Advances and Outlooks for Carbon Nanotube‐Based Thermoelectric Materials and Devices

Abstract

The unique structure of carbon nanotubes (CNTs) endows them with exceptional electrical and mechanical properties, along with a high surface area, making them highly beneficial for use as flexible, high‐performing thermoelectric materials. As a result, the application of CNTs in the thermoelectric field has become increasingly widespread. Considering the rapid advancements in this field, this review offers a timely overview of the most recent progress on CNT‐based thermoelectric materials and devices over the past five years. This review begins by introducing the fundamental concepts and thermoelectric mechanisms of CNT‐based thermoelectric materials. Then new strategies are explored to enhance their thermoelectric performance, focusing on doping and composites, while emphasizing the importance of CNT stability as a key research area. Additionally, the latest design concepts and expanded application scenarios for flexible and wearable CNTs‐based thermoelectric devices are summarized. Finally, the current challenges are addressed and future directions for the development of CNT‐based thermoelectric materials and devices are discussed.

The unique structure of carbon nanotubes (CNTs) offers excellent electrical, mechanical properties, and surface area, enhancing their use in flexible thermoelectric materials. This review highlights recent advances in CNT‐based thermoelectrics, focusing on doping, composites, stability improvement, and applications in wearable devices, while addressing current challenges and future directions for optimization.

1. Introduction

The development of human society is inextricably linked to energy, with the majority currently sourced from natural resources, particularly fossil fuels and coal. To mitigate the ongoing energy depletion crisis, extensive research over the past few decades has focused on harnessing green energy sources such as wind, solar, and hydropower for electricity generation. However, more than half of the energy generated in these processes is lost as heat, making the reuse of waste heat a critical area of research. Thermoelectric technology, which leverages the Seebeck effect to directly convert heat into electricity, has gained significant attention in this context.^{[ 1} , ² , ³ , ⁴ , ^{5 ]} Wearable and flexible thermoelectric generators (W‐ and F‐TEGs) have emerged as focal points of research,^{[ 6 ]} particularly due to the availability of constant heat sources (such as body heat) and the potential for miniaturization and wearability of thermoelectric materials and devices.^{[ 7 ]} These devices offer significant commercial potential as continuous power sources for medical sensors, human–machine interfaces, and connected devices.^{[ 8 ]} The basic structure of a TEG, illustrated in Figure 1a, consists of a closed‐loop circuit that connects a pair of p‐type (where holes are the majority carriers) and n‐type (where electrons are the majority carriers) materials.^{[ 1 ]} A single TEG generates power by utilizing the temperature difference (ΔT) across the thermoelectric materials, where carriers move from the hot side to the cold side, producing an output voltage (V) to power a load. A TEG can include multiple thermoelectric pairs, connected thermally in parallel and electrically in series, to multiply the Power output (P), as shown in Figure 1a. This enables TEGs to meet various power needs including energy storage in devices such as supercapacitors, regulation through voltage amplifiers, or direct power supply for low‐power electronics (such as sensors).^{[ 8 ]}

Figure 1.

Review of thermoelectric performance of carbon nanotube (CNT)‐based materials. a) Schematic diagram of thermoelectric devices with multiple p–n pairs (up panel) and single p–n pair (down panel). Here, SWCNTs is abbreviated from single‐walled carbon nanotubes, MWCNTs represents multiwalled carbon nanotubes. Diagram of thermoelectric device with multiple p–n pairs. Reproduced with permission.^{[ 275 ]} Copyright 2021, Wiley. Diagram of device with single p–n pair. Reproduced with permission.^{[ 118 ]} Copyright 2021, Royal Society of Chemistry. Structure of SWCNTs and MWCNTs. Reproduced with permission.^{[ 169 ]} Copyright 2022, De Gruyter. b) Summary of the power factor (S ² σ) as a function of absolute Seebeck coefficient value (|S|) for CNT films reported within five years.^{[ 68} , ⁷¹ , ⁷² , ⁷⁴ , ⁷⁷ , ⁸⁷ , ⁸⁸ , ⁹³ , ⁹⁸ , ¹⁰⁸ , ¹¹⁴ , ¹¹⁵ , ¹²⁰ , ¹³¹ , ¹³² , ¹³³ , ¹⁴² , ¹⁴⁵ , ¹⁵¹ , ¹⁶³ , ¹⁶⁸ , ¹⁸⁵ , ²⁰⁵ , ²⁰⁹ , ²²⁶ , ²⁴⁸ , ²⁵⁹ , ²⁶⁰ , ²⁶⁵ , ²⁶⁶ , ^{267 ]} c) Summary of S ² σ for CNT fibers reported within five years.^{[ 116} , ¹³⁵ , ¹⁵² , ¹⁶² , ²¹⁰ , ²²⁷ , ²³¹ , ²⁶⁵ , ²⁶⁹ , ²⁷⁰ , ²⁷¹ , ^{272 ]} d) Summary of the output power density (ω) of reported flexible thermoelectric generators (F‐TEGs) based on CNT films and fibers within five years.^{[ 41} , ⁷⁴ , ⁷⁷ , ⁹⁸ , ¹⁰² , ¹⁰⁴ , ¹⁰⁸ , ¹¹⁴ , ¹¹⁵ , ¹¹⁶ , ¹¹⁷ , ¹²⁰ , ¹³¹ , ¹³² , ¹³³ , ¹³⁵ , ¹⁴⁵ , ¹⁵² , ¹⁶² , ¹⁶³ , ¹⁶⁸ , ¹⁸⁵ , ²⁰³ , ²⁰⁵ , ²⁰⁹ , ²¹⁰ , ²²⁷ , ²³¹ , ²⁴² , ²⁴⁸ , ²⁶⁰ , ²⁶⁵ , ²⁶⁶ , ²⁶⁷ , ²⁷⁰ , ²⁷¹ , ^{274 ]}

Thermoelectric materials are the key determinant of a TEGs’ performance.^{[ 8 ]} Their efficiency is largely measured by the figure of merit, ZT, with higher ZT, corresponding to better P. ZT is defined by ZT = S ² σT/κ, where S represents the Seebeck coefficient, σ is the electrical conductivity, S ² σ is the power factor, T is the absolute temperature in Kelvin, and κ is the thermal conductivity.^{[ 9 ]} Typically, a higher S ² σ and lower κ result in a larger ZT. The total κ consists of both electronic thermal conductivity (κ _e) and lattice thermal conductivity (κ _l). However, optimizing ZT is challenging due to the strong coupling between parameters, such as σ, S, and κ _e.^{[ 10 ]} In general, the S can be expressed as $S=\frac{8{\pi }^2{k}_B^2}{3e{h}^2}{m}^{\ast }T{(\frac{\pi }{3n_c})}^{\frac{2}{3}}$, where k _B, e, h, m*, and n _c represent the Boltzmann constant, charge, Planck's constant, effective carrier mass, and carrier concentration, respectively.^{[ 10 ]} σ is defined as σ = n _c eµ, where µ is the carrier mobility, and κ _e = LσT, with L being the Lorenz constant.^{[ 10 ]} Thus, practical thermoelectric materials with high ZT require careful tuning of n _c to achieve large S ² σ, along with meticulous structural design to enhance the scattering of phonons of various wavelengths for low κ _l.^{[ 10 ]} However, these strategies may also scatter carriers, which can reduce both µ and σ. As a result, designing and optimizing thermoelectric materials remains a significant challenge.

Inorganic thermoelectric materials have been studied since the 1950s due to their relatively high ZT values.^{[ 11} , ^{12 ]} In the mid‐to‐high temperature range, materials such as GeTe,^{[ 13 ]} SnSe,^{[ 14 ]} PbTe,^{[ 15 ]} and Cu₂Se,^{[ 16 ]} have demonstrated ZT values as high as 2–3 between 600 and 1000 K.^{[ 8 ]} However, these materials typically show lower ZT values (<1) at near room temperature (below 350 K). Since W‐TEGs operate at near room temperature, developing thermoelectric materials with high ZT at this range has become a major research focus in recent years.^{[ 17 ]} Bismuth telluride‐based thermoelectric materials are commonly used for near‐room‐temperature applications, with ZT values exceeding 1.2 for n‐type materials and 1.5 for p‐type materials.^{[ 18} , ^{19 ]} However, these materials tend to be expensive, and their rigidity makes them difficult to apply in highly flexible TEGs unless produced as thin films or fibers,^{[ 20} , ²¹ , ^{22 ]} which can compromise their intrinsic high ZT. To overcome these limitations, recent research has focused on alternative near‐room‐temperature thermoelectric materials that offer higher ductility, such as n‐type Ag₂Q^{[ 23 ]} and p‐type AgCuQ (Q = S, Se, Te),^{[ 24 ]} as well as Mg₃(Bi, Sb)₂.^{[ 25} , ^{26 ]} These materials typically have greater plasticity and/or lower cost than bismuth telluride, while maintaining comparable ZT values. However, inorganic thermoelectric materials share common drawbacks,^{[ 27 ]} including high toxicity, high processing costs, and inherent rigidity, which limit their practicality in F‐TEGs.^{[ 28 ]} As a result, organic thermoelectric materials, such as conducting polymers (e.g., poly(3,4‐ethylenedioxythiophene):poly(styrene sulfonate), PEDOT:PSS),^{[ 29 ]} have recently attracted attention for their high flexibility and low κ. Nevertheless, their relatively low S and high cost still pose challenges for widespread practical applications.

In TEGs, the impact of κ in thermoelectric materials becomes less critical if the temperature differential between the hot and cold sides is well maintained. This is especially true for many TEGs, particularly F‐TEGs, which are designed to power low‐grade wearable electronic devices. For such applications, the focus shifts to identifying thermoelectric materials with high σ, low cost, and good flexibility. Carbon‐based materials are considered promising candidates due to their high σ, excellent processability, low cost, light weight, and environmentally friendly nature.^{[ 30} , ³¹ , ^{32 ]} They are also promising candidates for hybridizing with many other thermoelectric materials to improve the overall thermoelectric performance.^{[ 33} , ³⁴ , ³⁵ , ^{36 ]} Among carbon materials, carbon nanotubes (CNTs) have attracted significant attention since their discovery due to their unique 1D electron transport properties.^{[ 35} , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² , ⁵³ , ⁵⁴ , ⁵⁵ , ⁵⁶ , ⁵⁷ , ⁵⁸ , ⁵⁹ , ⁶⁰ , ⁶¹ , ⁶² , ⁶³ , ⁶⁴ , ⁶⁵ , ⁶⁶ , ⁶⁷ , ⁶⁸ , ⁶⁹ , ⁷⁰ , ⁷¹ , ⁷² , ⁷³ , ⁷⁴ , ⁷⁵ , ⁷⁶ , ⁷⁷ , ⁷⁸ , ⁷⁹ , ⁸⁰ , ⁸¹ , ⁸² , ⁸³ , ⁸⁴ , ⁸⁵ , ⁸⁶ , ⁸⁷ , ⁸⁸ , ⁸⁹ , ⁹⁰ , ⁹¹ , ⁹² , ⁹³ , ⁹⁴ , ⁹⁵ , ⁹⁶ , ⁹⁷ , ⁹⁸ , ⁹⁹ , ¹⁰⁰ , ¹⁰¹ , ¹⁰² , ¹⁰³ , ¹⁰⁴ , ¹⁰⁵ , ¹⁰⁶ , ¹⁰⁷ , ¹⁰⁸ , ¹⁰⁹ , ¹¹⁰ , ¹¹¹ , ¹¹² , ¹¹³ , ¹¹⁴ , ¹¹⁵ , ¹¹⁶ , ¹¹⁷ , ¹¹⁸ , ¹¹⁹ , ¹²⁰ , ¹²¹ , ¹²² , ¹²³ , ¹²⁴ , ¹²⁵ , ¹²⁶ , ¹²⁷ , ¹²⁸ , ¹²⁹ , ¹³⁰ , ¹³¹ , ¹³² , ¹³³ , ¹³⁴ , ¹³⁵ , ¹³⁶ , ¹³⁷ , ¹³⁸ , ¹³⁹ , ¹⁴⁰ , ¹⁴¹ , ¹⁴² , ¹⁴³ , ¹⁴⁴ , ¹⁴⁵ , ¹⁴⁶ , ¹⁴⁷ , ¹⁴⁸ , ¹⁴⁹ , ¹⁵⁰ , ¹⁵¹ , ¹⁵² , ¹⁵³ , ¹⁵⁴ , ¹⁵⁵ , ¹⁵⁶ , ¹⁵⁷ , ¹⁵⁸ , ¹⁵⁹ , ¹⁶⁰ , ¹⁶¹ , ¹⁶² , ¹⁶³ , ¹⁶⁴ , ¹⁶⁵ , ¹⁶⁶ , ¹⁶⁷ , ¹⁶⁸ , ¹⁶⁹ , ¹⁷⁰ , ¹⁷¹ , ¹⁷² , ¹⁷³ , ¹⁷⁴ , ¹⁷⁵ , ¹⁷⁶ , ¹⁷⁷ , ¹⁷⁸ , ¹⁷⁹ , ¹⁸⁰ , ¹⁸¹ , ¹⁸² , ¹⁸³ , ¹⁸⁴ , ¹⁸⁵ , ¹⁸⁶ , ¹⁸⁷ , ¹⁸⁸ , ¹⁸⁹ , ¹⁹⁰ , ¹⁹¹ , ¹⁹² , ¹⁹³ , ¹⁹⁴ , ¹⁹⁵ , ¹⁹⁶ , ¹⁹⁷ , ¹⁹⁸ , ¹⁹⁹ , ²⁰⁰ , ²⁰¹ , ²⁰² , ²⁰³ , ²⁰⁴ , ²⁰⁵ , ²⁰⁶ , ²⁰⁷ , ²⁰⁸ , ²⁰⁹ , ²¹⁰ , ²¹¹ , ²¹² , ²¹³ , ²¹⁴ , ²¹⁵ , ²¹⁶ , ²¹⁷ , ²¹⁸ , ²¹⁹ , ²²⁰ , ²²¹ , ²²² , ²²³ , ²²⁴ , ²²⁵ , ²²⁶ , ²²⁷ , ²²⁸ , ²²⁹ , ²³⁰ , ²³¹ , ²³² , ²³³ , ²³⁴ , ²³⁵ , ²³⁶ , ²³⁷ , ²³⁸ , ²³⁹ , ²⁴⁰ , ²⁴¹ , ²⁴² , ²⁴³ , ²⁴⁴ , ²⁴⁵ , ²⁴⁶ , ²⁴⁷ , ²⁴⁸ , ²⁴⁹ , ²⁵⁰ , ²⁵¹ , ²⁵² , ²⁵³ , ²⁵⁴ , ²⁵⁵ , ²⁵⁶ , ²⁵⁷ , ²⁵⁸ , ²⁵⁹ , ^{260 ]} CNTs are one of the allotropes of carbon, formed by rolling graphene sheets with covalent sp²‐hybridized C─C bonds into a honeycomb‐structured, 1D nanotube.^{[ 38 ]} Due to the unique quasi‐1D structure of CNTs, phonon transport is highly facilitated, resulting in CNTs possessing an exceptionally high κ of 2000–3000 W m⁻¹ K⁻¹.^{[ 261 ]} This unique structure also provides a high aspect ratio, promoting charge transport while limiting the Seebeck effect, leading to high σ and low S of <30 µV K⁻¹.^{[ 169} , ²⁰⁴ , ²⁶² , ²⁶³ , ^{264 ]} Over the decades, various strategies have been reported to enhance the thermoelectric performance of CNT films and fibers by reducing κ while increasing S and σ.^{[ 169} , ²⁰⁴ , ²⁶³ , ^{264 ]} Figure 1b summarizes the relationship between S ² σ and absolute Seebeck coefficient values (|S|) for CNT‐based films within the last five years.^{[ 68} , ⁷¹ , ⁷² , ⁷⁴ , ⁷⁷ , ⁸⁷ , ⁸⁸ , ⁹³ , ⁹⁸ , ¹⁰⁸ , ¹¹⁴ , ¹¹⁵ , ¹²⁰ , ¹³¹ , ¹³² , ¹³³ , ¹⁴² , ¹⁴⁵ , ¹⁵¹ , ¹⁶³ , ¹⁶⁸ , ¹⁸⁵ , ²⁰⁵ , ²⁰⁹ , ²²⁶ , ²⁴⁸ , ²⁵⁹ , ²⁶⁰ , ²⁶⁵ , ²⁶⁶ , ^{267 ]} Despite a series of effective treatments, pure CNT‐based films still exhibit relatively low S ² σ (<3 µW cm⁻¹ K⁻²) due to their low S. To further enhance their thermoelectric performance, treated CNTs are often combined with other functional materials, including inorganic nanomaterials, organic materials, and organic/inorganic hybrids.^{[ 118 ]} This approach potentially improves S and decouples S from σ by forming heterointerfaces that invoke an energy filtering effect,^{[ 268 ]} thereby filtering out low‐energy carriers. This process can lead to significant improvements in S ² σ. For instance, a high S ² σ value of 10.79 µW cm⁻¹ K⁻² was achieved by forming PANI/SWCNT films, where single‐walled CNTs (SWCNTs) are composited with polyaniline (PANI).^{[ 248 ]} Table 1 further summarizes the detailed thermoelectric performance of CNT‐based films over the last five years. Regarding CNT‐based thermoelectric fibers, Figure 1c provides an overview of the S ² σ of CNT fibers over the past five years.^{[ 116} , ¹³⁵ , ¹⁵² , ¹⁶² , ²¹⁰ , ²²⁷ , ²³¹ , ²⁶⁵ , ²⁶⁹ , ²⁷⁰ , ²⁷¹ , ^{272 ]} Due to their 1D characteristics, fibers are well‐suited for wearable or wearable thermoelectric textile‐type W‐TEGs,^{[ 273 ]} offering superior wearability compared to film materials. The optimization strategies are almost the same, for example, an S ² σ value of 34.25 µW cm⁻¹ K⁻² was achieved with Sb₂Te₃/CNT composite fibers.^{[ 265 ]} Table 2 provides a detailed summary of the thermoelectric performance of CNT‐based fibers over the past five years. Additionally, we review the performance of F‐TEGs based on CNT films and fibers. Figure 1d presents the reported power output density (ω) of CNT film‐ and fiber‐based F‐TEGs over the last five years.^{[ 41} , ⁷⁴ , ⁷⁷ , ⁹⁸ , ¹⁰² , ¹⁰⁴ , ¹⁰⁸ , ¹¹⁴ , ¹¹⁵ , ¹¹⁶ , ¹¹⁷ , ¹²⁰ , ¹³¹ , ¹³² , ¹³³ , ¹³⁵ , ¹⁴⁵ , ¹⁵² , ¹⁶² , ¹⁶³ , ¹⁶⁸ , ¹⁸⁵ , ²⁰³ , ²⁰⁵ , ²⁰⁹ , ²¹⁰ , ²²⁷ , ²³¹ , ²⁴² , ²⁴⁸ , ²⁶⁰ , ²⁶⁵ , ²⁶⁶ , ²⁶⁷ , ²⁷⁰ , ²⁷¹ , ^{274 ]} These results indicate that flexible thermoelectric materials and TEGs based on CNTs hold significant practical value.

