Carbon nanomaterials, including graphene, carbon nanotubes (CNTs) and fullerenes [Figure 1]^[33,34], have emerged as promising candidates for flexible thermoelectric materials due to their excellent electrical conductivity, mechanical properties, and thermal stability. In particular, graphene, as a two-dimensional material, has made remarkable progress in flexible thermoelectric device research in recent years due to its ultrahigh electrical conductivity and outstanding thermoelectric performance. CNTs, especially single-walled carbon nanotubes (SWCNTs) and multi-walled carbon nanotubes (MWCNTs), also play an important role in flexible thermoelectric material development due to their tunable conductivity and high flexibility. Fullerenes are carbon allotropes with cage-like structures. They are widely used in thermoelectric, electronics, and materials science for their unique properties.

Structural regulation strategies, such as doping, nanostructuring, and composite design, have significantly improved the thermoelectric performance of carbon nanomaterials. For example, graphene/CNT composites exhibit excellent power factors. These composites combine the high electrical conductivity of graphene with the excellent thermoelectric response of CNTs, fully leveraging the synergistic effect of the materials.

CNTs are tubular materials formed by carbon atoms linked via covalent bonds with sp² hybridization into hexagonal units. Carbon atoms form a conjugated electron cloud via a pair of unhybridized p electrons extending across the CNTs. Based on the number of layers in their tubular structure, CNTs can be divided into SWCNTs and MWCNTs^[35,36]. The application of CNTs in flexible thermoelectric devices has been extensively studied. In particular, for preparing thermoelectric aerogels, their excellent electrical conductivity and tunability make them ideal candidates.

Thermoelectric aerogels constructed from CNTs exhibit high conductivity and, owing to their thickness advantage, can directly capture heat energy along the temperature-gradient direction, making them ideal flexible thermoelectric materials. CNT aerogels can be constructed via a simple, rapid-evaporation process. For example, Lee et al. constructed three-dimensional MWCNT aerogels using a rapid solvent-evaporation method. The fabrication flowchart and various aerogel morphologies are shown in Figure 2A and B^[37]. Controlling the concentration of MWCNTs in the slurry enables the preparation of aerogels with different densities and pore diameters.

Although the porous structure of aerogels can effectively reduce thermal conductivity, it also results in relatively low electrical conductivity. In contrast, CNT films exhibit significantly higher electrical conductivity. In addition to high electrical conductivity, CNT films have a key advantage - large-scale production can be achieved via chemical vapor deposition (CVD), which lays an important foundation for the commercial application of thermoelectric materials. For example, Shi et al. fabricated Ag₂Se/MWCNTs composite films via screen printing. In these films, Ag₂Se nanoparticles bridge MWCNT networks, achieving a room-temperature power factor of 168 μW m^-1 K^-2 and a thermal conductivity of 0.28 W m^-1 K^-1. The composite retains 88% of its initial conductivity after 800 bending cycles (R = 5 mm), making it suitable for photothermoelectric integration in wearable devices^[38]. Sun et al. used thiophene as the carbon source, ferrocene as the catalyst, and nitrogen gas as the protective atmosphere, respectively, and synthesized MWCNTs in batches via a high-temperature floating CVD method [Figure 2C]. Large-area oriented CNT films were collected using a rotating roller [Figure 2D]. In Figure 2E, the average inner diameter of the MWCNTs is approximately 2 nm, and the outer diameter is approximately 5 nm. Raman spectroscopy confirms the structural characteristics of MWCNTs, and a low I_G/I_D ratio verifies their low carbon defect advantage [Figure 2F]^[39].

In 2004, physicists Geim and Novoselov from the University of Manchester, UK obtained single-layer graphene from graphite via mechanical exfoliation^[40,41], marking the birth of the first two-dimensional material and igniting a global research boom on two-dimensional materials. Graphene is a two-dimensional nanosheet formed by carbon atoms arranged in a hexagonal honeycomb structure via sp²-hybridized orbitals. The thickness of single-layer graphene is approximately 0.335 nm, approaching the thickness of a single carbon atom^[42,43]. In the six-membered carbon ring of graphene, each carbon atom is bonded to three neighboring carbon atoms via σ bonds. The π electrons not participating in σ bonding interact with those of other carbon atoms to form a large delocalized π bond, endowing graphene with excellent electrical properties by enabling electrons to move freely within its crystal structure. Owing to its long-range π-π conjugated structure, graphene exhibits numerous outstanding properties, including high chemical stability, high mechanical performance, and tunable thermal conductivity^[44,45]. Building on improved graphene preparation, tailoring, and modification methods, large-area, high-quality graphene with precisely controlled patterns and specific functional groups can be used to fabricate devices. These advances will promote in-depth research on graphene’s novel structures and electronic devices - such as magic-angle graphene and graphene-based single-molecule junctions - to facilitate the exploration of graphene’s novel properties and applications^[46].

