Advances in carbon‐based thermoelectric materials for high‐performance, flexible thermoelectric devices

Abstract

Waste energy harvesting can contribute to the increase of the efficiency of many industrial processes, which consume energy to produce valuable products. Among all the wasted energy, heat energy is the most abundant, existing in almost any situation. Thermoelectric devices have the capability to harvest and convert the thermal energy into electrical power via the Seebeck effect. With its simple operating principle, thermoelectric devices can be reliable even under the harshest environments, taking advantage of any type of heat source. As a result, various inorganic and organic materials are being explored as thermoelectric materials. Among the reported materials, carbon‐based materials are promising in terms of commericialization, due to their nontoxic and abundant nature, and solution processability. In particular, poly(3,4‐ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS), carbon nanotubes, and graphene are extensively studied as thermoelectric materials owing to their remarkable thermoelectric performance. Also, organic–inorganic hybrid halide perovskites show the potential to be used as future high‐performance thermoelectric materials. Here, the progess in carbon materials as thermoelectrics is reviewed in detail, focusing on four base materials (PEDOT:PSS, carbon nanotubes, graphene, and organic–inorganic hybrid halide perovskites). This review illuminates the potential of carbon‐based materials in the field of thermoelectrics and their application to next‐generation energy devices.

INTRODUCTION General background

Sustainable development is one of the greatest assignments of the 21st century, and the global community has set 17 goals to present a more sustainable future for next generations. Among the 17 goals, the seventh goal is to ensure access to affordable, reliable, sustainable, and modern energy. To offer clean and sustainable energy, the renewable energy market has grown substantially but still faces many challenges. The most important criterion for a commercially usable energy generation system is its efficiency. Unfortunately, in the entire energy cycle, from generation to consumption, it is estimated that nearly a third of the generated energy is wasted primarily in the form of heat, as shown in Figure 1A.¹

Figure 1. (A) 2019 Energy flow chart provided by Lawrence Livermore National Laboratory, representing ~67.5% waste heat.1 (B) Operating principle of thermoelectric devices. A cross-sectional illustration of a thermoelectric device with a single p, n couple, in which carriers drift from the hot side to the cold side, depending on the carrier type, and a typical thermoelectric device with multiple p, n couples connected electrically in series and thermally in parallel. Electrical power is generated from the device. (C) Reported peak dimensionless figure-of-merit (zT) values of thermoelectric materials. Materials were grouped by color, describing the temperature range in which performance is greatest. Blue, brown, green represent low, medium, high temperatures, respectively. Reproduced with permission: Copyright 2019, Elsevier.2 (D) Flexible thin-film thermoelectric device used to harvest human body heat at an air temperature of 15°C. Reproduced with permission: Copyright 2014, The Royal Society of Chemistry.3 (E) Interdependence of thermoelectric parameters with carrier density. This represents the greatest challenge in designing high-performance thermoelectric materials. Reproduced with permission: Copyright 2018, Wiley-VCH4

Thermoelectric (TE) devices have the capability to harvest the waste heat from industrial processes based on a phenomenon called the “Seebeck effect.” The Seebeck effect is a phenomenon in which a thermovoltage is formed when a temperature gradient is induced in conducting materials, as expressed in Equation (1). ΔV, ΔT, and α represent the magnitude of thermovoltage, applied temperature gradient, and Seebeck coefficient, respectively. [Image Omitted. See PDF]

As shown in Figure 1B, when two materials with different carrier types (p-type, n-type) are coupled in an electrically serial and thermally parallel configuration, the carriers drift from the hot side to the cold side, producing a significant amount of power output when matched with an appropriate load. The efficiency of the single TE couple can be derived from the following equations: [Image Omitted. See PDF] [Image Omitted. See PDF] [Image Omitted. See PDF]

The power output (P) of the single TE couple can be derived as the following equation: [Image Omitted. See PDF]where I (A), R_L (Ω), R_I (Ω), and α_pn (V·K⁻¹) represent the electrical current, load resistance, internal resistance of the single TE couple, and differential Seebeck coefficient, respectively. α_pn can be calculated from Equation (6) using the absolute Seebeck coefficients of each thermoelement (α_p and α_n): [Image Omitted. See PDF]

Maximum power output can be achieved when the load resistance is matched to the internal resistance, giving the relation between resistance and maximum power output from the following equation: [Image Omitted. See PDF]

zT is a dimensionless number that decides the maximum efficiency of a TE device operating at a certain temperature gradient (T_H and T_C), called the dimensionless TE device figure of merit. By assuming that both p-type and n-type materials of a TE device, or thermoelements, exhibit the same TE performance, a TE material figure of merit, zT, can be used to evaluate the applicability of a candidate material as a TE material.⁵ As shown in Equation (8), TE materials can be evaluated using the Seebeck coefficient (α), electrical conductivity (σ), thermal conductivity (κ), and absolute temperature (T). The values of α, σ, and κ are generally reported in units of µV·K⁻¹, S·cm⁻¹, and W·m⁻¹·K⁻¹, respectively. [Image Omitted. See PDF]

Due to the unique operating principle, TE devices do not require any moving parts, are simple to configure, and can be used in situations where other forms of energy generation are inapplicable. As a result, TE devices have been applied in deep space probes, using radioisotopes as heat sources.

Materials for carbon-based TE devices

So far in the TE society, inorganic semiconducting materials such as bismuth telluride and lead telluride are most extensively studied due to their degenerate nature, leading to their remarkable TE performance even in their pristine state, as shown in Figure 1C.² However, in real-world applications, heat sources for temperature gradients in TE devices are rarely planar, suggesting that material flexibility is a key factor in designing TE materials. Although inorganic flexible TE devices have been realized by fabricating the thermoelements in forms of thin films, thin-film-based TE devices are inadequate for industrial energy harvesting as large temperature gradients cannot be formed in the cross-plane direction (Figure 1D).³ Moreover, most elements comprising degenerate inorganic semiconductors (bismuth, antimony, tellurium, selenium, etc.) are rare-earth metals that are expensive and toxic, which are highly disadvantageous factors in producing large-area TE devices.

Several years ago, poly(3,4-ethylenedioxythiophene) (PEDOT) bound with tosylate anions reached a zT value of 0.25 at room temperature, comparable to inorganic semiconductors. Since then, conjugated polymers, such as polyaniline (PANI), poly(3-hexylthiophene), and PEDOT, have been studied as potential TE materials.⁶ Especially, poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS) has shown remarkable TE performance with the high electrical conductivity of ~1000 S·cm⁻¹ while maintaining a relatively low thermal conductivity of below 0.5 W·m⁻¹·K⁻¹.⁷

Low-dimensional carbon nanomaterials, graphene, and carbon nanotubes (CNTs) have also been proposed as potential TE materials owing to their excellent carrier conducting properties. Great TE power factors were estimated in single-walled CNTs and graphene both experimentally and theoretically, complementing and reinforcing the concept of low-dimensional nanomaterials as being more favorable over their bulk forms for high-performance TE materials, calculated in the work of M. S. Dresselhaus.^8-11

In addition to the predicted high TE performance of conjugated polymers, CNTs, and graphene, organic–inorganic hybrid halide perovskites have been also suggested as TE materials, due to their large Seebeck coefficient from degenerate conduction and valence bands and intrinsically low thermal conductivity, from first principle calculations and experiments.¹²

In terms of commercialization of TE devices, conjugated polymers, low-dimensional carbon nanomaterials, and organic–inorganic hybrid halide perovskites have several advantages over inorganic heavy metals with respect to their solution processability and abundant nature. Solution processability enables the fabrication of large-area TE devices using various coating methods, such as bar coating, spray coating, and inkjet printing, which ultimately reduces the production cost by eliminating the need for high-temperature and pressure conditions.^12,13 Also, solution processes are more favorable in forming nanocomposite materials, which is one of the most important strategies to maximize the TE performance.¹⁴

Conjugated polymers and low-dimensional carbon nanomaterials are also flexible, making them applicable to any situations through the conformability over nonplanar and irregularly shaped heat sources.¹⁵ Although halide perovskites are not flexible in their bulk form, they are able to be applied to flexible TE devices in forms of thin films.¹⁶

Lastly, the relatively low material cost per unit mass compared with inorganic heavy metals shows that the aforementioned materials will be able to be scalable into thick, bulk TE materials to guarantee the formation of large temperature gradients over the cross-plane direction, which will be critical in waste energy harvesting applications.¹⁷

However, organic TE materials are yet to replace the inorganic counterparts owing to their low zT value. Conjugated polymers show promisingly high electrical conductivity and low thermal conductivity, but the low Seebeck coefficient limits their TE performance. High TE power factor of single-walled carbon nanotubes (SWCNTs) was estimated, but due to the significantly high thermal conductivity, the overall zT value is quite low. In case of graphene, the semimetallic nature results in high electrical conductivity, but like most metals, the Seebeck coefficient is low and the thermal conductivity is high, making it hard for graphene to be directly applied to TE devices. For halide perovskites, although high Seebeck coefficients are observed, the low carrier concentration results in low electrical conductivity.

Challenges for carbon-based TE materials

The main reason that the obvious disadvantages in each material have not been completely overcome can be attributed to the trade-off relation between the individual TE parameters comprising the zT value. As shown in Figure 1E, the Seebeck coefficient generally decreases with increasing electrical conductivity upon increased carrier concentration via carrier doping.⁴ Also, the overall thermal conductivity increases with improved electrical conductivity due to the increasing electronic portion of thermal conductivity. The overall interdependence of TE properties is the greatest challenge in the enhancement of TE performance of carbon-based TE materials. Reliable evaluation of TE properties is also a major obstacle for carbon-based TE materials. As samples of carbon-based materials are often produced in the form of thin layers, precise determination of the TE properties can be challenging. The measurement of thermal conductivity is especially difficult, requiring delicate system design and calculations. Moreover, for materials in which carrier and heat transport is anisotropic, determination of both in-plane and cross-plane TE properties is critical for complete zT evaluation.¹⁸

Over the years, various strategies to improve the zT value of carbon-based TE materials have been proposed, such as chemical doping, nanostructuring, compositing, energy filtering, and so forth. As a result, noticeable improvements for carbon-based TE materials have been achieved, as shown in Figure 2. In this review, we illustrate the various strategies implemented for the TE property enhancement of carbon-based materials, mainly focusing on PEDOT:PSS, CNTs, graphene, and organic–inorganic hybrid halide perovskites. At the end of each material section, we introduce noteworthy examples of application to flexible TE devices. Cases in which delicate zT evaluation was successfully conducted and well demonstrated are also introduced to discuss the efforts the TE society must make for TE research to prosper even greater. All in all, we hope this review will be able to accurately describe the current prospects and challenges in carbon materials as TE materials, and guide researchers in the field of TEs toward designing high-performance, carbon-based TE materials.