Table 1.

Summary of thermoelectric properties of carbon nanotube (CNT)‐based films reported within five years. Here, the units for κ, σ, S, and S ² σ, are W m⁻¹ K⁻¹, S cm⁻¹, µV K⁻¹, and µW cm⁻¹ K⁻², respectively. Abbreviations: SWCNT, single‐walled carbon nanotube; FcMA, (dimethylamino) methyl group; MOF, metal–organic framework; CNT, carbon nanotube; SFX‐2, C₃₃H₂₁O; Spiro‐MeOTAD, (C₆₅H₄₀O₈N₄) _n ; MXene, Ti₃C₂T _x –H₂O; [HMIM][BF₄], 1‐hexyl‐3‐methylimidazolium tetrafluoroborate; CNTs, carbon nanotubes; PBDTT‐FTTE, (C₄₉H₅₇FO₂S₆) _n ; MWCNT, multiwalled carbon nanotube; P(BDTC), macrocyclic crown ether; nCB, nanocarbon black; TPP, riphenylphosphine; PYB, pyridineborane; DMSO, dimethyl sulfoxide; PEDOT:PSS, poly(3,4‐ethylene dioxythiophene): polystyrene sulfonate; PANI, emeraldine base polyaniline; PEDOT‐Tos, poly(3,4‐ethylenedioxytiophene)‐tosylate; a‐SWCNT, acidified‐single‐walled carbon nanotube; PSSH, polystyrene sulfonate; APA, aniline tetramer‐b‐polyethylene glycol‐b‐aniline tetramer; a‐TM, acid‐treatment of tourmaline; E7, a eutectic nematic mixture; CAS, sulfurochloridic acid; PEDOT, poly(3,4‐ethylenedioxytiophene); PE, polyethylene.

Material	Type	σ	S	S2σ	κ	ZT	Refs.
SWCNTs/nCB	p	1710	4644.2	151	–	–	[274]
CAS‐MWCNT	p	10 000	23	46.6	45.9	0.0304	[186]
PANI/SWCNT–DMSO	p	3980	55	10.79	–	–	[248]
PANI/SWCNT–DMSO	n	3980	−50	10.31	60	0.005	[248]
[HMIM][BF4]/CNTs	p	1154	81	7.62	3.84	0.060	[131]
CNT/Se/PEDOT:PSS	p	1825.02	66.38	5.842	5.64	0.034	[267]
SWCNT/FcMA	n	2674.86	−46.07	5.6754	52.98	0.0032	[77]
PEDOT:PSS/SWCNT	p	4717.8	32.6	5.0131	0.062	0.012	[185]
PEDOT:PSS/SWCNTs	p	2333.7	46.2	5.00	164.4	0.0009	[266]
SWCNT/E7	p	1665.5	50.48	4.2859	–	–	[87]
PEDOT:PSS/SWCNTs	n	1718.2	−49	4.1140	0.4	0.3	[145]
a‐TM/SWCNT	p	2837.4	34.2	3.35	–	–	[170]
SiO2@MoS2/SWCNT	p	1646.4	39.2	2.532	61.1	0.00123	[41]
Spiro‐MeOTAD/SWCNT	p	873.3	51.7	2.392	–	–	[168]
PEDOT:PSS sheet/SWCNTs	p	3085.4	26.95	2.24	–	–	[88]
SWCNTs/PYB	n	766.7	−48.5	2.238	40.1	0.00154	[117]
SFX‐2/SWCNT	p	762.2	53.4	2.186	–	–	[98]
SWCNT/MXene	p	1293.76	39.64	2.0329	9.764	0.0062	[132]
SWCNTs/DMSO	n	3490	−23.7	1.95	19.88	–	[163]
PEDOT:PSS/CNT	p	1602.6	33.4	1.827	–	–	[72]
PEDOT‐Tos/a‐SWCNT	p	4731.6	21.1	1.685	–	–	[74]
PE/PEDOT/SWCNT	p	459.82	47.58	1.5881	0.0879	–	[151]
SnSe nanobelt/SWCNT	p	586.37	49.3	1.45	14	0.003	[68]
P(BDTC)/SWCNT	p	864.8	39.9	1.377	–	–	[133]
PEDOT‐Tos/Te/SWCNTs	p	578.4	47.75	1.3191	–	–	[108]
CNTs/PEDOT	p	910.1	37.9	1.311	–	–	[226]
PEDOT:PSS/SWCNT	n	1410	−29.54	1.23	–	–	[205]
PSSH/SWCNT	p	3749	17.7	1.177	–	–	[259]
APA/SWCNT	n	–	−49.0	1.063	–	–	[93]
PEDOT:PSS/SWCNT	p	1514	25.3	0.96	–	–	[205]
MOF/CNT	n	347.4	−56.4	0.844	0.18	0.071	[260]
PEDOT:PSS/SWCNT	p	1562	21.9	0.739	0.3	0.07	[142]
PEDOT:PSS/SWCNT	p	1130	18.9	0.486	–	–	[120]
PBDTT‐FTTE /MWCNT	p	210.8	47.7	0.4821	–	–	[114]
PEDOT:PSS/SWCNT	p	510.6	20.18	0.2068	–	–	[71]
CNT/PEDOT:PSS	n	1938	−9.8	0.164	–	–	[209]
TPP/CNTs	n	48.1	−53.9	0.136	0.123	0.0091	[115]
MWCNT	p	84	19.8	0.0004	0.58	0.0000142	[64]

Table 2.

Summary of thermoelectric properties of CNT‐based fibers reported within five years. Here, the units for κ, σ, S, and S ² σ, are W m⁻¹ K⁻¹, S cm⁻¹, µV K⁻¹, and µW cm⁻¹ K⁻², respectively. Abbreviations: CNTFs, CNT‐based fibers; DWCNT, double‐walled carbon nanotube; PU, polyurethane; N‐DMBI, 4‐(1, 3‐dimethyl‐2, 3‐dihydro‐1H‐benzimidazole‐2‐yl) phenyl) dimethylamine; NBP, nickel‐backboned polymer; PEI, polyethylenimine; Au NPs, gold nanoparticles.

Material	Type	σ	S	S 2 σ	κ	ZT	Refs.
Sb2Te3/CNT	p	4620	85	34.25	–	–	[265]
CNT/Bi2Te3	n	4675	−80	27.30	–	–	[265]
CNT	p	6440.3	64.5	26.19	17	0.05	[270]
N‐DMBI/CNT	n	1182	−113	15.34	27.3	–	[227]
PEI/Au NPs/CNTFs	n	2100	−80	14.00	–	–	[231]
CNTs/PANI	p	3651	59.5	12.94	–	–	[271]
NBP/CNT	p	1236.85	76.26	71.948	–	–	[272]
PEDOT:PSS/CNT	p	2110	36.4	2.80	–	–	[162]
CNT‐modified	p	747	57	2.42	63.0	0.00115	[269]
DWCNT‐PU	p	285.1	72.5	1.497	–	–	[135]
CNT/PEDOT:PSS	p	26.3	44	0.0536	–	–	[152]
CNTs/PEDOT:PSS	p	14.62	57.27	0.0479	–	–	[210]
CNTFs	p	17.4	51.5	0.0459	4.17	0.0000016	[116]

To date, significant progress has been made in CNT‐based thermoelectric materials and devices, with various practical films and fibers emerging as promising candidates for F‐TEGs, particularly in wearable technologies. Given the rapid developments in this field, periodic reviews of developments in CNT‐based thermoelectric materials and devices are crucial for guiding future research. This review, drawing on the authors’ years of research and achievements in the field, aims to provide an overview of the latest progress in CNT‐based thermoelectric technology. It focuses on exploring underlying mechanisms, optimizing thermoelectric performance, integrating advanced devices, and identifying novel applications. Finally, we address the challenges and bottlenecks that remain in CNT‐based thermoelectric materials and devices and offer fresh perspectives for future research directions.

2. Fundamentals of Thermoelectric CNTs

Within the CNT family, SWCNTs stand out as particularly promising thermoelectric materials due to their higher intrinsic σ compared to double‐walled CNTs (DWCNTs) and multiwalled CNTs (MWCNTs). SWCNTs also offer greater ease of surface modification and functionalization.^{[ 276 ]} In recent years, various strategies have been developed to enhance the thermoelectric performance of CNTs by increasing the S and σ while reducing the κ.^{[ 204 ]} The physicochemical properties of CNTs are largely determined by the charge carrier type and density, which can be modulated through ionic and molecular doping or charge injection. Chemical doping and dedoping alter carrier characteristics, allowing the S of CNTs to shift from positive (p‐type) to negative (n‐type).^{[ 277 ]} CNTs support both p‐type and n‐type doping through interacting with molecules that have high electron affinity or low ionization potential.^{[ 278 ]} One of CNTs’ unique advantages is their ability to modulate n _c using charge‐transfer dopants. CNT networks typically have a high surface area, making them highly responsive to adsorbing redox molecules. They can be immersed in various solvents and dopant solutions without structural degradation.^{[ 279 ]} For instance, chemical doping with metals like sodium and potassium often produces n‐type CNTs. However, such doped CNTs are highly unstable, tending to revert to p‐type when exposed to atmospheric oxygen, making the search for stable and manageable n‐type dopants a continuing challenge.^{[ 280 ]} In addition to doping, forming composites is another effective strategy for improving the thermoelectric performance of CNTs. In CNT‐based thermoelectric composites, the large surface area of CNTs and their strong π–π conjugation effect with organic materials significantly enhance interfacial interactions between the phases. CNTs provide high µ and excellent mechanical properties, improving both the σ and flexibility of composite systems, while inorganic nanoparticles contribute high S and low κ. By carefully adjusting the ratio and structure of these composites, thermoelectric performance can be systematically optimized.

2.1. Structure

The unique electrical transport properties of CNTs stem from their distinctive structures, making it crucial to study these structures to develop strategies for enhancing their electrical performance. Taking SWCNTs as an example, they have a cylindrical shape formed by rolling a graphene sheet along a chiral vector.^{[ 281 ]} Similar to the electronic structure of graphene, SWCNTs possess four valence orbitals: 2s, 2p_x, 2p_y, and 2p_z. Electrons in two of the 2p orbitals form in‐plane C─C bonds with adjacent carbon atoms, while the remaining 2p orbital forms delocalized out‐of‐plane π‐bonds. The rolling direction of the graphene sheet is defined by the chiral vector (n, m), where n represents the tube diameter and m denotes the chiral angle (α).^{[ 281 ]} It is important to note that n here refers to the vector, not the carrier concentration (n _c). SWCNT structures are classified into three types based on the rolling angle and chiral indices: zigzag, armchair, and chiral nanotubes.^{[ 282 ]} According to the tight‐binding model, the parameter (n − m) determines whether a CNT exhibits metallic or semiconducting properties.^{[ 283 ]} Specifically, all armchair nanotubes are metallic, whereas zigzag and chiral nanotubes can exhibit either metallic or semiconducting behavior.^{[ 282 ]}

Figure 2a illustrates a segment of a nanotube, where a (9,4) graphene sheet is rolled up to display the chiral indices (n, m). The value of (n − m) mod 3 determines the characteristics of SWCNTs: SWCNTs are semiconductors when (n − m) mod 3 equals 1 or 2, and metallic when (n − m) mod 3 equals 0. In Figure 2a, white hexagons represent semiconducting SWCNTs, gray hexagons indicate metallic SWCNTs, and green hexagons mark typical commercial SWCNTs, which are dispersed in sodium dodecylbenzene sulfonate (SDBS) via the high‐pressure carbon monoxide process.^{[ 276 ]} Researchers have also employed the linearized augmented plane‐wave method to study ultrasmall radius tubes with a diameter of 0.4 nm, revealing that the (4,0) nanotube is a stable structure with exceptionally high n _c, as evidenced by the van Hove peaks near the Fermi level.^{[ 283 ]} Recent studies have focused on surface modifications of CNTs, such as doping, defect engineering, and surface functionalization, to further enhance the S ² σ and reduce κ.

