Thermal conductivity and thermoelectric properties in 3D macroscopic pure carbon nanotube materials

Sintered carbon nanotube (CNT) blocks and porous CNT sponges were prepared, and their thermoelectric properties were measured. The maximum dimensionless thermoelectric figure-of-merit, ZT, at room temperature of the sintered single-walled carbon nanotube (SWCNT) block is 9.34 × 10, which is twice higher than that of the sinteredmulti-walled carbonnanotube (MWCNT) block in thiswork and also higher than that of other sintered MWCNT blocks reported previously. In addition, the porous MWCNT sponge showed an ultra-low thermal conductivity of 0.021W/(mK) and significantly enhanced ZT value of 5.72× 10 at room temperature and 1 atm. This ZT value is higher than that of other 3D macroscopic pure CNT materials reported. The pronounced enhancement of the ZT in the porous MWCNT sponge is attributed to the ultra-low density, ultra-high porosity, and interconnected structure of the material, which lead to a fairly low thermal conductivity and better Seebeck coefficient. The finding of this work provides an understanding for exploring potential enhancementmechanisms and improving the thermoelectric properties of CNT-based thermoelec-


Introduction
Thermoelectric materials have attracted considerable attention because of their unique ability of direct conversion between thermal and electrical energy [1][2][3][4]. Performance of thermoelectric materials is given by the dimensionless figure-of-merit, ZT S σT k 2 = / , where T , S, σ, and k are the temperature, Seebeck coefficient, electrical conductivity, and thermal conductivity, respectively [5,6]. The carbon nanotubes (CNTs) have unique electrical conductivity, and their thermal conductivity can be tuned by varying the morphology and structure, thus have been considered as promising thermoelectric materials [7][8][9]. Prasher et al. [9] first measured the thermal conductivity and thermoelectric power of single-walled carbon nanotube (SWCNT) packed bed at a pressure of 20-90 psi. The ZT was estimated theoretically up to 0.0047. However, CNT packed bed is not suitable to be directly used as the thermoelectric materials because of its unassembled structure.
The dense CNT films, ropes, or fibers in microscale and nanoscale were found to have high ZT. These CNT materials have relatively high mass density, thus making a good power factor. Hone et al. [10] prepared aligned SWCNT films (thickness of 1.3 and 5 μm; mass density of 1.33 g/cm 3 ) and nanotube ropes by filtration/deposition from suspension in strong magnetic fields, and the measured ZT reached 1 × 10 −4 at room temperature. Zhou et al. [11] prepared a SWCNT interconnected film with a thickness of 600 nm. The grown SWCNT film exhibited an excellent power factor (2,482 μW/m K 2 ) at room temperature, and the ZT of the SWCNT fiber obtained by twisting these films reached 0.014.
As for CNT composite materials, Toshima et al. [12] prepared a flexible composite thermoelectric film based on CNT by complexing nickel-coordinating polymers of vinyl tetrathiol and polyvinyl chloride film. A ternary composite film with a conductivity of 630 S/cm and a thermal conductivity of only 0.07 W/(m K) was obtained.
The ZT of the composite film has reached 0.3 and is considered as one of the best performing CNT-based flexible films. Tan et al. [13] developed rGO/SWCNT hybrid aerogels with extreme low density 10 mg/cm 3 by constructing 3D double-interconnected network porous microstructures. The hybrid aerogel reveals very low thermal conductivity and a high figure-of-merit of 8.03 × 10 −3 . These CNT composite materials generally have extreme low density and high porosity, thus leading to a fairly low thermal conductivity. However, most of the experimental studies on the CNT composite based nanofilm were conducted with only the power factor focused on, primarily because of the difficulty to accurately measure the thermal conductivity of thin film samples [14][15][16].
So far, the reported ZT of 3D macroscopic pure CNT materials is fairly low. Jin et al. [17] measured the thermoelectric properties of macroscopic bundles of long multiwalled carbon nanotube (MWCNT) annealed in argon (Ar) at 2,800°C, the ZT of the MWCNT bulk reached 5.4 × 10 −5 at room temperature. Qin et al. [18] prepared bulk multi-walled CNT by spark plasma sintering (SPS). The thermoelectric properties of the bulk material were measured in the directions perpendicular and parallel to the pressure direction. Both the thermal conductivity and electrical conductivity show apparent anisotropy, but the thermoelectric power has close values in the two different directions and the ZT of SWCNT bulk reached 4.3 × 10 −5 at room temperature. Yang et al. [19] prepared aligned MWCNT bulk samples by SPS method with the ZT value of 2.58 × 10 −6 at room temperature. They also prepared and measured the SPS sintered multi-walled CNT buckypaper, and interestingly, the ZT at room temperature decreased from 4.07 × 10 −5 to 1.29 × 10 −5 when SPS temperatures increased from 900 to 1,500°C. Compared to the above 3D macroscopic pure CNT materials with high mass density, porous CNT sponges can exhibit better ZT. Chen et al. [20] prepared the MWCNT sponges by chemical vapor deposition (CVD) and found their ZT reached 3.85 × 10 −4 in standard atmosphere at room temperature.
In this study, we focused on the thermal conductivity and thermoelectric properties of 3D macroscopic pure CNT materials, including the SPS sintered SWCNT and MWCNT bulk materials and MWCNT sponge. The objective of our work is to provide a clear understanding of the thermal conductivity and thermoelectric properties for pure CNT bulk materials via our comprehensive experimental research. This study could provide help in exploring potential enhancement mechanisms and improving the thermoelectric properties of CNT-based thermoelectric composites.