Current forming processes can be used to prepare graphene into a series of macroscopic bulk materials. Based on structural and morphological differences, these macroscopic materials are mainly divided into three categories: 3D aerogels, 2D films, and 1D fibers. Graphene aerogels used in the thermoelectric field are mainly constructed via template methods, including polymer-template and ice-template methods. For example, Zhang et al. used polydimethylsiloxane (PDMS) as the skeleton and mixed bulk NaCl into a graphene/PDMS mixed solution to serve as a template. The characterization and properties of the thermoelectric (TE) sponges are shown in Figure 2G-L^[47]. Owing to the extremely high fracture strength of PDMS, the graphene-based aerogel was endowed with excellent mechanical properties. It could still effectively maintain good morphological integrity and resistance stability during repeated compression cycles.

Unlike the CVD method for preparing large-area CNT films, the mass preparation of graphene films mainly relies on slurry-forming methods. For example, Feng et al. used a graphene dispersion with a concentration of 25 mg mL^-1 as the slurry, cast it on a conveyor belt moving at constant speed, controlled the thickness of the gel-state slurry with a roller, and obtained large-size layered graphene films after heating, drying, and peeling [Figure 2M-O]^[48]. Scanning electron microscopy (SEM), Raman spectroscopy, and X-ray diffraction (XRD) confirm that graphene films prepared via the doctor-blading process exhibit good uniformity. Based on graphene’s conduction mechanism, the Wiedemann-Franz law, and Mott’s law, further optimization of carrier concentration can enable graphene’s room-temperature power factor to exceed 200 μW cm^-1 K^-2. To reduce thermal conductivity, Guo et al. prepared layered graphene oxide (GO) films with porous structures via the ice-template method and thermally reduced them in high-temperature N₂ and NH₃ atmospheres; N doping via NH₃ resulted in n-type conductive graphene materials. The continuous porous structure in graphene films can reduce the lattice thermal conductivity via phonon-scattering mechanisms^[49]. In addition, the porous structure increases the contact barrier between graphene nanosheets and enhances the energy-filtering effect, slightly increasing the Seebeck coefficient of graphene films. Furthermore, they assembled graphene films onto pre-stretched PDMS substrates. After release, graphene formed wrinkled structures on the PDMS film, thereby gaining stretchability. This pre-stretching process offers a new approach for the fabrication and structural design of flexible thermoelectric films.

Chalcogenide nanomaterials (e.g., Bi₂Te₃, PbTe) have long dominated the fields of thermoelectric power generation and refrigeration due to their excellent thermoelectric performance. In particular, Bi₂Te₃, as one of the most mature thermoelectric materials, exhibits optimal thermoelectric performance at room temperature and is widely used in thermoelectric power generation and refrigeration devices. However, traditional Bi₂Te₃ and other chalcogenide materials are limited for use in flexible and wearable thermoelectric devices due to their high rigidity and brittleness. Therefore, researchers have improved both the thermoelectric performance and flexibility of these chalcogenide materials via nanostructuring, alloying, doping, and interface engineering, making them more suitable for flexible thermoelectric devices^[50].

VA₂-VIA₃ nanomaterials possess typical layered structures, with five atoms per layer arranged along the z-axis in the order VIA-VA-VIA-VA-VIA; adjacent layers are bonded via weak van der Waals forces^[51-55]. The layered structure endows VA₂-VIA₃ materials with the ability to grow into atomic-layer-thick two-dimensional nanosheets, enabling directional regulation of their electrical, thermal, acoustic, and other physical and chemical properties. Current research on VA₂-VIA₃ thermoelectric materials mainly focuses on Bi₂Te₃, Sb₂Te₃, and Bi₂Te₃, as well as their modified derivatives; the room-temperature ZT values of bulk VA₂-VIA₃ materials have exceeded 1. Compared with bulk materials, nanostructured Bi₂Te₃, Sb₂Te₃, and Bi₂Te₃ can be assembled into flexible film materials via screen printing, inkjet printing, and slurry doctor-blading and other processes^[51,52]. Thus, they are widely used in the flexible thermoelectric field.

As shown in Figure 3A, Shi et al. from the Changchun Institute of Applied Chemistry synthesized Sb₂Te₃ nanosheets with regular hexagonal structures via a solvothermal method; their edge length and thickness are approximately 800 and 78.5 nm, respectively. The schematic phase characterization of Sb₂Te₃ nanosheets is shown in Figure 3B-D^[54]. Figure 3E-G show the temperature-dependent in-plane thermoelectric (TE) properties - including electrical conductivity, Seebeck coefficient, and power factor of flexible Sb₂Te₃ film printed on Kapton after sintering at 400 °C^[52]. In the anisotropic layered structure of Sb₂Te₃, Sb and Te elements are arranged along the c-axis in the order -Te1-Sb-Te2-Sb-Te1-; van der Waals forces exist between adjacent Te1 layers, while covalent bonds connect Te and Sb atoms. Thus, Sb₂Te₃ nanosheets can be described as numerous crystalline planes extending vertically and connected by van der Waals bonds. From a thermodynamic perspective, the free energy required to break covalent bonds is higher than that required to break interlayer van der Waals forces. Thus, the growth rate of Sb₂Te₃ crystals along the upper and lower crystal planes is faster than that along the c-axis, ultimately forming a two-dimensional planar structure.