Figure 2. Chronological evolution of the zT value of carbon-based materials. PEDOT:PSS, poly(3,4-ethylenedioxythiophene):polystyrene sulfonate

CARBON-BASED THERMOELECTRICS (TES) PEDOT:PSS PEDOT:PSS pre- and posttreatment strategies for TE property enhancement

Various methods have been utilized to enhance the aforementioned intrinsic TE properties of PEDOT:PSS, which is schematically demonstrated in Figure 3. The TE performance of treated PEDOT:PSS is summarized in Table 1. One of the earliest pre- and posttreatment approaches on PEDOT:PSS-based TE was reported by Luo et al.¹⁹ The effect of dimethyl sulfoxide (DMSO) addition and posttreatment with DMSO, and 1-ethyl-3-methylimidazolium tetrafluoroborate (EMIMBF₄) was investigated. DMSO posttreatment resulted in elongated PEDOT grains, whereas smaller and more circular PEDOT grains were observed for EMIMBF₄ posttreatment. The structural changes caused variations in α and σ, which resulted in an optimized power factor of 38.46 µW·m⁻¹·K⁻². Briefly after, Kim et al.⁷ reported a record-high room-temperature zT value of 0.42 for DMSO mixed, expanded graphene (EG)-treated PEDOT:PSS thin films. The doping process was aimed to optimize the carrier concentration and minimize the total dopant volume at the same time by removing non-ionized PSS from the PSS-doped PEDOT grains through EG dipping. The control of total dopant volume played a crucial role in TE property enhancement due to its exponential dependence on carrier mobility. As a result of the dedoping process, a power factor of 469 μW·m⁻¹·K⁻² was achieved.

Figure 3. Schematic illustration of the different strategies resulting in enhanced thermoelectric performance of PEDOT:PSS (poly(3,4-ethylenedioxythiophene):polystyrene sulfonate)

Table 1 Reported enhanced thermoelectric performance of PEDOT:PSS using various doping methods

Method	Seebeck coefficient (μV·K−1)	Electrical conductivity (S·cm−1)	Power factor (μW·m−1·K−2)	Thermal conductivity (W·m−1·K−1)	zT	Sample (thickness)a	References
EMIMBF4, DMSO drop			38.46	0.17	0.068 (300 K)	Film (nm)	[19]
EG dip			469	0.308	0.42	Film (nm)	[7]
NaOH addition			19.6			Film (μm)	[20]
DMSO/HZ mixture coating	~44	~578	112			Film (nm)	[21]
TSA addition, DMSO/HZ mixture coating	49.3	1310	318.4	0.30	0.31	Film (nm)	[22]
H2SO4, NaOH drop	39.2	2170	334	0.2–0.34	0.29–0.49	Film (nm)	[23]
TFMS–MeOH drop	21.9	2980	143	0.224	0.19	Film (nm)	[24]
Benzenesulfonic acid addition, DMSO/HZ drop	42.6	1119	203.1			Film (nm)	[25]
ZnCl2/DMF treatment	26.1	1400	98.2		0.125	Film (N/A)	[26]
NaHCO3 aqueous solution addition			100			Film (N/A)	[27]
0.1 M MAI/DMF, H2O (80:20, volume/volume) drop	28	1831	144	0.29	0.11	Film (N/A)	[28]
PEIE/EG dip before H2SO4 dip	21.9	2770.7	133.0			Film (nm)	[29]
EMIM-DCA coating after H2SO4, NaOH drop			754	0.2~0.5	0.75	Film (nm)	[30]
PSSH(PSSNa) coating after H2SO4 drop	43.5 (48.1)	2120 (1732)	401 (401)			Film (nm)	[31]
Sodium alkylsulfonate addition (alkyl chain length, 6)			700.2	0.85	0.25	Film (N/A)	[32]

Note: Thermoelectric properties not directly mentioned in the manuscript of each reference were estimated based on the plots of each parameter (α, σ, power factor, κ, zT); unless specified otherwise in parentheses, all values are reported at room temperature.

Abbreviations: DMF, N,N-dimethylformamide; DMSO, dimethyl sulfoxide; EG, enhanced graphene; EMIMBF4, 1-ethyl-3-methylimidazolium tetrafluoroborate; EMIM-DCA, 1-ethyl-3-methylimidazolium dicyanamide; MAI, methylammonium iodide; PEDOT:PSS, poly(3,4-ethylenedioxythiophene):polystyrene sulfonate; PEIE, polyethylenimine ethoxylated; PSSH, poly(4-styrenesulfonic acid); TFMS, trifluoromethanesulfonic acid; TSA, toluenesulfonic acid.

a

“nm” and “μm” categorize sample; nm: t < 1 μm and μm: 1 μm ≤ t < 1 mm.

Many other chemical dedoping methods have been reported afterward as a means to enhance the TE properties of pristine PEDOT:PSS films. A strong base, sodium hydroxide, was used as an additive to control the pH value of the PEDOT:PSS solution, which, in effect, acted as a facile approach to optimize the carrier concentration.²⁰ As hydrogen-bound PSS and sodium-bound PSS have different carrier-generating capabilities, NaOH could be added to transform certain portions of hydrogen-bound PSS to sodium-bound PSS to control the carrier concentration. Maximum power factor of ~19.6 μW·m⁻¹·K⁻² was achieved at pH 1.8.

Another approach to eliminate the un-ionized PSS within the PEDOT:PSS films using a strong reducing agent, hydrazine, was reported.²¹ Unlike the previous work using an ethylene glycol dipping process, chemical dedoping was controlled by varying the reducing agent concentration.⁷ It is proposed that the overcoating of a DMSO–hydrazine mixture induced transition of PEDOT redox state from (bi)polaron to neutral. The posttreatment method enabled penetration into the PEDOT and PSS chains for a large dedoping effect. Through hall effect measurement, it was confirmed that the DMSO–hydrazine mixture treatment decreased the carrier concentration while enhancing mobility. The optimized DMSO–hydrazine-treated PEDOT:PSS film exhibited a power factor of 112 μW·m⁻¹·K⁻² with corresponding α and σ of 44 μV·K⁻¹ and 578 S·cm⁻¹. Shortly after, a follow-up research was reported by the same group with an additional p-toluenesulfonic acid monohydrate (TSA) doping process before the DMSO–hydrazine treatment.²² The sequential pre- and posttreatment allow precise tuning of the PSS concentration and PEDOT redox state. As shown in Figure 4A, the DMSO/TSA-doped PEDOT:PSS film consists of polaron-state thiophene backbones in PEDOT grains, which exhibits increased σ due to the larger carrier concentration from 970 to 1260 S·cm⁻¹ with minuscule reduction in α. Sequential DMSO–hydrazine treatment allows the insulating PSS to be removed, resulting in the increase in α at the expense of σ decrease. At the optimum TSA doping and DMSO–hydrazine dedoping conditions, the TE properties were enhanced to α: 49.3 μV·K⁻¹ and σ: 1310 S·cm⁻¹ for an overall power factor of 318.4 μW·m⁻¹·K⁻². Furthermore, a decrease in thermal conductivity was observed mainly due to the decrease in sample density and specific heat capacity (C_p). The overall contributions resulted in a high zT value of 0.31 at room temperature.

Figure 4. (A) Chemical doping of DMSO/TSA-doped PEDOT:PSS film by polaron formation (left), and chemical dedoping of DMSO–hydrazine-treated PEDOT:PSS film by neutral redox state formation (right). Reproduced with permission: Copyright 2014, Royal Society of Chemistry.22 (B) Seebeck coefficient of various PEDOT:PSS/ionic liquid heterostructures by volume percent of ionic liquid in methanol. Reproduced with permission: Copyright 2018, Elsevier.30 (C) Schematic of ion and hole accumulation of PEDOT:PSS:PSSH (left) and PSSH/A-PEDOT:PSS films (right) under temperature gradient. Fermi–Dirac distribution of PSSH-treated PEDOT:PSS film and potential barrier produced from Soret effect of PSSH layer (bottom). Reproduced with permission: 2018, Royal Society of Chemistry.31 (D) Variations of binding energy (ΔE) upon addition of anionic dopants of different chain lengths. Reproduced with permission: 2019, Wiley-VCH.32 DMSO, dimethyl sulfoxide; PEDOT:PSS, poly(3,4-ethylenedioxythiophene):polystyrene sulfonate; PSSH, poly(4-styrenesulfonic acid); TSA, toluenesulfonic acid

The addition and/or treatment of PEDOT:PSS with acids have also proved to be effective in TE property improvement.^23–25 In a study conducted by Fan et al.,²³ spin-coated PEDOT:PSS films were treated with 1 M H₂SO₄ on a heated surface. Afterward, NaOH was dropped on the sample for further treatment. The acid treatment decreased the PSS content within the sample, whereas base treatment changed the oxidation states of PEDOT chains from bipolaron toward neutral. The acid treatment enhanced σ from the charge mobility increase, whereas α increased from lower carrier concentration. It was pointed out that multiple acid treatment before the base treatment could result in greater σ (3088 S·cm⁻¹). The optimized α and σ were 39.2 μV·K⁻¹ and 2170 S·cm⁻¹, respectively, contributing to an overall power factor of 334 μW·m⁻¹·K⁻². A superacid, trifluoromethanesulfonic acid, in methanol (TFMS–MeOH) was also used as a posttreatment method to treat PEDOT:PSS films.²⁴ As TFMS has a low pK_a value of −14.7 and its conjugate base (CF₃SO₃⁻) exhibits a non-nucleophilic characteristic, the superacid is expected to reduce the PSS concentration more easily as compared with other acids. The treatment of TFMS in MeOH enabled multiple effects, such as poly(4-styrenesulfonic acid) (PSSH) removal and better aligned and dense PEDOT chains, which made possible the enhancement of σ without large reduction in α. Upon superacid treatment, α was increased from 17.6 to 21.9 μV·K⁻¹ and σ from 0.7 to 2980 S·cm⁻¹, and a maximum power factor of 143 μW·m⁻¹·K⁻² was achieved. Accounting an out-of-plane thermal conductivity of 0.224 W·m⁻¹·K⁻¹, a zT value of 0.19 was estimated. In another work, various sulfonic acids were added to PEDOT:PSS solutions as dopants, and subsequent DMSO–hydrazine treatments were performed.²⁵ Sulfonic acids of different stereochemical and conjugated structures were tested, of which the highest σ of 1996 S·cm⁻¹ was observed when benzenesulfonic acid was used. Sulfonic acid addition improved σ through PEDOT–PSS phase separation, partial change in the PEDOT redox state, and chain conformation change. DMSO–hydrazine treatment was used to further tune the PEDOT redox state, raising α. As a result, the acid-doped and DMSO–hydrazine-treated PEDOT:PSS film showed peak α and σ of 42.6 μV·K⁻¹ and 1119 S·cm⁻¹, respectively. Finally, a maximum power factor of 203.1 μW·m⁻¹·K⁻² was reported.