Experiment
The SWCNT and MWCNT powders for the preparation of the sintered CNT blocks are purchased from Chengdu Organic Chemicals Co. Ltd (Chinese Academy of Sciences). The purity of SWCNT is greater than 95%, and the SWCNT has an average outer diameter of 1-2 nm and an average length of 1-3 μm. The MWCNT has an average outer diameter of 4-6 nm, an average length of 0.5-2 μm, and purity greater than 98%. It should be noted that the length of both the SWCNT and MWCNT of the purchased samples is much shorter than those in the literature (e.g., an average length of 5-15 μm in ref. [9]), since we attempted to investigate whether shorter length CNT can enhance the ZT by the reduction in thermal conductivity.
The thermal conductivity of the two CNT stacked states was measured using a transient hot-wire method (TC3000E, XIA TECH, CHN). The resistivity, Seebeck coefficient, and thermal conductivity of the SPS sintered CNT bulk samples were measured using a PPMS (Quantum Design, PPMS-9) at a temperature of 300 K. The conductivity and Seebeck coefficient of CNT sponge were measured using a SBA 458 Nemesis (Netzsch). The thermal diffusivity was measured using a laser flash method (Netzsch LFA 467) after spray coated with graphite to enhance emissivity. The heat capacities were measured using a differential scanning calorimeter (TA Q5000). The measurements were performed in Ar atmosphere under normal pressure around 1 atm. The thermal conductivity k of the samples was calculated using k DρC p = , where D is the thermal diffusivity, ρ is the mass density, and C p is the heat capacity. Because of the different instruments used in the measurements, resistivity and conductivity directly measured were given in this paper.

Preparation of SPS sintered SWCNT and MWCNT blocks
The CNT powder was stacked in a graphite mold (pore size ∼20 mm). Then, the powder was sintered using an SPS apparatus (SPS-20T-10, Shanghai Chen Hua Technology Co., China). A constant pressure of 40 MPa was maintained in all SPS experiments while changing the applied current density to achieve different SPS temperatures (T SPS ). We sintered the CNTs at 1,200, 1,400, 1,600, 1,800, and 2,000°C in vacuum for 5 min. The obtained typical sintered CNT blocks are shown in Figure 1a and b.

Morphological characterization of SPS sintered SWCNT and MWCNT blocks
The microscopic morphology of the CNT macroscopic material was characterized by scanning electron microscopy (Zeiss, Merlin compact). Here we take the bulk SWCNT materials sintered at 2,000°C as an example. Figure 1c shows the SEM image of the polished surface of the sintered bulk SWCNT sample. It can be observed that some covalently bonded junctions are formed because of the local high temperature to join SWCNT together [21]. Similarly, local welding of SWCNT (intertube bonds) also can be observed on the fracture surfaces of the bulk SWCNT materials, as shown in Figure 1d and e.

Preparation of CNT sponge
A MWCNT sponge using a CVD method by Gui et al. was prepared [22,23]. The main ingredient of the sponge is MWCNT, with purity of over 90%, which is consistent with previous reports. An MWCNT sponge with a density of 5.3 mg/cm 3 (Figure 2a and b) was prepared. The asgrown MWCNT sponge is shown as a black carpet with a thickness of 2.37 mm (Figure 2b).