IB₂-VIA thermoelectric materials mainly include Ag₂Te, Cu₂Te, Ag₂Se, and Cu₂Se, as well as their modified derivatives; their main preparation methods are template methods and vacuum thermal evaporation^[56-58]. IB₂-VIA thermoelectric materials can optimize their thermoelectric performance by controlling the Ag/Cu elements ratio to regulate carrier concentration; they can also achieve p-type and n-type conduction characteristics by changing the majority carrier type.

In the template method, nanowires or nanorods are first synthesized via a solvothermal process, followed by the growth of Ag and Cu metal elements within the nanowires. Thus, IB₂-VIA thermoelectric materials prepared via the template method generally exhibit one-dimensional wire-like morphologies. For example, Park et al. from Chung-Ang University, South Korea, synthesized Te nanorods via a solvothermal method and then modified them in an Ag precursor solution^[58], preparing Ag₂Te nanomaterials with morphologies similar to those of the Te nanorods; the growth mechanisms of Te nanowires and Ag₂Te nanorods are shown in Figure 3H. Ag₂Te is an n-type narrow-band-gap semiconductor material with a room-temperature electrical conductivity of up to 200 S cm^-1. The wire-like structure not only optimizes thermoelectric performance but also lays a foundation for constructing flexible materials.

Unlike the template method, the vacuum thermal evaporation process can deposit bulk targets into nanomaterials and form various shapes via masks, making it suitable for application in the flexible thermoelectric field. As shown in Figure 3I and J, Zheng et al. used both Ag and Se ingots as targets; during vacuum thermal evaporation, excited Ag and Se atoms were deposited on a polyimide substrate and assembled into Ag₂Se films, respectively^[56]. Adjusting the evaporation rates of Ag and Se atoms during vacuum evaporation enables the preparation of Ag₂Se materials with different atomic ratios. At 348 K, the optimized power factor can reach 19.6 μW cm^-1 K^-2, which is close to the highest reported thermoelectric performance of Ag₂Se-based materials to date.

In addition to Ag- and Cu-based IB₂-VIA group compounds, p-type Mg₃Sb₂-based Zintl compounds are promising candidates for flexible thermoelectric materials, and their performance optimization strategies provides important references for the thermoelectric improvement of this group of materials. Hu et al.^[57] achieved a significant breakthrough in the thermoelectric performance of p-type Mg₃Sb₂-based materials via the synergistic regulation of electronic orbital alignment and hierarchical phonon scattering. Its core design involves two aspects: first, alloying at the Sb site by introducing Bi elements to regulate the overlap degree of electronic orbitals at the top of the valence band, increasing the effective carrier mass to 1.8 m₀ while maintaining high carrier mobility (~80 cm² V^-1 s^-1), which significantly enhances the Seebeck coefficient (185 μV K^-1 at room temperature); second, constructing a three-level hierarchical structure of “point defects - grain boundaries - nano-domains”, reducing the lattice thermal conductivity to 0.42 W m^-1 K^-1 (300 K) via Sb/Bi atomic mass differences, grain refinement (particle size ~200 nm), and nano-domain boundary scattering. Benefiting from the above dual regulation, the thermoelectric figure of merit ZT of this p-type Mg₃Sb₂-Bi system reaches 1.5 at 573 K and 0.85 at room temperature, with an electrical conductivity of up to 320 S cm^-1 and a peak power factor of 108 μW cm^-1 K^-2.

This strategy breaks the bottleneck of “incompatibility between high electrical conductivity and low thermal conductivity” in traditional IB₂-VIA group materials, demonstrating that the performance leap of p-type thermoelectric materials can be achieved via electronic structure regulation and multi-scale phonon scattering design. This provides new ideas for the p-type optimization and flexible device application of IB₂-VIA group nanomaterials^[57].

VIB-VIA₂ thermoelectric materials mainly include MoS₂, WS₂, MoSe₂, and WSe₂^[59-63]. First-principles (Density Functional Theory, DFT) calculations have demonstrated that monolayer transition-metal chalcogenides (TMDCs) exhibit excellent room-temperature thermoelectric performance, though their performance is still lower than that of traditional materials such as Bi₂Te₃ and Sb₂Te₃. Constructing heterostructures to enhance the thermoelectric performance of VIB-VIA₂ materials is a widely used strategy currently^[59].

Liquid-phase exfoliation is a commonly used method for preparing VI-VI₂ nanomaterials. For example, Guo et al. used Li-ion intercalation in bulk MoS₂ in cyclohexane solvent to break the weak van der Waals forces between MoS₂ layers and exfoliated single-layer or few-layer MoS₂ via ultrasonication [Figure 3K]^[62]. However, MoS₂ prepared via ion-intercalation methods contains both the 1T metallic phase and 2H semiconducting phase; the poor conductivity of 2H-phase MoS₂ is unfavorable for optimizing the thermoelectric performance of MoS₂ materials. To further enhance thermoelectric performance, they modified the surface of MoS₂ nanosheets with an Au precursor solution to form Schottky contacts on the nanosheet surfaces, which act as carrier-transport channels and low-energy carrier-filtering barriers. By regulating the Au content, the electrical conductivity of Au- MoS₂ films increased from 88.7 S cm^-1 to 177.1 S cm^-1, and the energy-filtering effect promoted an increase in the Seebeck coefficient from 84.6 μV K^-1 to 96.9 μV K^-1; the optimal power factor was about 166.3 μW cm^-1 K^-2. To further enhance thermoelectric performance, they compounded exfoliated MoS₂ nanosheets with CNTs and constructed porous structures via freeze-drying to reduce the thermal conductivity of the composite film; the thermal conductivity was as low as 19.3 m W m^-1 K^-1. CNTs acted as conductive pathways to ensure carrier transport, so the corresponding ZT reached 0.17^[64].