The effects of inorganic and organic salt solution treatment on TE properties of PEDOT:PSS films were investigated.^26–28 Inorganic salt zinc chloride (ZnCl₂) in N,N-dimethylformamide (DMF) was used to treat PEDOT:PSS films.²⁶ ZnCl₂ and DMF caused synergistic enhancement in TE properties from PSS removal and PEDOT chain conformation change. Treating temperature and concentration were optimized to attain a power factor of 98.2 μW·m⁻¹·K⁻² and zT value of 0.125. The corresponding α and σ values are 26.1 μV·K⁻¹ and ~1400 S·cm⁻¹, respectively. Other inorganic salts, such as sodium hydrogen carbonate (NaHCO₃), sodium sulfite (Na₂SO₃), and sodium borohydride (NaBH₄), were also used.²⁷ NaHCO₃ served as acido-base dedopants, whereas Na₂SO₃ and NaBH₄ acted as redox dedopants. All approaches showed a decrease in charge concentration by changing the PEDOT oxidation states from bipolaron toward polaron or neutral. Combined with EG treatment, a power factor of up to 100 μW·m⁻¹·K⁻² was achieved using NaHCO₃. Treatment with an organic salt, methylammonium iodide (MAI), in DMF/H₂O was used as another method for TE property improvement.²⁸ Treating PEDOT:PSS films with 0.1 M MAI in 80% DMF–20% H₂O under elevated temperatures removed excess PSS and the chain conformation changed from a coiled structure to a linear structure, remarkably enhancing the carrier mobility, as well as a slight increase in α. The α and σ were enhanced to 28 μV·K⁻¹ and 1831 S·cm⁻¹, respectively, resulting in a power factor of 144 μW·m⁻¹·K⁻². The zT value reached 0.11 from a calculated κ of 0.29 W·m⁻¹·K⁻¹.

Rather than using toxic reductants, a nontoxic reductant, polyethylenimine ethoxylated (PEIE), in ethylene glycol was used before further acid treatment by H₂SO₄.²⁹ PEIE/EG dipping reduced the oxidation level in PEDOT:PSS films to enhance α. Subsequent H₂SO₄ treatment was performed to increase the doping level of PEDOT, with excess oxidation prevented by the previous PEIE/EG step. As a result, improved α and σ of 21.9 μV·K⁻¹ and 2770.7 S·cm⁻¹ were reached, respectively. A power factor of 133.0 μW·m⁻¹·K⁻¹ was reported.

A record-high TE performance of PEDOT:PSS, comparable to bismuth telluride (Bi₂Te₃), was reported by Fan et al.³⁰ using ionic liquid treatment. In this study, three types of ionic liquids (1-ethyl-3-methylimidazolium tetrafluoroborate [EMIM-BF₄], 1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide [EMIM-TFSI]), and 1-ethyl-3-methylimidazolium dicyanamide [EMIM-DCA]) were used to coat PEDOT:PSS films, which were treated sequentially beforehand with H₂SO₄ and NaOH. Overall, the ionic liquid treatment increased the Seebeck coefficient greatly while only slightly lowering the electrical conductivity, as shown in Figure 4B. The mechanism of Seebeck coefficient improvement was explained from the ion accumulation at the hot and cold sides when the PEDOT:PSS film was subjected to a temperature gradient. Upon ionic liquid treatment, ions can thermally diffuse to the hot and cold sides, and the accumulated ions can induce an electric field from the net charge. The electric field may either scatter the low-energy holes or increase the holes' energy. As the Seebeck coefficient depends on the mean charge carrier energy, the ion accumulation results in the Seebeck coefficient enhancement. Using a PEDOT:PSS/EMIM-DCA heterostructure, the maximum power factor of 754 μW·m⁻¹·K⁻² was achieved. Assuming a thermal conductivity of 0.2–0.5 W·m⁻¹·K⁻¹, a zT value of 0.75 was estimated.

Alternatively, polyelectrolytes were coated on PEDOT:PSS films to induce the Soret effect and thereby enhance the TE properties.³¹ Using PSSH as a polyelectrolyte, Soret effect causes protons to diffuse toward the cold end of the PSSH layer, whereas holes diffuse toward the cold end of the PEDOT:PSS layer (Figure 4C) under high relative humidity. As can be seen in Figure 4C, the proton accumulation in the PSSH layer produces potential barriers to filter out the low-energy holes, enhancing the Seebeck coefficient. At 100% relative humidity, α and σ became 43.5 μV·K⁻¹ and 2120 S·cm⁻¹, respectively, resulting in a power factor of 401 μW·m⁻¹·K⁻². Treatment with PSSNa also resulted in a high power factor of 401 μW·m⁻¹·K⁻² with α and σ of 48.1 μV·K⁻¹ and 1732 S·cm⁻¹, respectively.

In a more recent study, electrostatic interaction between conducting PEDOT and insulating PSS was systematically controlled using small-molecule anionic dopants, sodium alkyl sulfonates, to improve the TE properties of PEDOT:PSS films.³² The electrostatic interaction strength was controlled by adding anionic dopants of different alkyl chain lengths (N = 1, 3, 6, 8, 13, 18). It was argued from density functional theory calculations that the greater carbon chain lengths exhibited higher binding energies toward the PEDOT chain, owing to a higher level of polarization of the dopant. This causes larger affinity between anionic dopants and PEDOT chain as compared with PSS counterions. Once bound to the PEDOT chain, the bulky dopants occupy large volumes of the PEDOT chain, weakening the electrostatic interaction between PEDOT and PSS chains (Figure 4D). The reduced electrostatic interaction causes three main effects: (1) conformational changes of PEDOT:PSS from conventional core–shell to a fibrous network, (2) decrease of oxidation level, and (3) increased crystallinity of PEDOT chains and change in molecular ordering from benzoid to quinoid. As a result, α and σ were simultaneously enhanced through the combination of the above three effects on carrier mobility and concentration. Optimal TE properties were observed at N = 6 with a power factor of 700.2 μW·m⁻¹·K⁻² and a zT value of 0.25.

PEDOT:PSS nanocomposites for TE property enhancement Composites with inorganic materials

The TE performance of PEDOT:PSS/inorganic material composites is summarized in Table 2. As one of the earliest attempts to enhance the TE properties of PEDOT:PSS through composites, Du et al.³³ prepared a Bi₂Te₃-based alloy nanosheet dispersion to physically mix with a commercially available PEDOT:PSS solution. Drop-cast composite films with 4.1 wt% of nanosheet concentration exhibited the highest TE properties with a power factor of 32.26 μW·m⁻¹·K⁻².

Table 2 Reported enhanced thermoelectric performance of PEDOT:PSS in various composite systems

Material	Seebeck coefficient (μV·K−1)	Electrical conductivity (S·cm−1)	Power factor (μW·m−1·K−2)	Thermal conductivity (W·m−1·K−1)	zT	Sample (thickness)a	References
PEDOT:PSS/Bi2Te3 nanosheet			32.26			Film (μm)	[33]
PEDOT:PSS/Bi2Te3	48.24	1390	325.3	0.31	0.28	Film (μm)	[34]
PEDOT:PSS/Bi0.5Sb1.5Te3	49	1285	308	0.19	0.484	Film (μm)	[35]
PEDOT:PSS/Bi2Te3 nanowire	45	~1000	205	0.29	$\ge$0.2	Film (nm)	[36]
PEDOT:PSS-coated Te nanorod	114.97	214.86	284	0.22	0.39	Film (μm)	[37]
PEDOT:PSS/Te nanorod/SSWNT	118	139	206			Film (μm)	[38]
PEDOT:PSS/Te nanowire			145			Film (μm)	[39]
PEDOT:PSS/SnSe nanosheet			386			Film (N/A)	[40]
PEDOT:PSS/SnSe0.97Te0.03			130.3			Film (μm)	[41]
PEDOT:PSS/CuxSey			270.3, ~389.7 (at 418 K)			Film (μm)	[42]
PEDOT:PSS/CuCl2	−18,200	0.052	1700			Film (nm)	[43]
PEDOT:PSS/PANI-CSA multilayer			49	0.24	0.08 (400 K)	Film (nm)	[44]
PEDOT:PSS/PANI-CSA multilayer	44	1680	325			Film (nm)	[45]
PEDOT:PSS/graphene/CNT	23.2	689	37.08	0.360	0.031	Film (N/A)	[46]
PEDOT:PSS on RTCVD graphene	~23	1096	57.9			Film (nm)	[47]
PEDOT:PSS/liquid exfoliated graphene	~22.5	1060	53.3	0.3	0.05	Film (nm)	[48]
PEDOT:PSS/graphene quantum dot		14.6	71.72	~1.53			Film (μm)	[49]
PEDOT:PSS/SWCNT			464 (DMSO), 407 (FA)			Film (μm)	[50]
SWCNT/PEDOT:PSS-coated Te nanorod	56	332	104			Film (μm)	[51]
Carbon-coated CNT network coated with PEDOT:PSS	82.9	734.5	504.8			Film (μm)	[52]
SWCNT/PEDOT:PSS	55.6	1701	526	0.4~0.6	0.26~0.39	Film (N/A)	[53]

Note: Thermoelectric properties not directly mentioned in the manuscript of each reference were estimated based on the plots of each parameter (α, σ, power factor, κ, zT); unless specified otherwise in parentheses, all values are reported at room temperature.

Abbreviations: CNT, carbon nanotube; DMF, N,N-dimethylformamide; DMSO, dimethyl sulfoxide; FA, formic acid; PANI-CSA, camphorsulfonic acid-doped polyaniline; PEDOT:PSS, poly(3,4-ethylenedioxythiophene):polystyrene sulfonate; RTCVD, rapid thermal chemical vapor deposition; SWCNT, single-walled carbon nanotube.

a

“nm” and “μm” categorize sample; nm: t < 1 μm and μm: 1 μm ≤ t < 1 mm.

Regarding PEDOT:PSS composites with bismuth telluride (Bi₂Te₃) or its alloys, recent works have adopted various strategies for TE property enhancement, such as defect engineering, nanostructuring, and carrier filtering.^34–36 Goo et al.³⁴ proposed the creation of anti-site defects within Bi₂Te₃ crystals using proton irradiation. Proton irradiation implants H⁺ ions within Bi₂Te₃. This creates surface defects, which act as adsorption sites, resulting in enhanced electrostatic interaction between Bi₂Te₃ and PEDOT:PSS. Greater electrostatic interaction leads to reduced contact resistance between the interface of the inorganic and organic structures, and therefore Seebeck coefficient and electrical conductivity are simultaneously increased. Using Bi₂Te₃ particles irradiated with proton energies of >12 eV, α and σ were enhanced to 48.24 μV·K⁻¹ and 1390 S·cm⁻¹, respectively, for a power factor of 325.3 μW·m⁻¹·K⁻². With an out-of-plane thermal conductivity of 0.31 W·m^–1·K⁻¹, the zT value of PEDOT:PSS/Bi₂Te₃ composite film was 0.28.