Figure 2.

Structures and doping of CNT‐based thermoelectric materials. a) Illustration of the formation of SWCNTs from a graphene sheet by rolling it along roll‐up vector, determined by the chiral indices (0,0) to (n,m). Here, (n, m) represents the chiral indices. SDBS refers to sodium dodecylbenzenesulfonate. HiPCO stands for high‐pressure carbon monoxide. Reproduced with permission.^{[ 276 ]} Copyright 2018, Wiley. b,c) Schematic illustrations of the doping mechanism for n‐type and p‐type CNTs. d) Electronic band structure of n‐type and the p‐type CNTs. Here, E _g represents bandgap energy. E _f denotes Fermi energy level. Reproduced with permission.^{[ 102 ]} Copyright 2021, Elsevier. e,f) Diagrams of energy level of n‐type and p‐type SWCNTs‐based films. Here, LUMO is abbreviated from lowest unoccupied molecular orbital. HOMO is abbreviated from highest occupied molecular orbital. TCNQ denotes tetracyanoquinodimethane. F4TNQ represents 2,3,5,6‐tetrafluoro‐tetracyanoquino‐dimethane. CN6CP stands for hexacyano‐trimethylene‐cyclopropane. N‐DMBI is abbreviated from dimethyl‐2‐phenyl‐2,3‐dihydro‐1 H‐benzoimidazole. PEI denotes polyethyleneimine. DETA represents diethylenetriamine. Reproduced with permission.^{[ 163 ]} Copyright 2022, Elsevier.

2.2. Doping

Doping, which involves electron transfer between CNTs and dopants, is an effective approach to enhance the σ of CNTs. This process not only improves cycling performance but also enables partial or complete reversibility by allowing CNTs to accept electrons or holes from the dopants.^{[ 284 ]} Various doping strategies such as chemical, optical, interfacial, and electrochemical methods,^{[ 285 ]} have been explored based on different doping mechanisms. Among these, chemical doping is commonly used to modify n _c to achieve higher S ² σ.^{[ 285 ]} Figure 2b,c illustrates the fundamental principles of converting pristine CNTs into n‐type and p‐type CNTs.^{[ 102 ]} For n‐type CNTs, polyetherimide (PEI), which has a high density of amine (─NH─) groups with strong electron‐donating ability, transfers electrons to the CNTs through interactions with the π‐electron system and van der Waals forces. Importantly, this n‐type sample exhibits excellent temperature stability, ensuring consistent performance under a temperature gradient and enabling a reliable and well‐defined ZT value.^{[ 102 ]} For p‐type CNTs, due to the strong oxidizing ability, FeCl₃ serves as an electron acceptor to conduct redox reaction with CNTs, injecting holes into the CNTs during doping, with FeCl₃ being reduced to FeCl₂ in the process. Both n‐type and p‐type CNTs experience an increase in n _c, ultimately enhancing their S ² σ. The electronic band structures of n‐type and p‐type CNTs are depicted in Figure 2d. For pristine CNTs, the bandgap energy (E _g) and work function (W) are 0.642 and 4.53 eV, respectively. After PEI doping, the Fermi level (E _f) shifts upward, facilitating the transition of carriers from hole to electron carriers, resulting in n‐type CNTs. Conversely, during FeCl₃ doping, the E _f shifts closer to the valence band, increasing W to 5.03 eV and producing p‐type CNTs with holes as charge carriers. The choice of n‐type or p‐type dopants plays a critical role in determining the thermoelectric performance of CNTs, as illustrated in Figure 2e,f.^{[ 163 ]} For n‐type dopants, the doping ability is associated with the highest occupied molecular orbital (HOMO) energy, where a higher HOMO indicates stronger n‐type doping capability and results in higher S ² σ. For instance, CNTs doped with N‐DMBI (dimethyl‐2‐phenyl‐2,3‐dihydro‐1H‐benzimidazole) achieve the highest σ compared to those doped with PEI or diethylenetriamine (DETA), due to the hierarchy HOMO_N‐DMBI > HOMO_PEI > HOMO_DETA. Moreover, N‐DMBI forms good coverage on CNTs, significantly enhancing the air stability of the resulting composite films. Conversely, for p‐type dopants, the lowest unoccupied molecular orbital (LUMO) energy determines their doping ability, with a lower LUMO resulting in higher S ² σ. Among p‐type dopants, hexacyanocyclopropane (CN6CP) exhibits the best S ² σ compared to tetracyanoquinodimethane (TCNQ) and 2,3,5,6‐tetrafluorotetracyanoquinodimethane (F4TCNQ), following the hierarchy LUMO_CN6CP < LUMO_F4TCNQ < LUMO_TCNQ.

Several doping strategies have been reported to optimize CNT properties. For example, doping MWCNTs with benzodithiophene‐thienothiophene (BDT‐TTE) yielded an S ² σ value of 0.48 µW cm⁻¹ K⁻².^{[ 114 ]} Similarly, doping with spiro‐bifluorene derivatives increased the S ² σ by 108% compared to pristine CNTs.^{[ 168 ]} It is also well known that CNTs readily adsorb oxygen from the environment, leading to hole formation and p‐type thermoelectric characteristics. These results highlight the potential of CNTs for achieving enhanced performance tailored for real‐world applications. However, n‐type CNT films often suffer from instability and are easily converted to p‐type due to oxidation in atmospheric conditions, posing a challenge for practical use. To address these stability issues, various strategies, including hybrid doping techniques, surfactant functionalization, protective coatings, and synergistic doping approaches, have been developed. For example, PEI was employed to fabricate SWCNT/carboxymethyl cellulose films with long‐term stability, retaining their original S ² σ value for nearly a month by effectively preventing oxidation.^{[ 286 ]} The surfactant dimethyldioctadecylammonium chloride, which includes a hydrophobic segment, has been investigated for stabilizing SWCNTs by shielding them from exposure to the atmosphere. This approach effectively preserves the n‐type characteristics of SWCNTs, even under high‐temperature conditions.^{[ 287 ]} Poly(vinylpyrrolidone) and poly(vinylidene fluoride) coatings were applied to the surface of SWCNTs to protect them from oxidation, enabling the SWCNTs to maintain unchanged performance even after being stored in air for 30 days.^{[ 223 ]} Polyethylene glycol (PEG) was utilized to synergistically dope with SWCNTs, resulting in SWCNT/N,N‐dimethylferrocenemethylamine (FcMA)/PEG composites with superior air stability. This was attributed to stronger ability of PEG to bind with SWCNT defects compared to oxygen in the atmosphere.^{[ 288 ]} Typically, devices require both p‐type and n‐type materials. However, n‐type CNT materials lag their p‐type counterparts in performance and air stability, presenting a significant challenge in developing efficient n‐type CNTs for thermoelectric applications.

2.3. Defect Engineering

Defects have also been introduced to enhance the σ and S ² σ of CNTs. Generally, four types of defects are studied: topological, rehybridization, incomplete bonding, and doping with other atoms.^{[ 289 ]} Topological defects involve introducing ring sizes other than hexagons, while rehybridization refers to carbon atoms forming sp² and sp³ hybridized electrons. Incomplete bonding involves vacancies or dislocations within the carbon structure.^{[ 289 ]} Recently, the effects of carbon defects on the electronic structure of SWCNTs with three different chiral indices were calculated using first‐principles density functional theory.^{[ 234 ]} As shown in Figure 3a, when the chiral index (2n + m) mod 3 = 0, the conduction and valence bands overlap, resulting in a gapless, n‐type metallic property. In contrast, when (2n + m) mod 3 = 1 or 2, as seen in Figure 3b,c, the E _g between the valence and conduction bands exceeds 0.6 eV, displaying p‐type semiconductor characteristics. Figure 3d–f illustrates the calculations for SWCNTs with a single carbon vacancy defect for each chiral index (2n + m) mod 3 = 0, 1, or 2. Notably, for (2n + m) mod 3 = 1 or 2, the E _g decreases from above 0.6 to ≈0.3 eV, indicating enhanced carrier transitions, which in turn increases σ. Charge distribution functions and electron localization functions with and without carbon vacancy defects were also calculated for chiral indices (2n + m) mod 3 = 1 or 2, as shown in Figure 3g–j. An increment in electron density around the carbon defects was observed, suggesting a higher n _c. Overall, CNTs exhibit various types of defects, each having distinct impacts on their material performance. For practical applications, these defects should be thoroughly evaluated and strategically addressed to meet specific requirements in real‐world scenarios.

Figure 3.

Effects of carbon defects on the electronic structure for different (n, m) indices of SWCNT film. a–f) Illustration of the electronic band structure of pristine SWCNTs for (2n + m) mod 3 = 0 without carbon vacancy, (2n + m) mod 3 = 1 without carbon vacancy, (2n +m) mod 3 = 2 without carbon vacancy, (2n + m) mod 3 = 0 with carbon vacancy, (2n + m) mod 3 = 1 with carbon vacancy, and (2n + m) mod 3 = 2 with carbon vacancy. g) Diagram of charge density distribution for SWCNTs of type (2n + m) mod 3 = 1 without carbon vacancy (left panel) and with carbon vacancy (right panel). h) Schematic diagram of the electron localization function for SWCNTs of type (2n + m) mod 3 = 1 without carbon vacancy (left panel) and with carbon vacancy (right panel). i) Schematic diagram of the charge density distribution for SWCNTs of type (2n + m) mod 3 = 2 without carbon vacancy (left panel) and with carbon vacancy (right panel). j) Schematic diagram of the electron localization function for SWCNTs of type (2n + m) mod 3 = 2 without carbon vacancy (left panel) and with carbon vacancy (right panel). Here, the arrows point out carbon vacancies. Reproduced with permission.^{[ 234 ]} Copyright 2024, Nature Portfolio.

2.4. Surface Modification

In practical applications, such as during the film formation process, CNTs tend to aggregate into bundles due to van der Waals forces, leading to uneven dispersion. This issue is particularly challenging for CNT‐based composites, limiting their practical use in F‐TEGs.^{[ 65 ]} The most common and straightforward method for dispersing CNTs is sonication; however once sonication process stops, the CNTs tend to reaggregate.^{[ 290 ]} The aggregation is due to the surface chemistry of CNTs, making surface modification a widely used approach to enhance their dispersibility in solvents with varying polarities.^{[ 65 ]} The mechanism of surface modification involves binding or immobilizing active groups onto CNTs through covalent bond.^{[ 65 ]} A variety of active groups have been developed for surface modification, including ─NH₂ and carboxylic groups (─COOH). Researchers investigated functionalized CNTs using ─NH₂ groups from 11‐azido‐3,6,9‐trioxaundecan‐1‐amine and ─COOH groups from hydrogen peroxide, with the structures depicted in Figure 4a,b.^{[ 291 ]} Here, ─COOH and ─NH₂ groups are covalently bonded to CNTs through C─N and C─C bonds, resulting in a mixture with excellent chemical stability due to the anionic nature of ─COOH and the cationic nature of ─NH₂, respectively.^{[ 291 ]} Thermogravimetric analysis was conducted to monitor the specific groups attached to the CNT surface in the functionalized CNT‐COOH and CNT‐NH₂ samples, as shown in Figure 4c,d.^{[ 291 ]} Additionally, Figure 4e,f displays trace masses for H₂O (m/z = 18), NH₃ (m/z = 17), CO (m/z = 28), and CO₂ (m/z = 44) independently of temperature, indicating conversion of ─COOH and ─NH₂ to CO₂ and NH₃ during heating.^{[ 291 ]} After mixing functional CNT‐COOH and CNT‐NH₂, the S ² σ of p‐type CNT increased to 5.07 µW cm⁻¹ K⁻², while the S ² σ of n‐type improved to 1.71 µW cm⁻¹ K⁻². Importantly, mixing CNT‐COOH with CNT‐NH₂ has been observed to reduce κ while maintaining comparable σ. This is primarily attributed to the highly porous structure formed through hydrogen bonding between CNT‐COOH and CNT‐NH₂.^{[ 291 ]} Other functional groups, particularly sulfonic acid group (─HSO₃), have been explored for modifying CNT interfaces. For instance, the ─HSO₃ group in the dispersant polystyrene sulfonate (PSSH) has been used, followed by NaOH treatment, to enhance both σ and S ² σ.^{[ 259 ]} The mechanism of improved thermoelectric performance of the composite film is shown in Figure 4g. It was observed that PSSH binds to CNTs through π–π interactions between ─HSO₃ and C, forming a stable CNT solution. Excess insulating PSSH on SWCNT surfaces is then neutralized with NaOH, yielding sodium polystyrene sulfonate (PSSNa), achieving a σ as high as 3749 S cm⁻¹ and an S ² σ of 1.177 µW cm⁻¹ K⁻². Besides functional groups modified on the surface of CNTs, a range of surfactants has been employed to enhance the thermoelectric performance of CNT‐based materials.^{[ 269} , ²⁹² , ²⁹³ , ²⁹⁴ , ²⁹⁵ , ²⁹⁶ , ²⁹⁷ , ²⁹⁸ , ²⁹⁹ , ^{300 ]} Polyoxyethylene (50) stearyl ether has been reported to adsorb onto CNT bundles, thereby hindering heat transport and reducing κ, ultimately contributing to a higher ZT.^{[ 269 ]} Phosphonium salts were observed to form a dense coating on the surface of CNTs, significantly reducing their exposure to air, which is critically important for enhancing the air stability of n‐type CNTs.^{[ 292 ]} Additionally, phosphonium salts with longer hydrophobic alkyl chains have been shown to reduce CNT bundle sizes significantly, leading to dramatic improvements in σ. Similarly, ammonium halide salts have been reported as effective surfactants that wrap around CNTs to enhance the thermoelectric performance of CNT‐based materials.^{[ 294 ]} Interestingly, compared to single‐chain surfactants, gemini surfactants provide higher surface coverage, enabling more continuous carrier stability, which supports well‐defined thermoelectric output properties and enhanced air stability for extended operating durations.^{[ 294 ]} These results highlight that surface modification serves as a powerful strategy to synergistically influence multiple parameters of thermoelectric devices based on CNT materials. This approach should be meticulously optimized to achieve a high ZT, thereby enhancing the potential applications of CNT‐based materials in wearable electronics.

Figure 4.