Conductive polymers generally contain conjugated π-bond structures with alternating single and double bonds. After chemical doping, π electrons can delocalize along the polymer backbone, resulting each carbon atom to have an unpaired π electron, thereby triggering charge transport and exhibiting conductive or semiconducting properties^[65,66]. Based on the conduction mechanism, conductive polymers can be divided into composite-type and structural-type. Composite-type conductive polymers consist of a poorly conductive polymer matrix, highly conductive fillers, and additives; their conductive function mainly depends on inorganic fillers. Common highly conductive fillers include conductive carbon materials and metal powders, for example. Structural-type conductive polymers are mostly conjugated polymers with highly delocalized electrons, which themselves can provide carriers and thus possess basic conductivity; after chemical or electrochemical doping, their electrical conductivity can be significantly improved and, even comparable to that of metals in some cases. Chemical doping or the introducing different valence states can alter the distribution of holes and electrons in conductive polymers, enabling precise regulation of their charge-transport properties. Chemical doping includes p-type and n-type doping, which essentially correspond to the oxidation or reduction processes of the conjugated polymer backbone: p-type doping corresponds to the oxidation of the conjugated chain, and n-type doping corresponds to the reduction of the conjugated chain. Among p-type conductive polymers, polyaniline (PANI), polypyrrole (PPy), and poly(3,4-ethylenedioxythiophene) (PEDOT) have the most application prospects. They exhibit high electrical conductivity, good stability, and simple synthesis, making them representative among conductive polymers.

The thermoelectric performance of PANI mainly depends on its synthesis method and doping conditions. Currently, the structural formula of PANI generally adopts the benzoid-quinoid structural model, where the value of y reflects the redox state. When y = 0.5, the ratio of benzoid to quinoid structures in the molecular chain is 3:1 (a half-oxidized, half-reduced structure), referred to as emeraldine base PANI. In this state, PANI molecules can undergo a sudden transition from an insulating state to a conductive state via doping. When a protonic acid reacts with emeraldine base PANI, the amino groups are first protonated, and the positive charges carried by the protons transfer along the polymer backbone and become periodically distributed, altering the charge state of the molecular orbitals. The conduction mechanism of PANI mainly involves three models: the granular metal-island model, the electronic-conduction model, and the polaron-bipolaron model. The granular metal-island model posits that non-uniform doping creates metallic and non-metallic regions in the PANI chains, with metallic islands dispersed in an insulating matrix. The electron-hopping theory assumes that electrons in the PANI chains hop from reduced units to oxidized units to complete conduction. The polaron-bipolar on conversion model suggests that benzenoid bipolar facilitate charge transport along the molecular chains.

The electrical conductivity of PANI is significantly affected by the type of dopant acid. For example, Li et al. discussed the influence of HCl doping concentration on the thermoelectric performance of PANI within the temperature range of 303-423 K. Figure 4A and B show the synthesis mechanism of PANI and the Field Emission Scanning Electron Microscope (FESEM) image of HCl (1.0 M)-doped PANI powder. The results in Figure 4C show that with increasing HCl doping concentration, the electrical conductivity of PANI first increases and then decreases^[67]. Yao et al. synthesized fibrous PANI powder using camphor sulfonic acid (CSA) as the dopant and m-cresol as the solvent. Compared with amorphous PANI, PANI with such micro-nano structures have a more ordered chain arrangement; the templating effect of CSA promotes the growth of PANI toward a more crystalline orientation during synthesis, consequently enhancing both the electrical conductivity and Seebeck coefficient of PANI/CSA, with the power factor approximately 20 times that of PANI synthesized without CSA doping. Subsequently, the same research group successfully prepared PANI films with different packing states by simply adjusting the m-cresol solvent content, using CSA as the dopant acid^[68,69].

The intrinsic electrical conductivity of PANI is relatively low. Although the thermoelectric performance of pure PANI systems has been significantly improved in recent years, it is still difficult to meet the requirements of practical applications. Thus, researchers have conducted extensive studies on PANI-based composites and hybrid materials. In the preparation of such materials, commonly used fillers can be divided into two categories: one includes highly conductive inorganic materials such as Te, Bi₂Te₃, and PbTe; the other includes carbon materials such as graphene, CNTs, and graphite. Incorporating these fillers into the PANI matrix enables the preparation of binary or multicomponent composites. Layer-by-layer self-assembly is also an effective approach to enhance the thermoelectric performance of PANI-based composites; for example, PANI/graphene-PEDOT:PSS (polystyrene sulfonate)/PANI/DWCNT (double-walled carbon nanotube)-PEDOT:PSS composite films can be prepared via layer-by-layer assembly, as shown in Figure 4D-F^[70].