Another approach to fabricate high-performance PEDOT:PSS/Bi₂Te₃ composites is a Bi_0.5Sb_1.5Te₃ (BST) nanorod-forming process within a PEDOT:PSS solution, in combination with a two-step reduction process to maintain high electrical conductivity.³⁵ The first reduction using ascorbic acid lowered the reaction rate to grow the BST into one-dimensional crystals. The second reduction using sodium borohydride caused three-dimensional nanofiber structure growth. During the two-step reduction process, tellurium-dominant nanorods were converted to BST alloys, and oxidation of PEDOT:PSS occurred. Oxidation of PEDOT:PSS allowed the control of doping level for a higher electrical conductivity. Once the BST alloys were homogeneously dispersed into the composite film, PEDOT:PSS acted as conductive bridges between the BST nanofibrous network. Further optimization could be done through MeOH and DMSO treatment. At the optimized condition, the PEDOT:PSS/BST composite film exhibited α and σ of 49 μV·K⁻¹ and 1285 S·cm⁻¹, respectively. The optimized power factor was 308 μW·m⁻¹·K⁻², and with a calculated thermal conductivity of 0.19, the zT value of 0.484 was obtained at room temperature.

In another recent research, Bi₂Te₃ nanowire-based PEDOT:PSS nanocomposite films were fabricated with an attempt to apply carrier filtering as a strategy to enhance the TE properties using polar solvent vapor annealing.³⁶ PEDOT:PSS/Bi₂Te₃ nanowire solution was spin-coated on a glass substrate and annealed under DMSO vapor environment in a glass vial. The duration of DMSO vapor treatment controlled the PEDOT/PSS phase separation level, which results in PSS removal in the upper layer when continued for an extended time (>30 min). PSS removal decreases the PSS-to-PEDOT ratio, also lowering the work function of PEDOT:PSS. The work function difference between Bi₂Te₃ and PEDOT:PSS creates a barrier energy to scatter low-energy carriers, decreasing the mean carrier energy (Figure 5A). The energy filtering effect through PEDOT:PSS work function tuning simultaneously enhances the Seebeck coefficient and electrical conductivity, as seen in Figure 5B. At an optimized energy barrier level of 0.11 eV, corresponding to a polar solvent vapor annealing time of 120 min, α and power factor reached up to 45 ± 2.1 μV·K⁻¹ and 205 ± 17.803 μW·m⁻¹·K⁻², respectively. With a thermal conductivity of 0.29 W·m⁻¹·K⁻¹ assumed, zT value with a lower bound of 0.2 was estimated.

Figure 5. (A) Work function of poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS) as a function of PSS-to-PEDOT ratio (top), and resulting carrier filtering within the nanocomposite at the PEDOT:PSS/Bi2Te3 interface (bottom). Reproduced with permission: Copyright 2020, Elsevier.36 (B) Seebeck coefficient and barrier energy level (top), electrical conductivity (middle), and power factor (bottom) of nanocomposite as a function of polar solvent vapor annealing time. Reproduced with permission: Copyright 2020, Elsevier.36 (C) Thermoelectric performance of single-walled nanotube (SWNT)/PEDOT:PSS nanocomposites with 60 wt% SWNT as a function of NaOH concentration for treatment. Reproduced with permission: Copyright 2019, Elsevier53

Tellurium-based PEDOT:PSS composite TE materials have also shown great performance in multiple research studies.^37–39 Bae et al.³⁷ prepared tellurium/PEDOT:PSS hybrid composites in which nanorod-shaped tellurium was coated with a thin layer of PEDOT:PSS. The drop-cast Te/PEDOT:PSS composite film was posttreated with H₂SO₄ for further improvement of electrical conductivity through PSS removal. The H₂SO₄ concentration was controlled to control the conformational and compositional effects. At an optimized H₂SO₄ concentration of 80%, the power factor was 284 μW·m⁻¹·K⁻² with corresponding α and σ of 114.97 μV·K⁻¹ and 214.86 S·cm⁻¹, respectively. Using an estimated thermal conductivity of 0.22 W·m⁻¹·K⁻¹, the zT value of 0.39 was obtained. The same group reported telluride/PEDOT:PSS composite films in which nanocarbon material such as graphene nanoparticles (GNPs) or small-bundle single-walled carbon nanotubes (SSWNTs) was added to enhance the TE properties.³⁸ Films prepared with the addition of GNP clearly show flat-shaped flakes within the PEDOT:PSS/Te nanorod matrix, whereas addition of SSWNT shows interconnected networks between SSWNT and the PEDOT:PSS/Te nanorod matrix. The interconnected PEDOT:PSS/Te nanorod and SSWNT connections formed continuous pathways promoting electrical conduction through tunneling and hopping. With 0.3 wt% of SSWNTs added into the PEDOT:PSS/Te nanorod matrix, the power factor was optimized to 206 μW·m⁻¹·K⁻², with corresponding α and σ of 118 μV·K⁻¹ and 139 S·cm⁻¹, respectively.

To produce well-nanostructured composites, Sahu et al.³⁹ used sulfur anions as a linker molecule to atomically bind the PEDOT:PSS core–shell structure to inorganic nanostructures, such as nanowires, nanoplates, and nanocrystals. By modifying the surface of inorganic nanostructures with sulfur anions, the nanostructures are well dispersed into PEDOT:PSS aqueous dispersions, enabling the production of PEDOT:PSS-coated inorganic–organic hybrids maintaining the original crystalline properties. Using the strategy of surface modification on Te nanowire with an aspect ratio of ~1000, the PEDOT:PSS/Te nanowire composite exhibited a maximum power factor of 145 μW·m⁻¹·K⁻². Using Bi₂Te₃ as the inorganic material, an n-type nanocomposite with a maximum power factor of ~16 μW·m⁻¹·K⁻² was achieved.

Tin selenide-based composites have also been studied for the application to organic TEs.^40,41 Ju and Kim⁴⁰ prepared tin selenide nanosheets using a Li⁺ intercalation, subsequent exfoliation process from inorganic ingots, and they dispersed the synthesized nanosheets into the PEDOT:PSS solution. A maximum power factor of ~386 μW·m⁻¹·K⁻² was achieved at 5% DMSO addition and an SnSe nanosheet content of 20 wt%. A tin–selenide–telluride alloy nanosheet (SnSe_0.97Te_0.03)-based composite material was also used to improve the TE properties of PEDOT:PSS.⁴¹ As the content of the nanosheet increased, electrical conductivity decreased and Seebeck coefficient increased. Sn–Se–Te/PEDOT:PSS composite films with a nanosheet content of 15 wt% showed an improved power factor of ~130.3 μW·m⁻¹·K⁻².

Other inorganic materials, such as molybdenum sulfide, boron nitride, silicon dioxide, silicon nanoparticle, silicon carbide nanowire, two-dimensional molybdenum selenide, (Ca_{1 − x}Ag_x)₃Co₄O₉, tin sulfide nanobelt, and black phosphorus, have been used as a composite filler.^54–62 Lu et al.⁴² prepared PEDOT:PSS-coated copper selenide (Cu_xSe_y) nanowire composite films on a nylon porous membrane using vacuum-assisted filtration and cold pressing. The cold-pressed composite film forms a nanoporous network from the intertwining of nanowires, coming into contact to form nanowire–PEDOT:PSS–nanowire junctions. The PEDOT:PSS/Cu_xSe_y nanowire junction acts as an energy barrier to scatter the low-energy charge carriers, inducing the energy filtering effect. Overall, the PEDOT:PSS-coated nanowire exhibits lower carrier concentration but a higher carrier mobility. With a composite film starting from a Cu/Se nominal molar ratio of 3, the optimized power factor of 270.3 μW·m⁻¹·K⁻² at 300 K (~389.7 μW·m⁻¹·K⁻² at 418 K) was obtained.

Copper chloride (CuCl₂) was used to make metal–complex conductive polymer films for n-type organic TEs.⁴³ CuCl₂ was added to PEDOT:PSS solutions in different amounts, binding the copper ions (Cu²⁺) to the PSS in PEDOT:PSS films. This generates free anion channels within the wet PEDOT:PSS matrix, inducing an n-type ionic TE effect. At optimal CuCl₂ content of 40 wt%, the free Cl⁻ anions retain a high mobility through ionic channels, resulting in a high Soret effect. At a relative humidity of 80%, a high negative Seebeck coefficient of −18.2 mV·K⁻¹ and a resulting power factor of 1.7 mW·m⁻¹·K⁻² (1700 μW·m⁻¹·K⁻²) were achieved.

Composites with carbon materials

The TE performance of PEDOT:PSS/carbon material composites is summarized in Table 2. Camphorsulfonic acid-doped polyaniline (PANI-CSA) was used to make a multilayer structure with PEDOT:PSS via layer-by-layer (LbL) deposition using spin coating.⁴⁴ The PEDOT:PSS and PANI-CSA layer was deposited alternatively, with an approximate individual layer thickness of 20 nm. Increased stacking of PEDOT:PSS above PANI-CSA stretches the conductive PEDOT chains and PANI chains. The change in conformation increases the connectivity of conductive PEDOT grains, which promotes carrier transfer, with a negligible reduction in the Seebeck coefficient. The five-(PANI-CSA/PEDOT:PSS)-layer film results in a power factor of 49 μW·m⁻¹ ·K⁻² and an estimated zT value of 0.08 at 400 K with the assumption of a thermal conductivity value of 0.24 W m⁻¹·K⁻¹ from bulk PEDOT:PSS. The same group recently reported a further study of the PEDOT:PSS/PANI-CSA multilayer films by controlling the deposition conditions such as spin speed, number of layers, and individual layer thickness.⁴⁵ Conventionally available PANI-CSA was used as compared with the in-house-synthesized PANI-CSA used in their previous work.⁴⁴ With eight PANI-CSA/PEDOT:PSS layers deposited at 7000 rpm, and a single PANI-CSA/PEDOT:PSS layer thickness of ~40 nm, a power factor of 325 μW·m⁻¹·K⁻² was obtained with corresponding α and σ of 44 μV·K⁻¹ and 1680 S·cm⁻¹, respectively. The enhanced TE properties are attributed to the stretching of PEDOT chains, excess PSS removal of PEDOT:PSS, and conformational change from benzoid to quinoid.

Graphene was used for TE composite materials with PEDOT:PSS in forms of not only nanosheets, but also quantum dots and films prepared through rapid thermal chemical vapor deposition (RTCVD).^46–49 Yoo et al.⁴⁶ used graphene and CNTs to make PEDOT:PSS/graphene/CNT composites, which caused synergistic effects to enhance the TE properties. 3,4-Ethylenedioxythiophene (EDOT) monomers were polymerized in water with PSS-dispersed graphene and multiwalled carbon nanotubes (MWCNTs). The content of graphene and MWCNT was controlled to 5 wt% in total, with an equal weight ratio (1:1). At this composition, a maximum power factor of 37.08 μW·m⁻¹·K⁻² was obtained with corresponding α and σ of 23.2 μV·K⁻¹ and 689 S·cm⁻¹, respectively. With a thermal conductivity of 0.360 W·m⁻¹·K⁻¹, a zT value of 0.031 was achieved. The simultaneously enhanced Seebeck coefficient and electrical conductivity arise from stronger electrostatic interactions by forming composites with low-dimensional carbon materials. The increased level of interaction helps connect the conductive domains of the composite film. The same group used RTCVD as another approach for PEDOT:PSS/graphene hybrid films.⁴⁷ RTCVD graphene was transferred on a flexible polyethylene terephthalate (PET) substrate and treated with UV irradiation to produce hydrophilic surface characteristics. PEDOT:PSS was then spin-coated directly on the RTCVD graphene to obtain PEDOT:PSS/graphene hybrid films. The coating of PEDOT:PSS reduced the defects on the surface of RTCVD graphene. The spin speed of PEDOT:PSS was controlled to 3000 rpm, at which maximum power factor of 57.9 μW·m⁻¹·K⁻² was achieved with corresponding α and σ of 23 μV·K⁻¹ and 1096 S·cm⁻¹, respectively.