Interface engineering of CNT‐based thermoelectric materials. a,b) Schematic illustrations of CNT structures functionalized with ─COOH and ─NH₂ active groups. c,d) Thermogravimetric analysis showing the temperature difference of percent weight for CNT‐COOH and CNT‐NH₂. e) Intensity of CNT‐COOH for m/z = 18 (H₂O), 28 (CO), and 44 (CO₂) as a function of temperature. f) Intensity of CNT‐NH₂ for m/z = 17 (NH₃ and/or H₂O), 28 (CO), and 44 (CO₂ and/or HNCO). Reproduced with permission.^{[ 291 ]} Copyright 2023, American chemical society. g) Diagram showing the fundamentals of improving thermoelectric performance of PSSH/SWCNT hybrid films. Here, PSSH is abbreviated from poly(styrene sulfonic acid). PSSNa represents poly(styrenesulfonic acid sodium salt). Reproduced with permission.^{[ 259 ]} Copyright 2024, Elsevier.

2.5. Solvent Treatment

As mentioned earlier, untreated CNTs generally exhibit p‐type characteristics due to oxygen doping in air but can shift to n‐type when doping with an n‐type dopant. During this polarity shift, the solvent plays a key role in efficiently converting p‐type/n‐type CNTs to n‐type/p‐type. Various solvents have been used for this purpose, including H₂O, CH₂Cl₂, CHCl₃, o‐dichlorobenzene (ODCB), ethanol (EtOH), dimethyl sulfoxide (DMSO), and acetonitrile (ACN). The polarity of CNTs is influenced by the type of solvents used, as the polarity and geometry of the solvents affect the interactions between CNTs and surfactants, leading to variations in doping efficiency.^{[ 301 ]} One study compared the effects of solvents H₂O and DMSO on the thermoelectric performance of CNTs in the presence of the surfactant SDBS, as shown in Figure 5a–d.^{[ 163 ]} Using all‐atom molecular dynamics (MD), researchers investigated how different solvents (H₂O and DMSO) affect SDBS coating on CNTs. In water, SDBS wraps the SWCNTs in an ordered multilayer structure, with the alkyl chains near the SWCNTs and the sulfonate groups surrounded by water molecules. However, in DMSO, SDBS molecules randomly stack on the SWCNT surface, forming a porous structure that facilitates n‐type doping, effectively converting CNTs from p‐type to n‐type. Using coarse‐grained molecular dynamics, researchers predicted SDBS coverage morphology on SWCNTs in H₂O and DMSO. After simulation, in DMSO, three small, loosely packed aggregates formed, allowing DMSO molecules to permeate the porous aggregates. In contrast, SDBS‐coated SWCNTs in H₂O formed larger bundles, which blocked H₂O molecules from penetrating and prevented the presence of oxygen, thereby maintaining the p‐type characteristics of the CNTs.^{[ 163 ]} Due to the reduced intertube spacing in H₂O, p‐type SWCNTs achieved higher σ. However, SWCNTs in DMSO exhibited better n‐type thermoelectric performance because the porous structure in DMSO allowed n‐type dopants to enter more easily compared to the multilayer structure in H₂O.

Figure 5.

Solvent effects on polarity switching of CNT films. a–d) Diagrams for the aggregation state of SDBS and surrounding (20, 0) SWCNTs molecular under 200 ns in H₂O, 500 ns coarse‐grained (CG) simulation in H₂O, 200 ns in dimethyl sulfoxide (DMSO) and 500 ns CG simulation in DMSO. Reproduced with permission.^{[ 163 ]} Copyright 2022, Elsevier. e–j) Representative snapshots of SWCNT dispersions in PAN, PMMA, and a PAN3–PMMA1 mixed solutions. (e) PAN shows attraction to CNTs in DMSO and does not bind to CNTs in DMF and in DMAc. (g) PMMA shows strongest binding to CNTs in DMSO, consistent binding to CNTs in DMAc, and some binding to CNTs in DMF. i) A mixture of PAN and PMMA shows significant binding to CNTs in DMSO, some binding in DMAc, and no binding to CNTs in DMF. (f,h,j) Magnified insets from (e,g,i) show the local structure and polymer interactions at the CNT surface. Colored numbers refer to the indicated distances in nm. Snapshots are shown for 2 × 2 × 1 supercells containing 20 mer polymers after 10 ns NPT molecular dynamics in the dispersed state. Reproduced with permission.^{[ 302 ]} Copyright 2017, American Chemical Society.

Additionally, the solvent effects on CNT–polymer composites are often challenging to examine experimentally. Pramanik et al. conducted atomic‐level studies using molecular dynamics simulations to investigate the interactions and assembly mechanisms among CNTs, polymers, and solvents.^{[ 302 ]} It was found that the significant factors affecting interactions among CNTs, solvents, and polymers include van der Waals interactions, solvent polarity, and CH–π interactions. In pure solutions, planar molecules such as DMF and DMAc exhibit stronger van der Waals interactions with CNTs compared to polar DMSO. Among polymers, PMMA demonstrates stronger binding to CNTs than PAN. Consequently, for the CNT + polymer + solvent mixtures shown in Figure 5e–j, desirable CNT dispersion with robust polymer adsorption is more readily achieved, particularly with the PMMA + PAN mixture, which can be significantly adsorbed onto CNTs in DMSO. However, at high polymer concentrations, polymer adsorption onto CNTs decreases significantly due to polymer self‐aggregation.

This study clearly demonstrates the solvent effects on CNT dispersion as well as the interactions and polarity of CNT composite materials in solution at the atomic level. These findings provide valuable guidance for selecting suitable polymers and solvents to develop high‐performance CNT composite materials for thermoelectric devices.

2.6. Hybridization

As mentioned earlier, to further enhance the thermoelectric performance of CNTs, CNTs are often hybridized with other functional materials, including inorganic nanomaterials, organic materials, and organic/inorganic hybrids.^{[ 118 ]} The hybridization between CNTs and functional materials occurs through various interactions, including covalent bonding, noncovalent interactions, physical penetration, and more. This approach potentially improves S and decouples S from σ by forming heterointerfaces that trigger an energy filtering effect,^{[ 268 ]} which can filter out low‐energy carriers and improve the m ^*. This process can lead to significant improvements in S ² σ. The currently mature CNT‐based organic composite thermoelectric materials primarily include CNT/PEDOT:PSS, CNT/polythiophene (PEDOT), CNT/PANI, CNT/polypyrrole (PPy), etc.^{[ 276} , ^{303 ]} These organic conjugated materials typically exhibit low κ and do not disrupt the favorable electronic structure of CNTs. Instead, they enable strong interactions with CNTs, contributing to enhanced thermoelectric performance.^{[ 42 ]} Additionally, the combination of CNTs with metal–organic frameworks (MOFs) has recently gained significant attention.^{[ 42 ]} MOFs generally have a high Seebeck coefficient and low thermal conductivity, but often have a low intrinsic electric conductivity. These features are complementary with CNTs. Therefore, CNT/MOF composite films are a promising class of material for thermoelectric devices. One bottleneck of CNT/MOF composites is the strong aggregation of CNT and MOF themselves. Xue et al. developed in situ growth of MOFs on CNT to address this issue, achieving a record high electrical conductivity and ZT among MOF‐composite CNT materials. Besides, research on the composites of inorganic nanoparticles with CNTs has mainly focused on the integration of silver telluride, bismuth telluride, and antimony telluride, as well as other inorganic sulfides with CNTs.^{[ 304} , ³⁰⁵ , ^{306 ]} Inorganic thermoelectric materials typically exhibit high performance and low thermal conductivity but lack flexibility. Combining high‐performance inorganic materials with organic CNTs offers a balanced solution, achieving both performance and flexibility, making them highly promising for wearable electronics.

3. Progress in Improvement of Thermoelectric Performance

With a deepening understanding of the thermoelectric mechanisms in CNT‐based thermoelectric materials, various methods have been explored to enhance their thermoelectric performance. Techniques such as doping, surface modification, and compositing have proven effective in further increasing both σ and S. Additionally, modifying the E _f and reducing the κ of the original CNT‐based materials are beneficial for achieving high ZT values, thereby promoting their practical application in F‐TEGs.^{[ 66 ]} This part will introduce several advanced methods with fundamentals, including ionic liquids (ILs) treatment, and hybridizing functional organics, metal–organics, inorganic nanomaterials, and inorganic–organic composites.

3.1. ILs Treatment

ILs are liquid ionic compounds composed of organic cations and counter anions, which have excellent ionic transport property. In recent years, these materials have attracted widespread attention in CNT‐based thermoelectric materials for improved CNT dispersion and enhanced ions concentration to significantly enhance ionic conductivity.^{[ 307 ]} The CNT dispersion mechanism with ILs is mainly ascribed to the van der Waals interactions and cation–π interactions between CNT and ILs.^{[ 308 ]} These interactions can shield the CNT–CNT interactions for efficient dispersion. The conductivity mechanism involves charge transfer through cation–π, van der Waals, and Coulomb interactions between the IL and the π‐surface of CNTs.^{[ 131 ]} Generally, a double electric layer forms between the surface charge of the CNTs and the IL: the first layer consists of cations interacting with the π‐surface through chemical interactions, and the second layer is formed by ions attracted via van der Waals and Coulomb forces.^{[ 309 ]} The efficiency of IL treatment is thus determined by the surface properties of the CNTs, including charge state, diameter, curvature, and their binding capacity with the IL.^{[ 310 ]} MD simulations are commonly used to study surface characteristics of CNTs; for example, studies have demonstrated that larger CNT diameters enhance interaction with ILs, thereby improving thermoelectric performance.^{[ 310 ]} Other research shows that increasing the curvature of CNTs positively impacts the double electric layer.^{[ 311 ]} To investigate the effect of anion size on thermoelectric performance, SWCNTs/AuCl₃ ion‐exchange doping has been used, showing that larger anion sizes lead to higher σ.^{[ 258 ]} Based on the clear fundamental of the optimization of treatment efficiency of ILs for CNTs, a variety of strategies have been investigated in recent years.

Methylimidazolium‐based ILs are commonly used for treating CNTs. For instance, researchers investigated the synthesis process of 1‐hexyl‐3‐methylimidazolium tetrafluoroborate ([HMIM][BF₄]) as an IL for treating few‐walled CNTs (FWCNTs),^{[ 131 ]} as illustrated in Figure 6a. With the help of varying concentrations of SDBS surfactants, SDBS/FWCNT films were first obtained via vacuum filtration through a nylon membrane and were then washed multiple times with deionized water. These pristine SDBS/FWCNT films then underwent two separate treatments to produce p‐type and n‐type FWCNTs. For p‐type FWCNTs, pristine SDBS/FWCNT films with a low amount of SDBS were directly vacuum‐filtered while exposed to an IL solution (created by dissolving IL in dimethylformamide, DMF). To remove excess SDBS on the FWCNT surfaces, the films were then annealed at 120 °C and washed with high amounts of DMF. To create n‐type FWCNTs, the pristine SDBS/FWCNT films were initially subjected to DMF washing with 15 mL during the vacuum filtration process, followed by the same steps as those for synthesizing p‐type SDBS/FWCNT films. Energy level calculations were performed to investigate the fundamentals of [HMIM][BF₄] treatment. Specifically, the E _f of FWCNT films was observed to lie between 4.53 eV (the HOMO level of BF₄ ⁻) and 5.05 eV (the LUMO level of HMIM⁺), facilitating p‐doping or n‐doping based on energy level adjustments (Figure 6b).^{[ 131 ]} For p‐type doping, the electrons in the pristine FWCNT films jumped to the LUMO level of HMIM⁺, shifting from −4.72 eV to −5.05 eV, whereas for n‐type doping, after DMF washing, the electrons transferred from the HOMO level of BF₄ ⁻ to the FWCNTs, changing the energy level from −4.53 to −4.66 eV. To further understand the mechanism, they analyzed the FWCNT surface in negative and neutral states. As shown in Figure 6c, FWCNTs exhibited a negative surface with p‐type doping (using HMIM⁺) and a neutral state with n‐type doping (using BF₄ ⁻).^{[ 131 ]} Ultimately, treating with ILs resulted in p‐type SDBS/FWCNT films with a thickness of ≈2 µm, optimized S ² σ values from 4.00 to 7.62 µW cm⁻¹ K⁻², and reduced κ between 3.68 and 3.84 W m⁻¹ K⁻¹, yielding a ZT value of 0.06. By integrating p‐type SDBS/FWCNT films with n‐type CNT films in an F‐TEG, they achieved a V of 334 mV and a P of 6.75 µW. This study showcased the cationic and anionic doping capabilities of ILs on CNTs for achieving p‐ or n‐type thermoelectric performance. By varying the cations or anions of ILs with distinct doping abilities or diverse interactions with CNTs, ILs‐treated flexible CNT materials hold great potential to deliver enhanced thermoelectric performance, paving the way for advanced real‐world applications.

Figure 6.

Ionic‐liquid (IL) treatment on CNTs. a) Schematic diagram for synthesizing few‐walled CNT (FWCNT) papers with and without dimethylformamide (DMF) washing. b) Illustration of the energy levels for pristine FWCNT, D15 samples, and IL molecules. Here, pristine FWCNT represents the amount of SDBS is 36.2% in the FWCNT papers. D15 denotes the rationale for SDBS is 5.87% in the FWCNT papers. c) Diagrams of the cation and anion doping effects on FWCNTs with 1‐hexyl‐3‐methylimidazolium tetrafluoroborate ([HMIM][BF₄]) under different ration of SDBS. Reproduced with permission.^{[ 131 ]} Copyright 2022, Elsevier.

Various MD‐based strategies have been explored for dispersing CNTs and enhancing their thermoelectric performance. For instance, researchers have studied the dispersion mechanisms of SWCNTs in imidazolium‐based ILs using MD simulations, revealing that van der Waals interactions play a pivotal role in the dissolution process.^{[ 312 ]} Furthermore, researchers investigated the competitive hydrogen bonding around SWCNT surfaces using two imidazolium‐based ILs, 1‐ethyl‐3‐methylimidazolium tetrafluoroborate ([Emim][BF₄]) and 1‐butyl‐3‐methylimidazolium hexafluorophosphate ([Bmim][PF₆]). Their findings revealed that PF₆ ⁻ anions form denser aggregates around SWCNTs with larger diameters compared to BF₄ ⁻ anions.^{[ 313 ]} Research has also highlighted the thermoelectric improvements when treating CNTs with ILs. It was showed that adding ILs can synergistically increase both σ and the S in SWCNT–polyelectrolyte composites, achieving an optimized S ² σ exceeding 4.60 µW cm⁻¹ K⁻².^{[ 314 ]} Similarly, researchers reported that treating SWCNTs with 1‐ethyl‐3‐methylimidazolium bis(trifluoromethylsulfonyl) imide ([Emim][TFSI]) results in a synergistic enhancement of σ and S, increasing S ² σ by a factor of 10 compared to pristine SWCNTs.^{[ 315 ]} Therefore, ILs treatment has been recognized as the promising method for the enhancement of thermoelectric performance of CNTs, and more kinds of ILs are exploring for further improving the overall performance. Even more promising, ILs possess exceptional stretchability and biocompatibility, which can significantly broaden the application scope of ILs‐treated CNTs, particularly in the field of bioelectronics.