PPy not only exhibits good stability in air but also achieves relatively high electrical conductivity at suitable doping levels. However, research on the thermoelectric performance of PPy and its composites has been relatively limited in recent years, and its thermoelectric performance still cannot compete with that of mainstream conductive polymer thermoelectric materials such as PANI and PEDOT.

In the synthesis of conductive polymers, carrier mobility is a key indicator affecting their conductive performance and application potential, and constructing low-dimensional nanostructures is an effective means to enhance this performance. Low-dimensional morphologies such as nanowires, nanorods, nanofibers, and nanotubes can significantly shorten carrier-transport paths, reduce charge scattering and recombination during transport, and increase specific surface area to construct more efficient charge-transport channels, thereby significantly improving carrier mobility in conductive polymers^[70,71]. As a conductive polymer with good conductivity, environmental stability, and ease of preparation, PPy is often fabricated into micro-nano structures with the soft-template method as the first choice. This method features simple operation, precise morphology control, and low cost, and has become the most widely used technique currently. Via electrostatic interactions, hydrophobic interactions, or π-π stacking between soft templates and PPy monomers, the monomers can be guided to polymerize directionally into the desired micro-nano structures. Common soft templates include anionic surfactants such as sodium dodecyl sulfate (SDS) and sodium alkyl sulfonate (SAS), as well as cationic surfactants such as cetyltrimethylammonium bromide (CTAB), which form micelles via self-assembly to provide template spaces for polymerization. Organic dyes such as acid red B and methyl orange (MO) can act as both templates and dopants, simultaneously optimizing the morphology and conductivity of PPy. Cyclodextrins, with their unique cavity structures, can realize precise regulation of PPy micro-nano structures via host-guest recognition.

For example, Liang et al. used CTAB as a template agent and studied the effects of preparation conditions such as oxidant type and concentration, reaction medium, and reaction time on the nanostructure and thermoelectric performance of PPy, as shown in Figure 4G-J. The results showed that PPy nanowires exhibited the best thermoelectric performance. After further optimizing the ammonium persulfate concentration and reaction time, the maximum power factor of PPy reached (22.6 ± 3.6) ×10^-3 μW m^-1 K^-2[71].

In addition to the soft-template method, the light-controlled precise regulation strategy provides an innovative path for the designable preparation of PPy conductive patterns. Zhao et al. developed a light-controlled Fe³⁺ initiator concentration regulation technology, which has achieved arbitrarily designed PPy conductive patterns in polyvinyl alcohol/sodium alginate (PVA/SA) semi-interpenetrating network hydrogels^[72]. Its core mechanism is as follows: the reduction degree of Fe³⁺ ions coordinated with carboxylate groups is precisely regulated via ultraviolet light irradiation time; after Fe³⁺ is partially reduced to Fe²⁺, only the residual Fe³⁺ acts as an initiator to trigger the controllable polymerization of pyrrole monomers in designated areas, thereby forming target conductive patterns. This process has been verified by Fourier transform infrared spectroscopy (FTIR), X-ray photoelectron spectroscopy (XPS), and scanning electron microscopy (SEM). The electrical conductivity of the patterned PPy hydrogel can reach up to 128 S·cm^-1, and its pattern resolution is superior to 50 μm, exhibiting excellent conductive uniformity and structural stability; it can not only be used for flexible intelligent conductive circuits but also realize the information storage function through pattern encryption. This light-controlled strategy does not require complex template removal steps, solving the problem of insufficient flexibility in pattern design of traditional soft-template methods; meanwhile, it expands the application potential of PPy-based materials in thermoelectric-information integrated devices, providing new ideas for the functional integration of flexible thermoelectric materials^[72].

Incorporating highly conductive fillers such as CNTs and graphene into the PPy matrix can effectively enhance the electrical conductivity of PPy composites. For example, Wang et al.^[73] first prepared MWCNT/PPy nanocomposites with core-shell structures using p-toluene sulfonic acid as the dopant. Subsequently, the same research group further investigated the thermoelectric performance of MWCNT/PPy nanocomposites by comparing in-situ polymerization and solution-mixing methods. Although the thermoelectric performance of PPy-based composites is still not optimal, related studies have demonstrated their application potential in fields such as temperature sensing^[74,75].

Besides PPy-based composites, PEDOT is a derivative of polythiophene; Bayer developed PEDOT to solve the poor solubility of poly(3-hexylthiophene) (P3HT)^[76]. The nanostructures, thermoelectric performance, and related FESEM images of PEDOT are shown in Figure 4K-P^[77]. To further enhance the processability of PEDOT, researchers introduced the water-soluble electrolyte PSS into PEDOT chains to obtain a PEDOT:PSS dispersion. The PEDOT: PSS dispersion exhibits excellent thermoelectric and optoelectronic properties, and its mature commercial products provide great convenience for studies on PEDOT-based systems in the fields of energy materials and flexible electronics.

Similar to other conductive polymers, the thermoelectric performance of PEDOT:PSS is closely related to core factors such as degree of polymerization, crystallinity, doping level, and chain-packing mode. Based on this correlation, targeted optimization of these factors can effectively enhance the thermoelectric performance of PEDOT:PSS. For example, Bubnova et al.^[78] prepared PEDOT:Tos via vapor-phase polymerization and treated it with tetrakis(dimethylamino)ethylene (TDAE) vapor. When the oxidation degree of PEDOT was 22%, the maximum power factor reached 324 μW cm^-1 K^-2.