Liquid-exfoliated graphene was prepared for the fabrication of PEDOT:PSS/graphene composite films.⁴⁸ Exfoliated graphene organic solvent dispersion was mixed into PEDOT:PSS solutions, which was coated on PVDF porous membranes via vacuum filtration. The as-prepared film was further treated by immersion into hydrazine of different concentrations to enhance the TE properties. PEDOT:PSS wraps the graphene via strong π–π interaction, creating channels for carrier transport. Therefore, the addition of exfoliated graphene and hydrazine treatment lead to an increase in the Seebeck coefficient with a slight decrease in the electrical conductivity. For PEDOT:PSS/graphene composite films with 3 wt% of graphene, a power factor of 53.3 μW·m⁻¹·K⁻² was obtained with α and σ of 22.5 μV·K⁻¹ and 1060 S·cm⁻¹, respectively. Assuming a thermal conductivity of 0.3 W·m⁻¹·K⁻¹, a zT value of 0.05 is estimated.

Graphene quantum dots (GQDs) with the size of several nanometers were prepared to make composites with PEDOT:PSS.⁴⁹ GQDs were produced by chemically cleaving graphene oxides and then mixed into PEDOT:PSS solutions. As a result of stronger π–π interactions between PEDOT chains and graphene regions, electrical conductivity and Seebeck coefficient were increased simultaneously to 71.72 S·cm⁻¹ and 14.6 μV·K⁻¹, respectively, after optimization of GQD content.

CNTs were used for TE composite films with PEDOT:PSS in many research studies, such as n layered nanostructure of PEDOT:PSS on top of an SWCNT nanofilm, ethylene glycol treatment on double-walled carbon nanotube (DWCNT)/PEDOT:PSS nanocomposites, and hot pressing for densification of few-walled carbon nanotube (FWCNT; 2–4 carbon walls)/PEDOT:PSS films, to mention a few.^63–65 To further improve the TE properties of PEDOT:PSS/CNT composites and explain the origins of the enhanced properties, Hsu et al.⁵⁰ prepared PEDOT:PSS/SWCNT drop-cast films with different CNT concentrations and treated the as-prepared PEDOT:PSS/SWCNT films with polar solvents such as DMSO and formic acid (FA) via dipping. With the solvent treatment, electrical conductivity increased by a factor of ~330, whereas the Seebeck coefficient was maintained at a similar value. CNT addition also caused the increase in electrical conductivity. It is stated that upon CNT addition, carrier mobility increased greatly due to the exceptional carrier mobility of CNTs. The solvent dipping step facilitated PSS removal from the weakened electrostatic interaction between the PEDOT and PSS chains. Consequently, the maximum power factor was achieved at 6.7 wt% CNT for both solvents (464 μW·m⁻¹·K⁻² for DMSO, 407 μW·m⁻¹·K⁻² for FA).

SWCNT/PEDOT:PSS-coated Te nanorod composite films were prepared as another route to TE property enhancement.⁵¹ A PEDOT:PSS-coated Te nanorod dispersion was first prepared, and an SWCNT aqueous dispersion was mixed into the PEDOT:PSS-coated Te nanorod dispersion. The SWCNT/PEDOT:PSS-coated Te nanorod film was prepared by vacuum filtration on a porous nylon membrane and further treated by immersing the composite film in H₂SO₄ and heating on a hot plate. The addition of CNT increased the electrical conductivity and decreased the Seebeck coefficient. Treatment with H₂SO₄ also increased the electrical conductivity by PSS removal. H₂SO₄-treated SWCNT/PEDOT:PSS-coated Te nanorod composite films showed a maximum power factor of 104 μW·m⁻¹·K⁻² at 70 wt% CNT content with corresponding α and σ of 56 μV·K⁻¹ and 332 S·cm⁻¹, respectively.

In more recent studies, PEDOT:PSS/CNT composites exhibiting TE properties comparable to the highest levels achieved so far in organic TEs were reported.^52,53 To achieve high-performance TE composites, carbon-coated CNT networks were prepared using a floating catalyst chemical vapor deposition (FCCVD) method and further coated with PEDOT:PSS through dipping into a dilute PEDOT:PSS aqueous dispersion.⁵² Compared with CNT films prepared from membrane filtration of a CNT dispersion, continuous networks are formed directly without any surfactant or posttreatment that can act as insulating properties. By further coating with PEDOT:PSS, a composite with two heterojunctions (carbon/CNTs and carbon/PEDOT:PSS) was formed, which acted as interfacial energy barriers. As the work function of carbon, CNTs, and PEDOT:PSS differs, the level of energy barriers was controlled to induce the energy-filtering effect, in which low-energy carriers are scattered to increase the Seebeck coefficient. At a PEDOT:PSS dispersion dipping time of 20 min, the power factor was maximized to 504.8 μW·m⁻¹·K⁻² with corresponding α and σ of ~82.9 μV·K⁻¹ and ~734.5 S·cm⁻¹, respectively. Liu et al.⁵³ reported SWCNT/PEDOT:PSS films with a high TE power factor prepared with a facile DMSO addition and NaOH posttreatment process. SWCNTs were incorporated into PEDOT:PSS, which facilitated the creation of stable and effective conduction channels, simultaneously enhancing the Seebeck coefficient and electrical conductivity. To further improve the TE properties, drop-cast SWCNT/PEDOT:PSS films with 5% DMSO were treated with different concentrations of NaOH and annealed mildly. DMSO addition increased the carrier mobility by changing the polymer chain conformation from coiled to an extended structure, whereas further NaOH treatment optimized the carrier concentration to tune the Seebeck coefficient to a higher level. As a result, SWCNT/PEDOT:PSS composite films with a high power factor of ~526 μW·m⁻¹·K⁻² were obtained with α and σ of 55.6 μV·K⁻¹ and 1701 S·cm⁻¹, respectively (Figure 5C). Assuming a thermal conductivity of 0.4–0.6 W·m⁻¹·K⁻¹, a high zT value of 0.26–0.39 was estimated at room temperature.

PEDOT:PSS for flexible TE

Bae et al.⁶⁶ reported one of the first spray-printed flexible TE generators, which can be applied to commercial and large-scale flexible TE generators. PEDOT:PSS-coated Te–Bi₂Te₃ nano-barbell structures were prepared from a solution-phase reaction, which was spray-printed on flexible polyimide, as shown in Figure 6A. The power factor of hybrid films was optimized to 60.05 μW·m⁻¹·K⁻². The as-prepared flexible TE generator produced an open-circuit voltage of 1.54 mV using six legs at a ΔT of 10 K (Figure 6A). The same group also reported organic fiber-based TE generators from the synthesis of CNT/PEDOT:PSS composite fibers for the application to textile-based devices.⁶⁷ CNT/PEDOT:PSS fibers were synthesized by ball milling a dispersion of PEDOT:PSS pellets and SWCNTs to make a paste, and spinning the paste in methanol for coagulation. Solvent posttreatment methods were used to optimize the TE properties of p-type fibers and branched polyethylenimine (PEI) was used for the n-type doping. Afterward, a fiber-based TE generator was fabricated using p-type and n-type CNT/PEDOT:PSS fibers with optimized power factors of 83.2 ± 6.4 μW·m⁻¹·K⁻² and 113 ± 25 μW·m⁻¹·K⁻², respectively (Figure 6B). As shown in Figure 6B, a high output power of 0.430 μW was measured for the fiber-based TE generator assembled with 12 p–n junctions at a ΔT of 10 K. This fiber-based TE generator demonstrates the potential to be used for large-area curved heat sources and human body heat energy harvesting.

Figure 6. (A) Schematic of the spray-printing method and a resulting thermoelectric (TE) generator on a glass substrate, flexible polyimide film (top), and open-circuit TE voltage–ΔT plot of TE generator on glass and flexible polyimide film (bottom). Reproduced with permission: Copyright 2016, The Royal Society of Chemistry.66 (B) Photographic image of a carbon nanotube (CNT)/poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS) fiber-based TE generator (top), and output power–current and output voltage–current plot of a CNT/PEDOT:PSS fiber-based TE generator (bottom). Reproduced with permission: Copyright 2018, Elsevier67

Liu et al.⁶⁸ reported TE generators assembled with PEDOT:PSS fibers derived from a gelated PEDOT:PSS for the p-type component and CNT for the n-type component. PEDOT:PSS fibers were prepared by inducing gelation of PEDOT:PSS dispersions in a capillary tube with an inner diameter of 1.0 mm. The PEDOT:PSS fibers were further treated with ethylene glycol. The PEDOT:PSS fibers that were used for the p-type material were optimized to exhibit power factors of 4.77 μW·m⁻¹ K⁻². A fiber-based TE generator composed of five p–n couples was fabricated, which showed a large open-circuit voltage and power density of 20.69 mV and 481.17 μW·cm⁻², respectively, at a ΔT of 60 K. This study shows the capabilities of PEDOT:PSS fibers based on hydrogels to be used for fiber-based TE generators.

CNT CNT doping and treatment for TE property enhancement

The TE performance of treated CNT is summarized in Table 3. One of the earliest studies that substantially improved the TE properties of pristine CNTs was conducted by an argon plasma treatment.⁶⁹ A CNT aqueous dispersion was sprayed on glass slides that were treated with argon plasma in a vacuum chamber for different time periods. After washing in deionized water and sonication, the treated CNT solution was dried on a copper foil that was peeled off as a bucky paper. Compared with non-argon-irradiated samples in which the CNT walls were crystalline, argon-irradiated samples had more structural defects, and the amount of defects increased with longer argon irradiation. The main effect of defect creation was to increase the semiconducting property of CNT, as confirmed from hall coefficient measurement. Other minor effects such as increased phonon scattering and low-energy carrier scattering for a lower carrier concentration were stated. Overall, the maximum Seebeck coefficient increased, and the electrical conductivity decreased after argon plasma treatment. At a treatment time of 20 s, the power factor was optimized to >120 μW·m⁻¹·K⁻². A peak zT value of 0.4 at ~670 K was achieved with a thermal conductivity of 0.28–0.34 W·m⁻¹·K⁻¹ in the measurement temperature range. This value is comparable to that achieved on most of the inorganic oxide TE materials.