Structural and morphological characterizations of CNT sponge
The SEM image of the sponge is shown in Figure 2c and d shows more detailed microstructure of the sample. It can be observed that the CNT sponge is a 3D porous network structure in which CNTs are entangled with each other and self-assembled. We also observed that the CNTs had a very high void ratio. The CNT sponge has the characteristics of isotropy. SEM at higher magnifications shows that CNTs have uniform diameter with scarce iron catalyst remained. The average diameter of CNTs is around 30 nm. Figure 3a demonstrates the change in measured thermal conductivity k with mass density of SWCNT and MWCNT packed bed that can be typically considered as unconnected randomly oriented CNTs network. It can be seen that k for both the SWCNT and MWCNT packed bed increases linearly with the increase in the mass density, e.g., k ρ ∝ , and k of SWCNT increases more significantly with the mass density than those of the MWCNT packed bed. When ρ SWCNTs increased from 0.097 to 0.217 g/cm 3 , k SWCNTs was enhanced from 0.057 to 0.142 W/(m K). As ρ MWCNTs increases from 0.151 to 0.311 g/cm 3 , k MWCNTs was enhanced from 0.047 to 0.087 W/(m K). Here k SWCNTs was obviously higher than k MWCNTs at the same mass density, mainly because the average length of the SWCNT (1-3 μm) is larger than that of the MWCNT (0.5-2 μm). The thermal conductivity obtained for SWCNT and MWCNT packed bed is also less than the results obtained by the Prasher et al. [9], because of the shorter length of the tubes in this work. To the best of our knowledge, there exist no experimental studies on the mass density dependence of thermal conductivity of CNT packed bed. We thus resort to the theoretical studies of Volkov and Zhigilei [24,25] and Chalopin et al. [26] on the thermal conductivity of the CNT network as a function of mass density and tube length. The Chalopin's model is based on the assumption that the axial thermal conductivity of individual CNT is infinite, whereas Volkov's model [24] considered the thermal  conductivity of individual CNT as a finite value. The model predictions are shown in Figure 3b. It can be observed that the results by Volkov's model generally follow a linear relation k ρ ∝ and those of Chalopin's model follow a parabolic relation k ρ 2 ∝ when the length of the SWCNT is set as 1.5 μm. Therefore, a better agreement can be found between our experimental results and Volkov's model. However, it can also be observed that the predictions by both Volkov's model and Chalopin's model have much larger values than those of the experimental results. As mentioned by Volkov and Zhigilei [24,25], the theoretical predictions can often be one or two orders different from the experimental results. These theoretical models can thus provide a semi-quantitative analysis of the effect of the mass density and length tube on thermal conductivity of SWCNT networks consisted of pristine tubes. Figure 4(a-d) shows the thermal resistivity, Seebeck coefficient, resistivity, and calculated ZT values of the sintered CNT blocks at different SPS temperature (T SPS ). With the increase in T SPS , the thermal conductivity at room temperature for SWCNT samples increases from 0.89 to 1.34 W/(m K) and that of the MWCNT samples increases from 0.59 to 1.07 W/(m K). Such increase in thermal conductivity is supposed to result from the increase in the mass density. Moreover, with the increase in the T SPS , more covalently bonded junctions between the tubes are formed via heat welding, and the overall tubetube thermal resistance is decreased [27,28]. This could also explain that the thermal conductivity of the SPS sintered CNT blocks is significantly enhanced compared with the corresponding CNT packed bed. It should be noted that the thermal conductivities of the SPS sintered CNT blocks in this work are much lower than those in refs. [17][18][19]29,30] because our target is to improve the ZT value, and thus shorter CNTs were used.

