs due to sintering is not significant for SDBS/CNTs with anionic surfactants, and the effect is not dominant in terms of thermoelectric output. On the other hand, for CTAB/CNT with a cationic surfactant, the removal of the surfactant molecular adsorption layer and oxygen doping³⁴ by calcination resulted in a clear enhancement of the PF value as well as carrier conversion of CNTs.

Previous studies typically used complex methods to enhance the carrier properties of the film. An example of such method is the optimization of the doping level of CNTs by compounding with conductive polymers such as PEDOT-PSS^35,36 or dopant agents such as 2,3,5,6-tetrafluoro-7,7,8,8-tetracyanoquinodimethane³⁷ through methods such as layer-by-layer assembly.³⁸ In addition, because CNTs have a large specific surface area and are highly sensitive to dispersants that are mixed in the film, several reports have demonstrated the effects of different removal methods on thermoelectric properties.³³ The results of these screening tests show that the thermal decomposition of dispersive polymers and cationic surfactants, which are essential during film deposition, improves the p-type semiconductor properties of CNT films. In addition, this method may include a crucial theory wherein n-type CNTs with low thermoelectric properties can be custom-made as high-performance p-type materials by calcination.

The proposed method was found to be most effective for the properties of CNTs when using PγCyD, as shown in Fig. S2. On the other hand, when the firing temperature reached 360 °C, the PF value of the film decreased to 497 μW/m⁻¹ K⁻² owing to the dramatic decrease in 2224 S cm⁻¹ in the σ value. The graphite/disordered graphite (G/D) ratios, which indicate nanotube crystallinity, were measured for a series of samples by Raman spectroscopy (Fig. S5); the measured ratios decreased with increasing calcination temperature. Excessive temperature conditions led to the gradual generation of structural defects in the samples, which is detrimental to the electrical conductivity of CNTs, in addition to cavities and gaps in the film, which resulted from the limited one-dimensional carrier transport due to the complete loss of organic polymers on the surface of the nanotubes.^39–41 The results of pore size distribution analysis (Fig. S6) derived from the Barrett–Joyner–Halenda (BJH) equation based on nitrogen adsorption isotherm data show that the sample calcined at 340 °C has cavities in the mesopores (2–50 nm), while the sample calcined at 380 °C has a large number of micropores (<2 nm). This behavior is noteworthy because a significant change is observed in the porous texture of the samples between the firing temperatures of 340 and 380 °C. In the sample calcined at 380 °C, the pores in the region of CNT diameter less than 1.5 nm can be identified as sidewall defects that are caused by the removal of the organic polymer. Considering the series of results, the abnormal decrease in conductivity above 340 °C in this operation is associated with the formation of defects in the nanotube walls, which is caused by excessive calcination. Previous studies have shown that maintaining a relatively high level of CNT crystallinity is the key to fabricate p-type materials that retain a high σ value.^42,43 To minimize the heat-treatment-induced structural defects in the nanotubes in the above results, a lower PγCyD calcining temperature was investigated by adding palladium(II) acetate as an oxidation catalyst for polymer pyrolysis.^27,28,44 TG–DTA measurements (Fig. S7) of the PγCyD/CNT under air circulation showed that 52.3 wt. % of the organic polymer decomposed in the temperature range of 287–462 °C. In contrast, for Pd–PγCyD/CNT, 52.4 wt. % of the organic polymer decomposed in a lower temperature range of 227–389 °C and a clear exothermic DTA peak was simultaneously observed. Therefore, the Pd catalyst promoted the oxidation of the PγCyD molecular layer on the nanotube. Figure 3 shows the temperature dependence of (a) S, (b) σ, and (c) thermoelectric PF for Pd–PγCyD/CNT films calcined at various temperatures. The S value of the unfired Pd–PγCyD/CNT films was 62.1 μV K⁻¹, and no effective improvement was observed with heat treatment. In general, the S value reflects the slope of the density of states (DOS) around the Fermi level.⁴⁵ For the heat-treatment temperature range of 270–340 °C, the S values of Pd–PγCyD/CNT films ranged from 60.6 to 63.5 μV K⁻¹, suggesting that the DOS slope around the Fermi level of the CNT matrix material is not significantly affected by heat treatment under these temperature conditions. On the other hand, the σ value behaves quite differently from the S value. The σ value of the Pd–PγCyD/CNT films increased monotonically with heat-treatment temperature increasing up to 250 °C and subsequently decreased. The addition of the Pd catalyst decreases the temperature at which the maximum σ value occurs by 90 °C; the maximum σ value without the Pd catalyst occurs at 340 °C, as shown in Fig. 2(b). The added Pd catalyst plays an important role in accelerating the oxidative thermal decomposition of the organic polymer.^27,28 Consequently, the PF value of Pd–PγCyD/CNT was optimized to 570 μW m⁻¹ K⁻² when the heat-treatment temperature was 270 °C. Fortunately, continuous in situ measurements (Fig. S8) showed that the thermoelectric output performance did not degrade even after three cycles of measurements in the temperature range of 57–150 °C. The G/D ratios measured using Raman spectroscopy (Fig. S9) and SEM observations (Fig. S10) before and after calcination revealed that the Pd catalyst effectively removed the wrapped PγCyD from the nanotube surface without causing structural defects in the CNTs due to the calcination, as expected.