Table 3 Reported enhanced thermoelectric performance of CNT using various doping methods

Method	Seebeck coefficient (μV·K−1)	Electrical conductivity (S·cm−1)	Power factor (μW·m−1·K−2)	Thermal conductivity (W·m−1·K−1)	zT	Sample (thickness/diameter)a	References
Argon plasma treatment	~350 (670 K)	~9.90 (670 K)	>120	0.28–0.34 (290–675 K)	0.4 (670 K)	Film (μm)	[69]
Oxidative doping of enriched semiconducting CNT networks			~340			Film (nm)	[70]
Purification of enriched semiconducting CNT networks using trifluoroacetic acid			~398			Film (nm)	[71]
Triphenylmethane carbinol base-doped SWCNT	−59	497	172			Film (μm)	[72]
SWCNT doped with cetyltrimethylammonium bromide	~−47	~850	185.7			Film (μm)	[73]
BV-doped or OA-doped n-type, p-type PT:PFPD semiconducting SWCNT			>700	2.38	~0.12	Film (nm)	[74]
Highly aligned metallic SWCNT films with Fermi energy tuning via electrolyte gating			~300			Film (nm)	[75]

Note: Thermoelectric properties not directly mentioned in the manuscript of each reference were estimated based on the plots of each parameter (α, σ, power factor, κ, zT); unless specified otherwise in parentheses, all values are reported at room temperature.

Abbreviations: BV, benzyl viologen; CNT, carbon nanotube; OA, triethyloxonium hexachloroantimonate; PT:PFPD, Plasma torch:poly[(9,9-di-n-dodecyl2,7-fluorendiyl-dimethine)-(1,4-phenylene-dinitrilomethine); SWCNT, single-walled carbon nanotube.

a

“nm” and “μm” categorize sample; nm: t < 1 μm and μm: 1 μm ≤ t < 1 mm.

Avery et al.⁷⁰ prepared a semiconducting CNT network through CNT enrichment and further optimized the TE properties by introducing dopants. Enriched semiconducting CNTs were obtained by dissolving SWCNT source materials in various fluorene-based polymer solutions. Each fluorene-based polymer selectively wraps semiconducting SWCNTs of different structures and chirality, allowing the acquisition of semiconducting SWCNT dispersions without any metallic SWCNTs detectable through absorbance or Raman spectroscopy. The acquired semiconducting SWCNT dispersions were coated using ultrasonic spray to obtain CNT networks and subsequently heavily doped with a strong oxidant, triethyloxonium hexachloroantimonate (OA), by soaking. The concentration of oxidative dopant solution was controlled to tune the doping level and optimize the TE properties. The p-type dopant treatment controlled the semiconducting SWCNT carrier concentration and Fermi energy, affecting the Seebeck coefficient, electrical conductivity, and thermal conductivity simultaneously. As shown in Figure 7B, semiconducting CNT networks are able to reach power factors as high as ~340 μW·m⁻¹·K⁻², a record-high value for SWCNT networks. The same group further improved the TE properties of semiconducting SWCNTs by removing the wrapping polymer used for the enrichment of semiconducting SWCNTs, by using a supramolecular polymer solution for the wrapping polymer, and treated further with trifluoroacetic acid (TFA) for polymer removal.⁷¹ Compared with the case when poly[(9,9-dioctylfluorenyl-2,7-diyl)-alt-co-(6,6ʹ-)] (PFO-BPy) was used with a carbon source of laser vaporization-prepared SWCNT dispersion in the previous study, the modified process resulted in an increase in carrier mobility and average CNT bundle size, therefore enhancing TE properties. As a result, a power factor as high as 398 μW·m⁻¹·K⁻² was obtained, which is the best single-component CNT network reported for TEs.

Figure 7. (A) Thermopower (top) and power factor (bottom) of various polyfluorene/semiconducting single-walled carbon nanotube (SWCNT) thin films. Reproduced with permission: Copyright 2016, Nature Publishing Group.70 (B) Maximum power factor as a function of maximum electrical conductivity for various semiconducting SWCNT networks. Blue ovals and orange ovals represent semiconducting SWCNT networks utilizing cleavable polymers and non-cleavable polymers, respectively. Reproduced with permission: Copyright 2017, The Royal Society of Chemistry.74 (C) Seebeck coefficient (left) and power factor (right) as a function of gate voltage for SWCNTs with increasing metallic SWCNT content from #1 to #4. #5 represents aligned metallic SWCNT films. Schematic diagram of electrolyte gating setup (bottom). Reproduced with permission: Copyright 2019, American Chemical Society75

N-type doping of CNTs has also been studied over the years using both molecular and polymeric dopants, such as PEI, phosphine-based aromatic dopants, cobaltocene, and 2-aryl-1,3-dimethyl-2,3-dihydro-1H-benzo[d]imidazole (DMBI) derivative organic dopants.^76–78 Nonoguchi et al.⁷² prepared n-type CNT TE materials with water processability that are air-stable for months using a triarylmethane derivative. When the pH value of an aqueous solution of malachite green (MG), which is a hydrochloric acid salt, was tuned to above 11.6, the MG is transformed to its neutral counterparts, triphenylmethane carbinol base (TPM-CB) nanoparticles, exhibiting electron-donating properties at the CNT/nanoparticle interface for n-type CNTs. CNT bucky papers were processed into n-type by dipping in the pH-controlled TPM-CB aqueous solution. The TPM-CB nanoparticles donate electrons to the SWCNTs, which induces the formation of cationic ammoniums in the nanoparticles, which work as stabilizers for their ionic pair, n-type SWCNTs. As a result, TPM-CB-doped SWCNT bucky papers exhibit n-type TE properties, which were maximized at a pH 12 with a power factor of 172 μW·m⁻¹·K⁻² and corresponding α and σ of −59·μV·K⁻¹ and 497 S·cm⁻¹, respectively. Air stability of up to 796 h was also confirmed for the optimized n-type SWCNT samples. The cationic charges that worked as stabilizers for the negative charges of n-type SWCNTs were stable in air. In engineering aspects, this proved to be an attractive process for large-scale n-type CNT TE material production.

A common alkylammonium cationic surfactant, cetyltrimethylammonium bromide (CTAB), was also used for the preparation of n-type SWCNT films using a simple solution mixing process.⁷³ N-type SWCNT films were prepared by mixing SWCNT and a cationic surfactant into a dispersion medium of choice (H₂O, DMF, or petroleum ether [PE]) and vacuum filtration on a filter paper. The optimal TE properties of n-type SWCNTs were searched by tuning the dopant concentration, dispersion medium, alkyl chain length, and type of the anion of the cationic surfactant. After only 0.5 h of doping treatment, pristine p-type SWCNTs were readily converted into n-type and reached a high power factor of 185.7 ± 8.5 μW·m⁻¹·K⁻² using a SWCNT:CTAB mass ratio of 10:5 in DMF. The optimized n-type SWCNT films retained their TE performance in air for 100 h with only slightly reduced Seebeck coefficient.

N-type conversion of semiconducting SWCNTs (s-SWCNTs) was also reported by Macleod et al.⁷⁴ to produce pure s-SWCNTs of both carrier types, integrating ink chemistry, solid-state polymer removal, and doping strategies. To readily remove the polymers used for the enrichment of s-SWCNTs, cleavable fluorene-based polymers such as SMP and PF-PD were used. The polymer-wrapped s-SWCNT films prepared through ultrasonic spraying were dipped into a mixture of TFA and toluene to remove the fluorene-based polymers. P-type and n-type doping was conducted by immersing the polymer-removed s-SWCNT films into solutions of OA in dichloroethane (DCE), and either potassium–crown ether complex or benzyl viologen (BV), respectively. By controlling the doping level, power factors exceeding 700 μW·m⁻¹·K⁻² were obtained for both p-type and n-type s-SWCNT networks, which is among the highest reported for carbon-based TE materials (Figure 7B). The greatly improved TE properties stem from the removal of sorting polymers, minimization of s-SWCNT bundle size, and the control of s-SWCNT diameters. By atomic layer deposition of 50-nm-thick Al₂O₃, the p-type and n-type s-SWCNT networks maintained their TE performance for over 300 h with <5% reduction in power factor.

In a recent study, metallic SWCNTs have been shown to exhibit greater TE properties as compared with highly enriched semiconducting SWCNTs by solving the typical trade-off relation of the Seebeck coefficient and electrical conductivity through the tuning of the Fermi energy with electrolyte gating.⁷⁵ The electrons and holes were injected via the double-layer formation, controlled with the gate voltage, V_G. Highly enriched metallic and semiconducting SWCNTs were prepared, and samples were prepared with varying ratios of metallic and semiconducting SWCNTs. Vacuum filtration was also utilized to produce highly aligned metallic SWCNT films. As shown in Figure 7C, different gate voltages tuned the Fermi energy of the samples, crossing into both p-type and n-type regimes. For highly aligned metallic SWCNT film samples, a maximum power factor of ~300 μW·m⁻¹·K⁻² was obtained when the Fermi energy was tuned to the position where it coincides with the first van Hove singularity (vHs). In the vHs region, the Seebeck coefficient and electrical conductivity of metallic SWCNTs were confirmed to increase simultaneously as compared with the semiconducting SWCNTs, in which the same traditional TE property trade-off phenomenon was observed. Consequently, this study demonstrates how metallic SWCNTs can also be potential candidates for TE materials.

CNT composites for TE property enhancement Composites with inorganic materials

The TE performance of CNT/inorganic material composites is summarized in Table 4. As an early attempt to apply CNTs to the field of TEs, a ceramic material, called 3Y-TZP, was used for CNT/ceramic composites, achieving zT values of ~0.02.⁹² Over the years, inorganic materials well known for their TE performance were used to fabricate composites with CNT, such as silver telluride, bismuth telluride, lead telluride, telluride, and cobalt-based ceramics.^93–96 Nunna et al.⁷⁹ prepared Cu₂Se-based composite materials with high TE performance in high-temperature regions by dispersing CNTs into the Cu₂Se matrix. The CNTs were mixed into Cu and Se powders using a ball-milling method. Due to the high reactivity of CNTs with Cu and Se from their free π electrons, Cu atoms adhered to the surface of CNTs and formed nanometer-sized grains in Cu₂Se upon reaction with Se. The Seebeck coefficient of the hybrid material increases as temperature is elevated through the phase transition from α-phase to β-phase, whereas the electrical conductivity reduces with increased temperature, following the trade-off relation in a typical heavily doped semiconductor. After the formation of hybrids with CNTs, the power factor decreases as compared with the pristine Cu₂Se. However, the thermal conductivity reduces to 0.4 W·m⁻¹·K⁻¹ for Cu₂Se/0.75 wt% CNT samples, which is a 50% decrease compared with the pristine Cu₂Se at 1000 K. This reduction in thermal conductivity is attributed to the enhanced scattering of phonons at the Cu₂Se/CNT interface. As a result, a peak zT value of 2.4 is reached at 1000 K.