3.2. Doping/Hybridizing with Organics

Although CNTs inherently possess a high σ, their κ is significantly higher than that of organic thermoelectric materials, resulting in a lower ZT.^{[ 316 ]} Organic materials, however, offer advantages like inherently low κ, light weight, flexibility, and cost‐effectiveness, making them promising candidates for CNT/organic composite materials. Importantly, organic materials show great promise for CNT dispersion,^{[ 317 ]} with conjugated polymers emerging as an efficient method due to their ability to disperse CNTs via a polymer wrapping mechanism on the CNT surface.^{[ 318 ]} Conjugated small molecules or polymers facilitate this process by leveraging their unique π‐conjugated backbones to interact effectively with the graphene‐like surface of CNTs. Additionally, organic compounds can act as effective CNT dopants.^{[ 150 ]} Generally, organic molecules attach to CNT surfaces through π–π stacking interactions.^{[ 319 ]} To date, the fundamental principles of organic compound interactions with CNTs have been extensively studied. The mechanisms differ for p‐type and n‐type organic dopants: in p‐type doping, where the work function of the dopant exceeds that of CNTs, electrons are transferred from the CNT to the dopant, increasing hole concentration and shifting the E _f of the CNT toward the HOMO.^{[ 320 ]} For n‐type doping, electron‐rich dopants transfer electrons to the CNT, enhancing electron density and moving the E _f toward the LUMO.^{[ 320 ]} Based on these mechanisms, a variety of strategies have been reported.

Several strategies have reported on both n‐type and p‐type hybrid CNT/organic materials, achieving substantial improvements in S ² σ. For instance, researchers demonstrated remarkably high S ² σ values, reaching 18 µW cm⁻¹ K⁻² for p‐type MWCNT/TCNQ and 10 µW cm⁻¹ K⁻² for n‐type MWCNT/PEI.^{[ 150 ]} Figure  7a shows a schematic of the synthesis process for MWCNT films doped with n‐type PEI and p‐type TCNQ. Typically, MWCNT/TCNQ and MWCNT/PEI films are prepared in three steps: wind‐speed control, doping, and cold‐pressing. MWCNT aerogels were produced via a floating chemical vapor deposition method, where adjusting wind speed optimized the alignment of the MWCNT aerogels. After identifying the optimal wind speed, different concentrations of n‐type PEI and p‐type TCNQ dopants were sprayed onto the MWCNT aerogel surface, forming MWCNT/TCNQ and MWCNT/PEI films, as illustrated in Figure 7b.^{[ 150 ]} Cold‐pressing was subsequently applied to further enhance S ² σ by increasing the density of the MWCNTs, thus synergistically increasing σ while maintaining S. The structural changes of MWCNTs before and after cold‐pressing are depicted in Figure 7c.^{[ 150 ]} Figure 7d shows an image of the F‐TEG, with an inset of a single F‐TEG leg, and Figure 7e compares ω values for various CNT/organic‐based F‐TEGs.^{[ 150 ]} By employing the synthesized MWCNT/TCNQ and MWCNT/PEI films as p–n legs in a Lego‐like TEG, a device was constructed to generate electricity from a hot plate at 80 °C, achieving a high ω of 2700 µW cm⁻². Additionally, both the p‐ and n‐type composite films demonstrate excellent air stability for over 60 days, which can be attributed to the stability of the dopants and the compact packing of the samples achieved through pressing.^{[ 150 ]} However, to achieve a high ZT, doping‐related CNT materials need to address the challenge of reducing κ, particularly after cold‐pressing, which enhances the S ² σ but fails to lower the κ.^{[ 150 ]}

Figure 7.

Illustration of doping/hybridizing organics to CNTs. a) Schematic of the synthesis process of double‐walled carbon nanotube (DWCNT) with n‐type dopant polyethyleneimine (PEI) and p‐type dopant TCNQ. b) Schematic diagrams depicting the structure of TCNQ‐doped MWCNT (left) and PEI‐doped MWCNT (right). c) Schematic diagram showing the structural changes in MWCNT before and after cold pressing. d) Image of an F‐TEG device, with a close‐up inset of a single F‐TEG leg. e) Comparison chart of ω for the reported CNT/organic‐based F‐TEGs. Reproduced with permission.^{[ 150 ]} Copyright 2022, Wiley.

Numerous research groups have reported diverse strategies to optimize the S ² σ of p‐type or n‐type CNT/organic hybrid films. For instance, researchers made p‐type MWCNT/PBDTT‐FTTE films. Following a doping process with FeCl₃, these hybrid films showed an enhanced S ² σ of 0.4821 µW cm⁻¹ K⁻².^{[ 114 ]} This improvement is mainly from two aspects: on the one hand, the wrapping of PBDTT‐FTTE on the surface of MWCNT is evidenced by several measurements, delivering a homogeneous dispersion of MWCNT/PBDTT‐FTTE solution in o‐DCB. On the other hand, doping delivered a realignment of Fermi level that the work function of doped composite film is reduced from −4.845 to −4.969 eV indicating that the Fermi level of doped composite film is shifting closer to the valence band. An 181% improvement of S ² σ reaching to 13.77 µW cm⁻¹ K⁻² in p‐type SWCNT/macrocyclic crown ether (P(BDTC)) films was reported compared with the counter polymer with linear ether chains.^{[ 133 ]} This enhancement primarily stems from side chain engineering, where crown chains increase steric hindrance, preventing polymer backbones from self‐aggregating. This allows the polymer to interact more effectively with CNTs, improving dispersion. Other studies have reported the design of a series of spiro‐bifluorene derivatives combined with SWCNTs, demonstrating that the introduction of oxygen and sulfur atoms adjusts the HOMO levels of the organic molecules to better align with the Fermi level of SWCNTs. Notably, for Spiro‐MeOTAD with oxygen atoms, the interfacial interaction and electron transfer with SWCNTs were experimentally verified. As a result, the Spiro‐MeOTAD/SWCNT film achieved the highest S ² σ of 2.39 µW cm⁻¹ K⁻² compared to the other two counterpart polymers.^{[ 168 ]} One group investigated the relationship between the intertube junctions of conjugated polymers and the thermoelectric performance of CNTs/polymer hybrids.^{[ 212 ]} Their findings indicated that the isotropic and continuous distribution of the conjugated backbone determines the transportation efficiency of charge carriers at intertube junctions. By comparing two conjugated polymers, IDTBT and PBTTT, it was found that while both polymers exhibit similar dispersion abilities for CNTs, the amorphous IDTBT, which does not self‐aggregate on the surface of CNTs, facilitates charge carrier transport at intertube junctions. This results in higher σ compared to PBTTT, which tends to self‐aggregate on the CNT surface, thereby suppressing charge carrier transport between neighboring CNTs. Although the structure of IDTBT facilitates phonon propagation, leading to a slightly higher κ, the IDTBT/CNT composite achieves a higher ZT of 0.043 due to its superior S ² σ, compared to the PBTTT/CNT composite (ZT = 0.019).^{[ 321 ]} Others explored the cross‐linking effects on SWCNTs/polymer composites, demonstrating that the use of Grubbs reagent enables covalent cross‐linking of exo‐olefine on the side chains of conjugated polymers, which negatively impacted σ, leading to reduced charge transport due to the formation of a more closely cross‐linked network.^{[ 321 ]} This effect might be good for future stretchable thermoelectric devices, but need to consider the balance of stretchability and performance upon practical applications. Some researchers applied multiple interface engineering techniques to enhance the S ² σ of CNTs/PANI, achieving a value of 4.07 µW cm⁻¹ K⁻².^{[ 322 ]} They showed that dedoping and redoping processes could improve the charge filtering effect, resulting in increased σ while maintaining S. Typically, the tensile properties of CNTs are improved by incorporating them with polymers through interfacial interactions, cross‐linking, and functionalization. Specifically, load transfer at the interface is enhanced by the interfacial interactions between CNTs and polymers, which facilitate the packing and orientation of polymer chains, thereby improving the tensile strength.^{[ 323 ]} Cross‐linking effectively reduces sliding between CNTs by introducing covalent bonds, thereby enhancing tensile strength.^{[ 324 ]} Functionalizing CNTs with polymers increases Young's modulus through the incorporation of polymer chains, resulting in enhanced tensile properties.^{[ 325 ]} For instance, Zheng et al. studied the tensile properties of SWCNT/polymer composites and found that a ladder‐type conjugated polymer, TPT‐TT, with superior backbone planarity enables efficient encasing on SWCNT surfaces, leading to more uniform dispersion and resulting in a desirable nanocomposite film with a high S ² σ of 6.788 µW cm⁻¹ K⁻².^{[ 326 ]} Importantly, by mixing the SWCNT/TPT‐TT composite with styrene–ethylene–butylene–styrene (SEBS) elastomer and carefully controlling the weight percentage of SEBS, well‐defined entangled SEBS domains are formed, which maintain the integrity of the rigid SWCNT/TPT‐TT composite while significantly enhancing its stretchability. At 25 wt% SEBS, a balance between stretchability and power factor was achieved, offering a practical approach to developing stretchable thermoelectric devices for wearable applications. Despite conjugated polymers being recognized as promising materials for fabricating CNTs/polymer composites with improved S ² σ, the mechanisms underlying these enhancements are diverse. Further studies focusing on the interaction mechanisms between CNTs and conjugated polymers are necessary to derive university principles for high‐performance CNT/polymer composite thermoelectric materials.

For n‐type CNT/organic films, researchers investigated ferrocene derivatives doped with PEI/SWCNT, resulting in an S ² σ of 1.822 µW cm⁻¹ K⁻², attributed to an increase in S to −64.28 µV K⁻¹.^{[ 327 ]} More importantly, the strong interactions between glycol‐substituted ferrocene and PEI enable ferrocene to attach to the surface of CNTs, forming a thorn‐like structure that effectively shields the CNT composite film from water and oxygen, thereby significantly enhancing the air stability of the resultant devices.^{[ 327 ]} The incorporation of ferrocene derivatives significantly reduced the κ of the resultant composite films from 60 to 10 W m⁻¹ K⁻¹, resulting in an improved ZT value of 4.69 × 10⁻³.^{[ 327 ]} Another report indicated that both the NMP solvent and FcMA can transfer electrons to SWCNT, which shifts the Fermi level up for a dramatic increment of conductivity. The fabricated n‐type hybrid SWCNT/FeMA films achieved an S ² σ as high as 5.67 µW cm⁻¹ K⁻². Additionally, phonon scattering at the interface formed between SWCNTs and FcMA results in a slightly reduced κ, leading to an improved ZT value of 3.0 × 10⁻³.^{[ 77 ]} The air stability of SWCNTs is also improved compared to pure SWCNTs due to the reducing ability of FeMA, which mitigates the oxidation of SWCNTs in air. In contrast, the n‐type properties of pure SWCNTs quickly transition to p‐type characteristics within 1 h of air exposure. Additionally, other researchers have examined the effects of borane‐nitrogen derivatives as n‐type dopants on the thermoelectric performance of SWCNT composites. Their findings indicate that SWCNT/pyridineborane (PYB) films, with a planar structure and low steric hindrance, exhibit an improved S ² σ of 2.238 µW cm⁻¹ K⁻² and enhanced air stability compared to the commonly used SWCNT/NaNH₄ composite films.^{[ 117 ]} However, it is unfortunate that the utilization of PYB did not reduce the κ. Notably, this marks the first instance of using borane‐nitrogen derivatives as n‐type dopants for SWCNTs.^{[ 117 ]} Considering the fine tunability of reducing capability and multiple structural modifications of borane‐nitrogen derivatives, it is expected that more borane‐nitrogen derivatives can be developed to boost the n‐type thermoelectric performance of SWCNT composites.

Another class of organic materials is conducting polymers (CPs), which are a type of organic macromolecule constructed from one or more repeating units connected by covalent and π bonds along the backbone, which creates a delocalized π‐electron cloud around the chain.^{[ 328 ]} Over the decades, CPs have been recognized as promising thermoelectric materials due to their advantages, such as lower κ, high machinability, and solution processability.^{[ 145 ]} Researchers have shown significant interest in investigating CNTs/CPs composites, which benefit from the combination of lower κ from CPs, along with the higher σ from CNTs.^{[ 329 ]} In CNTs/CPs composites, CPs typically wrap around the surface of CNTs through π–π interactions; for example, long strip PEDOT:PSS can wrap around the surface of SWCNTs, forming hybrids. Additionally, Coulombic interactions arise from charge carriers transferred from CPs to CNTs, resulting in increased σ and S due to a shifted E _f of the hybrids.^{[ 171 ]} With the assistance of dopants, these Coulombic interactions can be further enhanced by doping radical cations of CPs with anions of dopants, leading to an improvement in S ² σ. Based on this mechanism, various strategies have been explored to enhance the performance of CNTs/CPs composites.^{[ 40 ]}

Researchers investigated a nonsolvent secondary doping method at high velocity. They utilized PEDOT:PSS as the CP to fabricate SWCNT/PEDOT:PSS films with an extremely high conductive pathway. The synthesis process, including dispersion and cold pressing, is illustrated in Figure 8a.^{[ 185 ]} During the dispersion step, PEDOT:PSS was mixed with dopants, and the mixture was introduced into an SWCNT dispersion using a syringe to obtain a well‐mixed solution. During this process, since SWCNT is dispersed in nonsolvent, the addition of PEDOT:PSS will have strong physical interactions (such as shearing and colliding) with SWCNT to form different morphologies. It was proved that there will be an optimal SWCNT wt% upon applying different nonsolvents (in the current case, 80% for DMSO and 40% for DMF). Subsequent filtration, washing, and drying completed the fabrication of the SWCNTs/PEDOT:PSS films. The synthesized hybrids then underwent cold pressing at 40 MPa for 10 min to further enhance S ² σ. As mentioned above, with varying wt% SWCNTs, the morphologies are different. Therefore, the thermoelectric performance of SWCNT/PEDOT:PSS is closely correlated with morphology. Taking the formation of the SWCNT/long strip PEDOT:PSS hybrid films shown in Figure 8b as an example,^{[ 185 ]} as depicted in Figure 8c, after doping, PEDOT aggregated, and excess insulating PSS was removed, creating charge percolation channels that resulted in higher σ, with holes in PEDOT:PSS primarily conducting through interchain jumps.^{[ 185 ]} As shown in Figure 8d, the PEDOT:PSS surface underwent deformation when dissolved in highly polar dopants, leading to varied long strip and irregular multibranched morphologies.^{[ 185 ]} Careful optimization of SWCNT concentrations revealed fluctuations in κ. Ultimately, at 40 wt% SWCNTs, the SWCNTs/PEDOT:PSS composite achieved the best S ² σ of 5.0132 µW cm⁻¹ K⁻² with a high σ of 4717 S cm⁻¹, contributing to a maximum ZT value of 0.012.^{[ 185 ]} To ensure the long‐term operation of the TEGs, they also assessed the flexibility of the SWCNTs/PEDOT:PSS hybrids by monitoring real‐time resistance changes. As illustrated in Figure 8e, after bending 1000 cycles with ≈45% strain, the resistance change remained within 5%, indicating that the SWCNTs/PEDOT:PSS hybrids exhibited excellent mechanical stability for devices.^{[ 185 ]} The V and loading current (I) of a single TEG were monitored as a function of ΔT, as shown in Figure 8f.^{[ 185 ]} The stabilized V and I were 3.8 mV and 315 µA at a ΔT of 23 K, respectively. Finally, they assembled prototype TEGs by linking 6 legs of p‐type SWCNTs/PEDOT:PSS hybrids with polyimide (PI) substrate using silver paste and copper foil, as depicted in Figure 8g. The maximum P achieved was 4.416 µW at a ΔT of 58 K, resulting from V and I value of 7.3 mV and 605 µA, respectively. Overall, this method effectively prevents the self‐aggregation of polymers, which is particularly important for polymers with strong π‐stacking properties. By wrapping around the SWCNTs, the high π‐stacking of the polymer subsequently enhances the σ of the composite film. Hence, this method is also well‐suited for fabricating SWCNT composite films with pure polymers beyond PEDOT:PSS.