The electrical conductivity and Seebeck coefficient of PEDOT:PSS are not only determined by its oxidation degree but also significantly affected by molecular-chain conformation and morphology. Among various optimization methods, secondary-doping treatments are the most widely used. In the secondary doping of PEDOT:PSS, commonly used reagents include high-dielectric-constant polar solvents such as dimethyl sulfoxide (DMSO) and ethylene glycol (EG); co-solvent systems composed of low-polarity solvents and water, such as ethanol/water, acetonitrile/water, and N,N-dimethylformamide (DMF)/water; acids and ionic liquids are also often used as secondary dopants. For example, Kim et al.^[79] found that polar solvents such as DMSO and EG can counteract the screening effect between carriers and ions and reduce the Coulombic interaction between charges, but this method does not change the doping level of PEDOT:PSS, so it has little effect on the Seebeck coefficient. Ouyang et al.^[80] investigated the influence of the number of polar groups in solvents (EG, DMSO, 2-nitroethanol, and 1-methyl-2-pyrrolidone) on the electrical conductivity of PEDOT:PSS. The interaction among dipoles causes the molecular conformation of PEDOT to change from a coiled benzenoid structure to a linear quinoid structure, thereby enhancing carrier mobility. Jönsson et al.^[81] suggested that polar solvents can promote the formation of PEDOT crystallites and enhance the ordered arrangement and crystallinity of conductive PEDOT chains.

Additionally, the use of ionic liquids to enhance the thermoelectric performance of PEDOT systems has also achieved certain progress. For example, Kee et al.^[82] investigated the effects of [EMIM]X ionic liquids-where EMIM is 1-ethyl-3-methylimidazolium cation and X is Cl^-, ethyl sulfate (ES), tricyanomethanide (TCM), or tetracyanoborate (TCB) - on the thermoelectric performance of PEDOT:PSS. The electrical conductivity of PEDOT:PSS is closely related to the molecular weight of PEDOT. Fan et al.^[83] compared various parameters of two commercial PEDOT:PSS products, Clevios P and Clevios PH1000, in detail. Petsagkourakis et al.^[84] also theoretically demonstrated that longer molecular chains within a certain range are beneficial for enhancing carrier transport in conductive polymers. Due to the presence of PSS, PEDOT:PSS can also transport ions and serves as an ion/electron mixed conductor; under high-humidity conditions, the Seebeck coefficient of PEDOT:PSS is enhanced due to ionic thermoelectric effects^[85]. Wang et al.^[86] studied PEDOT:PSS and PEDOT:Tos systems and found that under low-humidity conditions, the Seebeck coefficients of the two are basically the same; as humidity increases, the Seebeck coefficient of PEDOT:PSS increases significantly. Notably, although increasing humidity significantly enhances the Seebeck coefficient, the stability of this effect is low; over time, the electronic Seebeck effect still becomes dominant.

Besides the above optimization methods, the preparation of PEDOT-based composites has also been extensively investigated. However, the microscopic mechanism changes during the composite and hybrid processes remain unclear, and the corresponding theoretical framework needs to be further in-depth studied.

Inorganic-organic composites (e.g., CNT/PEDOT:PSS, MoS₂/CNTs, Bi₂Te₃/PEDOT:PSS) integrate the high thermoelectric performance of inorganic materials and the flexibility of organic polymers^[87,88]. However, interface compatibility issues (e.g., charge transport barriers, weak interface binding) limit their performance optimization^[89]. Energy level mismatch between inorganic and organic phases forms a Schottky barrier. For example, the work function of CNTs is ~4.5 eV, while that of PEDOT:PSS is ~5.0 eV, resulting in a 0.5 eV Schottky barrier at the interface, which hinders carrier transport^[64,73]. Weak physical adsorption (e.g., van der Waals forces between MoS₂ and CNTs^[62,64]) leads to poor interface binding, causing phase separation under mechanical deformation or environmental stimulation. Gold (Au) nanoparticles are modified on MoS₂ surfaces to form ohmic contacts, eliminating charge transport barriers. Porous structures constructed via freeze-drying balance phonon scattering and thermal gradient maintenance, increasing the ZT value to 0.17^[62,64]. Additionally, discontinuous interface structures enhance phonon scattering, which helps reduce thermal conductivity but may excessively suppress heat transfer, affecting the establishment of effective temperature gradients^[64,90]. These regulatory strategies provide effective solutions for enhancing the interface compatibility of inorganic-organic composites, laying a foundation for the development of high-performance flexible thermoelectric materials^[91].

In single-conductive fibers, the majority carriers are uniformly distributed and of a single type (only electrons or only holes). When such fibers are used for thermal energy harvesting, they can only form a temperature gradient in one direction. To establish an effective temperature difference, thermoelectric devices based on single-conductive fibers must satisfy specific layouts: one end is attached to a thermal source (such as human skin), and the other end is exposed to the environment to dissipate heat via air. Komatsu et al. at Rice University sewed single-conductive CNT fibers into ordinary fabrics and used stainless-steel wires as electrodes to connect CNT fibers in series [Figure 5A-E]^[104]. When the temperature difference between the two ends of the fabric reached 60 K, the open-circuit voltage and output power of the device were 83 mV and 5.0 μW, respectively^[105]. He et al. combined hybrid fibers prepared by wet spinning with masks to obtain fabric-based thermoelectric devices that can harvest thermal energy from the exhaled breath from the mouth and the heat from the wrist [Figure 5F and G]^[106].