Table 4 Reported enhanced thermoelectric performance of CNT in various composite systems

Material	Seebeck coefficient (μV·K−1)	Electrical conductivity (S·cm−1)	Power factor (μW·m−1·K−2)	Thermal conductivity (W·m−1·K−1)	zT	Sample (thickness/diameter)a	References
Cu2Se/CNT	~300 (1000 K)			0.4 (1000 K)	2.4 (1000 K)	Pellet (mm)	[79]
MgAg0.97Sb0.99/CNT					0.83 (300 K), 1.05 (375 K)	Pellet (mm)	[80]
Lu0.1Bi1.9Te3/CNT	~–137 (423 K)		~1950 (423 K)	~0.8 (423 K)	1.05 (423 K)	Wafer (mm)	[81]
PANI-coated CNT	~29	~60	5.0433	~0.5		Pellet (nm)	[82]
PANI/graphene/PANI/DWCNT multilayer	130	1080	1825			Film (nm)	[83]
CNT/PANI	~46	~1900	401			Film (μm)	[84]
PEDOT:PSS/CNT	~41	~950	~160	0.2–0.4		Film (μm)	[85]
PEDOT/MWCNT	~27	2100	155			Film (μm)	[86]
PEDOT:PF6/SWCNT	~53	~900	253.7			Film (μm)	[87]
PEDOT:PF6/a-SWCNT			350			Film (μm)	[88]
PEDOT-coated CNT	~48	~680	157			Film (nm)	[89]
PEDOT/CNT	~−1000	~5	1050	~0.67	~0.5	Film (nm)	[90]
P3HT/CNTF	91.9 (344.15 K)	130.2 (344.15 K)	110.0 (344.15 K)			Film (nm)	[91]

Note: Thermoelectric properties not directly mentioned in the manuscript of each reference were estimated based on the plots of each parameter (α, σ, power factor, κ, zT); unless specified otherwise in parentheses, all values are reported at room temperature.

Abbreviations: a-SWCNT, acid-doped single-walled carbon nanotubes; CNT, carbon nanotube; MWCNT, multiwalled carbon nanotube; PANI, polyaniline; PEDOT:PSS, poly(3,4-ethylenedioxythiophene):polystyrene sulfonate; SWCNT, single-walled carbon nanotube.

a

“nm” and “μm” categorize sample; nm: t < 1 μm and μm: 1 μm ≤ t < 1 mm, and mm: 1 mm ≤ t.

Aside from high-temperature applications, improved TE performance for CNT/inorganic material composites has been also reported for room temperature and mid-temperature uses.^80,81 Lei et al.⁸⁰ prepared α-MgAgSb alloys with MWCNTs inserted into the matrix for efficient TE materials at room temperature. The MWCNTs exhibited metallic properties, enhancing the carrier transport capabilities, while scattering phonons at the α-MgAgSb/CNT interface. MgAg_0.97Sb_0.99/CNT hybrid materials with 0.1 wt% of CNT show improved electrical conductivity and reduced thermal conductivity, resulting in a zT value of 0.83 at 300 K and 1.05 at 375 K. Cao et al.⁸¹ prepared Lu-doped Lu_0.1Bi_1.9Te₃/CNT composites by hot pressing the CNT into bulk powder samples. The CNT functioned as conductive fillers to increase the carrier mobility and decrease the thermal conductivity. Lu_0.1Bi_1.9Te₃ samples with 0.05 wt% CNT show a zT value of 1.05 at 423 K.

Composites with PANI

The TE performance of CNT/PANI composites is summarized in Table 4. One of the earliest studies of CNT composites with PANI was conducted by Meng et al.,⁸² in which PANI was uniformly coated on CNTs by in situ polymerization methods. The PANI coating increased the Seebeck coefficient by inducing nanometer-sized energy barriers to filter out the carriers with low energy and decreased the Seebeck coefficient with a PANI content above 15.8 wt%. Electrical conductivity monotonically decreased due to the inferior conductivity of PANI as compared with CNT. The maximum power factor value was achieved at 15.8 wt% of PANI, reaching 5.0433 μW·m⁻¹·K⁻². CNT/PANI composites also show some promise due to the relatively low thermal conductivity (0.45 W·m⁻¹·K⁻¹) as compared with pristine CNTs.⁹⁷

CNT/PANI composites exhibiting super low thermal conductivity were reported by Chen et al.⁹⁸ for TE applications. In this study, MWCNTs were prepared into three-dimensional (3D) CNT networks using CVD techniques, and the produced bulk CNT network was used as a template for the in situ polymerization of PANI. The CNT network maintains a sponge-like structure with a great number of pores leading to its super low thermal conductivity of 0.035 W·m⁻¹·K⁻¹. Upon forming a composite with PANI, the TE properties were enhanced greatly to reach a room temperature zT value of 0.0022 with corresponding α and σ of 23.3 μV·K⁻¹ and 40.35 S·cm⁻¹, respectively. The thermal conductivity was decreased to 0.29 W·m⁻¹·K⁻¹. The enhanced Seebeck coefficient arises from the numerous interfaces created by the π-bond interaction between PANI and CNTs. The 3D CNT network and its subsequent usage as templates for TE materials solve the aggregation of CNTs and provide a new approach to fabricate CNT composites with enhanced TE properties and flexibility.

Cho et al.⁸³ prepared well-dispersed polymer/carbon nanocomposites by stacking layers in the order of PANI, graphene, PANI, and DWCNTs having diameters of 2–3 nm, using an LbL deposition approach. Through the repeated deposition of quadlayer (QL) films in the order of PANI, graphene, PANI, and DWCNTs, all constituent chains and plates show a structural change, resulting in enhanced TE properties. PANI chains readily cover the DWCNTs, offering conductive networks that are well aligned. The large surface area of graphene serves to connect these PANI/DWCNT domains. With increased number of cycles of QL deposition, a 3D conductive polymer/carbon nanocomposite film with charge-transfer capabilities is formed. A maximum power factor of 1825 μW·m⁻¹·K⁻² is achieved for 40 QL films with α and σ of 130 μV·K⁻¹ and 1080 S·cm⁻¹, respectively. The increased electrical conductivity and Seebeck coefficient are attributed to the bridging of graphene layers with PANI-coated DWCNTs and the improved carrier mobility from the strong π–π interaction between PANI and DWCNTs. This study provided a new way to prepare nanocomposites with high TE performance, rivaling inorganic semiconducting materials.

In a more recent study, the TE properties of CNT/PANI composites were improved via tuning the interfacial interaction with amine functionalization of CNTs and controlling the camphorsulfonic acid (CSA) doping level of PANI.⁸⁴ The amine-functionalized CNT/PANI composites were prepared through in situ polymerization of PANI with a solution containing amine-functionalized CNTs and aniline monomers. The product was produced into powders through freeze-drying and was doped with CSA in m-cresol in different molar ratios. The resultant mixture was casted on a glass slide to produce amine-functionalized CNT/PANI composite films. The CNTs and PANI chains form well-connected conductive structures from the strong π–π interactions, resulting in TE properties that are superior to pristine CNTs or PANI. The electrical conductivity and Seebeck coefficient were improved monotonically, reaching the maximum value at 94 wt% amine-functionalized CNTs and decreasing afterward. The effect of enhanced TE properties stems from the highly conductive network due to the homogeneously coated PANI on amine-functionalized CNTs and filtering of low-energy carriers at the CNT/PANI interface. A power factor of 273 μW·m⁻¹·K⁻² was achieved solely for the effect of CNT/PANI composition with corresponding α and σ of ~37.0 μV·K⁻¹ and ~2012 S·cm⁻¹, respectively. After controlling the PANI/CSA mole ratio to 9:1, a maximum power factor of 401 μW·m⁻¹·K⁻² was achieved, which is one of the highest reported for organic TEs (Figure 8A).

Figure 8. (A) Electrical conductivity, Seebeck coefficient (top), and power factor (bottom) of 94 wt% A-CNT/PANI composites as a function of polyaniline (PANI)/camphorsulfonic acid (CSA) molar ratio. Reproduced with permission: Copyright 2018, Elsevier.84 (B) Schematic illustration of poly(3,4-ethylenedioxythiophene) (PEDOT):PF6/single-walled carbon nanotube (SWCNT) composite preparation procedure via dynamic three-phase interfacial electropolymerization (top) and optimized thermoelectric properties of PEDOT:PF6/SWCNT as a function of SWCNT content (bottom). Reproduced with permission: Copyright 2018, The Royal Society of Chemistry87

Composites with PEDOT

The TE performance of CNT/PEDOT composites is summarized in Table 4. Yu et al.⁸⁵ prepared PEDOT:PSS/CNT composite films by coating the CNTs with PEDOT:PSS particles to produce PEDOT:PSS/CNT junctions, enabling the electron transport but discouraging the phonon transport. The Seebeck coefficient is suggested to be maintained through the energy filtering effect of low-energy carriers, whereas the excellent electronic properties of CNTs enhance the electrical conductivity. The thermal conductivity remains low due to the intrinsically low thermal conductivity of conducting polymers. At a CNT content of 60 wt%, the PEDOT:PSS/CNT composite shows an optimized power factor of ~160 μW·m⁻¹·K⁻².

A multilayer system using nanometer-thick layers of MWCNT, on which PEDOT was electropolymerized, was prepared for TE applications by the same group.⁸⁶ To ensure the homogeneity of PEDOT and low contact resistance between the MWCNTs and PEDOT chains, an LbL assembly method was used to stack the MWCNTs into a film, and the MWCNT film was used as a working electrode for an in situ electrochemical polymerization process. The best TE performance was observed for 30 bilayers of MWCNTs stabilized with poly(diallyl dimethylammonium chloride) and sodium deoxycholate, respectively, and 30 min of polymerization, reaching a power factor of 155 μW·m⁻¹·K⁻² with a remarkable electrical conductivity of 2100 S·cm⁻¹. The improved TE properties stem from the connectivity of conductive MWCNTs and PEDOT chains.