Figure 8.

Effects of introducing poly(3,4‐ethylenedioxythiophene)/poly(styrenesulfonate) (PEDOT:PSS) in SWCNT‐based hybrid films. a) Schematic diagram of the synthesis process of SWCNT/PEDOT:PSS film. b) Schematic of formation process of SWCNT/long strip PEDOT:PSS hybrid films. c) Diagram of morphological models for PEDOT:PSS before and after doping. Here, h ⁺ represents hole carriers. d) Schematic illustration of morphology change process for PEDOT:PSS from original state to long and multibranched strip. e) Resistance and strain as a function of number of cycles of SWCNT/PEDOT:PSS leg. Here, ΔR/R ₀ denotes resistance change divided by original resistance. f) Voltage (V), temperature difference (ΔT), and current (I) as a function of time for SWCNT/PEDOT:PSS‐based F‐TEGs. g) Diagram and image of F‐TEG device. Here, PI represents polyimide. Reproduced with permission.^{[ 185 ]} Copyright 2023, American Chemical Society.

Researchers have also fabricated p‐type SWCNTs/PEDOT:PSS hybrids using secondary doping methods, including DMSO doping and sodium borohydride (NaBH₄) dedoping. The mechanism of each step is illustrated in Figure 9a.^{[ 145 ]} During DMSO doping, the interaction between DMSO and PEDOT:PSS leads to the extension of the dominant PEDOT chains by removing excess insulating PSS. This process creates channels for charge carriers, resulting in an increase in σ from 5.3 to 775 S cm⁻¹. With the addition of SWCNTs, further insulating PSS is removed from the PEDOT:PSS matrix, establishing 3D conductive paths between the SWCNTs and PEDOT:PSS. This provides additional conductive channels for charge carriers, resulting in an increase in σ to 2611 S cm⁻¹, while the S remains almost unchanged. During the NaBH₄ doping process, S is increased from 16.8 to 49 µV K⁻¹ by reducing the oxidation level of PEDOT:PSS while maintaining the structure of the SWCNTs. This leads to an enhancement of S ² σ to 4.11 µW cm⁻¹ K⁻², benefiting from σ of 1718 S cm⁻¹ and S of 49 µV K⁻¹, respectively. It is worth to note that during both the DMSO doping and NaBH₄ dedoping processes, the dispersion of SWCNTs in PEDOT:PSS is observed to keep well that is the prerequisite for continueous increment of thermoelectric performance. Due to the low κ of the PEDOT:PSS matrix, the κ of the PEDOT:PSS/CNT composite film was experimentally measured to be only 0.4 W m⁻¹ K⁻¹, resulting in a ZT value of ≈0.3.^{[ 145 ]} The researchers fabricated homemade parallel TEGs with 8 legs using the synthesized SWCNTs/PEDOT:PSS hybrids, achieving a P of 0.391 µW at a ΔT of 20 K.^{[ 145 ]}

Figure 9.

Effects of doping on SWCNT/PEDOT:PSS hybrid films. a) Illustration of the mechanism for SWCNT/PEDOT:PSS hybrid films. Reproduced with permission.^{[ 145 ]} Copyright 2022, Elsevier. b,c) Schematic diagrams of F‐TEG in vertical and horizontal direction based on SWCNT/PEDOT:PSS hybrid films with same p–n pairs. d) Diagram of a typical F‐TEG process, starting from one single leg (left) to multiple legs (middle) to fabricate an F‐TEG device (right). Reproduced with permission.^{[ 266 ]} Copyright 2024, Elsevier.

The same group conducted further research to improve S ² σ in SWCNTs/PEDOT:PSS hybrids by using a binary dopant combination of sodium bicarbonate (NaHCO₃) and ethylene glycol (EG), achieving higher S ² σ values exceeding 5.00 µW cm⁻¹ K⁻².^{[ 266 ]} Similarly, after binary doping, the dispersion of SWCNT in PEDOT:PSS is observed to keep well. Differently, in this work, they successfully assembled vertical F‐TEGs with enhanced thermoelectric performance. As illustrated in Figure 9b,c, they compared simulations of heat flow between parallel and vertical F‐TEGs under the same thermoelectric leg conditions.^{[ 266 ]} The results indicate that vertical F‐TEGs can maintain a larger ΔT with less heat loss, leading to a higher P of 2.6 µW at a ΔT of 48 K. Typically, a single F‐TEG leg was fabricated by securing the thermal legs onto supporting foam, and the p–n legs were connected using conductive silver adhesive. The structure is shown in the left panel of Figure 9d.^{[ 266 ]} The middle panel of Figure 9d illustrates the formation of the F‐TEGs from a single leg; specifically, a single leg pair was interconnected with copper wires to create a closed circuit. The bottom of the F‐TEGs was protected by covering it with a PI substrate, and the digital image of the assembled F‐TEGs is presented in the right panel of Figure 9d.^{[ 266 ]} The F‐TEG exhibits a competitive normalized power density (ω _n) of over 2.5 µW cm⁻² K⁻². Beyond the highlighted advantages of this work in thermoelectric devices, the vertical device structure holds significant promise for the development of future microscale devices, enabling high‐density integration and facilitating more cutting‐edge applications.

In conclusion, significant progress has been made in the organic doping and hybridization of CNTs in recent years. The underlying mechanisms are well‐understood, and numerous strategies have been explored to develop both n‐type and p‐type CNT/organic hybrid films, with F‐TEGs assembled to advance practical applications in wearable and implantable devices. However, achieving simultaneous improvement in σ and S remains challenging due to the coupling interactions within hybrid CNT/organic films. Leveraging the multiple structural modifications of organic materials and their inherent advantages for thermoelectric devices, CNT/organic composites are anticipated to become a key focus for advancing thermoelectric performance to levels comparable with inorganic materials, while offering additional desirable features such as stretchability and implantability for wearable bioelectronic applications.

3.3. Hybridizing Metal–Organic Materials

Metal–organic materials, such as MOFs, are microporous crystalline materials composed of metal ions and organic ligands. The shape and size of the pores are determined by the properties of the metal ions and organic ligands used.^{[ 150 ]} These composite materials have garnered considerable interest as promising thermoelectric materials due to their high surface area, the presence of active sites, tunable structure, and inherently low κ.^{[ 330 ]} However, the weak overlap between the valence orbitals of the metal ions and the π orbitals of the organic ligands leads to low σ, which limits their practical application in commercial thermoelectric materials.^{[ 331 ]} Consequently, researchers have focused on mechanisms to improve the S ² σ of CNT/MOF hybrids. For example, oxygen adsorption on the porous surface of MOFs has been reported.^{[ 332 ]} Based on this finding, it has been demonstrated that MOFs can act as oxygen storage centers, enabling CNTs to undergo oxygen p‐type doping, which shifts the E _f of CNTs toward the HOMO, thus inducing p‐type doping. Various strategies have been explored to enhance the S ² σ of CNT/MOF hybrids based on these established principles.^{[ 260 ]}

Some reported strategies have focused on establishing the relationship between the metal–organic structure and thermoelectric performance by adjusting the ratio of CNTs to metal–organic materials in CNT/MOF hybrids. For instance, researchers investigated the correlation between MOFs and the S ² σ of CNT/MOFs by experimenting with different ratios of the two components.^{[ 260 ]} Targeted CNT/MOFs were synthesized through two simple steps, as illustrated in Figure 10a.^{[ 260 ]} First, in a dissolution step assisted by SDBS, pristine SWCNTs were sonicated to obtain a well‐dispersed SWCNT solution. Then, varying amounts of synthesized Al(OH)(O₂C)–C₆H₄–(CO₂) as MOF powders were added to the SWCNT solution, followed by a second sonication to form a uniform MOF/CNT mixture. In the next step, the mixture was dried at 60 °C overnight, producing freestanding CNT/MOF films ≈25–30 µm thick. This process yielded two pore geometries and sizes in CNT/MOF films, specifically MIL‐68(Al) and MIL‐53(Al) films. MIL‐68(Al) exhibited a hexagonal structure with pore sizes of ≈13.9 Å and smaller triangular pores ≈5.0 Å, while MIL‐53(Al) featured diamond‐shaped pores ≈6.2 Å, as shown in Figure 10b.^{[ 260 ]}

Figure 10.

Effects of introducing metal–organic framework (MOF) in CNT‐based hybrid films. a) Schematic of the synthesis process for MOF/CNT hybrid films. b) Illustration of two types of MOFs framework. Here, MIL‐68(Al) represents hexagonal pores with a size of ≈13.9 Å. MIL‐53(Al) denotes rhombus‐shaped pores with sizes of ≈6.2 Å. c) Diagram showing the oxygen adsorption process in MOFs and the resulting n‐to‐p type inversion phenomenon in CNTs. d) Comparison of ZT values as a function of κ in porous‐based thermoelectric CNTs. e,f) Diagram and image of flexible paralleled TEG with seven p–n pairs on polyethylene terephthalate (PET) substrate. g) Image of an F‐TEG device in a perpendicular configuration on a corrugated fiberboard substrate. Reproduced with permission.^{[ 260 ]} Copyright 2024, Elsevier.

Upon comparing the oxygen adsorption capabilities of MIL‐53(Al) and MIL‐68(Al) films, researchers found that MIL‐53(Al) with smaller pore sizes had higher oxygen content, resulting in greater p‐type doping efficiency and higher electrical conductivity. Additionally, they explored the fundamentals of MIL‐68(Al) films, illustrated in Figure 10c, where electrons were transferred from oxygen absorbed and uniformly stored in the MOFs to the CNTs, inducing p‐type doping.^{[ 260 ]} Figure 10d shows a comparison of ZT values as a function of κ for different porous‐based hybrid films.^{[ 260 ]} The MIL‐68(Al)/CNT composite film achieved an exceptionally low κ, resulting in a high ZT value. The reduced κ is primarily attributed to the hexagonal pore shape and larger pore size of MIL‐68(Al), as opposed to the smaller rhombus‐shaped pores of MIL‐53(Al). This structural feature facilitates greater gas–gas interactions in the MIL‐68(Al)/CNT film, effectively lowering κ compared to the gas–crystal interactions observed in the MIL‐53(Al)/CNT film.^{[ 260 ]} They then fabricated prototype F‐TEGs using seven p–n legs arranged in parallel on a polyethylene terephthalate substrate and perpendicularly on a corrugated fiberboard substrate, with images of the configurations shown in Figure 10e–g.^{[ 260 ]} A maximum P of ≈0.1655 µW was achieved at a ΔT of 20 K. MOFs exhibit varied porosities that influence their oxygen adsorption capabilities, making the relationship between MOF porosity and oxygen adsorption a key screening criterion for identifying promising MOFs to achieve high thermoelectric performance in CNT/MOF composite films.

Different strategies have been reported to optimize the S ² σ of CNT/MOF films. For instance, researchers have analyzed the charge transport mechanism in CNT/MOF hybrids to improve thermoelectric performance. In one study, Ni(2,3,6,7,10,11‐hexaiminotriphenylene)₂ (Ni₃(HITP)₂) MOFs with CNT fillers demonstrated a significantly enhanced ZT of 1.42 × 10⁻², approximately 270 times higher than pristine MOFs, due to reduction of thermal conductivity and improvements in both the S of 40 µV K⁻¹ and σ of 150 S cm⁻¹.^{[ 119 ]} Notably, the low κ of the composite film is primarily attributed to the MOF itself rather than the CNT–MOF–CNT junctions.^{[ 119 ]} The air stability of SWCNTs is also improved compared to pure SWCNTs due to the reducing ability of FeMA, which mitigates the oxidation of SWCNTs in air. In contrast, the n‐type properties of pure SWCNTs quickly transition to p‐type characteristics within 1 h of air exposure. Additionally, this MOF material can be combined with either p‐ or n‐type SWCNTs to produce p‐ and n‐type SWCNT/MOF composite films, respectively. Due to the exceptional air stability of Ni₃(HITP)₂, neither the σ nor the S of these composite films showed any degradation after 30 days of air exposure, demonstrating their outstanding stability.^{[ 119 ]} Other researchers have examined the effects of borane‐nitrogen derivatives as n‐type dopants on the thermoelectric performance of SWCNT composites. Their findings indicate that SWCNT/PYB films, with a planar structure and low steric hindrance, exhibit an improved S ² σ of 2.238 µW cm⁻¹ K⁻² and enhanced air stability compared to the commonly used SWCNT/NaNH₄ composite films.^{[ 117 ]} However, it is worth noting that upon making p‐ and n‐type composite films, the p‐type composite achieved better thermoelectric performance than n‐type one. This is mainly because the electrostatic interactions between n‐type Ni₃(HITP)₂ and n‐type PEI–CNT are weaker than that between n‐type Ni₃(HITP)₂ and p‐type CNT. Therefore, the CNT dispersion is better in p‐type composite than in n‐type composite. Other groups developed a ZIF‐67/CNT hybrid via an in situ growth process with the zeolitic imidazolate framework (ZIF‐67) as a dopant. It was found that the κ of the ZIF‐67/CNT composite is dominated by phonons at a low CNT ratio and by electrons at a high CNT ratio. Careful optimization of the CNT content resulted in a ZT of 2 × 10⁻² at room temperature, primarily attributed to an increased σ of ≈825.7 S cm⁻¹ and a reduced κ of 4.1 W m⁻¹ K⁻¹ at a CNT:ZIF‐67 ratio of 0.25.^{[ 42 ]} The in situ growth process might negate the poor electrical conductivity of MOF itself. However, this work can be more complementary if the performance can be compared with a control ZIF‐67/CNT composite film made by simply mixing the CNT and ZIF‐67. A similar work was done that flexible hybrids were synthesized by combining Ni‐1,2,5,6,9,10‐triphenylenehexathiol with SWCNTs through covalent grafting, which resulted in a ZT as high as 3.7 × 10⁻² at room temperature.^{[ 159 ]} The difference with in situ growth process is that the content of SWCNT is only 4 wt%, which indicates that this method is a promising method to compensate the electric conductivity by utilizing SWCNT but keep low thermal conductivity and high Seebeck of MOF itself.

In conclusion, doping CNTs with various types of metal–organic materials has significantly advanced the optimization of thermoelectric performance. However, the ZT values remain relatively low, and the fundamental mechanisms underlying the doping of metal–organic materials are still not fully understood.