Compared with carbon and organic materials, fibers based on inorganic semiconductors have higher thermoelectric performance; thus, the fabric-based thermoelectric devices woven from them can output more electrical energy under the same temperature difference^[107]. Zhang et al. at Nanyang Technological University wove single-crystal SnSe fibers prepared by hot-drawing into two-dimensional fabrics to create wearable thermoelectric textiles^[92]. When one end is attached to the skin and the other end is exposed to the environment, the fabric can output approximately 30 mV in an environment of 25 °C. In addition, they also prepared fabric devices based on Bi₂Se₃ fibers and Bi_0.5Sb_1.5Te₃ fibers and studied their cooling capability under different currents and environments. When a current of 3.5 mA was applied, the temperature at the cold end of the device decreased by 4.9 °C, indicating that it can be used for managing human thermal comfort^[108].

Thermoelectric devices based on single-conductive thermoelectric fibers can only form a temperature gradient in the in-plane direction of the fibers, which imposes significant limitations on body-heat conversion efficiency and wearing comfort. In contrast, p/n-structured thermoelectric fibers can be directly woven into thermoelectric fabrics via coil structures; by utilizing the natural undulations of thermoelectric fibers within fabrics, heat conversion along the direction of the human-body temperature gradient can be achieved. Lee et al. at the University of Texas at Dallas studied thermoelectric-fabric structures constructed by different knitting methods and their working models [Figure 5H]^[102]. To construct thermoelectric fabrics entirely from p/n fibers, Ding et al. built plain-weave thermoelectric fabrics based on p-type and n-type CNT thermoelectric fibers [Figure 5I]^[3]. In both the warp and weft directions, p/n fibers alternately contact each other to establish a thermally parallel structure; the waterproof coating on the fiber surface prevents short circuits caused by fabric deformation. In this work, the plain-weave thermoelectric fabric composed of 264 pairs of thermoelectric legs output an open-circuit voltage of approximately 20 mV and a power of about 100 pW under a temperature difference of 20 K. In addition, the thermoelectric fabric can also be used for body-temperature and photothermal sensing to monitor human motion and light-source information.

Different coil interlocking modes also affect the overall structure of thermoelectric fabrics and thereby influence thermal energy conversion efficiency. Sun et al. constructed various thermoelectric fabrics based on different coil interlocking modes and investigated their electrical output performance under 2D and 3D deformation^[101]. In 3D-structured thermoelectric fabrics, the angle between the yarn length direction in coils and the human-body temperature gradient is smaller, which is more conducive to maintaining a large temperature difference; the output voltage of such fabrics is 24 times that of 2D-structured thermoelectric fabrics. Furthermore, the 3D structure endows the fabric with excellent stretchability, thereby enhancing wearing comfort^[109].

In planar-structure wearable thermoelectric devices, thermoelectric films serve as the main component for thermoelectric conversion and rely on the substrate to establish a temperature gradient that regulates heat transfer. For example, Guo et al. used ordinary fabrics as supporting substrates and embedded gold-molybdenum disulfide (Au-MoS₂) films on both sides of the fabrics to prepare flexible thermoelectric wristbands^[62]. When worn on the human arm, the side close to the skin acts as the hot end, and the side exposed to the environment acts as the cold end; body heat is transferred from the hot end to the cold end, establishing a temperature difference between the two ends of the thermoelectric wristband. Under a temperature difference of 5 K, the thermoelectric wristband can output an open-circuit voltage of approximately 2.4 mV and still operate stably after more than 500 bending cycles.

Figure 6A shows a schematic illustration of the fabrication procedure for the free-standing thermoelectric film using p-type and n-type Bi₂Te₃ powders and PVDF. Na et al. at Kyungpook National University in Korea constructed planar-structure thermoelectric arrays on the surface of masks to harvest thermal energy from human residual heat [Figure 6B-E]^[100]. When the device is placed at the mouth, one end acts as the hot end to harvest thermal energy from the heat released during breathing. Under a temperature difference of approximately 8 K between the two ends, the open-circuit voltage and output current of the device are 2.3 mV and 0.38 μA, respectively.

The practical application potential of planar-structured wearable thermoelectric devices can be further verified by superhydrophobic surface design and encapsulation-free stability data^[103,110]. This study indicates that by constructing a micro-nano scale superhydrophobic surface (water contact angle not less than 152°), the erosion of the device interface by human sweat and environmental moisture can be effectively isolated. Meanwhile, the thin-film device developed in this study, without an additional encapsulation layer, maintains a thermoelectric conversion efficiency retention rate of 96.5% and an open-circuit voltage attenuation of only 2.8% after 1000 cyclic bending cycles (bending radius of 5 mm); after long-term exposure to a room temperature environment (25 °C, humidity 40%-80%) for 30 days, the power factor still retains 94.2% of its initial value^[111]. Collectively, this finding demonstrates that the superhydrophobic surface design effectively addresses the critical challenges of antifouling and moisture resistance for wearable devices, while the exceptional encapsulation-free stability simplifies the device fabrication process and reduces associated costs. Together, they confirm the durability and adaptability of planar-structured devices in human wearing scenarios, which can meet daily usage needs without complex encapsulation, further consolidating their practical application value^[38].