Recently, the use of the energy filtering effect,⁴⁶ which improves the S value of CNTs without a significant loss in σ, has gained significant attention mainly due to its ability to improve the performance of CNT thermoelectric materials.⁴⁷ The formation of nanotube/carbide heterostructures is necessary to achieve the desired functionality, but it has been a vexing problem. Its optimal library of additives (urea and glucose mixtures⁴⁷ and metal-organic framework⁴⁸) is limited, and the preparation is restricted to harsh conditions (400–800 °C under inert gas distribution conditions such as N₂). Table S3 shows the results of this study and a comparison of the preparation conditions with the results of typical properties of CNT thermoelectric materials employing heat treatment. The thermoelectric performance of Pd–PγCyD/CNT films calcined at 270 °C in the present experiment is among the most ideal low-temperature performances of CNT-based composites reported so far, which is a simple strategy. Compared to the approach of inert gas sintering to optimize the energy filtering effect, the atmospheric condition heat treatment in this experiment could be an alternative strategy that can help overcome this drawback, thus warranting future efforts in this research.

To verify the potential of Pd–PγCyD/CNT films calcined at 270 °C with excellent performance and thermal stability, a typical unileg thermoelectric module for environmental power generation was prepared and evaluated. Ten pieces of a film strip, which were 6 cm long and 0.5 cm wide, were connected in series using Ag paste electrodes on a polyimide substrate [Figs. S11(a) and S11(b)]. Figure S12 shows a schematic of the module evaluation. The hot end of the device was heated on a heating plate, and the cold end was placed in air. The temperatures of the hot (T_hot) and cold (T_cold) ends were monitored using a data acquisition and logging a multimeter system. The value of the temperature difference ΔT (=T_hot − T_cold) was controlled by heating the hot plate to increase the temperature of the hot end. Figure 4(a) shows the voltage generated at each ΔT. V_TH was calculated as follows:

where n and S_p are the leg number and Seebeck coefficient of the CNT films, respectively. The V_AC of the fabricated device shows a good linear relationship with ΔT, and its value is very close to V_TH. The output power–output current and output voltage–output current curves of the CNT base module at ΔT = 15, 30, 45, 60, and 75 °C are shown in Fig. 4(b). The output power–output current curve is parabolic, and the maximum output powers at ΔT = 15, 30, 45, 60, and 75 °C were approximately 0.37, 1.50, 3.45, 6.25, and 10.3 μW, respectively. The optimal assembly of this material easily yielded a power greater than 10 μW from simple and general modules, demonstrating the potential of this material to stimulate rapid progress in this cutting-edge field. By calculating the mass (0.044 g) of Pd, residual carbon, and CNTs of the film to be fabricated, the maximum power density of this device was found to be 235 μW g⁻¹ at ΔT = 75 °C, which is comparable to the maximum power density of some of the best single-component organic thermoelectric devices on flexible materials demonstrated to date.^49–51 

In summary, we developed a method to enhance the p-type properties of CNT buckypaper films by sintering the supramolecular compound PγCyD on CNTs, which was used as a nanotube solubilizer, under atmospheric conditions. This method can be adapted for various macromolecules other than PγCyD, and it recovers the electrical conductivity of CNTs coated with insulating organic materials by pyrolyzing the insulating macromolecules with atmospheric oxygen through oxidation. Further, doping with the oxidation catalyst Pd decreases the calcination temperature that yields the optimal CNT thermoelectric properties by 90 °C to accelerate the burning of PγCyD. The PF values of Pd–PγCyD/CNT were found to be optimized from 230 to 570 μW m⁻¹ K⁻² when the calcination temperature was increased by 270 °C (Fig. 5). This method is a valuable example of interface engineering in which the organic thermoelectric properties of CNT films can be activated under air flow or at temperatures below 300 °C. A typical module with 10 legs incorporating high-performance Pd–PγCyD/CNTs showed a voltage response at ΔT = 75 °C and achieved an excellent power output of 10.3 μW. In this study, CNTs with relaxed semiconducting and metallic components were used; however, the use of semiconducting CNTs with high S values can accelerate the study of physical properties of thermoelectric materials and should be explored in the future. Our results are expected to be useful not only in the design of thermoelectric devices but also in designs that require highly conductive and enriched CNT thin films, such as photovoltaic solar cells and field-effect transistors. We believe that the proposed method will be one of the key technologies in the fabrication of CNT-based materials and will ultimately advance our society toward an environmentally sustainable world.

See the supplementary material for the experimental overview, characterization, SEM images of a polymer-free CNT film and Pd- Pd-PγCyD/CNT film, TGA curves, BET Surface areas, thermoelectric properties of various polymer or surfactant/CNT films, Raman spectra, BJH pore size distribution, TG-DTA data, continuous in situ thermoelectric measurements, and photograph and evaluation of fabricated device.

This study was supported in part by KAKENHI grants [Nos. 18K14017 (to S.H.) and 19K05633 (to Y.S.)] from JSPS as well as a research grant from the Steel Foundation for Environmental Protection Technology.