3.4. Hybridizing Inorganics

Inorganic thermoelectric materials typically exhibit higher S and lower κ.^{[ 27 ]} To enhance overall thermoelectric performance, CNTs/inorganic composites have been developed with improved S ² σ, where CNTs often act as structural supports while providing flexibility.^{[ 333 ]} The enhancement in CNTs/inorganic hybrids arises from charge carriers transfer between the inorganic materials and the CNTs, which alters the E _f of the composites. To maintain high flexibility in the hybrid materials, these inorganics should be made into nanosized particles, such as ball‐milling.

Till now, various strategies have been reported to achieve CNTs/inorganic hybrids with enhanced S ² σ. For instance, researchers demonstrated that neighboring Sb₂Te₃ donates electrons to SWCNTs, achieving n‐type doping and fabricating flexible SWCNTs/Sb₂Te₃ hybrids with an exceptionally high S ² σ of 24.4 µW cm⁻¹ K⁻².^{[ 333 ]} They proposed a simple process for obtaining SWCNTs/Sb₂Te₃ by recrystallizing Sb₂Te₃ onto the surface of SWCNTs. Additionally, researchers indicated that electrons could be transported from Bi₂Te₃ to SWCNTs to create highly ordered SWCNTs/Bi₂Te₃ composites with excellent mechanical properties and an improved S ² σ of 16 µW cm⁻¹ K⁻². This improvement was attributed to interface engineering, nanopore array structures, and crystal orientation effects.^{[ 305 ]} These effects are attributed to the layered‐structure design strategy of Bi₂Te₃ on SWCNTs, which enables strong phonon scattering to significantly reduce κ, achieving a high ZT of 0.9. Additionally, this design delivers excellent temperature stability and reproducibility, highlighting its potential for practical device prototypes.^{[ 305 ]} Another study proposed a chemically exfoliated process to synthesize SWCNTs/SnSe films using a simple mixed solution and subsequent filtration method, achieving an S ² σ of 1.45 µW cm⁻¹ K⁻² at room temperature.^{[ 68 ]} The incorporation of SnSe enhanced interfacial phonon scattering, reducing the κ from 24 W m⁻¹ K⁻¹ (SWCNT) to 14 W m⁻¹ K⁻¹ (SWCNT/SnSe composite), which improved the ZT value from 0.0006 to 0.003.^{[ 68 ]} In addition to intermetallic compounds, monometals have also been investigated for doping CNTs to improve S ² σ. For example, researchers decorated SWCNTs with iron (Fe), resulting in an S ² σ increase of 140% compared to pristine SWCNTs through spontaneous reduction.^{[ 320 ]} Furthermore, metal ions such as Cu²⁺ were utilized to dope aniline tetramer (ANIT)/SWCNT composites, achieving an S ² σ of 1.453 µW cm⁻¹ K⁻² at room temperature, due to optimized σ of 560.1 S cm⁻¹ and S of 51.0 µV K⁻¹.^{[ 255 ]} A synergistic effect is observed where ANIT attaches to the surface of CNT via π–π interactions, enhancing CNT dispersion for improved σ, while the metal ions effectively dope the composite to optimize n, leading to an increased S. SWCNTs and inorganic thermoelectric materials inherently exhibit complementary characteristics, particularly in terms of S, κ, and flexibility. Achieving desirable thermoelectric parameters for a sufficiently high ZT with excellent flexibility requires careful optimization of the nanostructures formed between CNT and inorganic materials, paving the way for practical wearable applications.

3.5. Complex Hybridization

Generally, to enhance the thermoelectric performance and flexibility of CNT composite materials, multiple materials can be combined to leverage the advantages of each.^{[ 334 ]} Researchers have shown increasing interest in investigating the mechanisms behind optimized ZT for inorganic‐organic/CNT hybrids. Percolation theory has been primarily applied to explain the mechanism of improved thermoelectric performance; when the ratio of inorganic to organic content approaches the threshold, both σ and κ undergo dramatic changes in inorganic–organic/CNT hybrids.^{[ 335 ]} Researchers have explored phonon and band engineering to achieve optimized ZT, demonstrating that the κ _l significantly decreases due to various mean free paths that effectively scatter phonons.^{[ 336 ]} It has been shown that superlattice effects play a crucial role in thermoelectric performance, as increasing the thickness of superlattices can enhance S and reduce κ, resulting in a higher ZT value.^{[ 337 ]} Additionally, researchers indicated that interface thermal resistance and the density of clusters in inorganic–organic/CNT composites contribute to lower κ _l, as evidenced by scanning thermal microscopy. This is primarily due to the presence of multiple interfaces in a ternary system, which increases the interface intensity and enhances phonon scattering and damping, resulting in reduced κ.^{[ 92 ]} However, in this ternary system, conductive layers form between the inorganic–organic and organic–CNT interfaces, leading to synergistic improvements in σ and S.^{[ 92 ]} Overall, an ZT of 0.24 at 410 K was achieved.

Based on these fundamentals, various strategies have been reported. For instance, researchers fabricated CNTs/Se/PEDOT:PSS ternary composites achieving an S ² σ of 5.8419 µW cm⁻¹ K⁻², benefiting from the formation of charge transfer pathways between PEDOT:PSS and Se or CNTs.^{[ 267 ]} Another study utilized poly(3,4‐ethylenedioxythiophene)‐tosylate (PEDOT‐Tos) as a directing agent for tellurium (Te) and SWCNTs as the framework to construct flexible SWCNTs/Te/PEDOT‐Tos hybrids, resulting in an S ² σ of 1.319 µW cm⁻¹ K⁻².^{[ 108 ]} They also reported the development of spiral architecture 3D TEGs using these ternary composites, achieving a P of 7.04 µW with 10 legs.

Clearly, the inorganic–organic cohybridization with CNTs has made significant progress. However, understanding the mechanisms behind the enhancement of thermoelectric performance in such complex hybrids remains a challenge. With a deep understanding of the mechanisms to harness the advantages of each component in a complex hybrid, careful optimization of component proportions, and material preparation techniques is expected to achieve breakthroughs in ZT values, enabling practical advancements in wearable applications.

4. Design of CNT‐Based F‐TEGs

F‐TEGs are self‐powered devices that directly convert low‐grade waste heat into electrical energy based on the Seebeck effect. They offer advantages such as high mechanical performance, easy maintenance, high reliability, and environmental friendliness.^{[ 338 ]} Generally, there are two types of CNT‐based F‐TEGs: one is film‐based, composed of 2D CNT films, and the other is fiber‐based, created by weaving CNT thermoelectric wires/fibers or coating CNTs onto flexible, wearable fabric substrates.^{[ 29} , ⁹⁰ , ^{339 ]} Overall, CNT‐based F‐TEGs are promising for wearable electronics due to their promising thermoelectric P, outstanding flexibility, scalability, and good air stability.^{[ 115 ]} Table  3 summarizes the performance of CNT‐based F‐TEGs reported over the past five years. Currently, researchers are exploring various strategies and unique device design approaches to enhance the thermoelectric performance of CNT‐based F‐TEGs and expand their application scenarios.^{[ 77 ]}

Table 3.

A summary of the performance of CNT‐based F‐TEGs reported within five years. Here, the units for p–n leg, V, P, ΔT, and ω are pairs, mV, µW, K, and µW cm⁻², respectively.

Material	p–n leg	V	P	ΔT	ω	Refs.
PEDOT‐Tos/Te/SWCNTs	5	31.26	9.59	60	461.33	[108]
PEDOT:PSS/SWCNTs	9	23.2	361	48	361	[266]
SWCNTs/Ncb	20	99.32	2401.9	25	120.095	[274]
Bi2Te3‐CNT	150	338	0.1596	20	38	[265]
PEI/Au NPs/CNTFs	72	218.16	6.1	30	31.7	[231]
CNT	10	1050	950	39	15.833	[102]
CNT	76	140	144	52	7.579	[242]
PEDOT:PSS/SWCNT	6	7.3	4.416	58	4.907	[185]
PEDOT‐Tos/SWCNT	10	28	1.967	60	3.073	[104]
CNT yarns	40	58.1	6.9	30	1.03	[270]
[HMIM][BF4]/CNTs	77	334	6.75	30	0.365	[131]
PANI/SWCNT–DMSO	4	4.4	0.101	35	0.1804	[248]
P(BDTC)/SWCNT	10	23.4	0.328	70	0.1639	[133]
SWCNT/MXene	10	41.48	1.54	117.3	0.154	[132]
SWCNT/FcMA	5	22.7	0.75	54.1	0.15	[77]
PEDOT‐Tos/a‐SWCNT	6	4.68	0.5789	60	0.1286	[74]
SFX‐2/SWCNT	10	28.6	1.0	60	0.1	[98]
PEDOT:PSS/SWCNT	6	6.6	1.2	60	0.1	[120]
PEDOT:PSS/SWCNT	10	36.58	9.45	70	0.069	[205]
PEDOT:PSS/SWCNTs	8	6.34	0.391	20	0.0611	[145]
SWCNTs/DMSO	5	16	0.4374	60	0.0583	[163]
CNT/Se/PEDOT:PSS	5	3.95	0.2068	44.7	0.0517	[267]
PBDTT‐FTTE/MWCNT	7	9.12	0.007	65	0.029	[114]
CNTs/PEDOT:PSS	75	96.95	1.309	32	0.028	[210]
CNTs/PANI	40	26.2	0.377	10	0.0236	[271]
PEDOT:PSS/CNT	5	4.76	0.00041	10	0.0182	[162]
DWCNT–PU fiber	108	218	16.9	30	0.0112	[135]
Spiro‐MeOTAD/SWCNT	6	9.77	0.0573	30	0.0096	[168]
MXene/CNT/PEDOT:PSS	4	0.04	0.0032	30	0.0004	[209]
Bi2Te3/CNT	5	1.8	0.0016	303	0.00018	[203]
N‐DMBI/CNT	20	63.7	0.0057	328	0.00007	[227]
MOF/CNT	7	11.2	0.165	20	–	[260]
SiO2@MoS2/SWCNT	5	11.2	1.1	50	–	[41]
TPP/CNTs	5	1.07	0.349	298	–	[115]
SWCNTs/PYB	5	28.8	1.15	66	–	[117]
CNTFs	4	6.46	0.026	33.4	–	[116]
CNT/PEDOT:PSS	8	10.5	0.00678	33	–	[152]

4.1. Film‐Based F‐TEGs

Using CNT films for the design of F‐TEGs offers several advantages, such as ease of assembly, minimal performance loss, and strong stability. Previously, we introduced some straightforward design concepts for CNT film‐based F‐TEGs (Figure 10). To investigate CNT‐based films for F‐TEGs with continuous power sources, researchers fabricated a CNT‐based clay that utilizes the human body heat as a permanent power source, aided by a viscous additive.^{[ 115 ]} This clay can be treated as a thick film. Specifically, a nonionic surfactant, Triton X‐100, was used as a viscous additive and dispersant for CNTs, resulting in a CNT‐based clay, as shown in Figure 11a.^{[ 115 ]} They also produced p‐type and n‐type SWCNT clays by adding 5 wt% p‐type dopant TCNQ and 1 wt% n‐type dopant triphenylphosphine. Wearable devices require CNT‐based F‐TEGs that exhibit excellent elongation and remolding properties to adhere effectively to human skin. As illustrated in Figure 11b, the elongation was measured by wrapping SWCNT clays in an elastic sponge wall between elastomer films. Upon elongation from 0% to 100%, both σ and S remained stable, indicating that the SWCNT clays exhibited stable S ² σ during elongation deformation.^{[ 115 ]} Figure 11c shows an image of the F‐TEG skin patch based on CNT clay along with a schematic of its front and back sides.^{[ 115 ]} The researchers then utilized the p‐type and n‐type SWCNT clays to fabricate TEGs, as depicted in Figure 11c.^{[ 115 ]} Five pairs of p‐type and n‐type SWCNT clays were attached to a hexagonal elastic sponge mold using double‐sided adhesive. Copper wires connected the p‐type SWCNT clays with the n‐type SWCNT clays for thermoelectric testing, and biocompatible elastomers served as protective layers on both sides of the assembled SWCNT clay devices. To evaluate the actual thermoelectric performance, the fabricated SWCNT clay devices were bent on the wrist or neck, leveraging the temperature gradient between human skin (≈36 °C) and the surrounding air (≈25 °C), achieving a stable P of 0.18 µW. Figure 11d presents a schematic diagram of the remolding process for CNT clay.^{[ 115 ]} During the remolding test, the SWCNT clays could return to their original shape after significant cutting and kneading. After being remolded 50 times, both σ and S showed minimal change, demonstrating stable S ² σ during the remolding process. To evaluate the stability of the thermoelectric performance, as shown in Figure 11e, a single SWCNT clay device was stored in atmospheric conditions for one year, retaining ≈80% of its original P. This stability is attributed to the presence of a viscous additive, which acts similarly to encapsulation, protecting both p‐type and n‐type SWCNTs from air exposure.^{[ 115 ]} This work demonstrates the feasibility of F‐TEGs as e‐skin from several inspiring aspects, including reconfigurability, skin adaptability, biocompatibility, skin adhension, stretchability, repairability, and air stability. Promisingly, the thermoelectric performance is highly tolurent upon extreme mechanical deformation conditions, highlighting their potential in realistic applications.

Figure 11.

Deviceization of CNT film‐based materials. a) Schematic diagram of CNT clay as a thick film. b) Images of CNT clay before and after elongation. c) An image of F‐TEG skin patch based on CNT clay and schematic of front and back side. d) Schematic diagrams of remolding process of CNT clay. e) Power output (P) as a function of hot region temperature for the F‐TEG skin patch, both without storage and after one year of storage. Reproduced with permission.^{[ 115 ]} Copyright 2021, Elsevier.

To enhance the thermoelectric performance of CNT‐based film F‐TEGs, novel topological designs have been proposed. Researchers introduced an innovative TEG structure that incorporates intrinsically built‐in fins as heat sinks to efficiently disperse heat. Conventional heat sinks are often bulky, rigid, and heavy, which limits their practical application.^{[ 242 ]} The structure of a single F‐TEG is illustrated in the left panel of Figure 12a.^{[ 242 ]} From top to bottom, the architecture consists of top fins, two interconnects with thermoelectric materials inserted in the middle, and a bottom interconnect covered by thermal interface materials, all attached to a Kapton substrate. Typically, the fins are formed by adhering copper interconnects to the substrate, with each interconnect containing plated through holes to ensure electrical connection on both sides of the substrate. The thermal interface materials provide excellent thermal coupling with the metal base. Using physics‐based models and simulations via COMSOL modeling, the thermal distribution of the TEGs was analyzed, demonstrating that heat flows along the length of the fins on the Kapton substrate, as shown in the right panel of Figure 12a. The 2D TEGs can be folded into a 3D hexagonal honeycomb structure, either manually or with the assistance of oriented polymers. The digital images of the top view and left view of the TEGs are presented in the upper and lower panels of Figure 12b, respectively.^{[ 242 ]} This built‐in heat sink is believed to benefit more convenient integration of self‐powered F‐TEGs into realistic wearable applications.

Figure 12.