Measurement results and discussion
The Seebeck coefficient S of the SWCNT samples is higher than that of the MWCNT samples at different T SPS , which is consistent with the results of the annealed dense and thick SWCNT films by Hone et al. [10] and the SPS sintered bulk MWCNT by Qin et al. [18]. The measured resistivity of both SWCNT and MWCNT samples shows a significant drop at T SPS of 1,400°C, followed by a plateau when T SPS was above 1,400°C. The lowest resistivity of sintered SWCNT samples is 2.89 × 10 −4 Ωm which is two orders higher than the resistivity value of about 1 × 10 −6 Ωm for the aligned SWCNT [10], attributed to the characteristic of the 3D random distribution of the sintered SWCNT samples. Meanwhile, the lowest resistivity of MWCNT samples is 3.08 × 10 −4 Ωm, which is about two times higher than that of the previously reported sintered MWCNT samples [19], possibly attributed to the shorter length of MWCNT. The ZT value of MWCNT block is about 3.03 × 10 −5 which is equivalent to the ZT value of previously reported MWCNT sintered block [19]. However, the sintered SWCNT block has a maximum ZT value of 9.34 × 10 −5 , which is much higher than the MWCNT sintered block in this work or that reported before. Overall, it is shown that SWCNT has better performance compared to MWCNT. Figure 5(a-d) shows the thermal conductivity, Seebeck coefficient, electrical conductivity, and calculated ZT of the CNT sponge as a function of temperature, respectively, and compared with the thermoelectric properties of MWCNT sponges reported by Chen et al. [20]. The ultrahigh porosity of the CNT sponge is responsible for the low thermal conductivity and good thermal insulation property. As shown in Figure 5a, the MWCNT sponges in this work exhibit an ultra-low thermal conductivity value of 0.021 W/(m K) which is 67% lower than that of the MWCNT sponges by Chen et al. [20] measured at a temperature of 300 K and 1 atm. The Seebeck coefficient is positive and increases as temperature increases, which is in agreement with those of CNT samples in other literature [19,20]. The positive Seebeck coefficient can be explained by hole-type majority carriers and the hole doping effect [20,31,32]. Figure 5(c) shows the comparison of the electrical conductivity of the sponges. It can be seen that the conductivity of our sponge is about 30% less than that by Chen et al. [20] because the mass density of our sponge is only about half of the latter. As shown in Figure 5(d), the ZT of the sponge shows a significant enhancement with a value of 5.72 × 10 −4 at room temperature and 1 atm. In fact, the power factor, S σ 2 , of the MWCNT sponge in this work and that by Chen et al. [20] is comparable. This can be understood by the low density of MWCNT sponges which is about half of that by Chen et al. [20], which leads to a decrease in thermal conductivity which decreases with the mass density [24][25][26] and thus enhanced ZT values. The comparison of ZT values between the results of this work and other 3D macroscopic pure CNT materials [10,[17][18][19][20] is shown in Figure 6. The maximum ZT value of the MWCNT sintered block in this work is about 3.03 × 10 −5 , equivalent to the previously reported the MWCNT sintered block. The ZT value of the SWCNT sintered block is about three times as high as that of the MWCNT sintered block, higher than that of other CNT block materials. The ZT value of the MWCNT sponge prepared in this work reaches a maximum of 5.72 × 10 −4 at room temperature and 1 atm, superior to other 3D macroscopic pure CNT materials and improved by an order of magnitude. However, it is worth mentioning that although the CNT sponge prepared in this paper has ultra-low thermal conductivity, the Seebeck coefficient and the conductivity are not superior.
The sintered CNT bulk materials and CNT sponge represented two different routes in enhancing the thermoelectric figure of merit for 3D macroscopic pure CNT materials. The former attempts to improve ZT by increasing σ, while the latter by decreasing k. The mass density of MWCNT sponge was decreased by 322 times as compared with the sintered MWCNT blocks at T SPS = 2,000°C. Although this resulted in a 38.9 times reduction in electrical conductivity for the MWCNT sponge, a 50.9 times reduction in thermal conductivity and a 2.81 times increase in Seebeck coefficient are obtained, which leads to a much higher ZT than that of sintered MWCNT blocks. This suggested that CNT sponge with ultralow thermal conductivity and good Seebeck coefficient can be a promising candidate for CNT-based thermoelectric composites.

Conclusion
In summary, the thermal conductivity and thermoelectric properties of sintered CNTs and porous CNT sponges were investigated. It is found that the maximum ZT at room temperature of the SPS sintered SWCNT blocks is 9.33 × 10 −5 , twice higher than that of the sintered MWCNT blocks in this work and also higher than that of other sintered MWCNT blocks reported previously. Therefore, SWCNT block is a better alternative than its MWCNT counterpart. Moreover, MWCNT sponge with super low mass density is prepared and measured. It was found that MWCNT sponges showed an ultra-low thermal conductivity of 0.021 W/(m K) and significantly enhanced ZT value of 5.72 × 10 −4 at room temperature and 1 atm, which is higher than other reported 3D macroscopic pure CNT materials. The significant enhancement in ZT of the porous MWCNT sponges is attributed to the ultra-low density, ultra-high porosity, and interconnected structure of material, which lead to a fairly low thermal conductivity and better Seebeck coefficient. This work presents two different routes in enhancing the thermoelectric properties of CNT-based materials, and the finding provides a crucial understanding in exploring potential enhancement mechanisms.
Another contribution of this study is to bring readers a clear understanding of the thermal conductivity and thermoelectric properties for 3D macroscopic pure CNT materials via our comprehensive experimental research. Although many scholars have expected that the pure CNT bulk materials can be as promising thermoelectric materials, our work shows clearly that the ZT values achieved in 3D macroscopic pure CNT materials are still poor (on the level of 10 −4 even it is significant enhanced) compared to the recent progress of thermoelectric field [33][34][35][36][37][38][39][40][41][42][43][44][45][46][47][48][49].