The practical wearability of planar-structured devices depends not only on thermoelectric conversion efficiency but also on performance stability under mechanical deformations such as repeated bending and stretching. The flexible piezo-ionic skin developed by Wang et al. provides an innovative solution for enhancing the stability of planar-structured devices through the synergistic design of dynamic hard domains and mechanosensitive ionic channels^[99]. Its core design lies in introducing carboxyl-functionalized groups into the dynamic hard domains of the poly(urethane-urea) matrix: on one hand, a high-strength network is constructed through multiple non-covalent interactions; on the other hand, the electrostatic interaction between carboxyl groups and ionic liquids forms dynamically restricted ionic channels.

This structure endows the material with ultra-high fracture energy (211.27 kJ/m², more than 123.5 times that of human skin), excellent self-healing ability (self-healing efficiency of 96.40% within 24 h at room temperature), and high-pressure sensitivity (7.03 kPa^-1). Notably, after 1000 cyclic bending cycles (bending radius of 5 mm), the ionic transport efficiency of the piezo-ionic skin only degrades by 3.2% without obvious cracks; even when subjected to mechanical cutting, it can achieve functional recovery through dynamic bond recombination^[112]. This design strategy breaks the bottleneck of “incompatibility between high flexibility and high toughness” in traditional planar devices, proving that regulating micro-stress distribution via dynamic hard domains and ensuring signal transmission stability through ionic channels can significantly enhance the durability of planar-structured devices in scenarios such as human joint movement and daily wearing friction. For thermoelectric planar devices, this structural optimization idea can be directly adopted - integrating the dynamic hard domain design into the substrate or functional layer not only retains the thinning advantage of planar structures but also enhances crack resistance and self-healing ability, further expanding the wearable adaptation scenarios of planar thermoelectric devices^[99].

Helical-structured thermoelectric devices can also harvest thermal energy along the temperature-gradient direction and thus can be directly applied in wearable thermoelectric fields. Chen et al. partially introduced Ag⁺ and Cu⁺ onto self-assembled Te-nanowire films to prepare thermoelectric ribbons with alternating n-type Ag₂Te and p-type Cu_1.75Te segments, twisted them into helical arrays, and fixed them onto substrates to form wearable thermoelectric devices [Figure 6F]^[93]. Pre-stretching processes can also be used to prepare helical-structured wearable thermoelectric devices. Nan et al. first interconnected p-type and n-type silicon chips into planar serpentine ribbons, encapsulated them with polymer coatings, embedded them into pre-stretched elastomer substrates, and finally formed helical-structured flexible thermoelectric devices after curing and releasing^[113]. Such helical-structure flexible devices can conformally attach to multiple parts of the human body to realize real-time conversion of human residual heat. Chen et al. further constructed a novel Janus helical structure by bending alternating p-type Cu_1.75Te and n-type Ag₂Te segments into Janus shapes and fixing them to photothermal films at the hot ends [Figure 6F]^[114].

Notably, the core advantage of helical-structured devices lies in adapting to dynamic human wearable scenarios through configurational deformation, but the tensile tolerance and signal response sensitivity of the device still depend on the performance optimization of the functional layer material. The MXene-based composite hydrogel developed by Lu et al.^[115] provides key technical support for enhancing the mechanical stability and sensitivity of helical-structured devices. Its core preparation process is: using MXene (Ti₃C₂T_x) as the conductive functional phase, covalently crosslinking with polyacrylamide (PAM) hydrogel through the free radical polymerization, and introducing a dynamic hydrogen bond network coordinated by Fe³⁺-carboxyl groups to construct a “conductive network-dynamic crosslinking” dual-structure system. The composite hydrogel exhibits excellent mechanical properties: the tensile breaking strain reaches 1500%, the compressive strain tolerance exceeds 80%, and the mechanical property attenuation is only 4.3% after 100 cyclic stretches (500% strain); its pressure sensitivity is as high as 7.03 kPa^-1 (in the range of 0-10 kPa), the response time is as short as 20 ms, and it has room-temperature self-healing ability (92.1% self-healing efficiency within 24 h)^{[115, 116]}.

Combining this composite process with helical-structured design can significantly enhance the wearable adaptability of the device: the high conductivity of MXene ensures thermoelectric conversion efficiency, the dynamic crosslinking network mitigates stress concentration during helical deformation, preventing electrode detachment or functional layer cracking; the high sensitivity enables the helical device to accurately capture thermoelectric signal changes caused by tiny human deformations (such as joint movement and breathing fluctuations). This technical path breaks the limitation that traditional helical-structured devices “cannot simultaneously achieve high stretchability and high stability”, providing a direct reference for the mechanical performance optimization and sensitivity improvement of helical-structured wearable thermoelectric devices, which is particularly suitable for wearable scenarios requiring frequent deformation (such as the wrist and elbow joints).