Strategies for fabricating F‐TEG device with a built‐in heat sink. a) Structure and thermal distribution of the F‐TEG device. b) Images of F‐TEG device from top view (up panel) and left view (down panel). c) Comparison of power density between Bi₂Te₃‐based and CNT‐based devices. d) Images of the designed F‐TEG device with no fins (left‐up panel), small fins (left‐middle panel), large fins (left‐down panel), along with the maximum power factor (P _max) as a function of ΔT for the three F‐TEG devices. Reproduced with permission.^{[ 242 ]} Copyright 2024, Elsevier.

Using this proposed architecture, researchers fabricated CNT‐based and Bi₂Te₃‐based TEGs, each consisting of 76 p–n pairs, and evaluated their performance. As depicted in Figure 12c, both types of TEGs demonstrated the highest ω _n values of 2 µW cm⁻² K⁻² for CNT‐based TEGs and a competitive ω _n of 0.68 µW cm⁻² K⁻² for Bi₂Te₃‐based TEGs, indicating that the proposed architecture effectively enhances thermoelectric performance.^{[ 242 ]} Notably, the thermoelectric performance of the CNT‐based F‐TEGs exceeded that of the Bi₂Te₃‐based TEGs by more than three times, highlighting their improved robustness under thermal cycling. To further investigate, the researchers compared the thermoelectric performance across different lengths of the fins, as illustrated in Figure 12d.^{[ 242 ]} Their findings suggest that increasing the length of the fins optimizes S ² σ for the CNT‐based F‐TEGs, as shown in Figure 12e. This improvement in the contact area between the F‐TEGs and the surrounding air reduces heat loss at the interface and maintains a higher ΔT across the thermoelectric materials. This innovative device structure provides significant insights into the design of future foldable and portable thermoelectric devices, enabling the seamless integration of diverse thermoelectric materials (organic, inorganic, etc.) to fully leverage the benefits of both p‐ and n‐type legs for practical applications.

Improving the fineness of p–n patterning in smaller sizes to increase the number of p–n thermocouples can significantly enhance thermoelectric performance of TEGs. Typically, conventional CNT‐based F‐TEGs are constructed using several pairs of p–n CNT films, which generate V due to thermal gradients based on the Seebeck effect, as illustrated in Figure 13a. The V can be defined as V = N × S _p–n × ΔT, where S _p–n denotes Seebeck coefficient of p–n CNT films, N represents the number of p–n patterning. In addition to the inherent S and thermal properties, the number of p–n pairs plays a crucial role in enhancing the P of CNT‐based F‐TEGs.

Figure 13.

Advances in the preparation of p–n CNT‐film patterning. a) Image of a conventional F‐TEGs and a diagram of a single p–n pair F‐TEGs. b) Schematic diagram of preparing films with dipping (left panel), printing (middle panel), and aerosol doping (right panel) methods to generate different sizes of p–n pattering resolutions. c) Doping aerosol system for CNT films. d) Photograph of an F‐TEG device (left panel) and diagram of multiple p–n pairs F‐TEG device (right panel). Reproduced with permission.^{[ 341 ]} Copyright 2024, American Chemical Society.

Researchers have made considerable progress in improving the fineness of p–n patterning. Figure 13b compares the p–n patterning resolutions of CNT films produced using three different methods. One study reported that a coarse p–n patterning can be produced by dipping which had a resolution of ≈1 mm due to capillary action separating it from the dopants.^{[ 340 ]} In the same year, researchers achieved high‐performance CNT‐based F‐TEGs by utilizing printing techniques, successfully reducing the resolution of p–n patterning to 0.1 mm.^{[ 144 ]} However, these methods cannot satisfy microscale p–n control technology. Therefore, researchers proposed an aerosol doping system to create p–n patterning with a resolution of only 1 µm.^{[ 341 ]} The aerosol doping system, illustrated in Figure 13c, involves heating pristine p‐type CNT films and exposing them to n‐type dopant aerosols with diameters under 3 µm, produced by ultrasonic nebulization and air separation. The smaller n‐type dopant aerosols return to the solution reservoir via a cold trap chiller, while larger aerosols are directly recycled to the same reservoir. Using a PI substrate, the finely resolved CNT films fabricated through this aerosol doping system were then utilized to create F‐TEGs. The digital image shown in Figure 13d demonstrates 25 p–n patterns arranged linearly, measuring 11 mm × 11 mm × 6 µm, with a thickness composed of a 5 µm PI substrate and 1 µm CNT films. This configuration achieved V in the range of several tens of microvolts and a thermoelectric conversion capacity of 2000 µV K⁻¹. To enhance the spatiotemporal resolution of thermoelectric devices, Chen et al. developed jointless p–n modules through selective filtration.^{[ 342 ]} When this module was used as a temperature sensor, it shows a quick response, a less signal variation, and a linearly matched temperature–voltage relationship, making it a desirable biomimetic thermoreceptor for robots. These results demonstrate that there is still significant potential for further expansion in the design and application of CNT film‐based F‐TEGs. Developing efficient strategies to enhance the precision of p–n patterns at small scales while maintaining well‐defined thermoelectric performance remains highly challenging but is urgently needed for advanced applications, such as multidevice integration.

4.2. Fiber‐Based F‐TEGs

For 1D CNT thermoelectric fibers, woven devices made from these fibers can be directly integrated with clothing and other materials, offering enhanced flexibility, wearability, and adaptability.^{[ 273 ]} Although their performance is often low, this significantly broadens their application scenarios. To explore CNT fiber‐based TEG with scalable and excellent thermoelectric performance, researchers fabricated a cotton‐based textile generator.^{[ 210 ]} The synthesis process is illustrated in Figure 14a, where pretreated cotton yarns underwent a dipping process to allow SWCNTs to penetrate the gaps between the fibers through sonicated dispersion and capillary action. Subsequently, the coated CNT yarns were dip‐coated with MWCNTs/waterborn polyurethane (WPU) to achieve a uniform layer of CNTs on the yarn surface. The uniform CNT yarns were then wrapped in polytetrafluoroethylene with silver paste applied to the sides for electron conduction; one side was immersed in a PEDOT:PSS solution to create p‐type segment CNT yarns, while the other side was treated with a PEI/ethanol solution to form n‐type segments. Through sewing, the fabricated 2D yarns were transformed into 3D warp‐knitted spacer fabric. During this process, dip‐coating played a crucial role in enhancing the tensile strength and coating stability of the CNT yarns. As shown in Figure 14b, after removing CNT yarns with and without dip‐coating from the nonwoven fabric, the dip‐coated yarns left significantly less residue, indicating good adhesion to the cotton matrix. After dip‐coating, the tensile strength of the yarns nearly returned to that of the original cotton, as illustrated in Figure 14c. The F‐TEGs fabricated with 75 p–n patterns exhibited a V of 37 mV and a ω of 0.009574 µW cm⁻² when placed on a human arm, as shown in the digital image in Figure 14d. To boost the real‐world applications of F‐TEGs, the procedure this work provided is facile and scalable with excellent and steady‐output TE performance, being quite beneficial. Once the device performance is improved to a higher value to enable realistic application without a voltage amplifier, its applications scale will be highly broadened.

Figure 14.

A strategy for fabricating thermoelectric textile based on CNTs yarns. a) Schematic showing the preparation process of CNTs yarns and the fabrication process of thermoelectric textiles. b) Images of the thermoelectric textile from both top view (up panel) and side view (down panel), along with an adhesive test image after two coating steps (inset image). c) Stress–strain curves after various treatment steps, with an SEM image of cotton yarns coated with MWCNTs/WPU (inset image). d) Thermoelectric textiles worn on a human arm (inset image) and a thermal photograph of the textile. Reproduced with permission.^{[ 210 ]} Copyright 2023, Springer Nature.

To enhance thermoelectric performance, researchers coated microencapsulated phase change materials (MPCMs) onto the hot side of the CNT‐based fiber F‐TEGs, creating phase‐transition‐promoted F‐TEGs that generated over 25% more energy from the human body.^{[ 231 ]} The structure, illustrated in Figure 15 , features 72 p–n segments sewn together to fabricate the CNT‐based fiber F‐TEGs. The p‐type legs were synthesized by modifying one side of the CNTs with gold nanoparticles, while the n‐type legs were formed by immersing the other side of the CNTs in PEI. A P of 0.270 µW can be achieved at a ΔT of 66 K, indicating potential for application on wearable electronics. MPCM is particularly advantageous for the efficient utilization of waste heat, precise temperature control, and electronic cooling. Moreover, careful optimization of MPCM can significantly enhance encapsulation efficiency, a critical step toward enabling practical wearable applications, especially for air‐sensitive n‐type thermoelectric materials.

Figure 15.

Schematic of F‐TEG device based on CNT fibers with MPCM. Here, MPCM denotes microencapsulated phase change materials. Reproduced with permission.^{[ 231 ]} Copyright 2024, American Chemical Society.

5. Challenge and Outlooks

In recent years, research on CNT‐based thermoelectric materials has grown exponentially, driven by continuous advancements in optimization techniques and fabrication processes. The preparation of both p‐type and n‐type CNTs is currently achieved through two primary approaches: doping and composite methods. While p‐type doping has reached a more advanced stage, leading to higher thermoelectric performance, n‐type doping, especially with alkali metal dopants, has shown fewer promising results. At present, organic and ionic dopants are predominantly used to modify CNTs. Additionally, the composite method for enhancing CNT thermoelectric materials combines the strengths of various materials, utilizing the high S of inorganic materials and the low κ of organic materials, to improve overall thermoelectric performance. Although numerous studies have been conducted on CNT‐based composite thermoelectric materials, achieving a significant breakthrough in thermoelectric performance by effectively combining the advantages of each component remains a considerable challenge. On the one hand, the combination of low‐dimensional and multidimensional nanomaterials with CNTs offers the potential to achieve higher ZT values. On the other hand, optimizing device structures based on material enhancements is essential for achieving better performance at the module level. This dual approach—focusing on both material integration and device design—will be the key to advancing the field.

Significant progress has been made in the development of CNT‐based thermoelectric materials, whether through electron transfer doping or the incorporation of other high‐performance materials. However, there remains considerable room for improvement, particularly for n‐type CNT‐based thermoelectric materials. First, the poor air stability of these materials poses a major challenge, hindering their practical application. It is essential not to focus solely on pursuing high S ² σ values, but also to develop new strategies that enhance both air stability and thermal stability to prevent oxidation. Second, the interactions at the interface between CNTs and other materials—whether inorganic or organic, in doping or composite structures—can significantly influence performance. In‐depth molecular‐level research on these interactions will be critical for optimizing overall performance. Finally, the exploration of environmentally friendly CNT‐based thermoelectric materials and devices has been limited, underscoring the need for further research into high‐performance thermoelectric materials that also prioritize environmental friendliness.

To advance CNT‐based thermoelectric materials, efforts must be directed toward scaling them up to practical dimensions (Figure 16 ). While CNT production has already reached commercial scale, the preparation methods discussed in this review remain largely confined to laboratory settings. Future research should prioritize the ongoing improvement of equipment and process flows to meet the size and scalability requirements for commercially viable CNT‐based thermoelectric materials. Achieving this will be key to enabling them to perform effectively in real‐world applications. The outlooks to CNT‐based thermoelectric materials and devices include the following aspects.

1. Structural optimization: To enhance the thermoelectric performance of CNT materials, including films and fiber‐based materials for F‐TEGs, the structure of CNTs can be further optimized. The aim is to significantly improve σ and S while noticeably reducing κ, all without compromising the fundamental physical properties of CNTs (such as strength and flexibility). This optimization can be based on defect engineering, modifying the carbon atoms and carbon bonds on the CNT surface. Functionalization, involving the design of more functional groups, can facilitate the synergistic improvement of thermoelectric properties and achieve n‐type conversion to further enhance performance. For specific material types (films, fibers, etc.), mechanical densification (e.g., cold and hot pressing), and polymer infiltration can further improve the density of CNT‐based films and fibers, enhancing σ and structural stability. Recently, metal encapsulation has proven to be an effective method to further improve the performance of fiber‐based CNTs.

2. Role expansion: CNTs can be expanded for various thermoelectric applications. This expansion not only includes their use as composite materials with other bulk or flexible thermoelectric materials but also extends to new configurations integrated with advanced manufacturing processes. Examples include 3D porous CNT sponges, CNT sponge/carbon fiber multiscale structures, and even CNT sponge/CNT fiber frameworks for thermoelectric CNT‐based nanocomposites, which enhance both interface/matrix and σ. CNT fibers (e.g., twisted, coiled, or infiltrated) and their fabrics can broaden their applications in thermoelectric energy harvesting and cooling. Functional systems that integrate thermoelectric capabilities can be developed based on fibers, textiles, and 3D porous networks/composites, which may also improve the flexibility and fatigue resistance of wearable devices.

3. Device design: The high σ of CNTs grants them an exceptional S ² σ, comparable to traditional bulk inorganic thermoelectric materials. For flexible device applications, if the temperature gradient across the device can be maintained consistently, the disadvantage of high κ is mitigated. In this case, high σ enables fast device response times, high S ² σ supports strong thermoelectric conversion capacity, and the flexibility, stability, and biocompatibility of CNTs improve the design concept and ease of use. Additionally, the controllable cost of CNTs presents promising prospects for industrial production and commercialization. Therefore, the design of F‐TEGs based on CNTs should leverage the unique properties of CNT materials to maximize their value.

4. Application innovation: As mentioned, applications for CNT‐based thermoelectric materials and devices are still emerging, but their potential use cases are vast. CNTs are a diverse class of materials with no conclusive evidence of toxicity for most types. For CNTs with potential risks, appropriate chemical functionalization can reduce their hazard. CNTs are typically tightly entangled by van der Waals interactions, forming stable structures that are not easily shed, inhaled, or able to penetrate the skin. Moreover, CNTs exhibit chemical stability and inertness within a wide voltage range and with mild chemical agents, making them ideal for sustainable power sources in wearable electronics. They can be integrated into electronic skin to enable multifunctional human–machine interaction and sustainable applications.

Figure 16.

Outlooks for CNT‐based thermoelectric materials and devices.

Conflict of Interest

The authors declare no conflict of interest.

Biographies

Xiao‐Lei Shi is an ARC DECRA fellow/senior lecturer at the Queensland University of Technology. He received his Bachelor's and Master's degrees from the University of Science and Technology Beijing in 2008 and 2011, respectively, and earned his Ph.D. from the University of Queensland in 2019. His research mainly focuses on high‐performance thermoelectric materials and devices.

Zhi‐Gang Chen is a founding director for the ARC Research Hub in Zero‐emission Power Generation for Carbon Neutrality (ZeroPC), and a Capacity Building Professor in Energy materials at the Queensland University of Technology. He received his Ph.D. in materials science and engineering from the Institute of Metal Research, Chinese Academy of Sciences, in 2008. His expertise is in sustainable functional materials, thermoelectrics, advanced manufacturing, and advanced microscopy.

Zhou S., Shi X.‐L., Li L., Liu Q., Hu B., Chen W., Zhang C., Liu Q., Chen Z.‐G., Advances and Outlooks for Carbon Nanotube‐Based Thermoelectric Materials and Devices. Adv. Mater. 2025, 37, 2500947. 10.1002/adma.202500947