To achieve PEDOT/SWCNT nanocomposites, three-phase interfacial electropolymerization was used as a route.⁸⁷ For the fabrication process, as shown in Figure 8B, PEDOT:PF₆ film was first prepared into a film and was crushed into an ethanol solution with SWCNTs, which was vacuum-filtered on porous nylon membranes. The composite films consist of well-dispersed PEDOT chains and SWCNTs due to the strong interaction between the two constituents. The PEDOT chains are also transformed from a coiled structure to an extended structure, facilitating charge transport. With increased SWCNT content, the electrical conductivity increases rapidly and decreases after 40 wt% SWCNTs, whereas the Seebeck coefficient continues to increase, reaching its maximum at 90 wt% SWCNTs. As a result, the power factor reaches its optimal value of 253.7 ± 10.4 μW·m⁻¹·K⁻² at 80 wt% SWCNT loading (Figure 8B). The same group further improved the TE properties of PEDOT/SWCNT composites using acid-doped SWCNTs (a-SWCNTs) and also controlling the type and content of PEDOT counterions.⁸⁸ Acid doping of SWCNT was done by mixing and refluxing the SWCNTs in a mixed-acid solution of sulfuric acid and nitric acid. The a-SWCNTs were vacuum-filtered into a film and a PEDOT:PF₆ film was prepared with three-phase electropolymerization, as done in the previous study.⁸⁷ The prepared PEDOT:PF₆ and a-SWCNT films were both crushed and mixed into an ethanol solution and vacuum-filtered on a nylon membrane. The acid doping of SWCNTs induces charged holes in the SWCNTs from the transfer of negative charges of the SWCNTs to the acids. This acid functionalization increases the carrier concentration and mobility and reduces the contact resistance between adjacent SWCNTs. As a result, the electrical conductivity increases with larger a-SWCNT content, and the Seebeck coefficient also increases. The peak power factor of 350.0 ± 47.6 μW·m⁻¹·K⁻² is reached at room temperature for PEDOT:PF₆/a-SWCNT films with 80 wt% a-SWCNTs.

Wang et al.⁸⁹ fabricated core–shell nanostructured PEDOT/CNT composites to achieve high TE power factors by coating the CNT bundles with PEDOT via an in situ chemical oxidation polymerization method. A PEDOT/CNT nanocomposite dispersion in methanol was obtained, and vacuum filtration was used to prepare PEDOT/CNT nanocomposite films. The strong interactions between the PEDOT chains and CNTs facilitate the formation of an organized conductive network. The TE power factor reached its highest value of 157 μW·m⁻¹·K⁻² at a CNT content of 67 wt%.

N-type TE materials were also prepared from PEDOT/CNT composites using tetrakis(dimethylamino)ethylene (TDAE) treatment.⁹⁰ CNT films were first prepared through spray coating on glass substrates and were immersed into EDOT monomer solution with FeCl₃ for in situ polymerization, forming PEDOT/CNT composite films. The CNT content within the composite films was controlled by differing the spray time of CNTs. TDAE treatment was conducted with the PEDOT/CNT composite placed at the top of a closure, and the TDAE was vaporized inside to treat the prepared sample. Upon TDAE treatment, the p-type characteristics of PEDOT/CNT films changed to n-type from electron injection. The carrier concentration of PEDOT chains was reduced, whereas the carrier mobility was improved from the high carrier mobility of CNTs after TDAE treatment. At a 10.7% concentration of CNTs, a high power factor of 1050 μW·m⁻¹·K⁻² was obtained. Combined with a low thermal conductivity of ~0.67 W·m⁻¹·K⁻¹, a large zT value of ~0.5 was reported at room temperature.

Composites with poly(3-hexylthiophene)

Very recently, Mardi et al.⁹¹ prepared nanocomposite films consisting of carbon nanotube forest (CNTF) and poly(3-hexylithiophene) (P3HT). CNTF acted as a filler and improved the electrical conductivity of P3HT by forming a well-interconnected network. The CNTF was transferred on a glass substrate, and the P3HT was spin-coated by the transferred CNTF. The TE properties were further optimized by doping the P3HT with lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) and tert-butylpyridine (TBP) with certain concentrations. Peak TE performance was obtained for P3HT/CNTF composite films with 24 μL LiTFSI/mL P3HT in solvent and 22.8 μL TBP/mL P3HT in solvent. The TE power factor of 110.0 μW·m⁻¹·K⁻² was obtained at 344.15 K, with corresponding α and σ of 91.9 μV·K⁻¹ and 130.2 S·cm⁻¹. This study is another example that has successfully demonstrated the importance of carbon-based materials for TE materials.

CNTs for flexible TE

Over the years, p-type and n-type CNT-based TE materials have been reported, introducing various molecular dopants, treatment methods, and polymers.^{72–74,76–78} However, there have not been many reports of TE generators made of both p-type and n-type CNT-based legs exhibiting high performance, as well as good flexibility and stability. Flexible CNT-based generators with high power density were prepared by An et al.,⁹⁹ consisting of p-type legs doped with 2,3,5,6,-tetrafluoro-7,7,8,8-tetracyanoquinodimethane (F4TCNQ) and n-type legs doped with BV dichloride. A single p-type and n-type leg had a web-like morphology and exhibited high power factors of 2252 and 3103 μW·m⁻¹·K⁻². The freestanding p-type and n-type CNT webs were attached on a flexible PET substrate and connected electrically in series using silver paste (Figure 9A). A flexible TE generator with 10 p–n couples was fabricated producing power of 7.1 μW at ΔT = 20 K, corresponding to output per weight of 28.3 μW·g⁻¹ and output per occupying area of 2.0 μW·cm⁻² (Figure 9A). When the area was normalized to the leg cross-sectional area, a high power density of 1180 μW·cm⁻² was obtained. The CNT webs were also stable after long ambient exposure and repeated folding and crumpling cycles.

Figure 9. (A) Photographic image of a flexible thermoelectric device (TE) device (left). Power output–current and voltage–current plots of 20-leg flexible TE device using F4TCNQ-doped CNT webs as p-type, and BV-doped CNT webs as n-type. ∆T = 20°C (top right). Maximum power density depending on temperature gradient (bottom right). Reproduced with permission: Copyright 2017, The Royal Society of Chemistry.99 (B) Photographic image of a carbon nanotube (CNT) stripe composed of three p–n junctions (left) and voltage–current and power–current plot of TE device with a TH of 330 K and ∆T of 27.5 K (right). Reproduced with permission: Copyright 2017, Nature Publishing Group.100 (C) Power density of a flexible TE device with varying number of p–n junctions at ∆T = 5 K (left). Human body heat harvesting of flexible TE device (right). Reproduced with permission: Copyright 2017, American Chemical Society.101 (D) Schematic illustration of TE device fabrication procedure using p-type and n-type CNT foams (top left). Bending of TE device demonstrating the flexible nature of the device (top right). Voltage–current and power–current plots of TE device with different ∆T (bottom left). Output power and power density (per area, weight) with different ∆T (bottom right). Reproduced with permission: Copyright 2019, Wiley-VCH102

A flexible and compact TE generator was prepared by Zhou et al.,¹⁰⁰ which showed high performance, air stability, and high scalability. In the study, SWCNT films were synthesized directly using FCCVD and transferred on a PET substrate. Controlling the FCCVD parameters, p-type SWCNTs with power factors of 1840 μW·m⁻¹·K⁻² were fabricated. For n-type counterparts, branched PEI was drop-casted on the sample, exhibiting power factors of ~1500 μW·m⁻¹·K⁻². To create TE generators, SWCNT films were stacked to produce large-area, thick SWCNT films, which were cut into 96 mm × 10 mm stripes. The cut SWCNT stripe was transferred on a PET substrate and shaded with adhesive tapes to mask regions of the stripe that are aimed for p-type legs. The unmasked regions were doped with PEI to produce n-type legs. Silver paste was applied to both ends of the stripe for use. A compact structure was created by simply folding the PET substrate along the edges of the adhesive tape, serving as the boundary marks for the p-type and n-type legs (Figure 9B). The TE generator consists of three p–n couples, creating an open-circuit voltage V_oc of 11.3 mV and short-circuit current I_sc of 0.9 mA at ΔT of 27.5 K (Figure 9B). The maximum power of 2.51 μW and a small internal resistance R₀ of 12.5 Ω were obtained. Normalized to the cross-sectional area of the generator, power density of 167 μW·cm⁻² was reported. This TE generator with a compact structure shows how large-scale TE generators can be fabricated by eliminating the needs for metal contacts and improving the bulky structure in conventional π-type configurations.

All-carbon nanotube yarn (CNTY)-based flexible TE generators were reported by Choi et al.,¹⁰¹ which showed exceptional power density of 697 μW·g⁻¹ at ΔT of 40 K. The CNTYs were synthesized using a floating catalyst method and wound around a polydimethylsiloxane (PDMS) support with dimensions of 41,080 mm³. The CNTYs were doped with FeCl₃ and PEI solutions for p-type and n-type doped regions, respectively, and CNTYs between the p-type and n-type regions were left undoped to be used as electrodes, eliminating the need for metal contacts and minimizing the contact resistance. A TE generator comprised of 60 p–n pairs was fabricated using p- and n-doped CNTYs with high power factors of 2387 and 2456 μW·m⁻¹·K⁻². The TE generator exhibits high power density and good flexibility, and it is able to harvest energy from the temperature gradient produced by body heat (Figure 9C). This type of TE generator shows promise for flexible and wearable energy harvesting with great scalability.

More recently, a flexible TE generator utilizing a conventional π-leg configuration was prepared by Lee et al.¹⁰² by using porous CNT foams in a PDMS mold. SWCNT slurries were doped with FeCl₃ and BV for p- and n-type doping, respectively. The doped SWCNT slurries were poured into a PDMS mold and subjected to reduced pressure to form porous foams. Silver paste was applied to connect the p-type and n-type foams. The TE generator made of CNT foams shows good flexibility without a significant increase in internal resistance after repeated bending cycles. The bulk structure and large porosity help retain the applied vertical-direction temperature gradients, exhibiting a maximum power output of 1.5 μW at a ΔT of 13.9 K. A high output power by weight of 82 μW·g⁻¹ is obtained, accordingly (Figure 9D). This method of flexible TE generator fabrication can solve the problem of bulk scalability that most organic TE materials encounter. Moreover, the CNT foam can be applied to various heat sources owing to its free shaping capability.

Graphene Graphene doping and treatment for TE property enhancement

The TE performance of treated graphene is summarized in Table 5. One of the early studies related to TE property enhancement was conducted by Xiao et al.¹⁰³ They prepared few-layer graphene (FLG) films on Cu foils by a CVD method, and the films were further treated directly with oxygen plasma for 10, 15, and 20 s. As shown in Figure 10A,B, the pristine FLG film exhibits an ordered lattice structure. However, oxygen plasma treatment caused the FLG film to take a more disordered lattice structure with higher structural defects that induce π–π* gap opening, thereby improving TE properties of FLG films. Also, 15 s of oxygen plasma treatment showed the optimal TE performance of FLG films; Seebeck coefficient was improved from ~80 to ~700 μV·K⁻¹ after 15 s of oxygen plasma treatment of pristine FLG films, resulting in a power factor of ~4.5 × 10⁻³ W·m⁻¹·K⁻².

Table 5 Reported enhanced thermoelectric performance of graphene using various doping methods

Method	Seebeck coefficient (μV·K−1)	Electrical conductivity (S·cm−1)	Power factor (μW·m−1·K−2)	Thermal conductivity (W·m−1·K−1)	zT	Sample (thickness/diameter)a	References
Oxygen plasma treatment of few-layer graphene	~700	~100	4500			Film (nm)	[103]
Annealing under N2 atmosphere and PEI doping of rGO	−16.3	2.83	0.0752			Film (nm)	[104]
Bromine doping of graphene fibers	~35	~5100	624	~88	~2.13 × 10−3, 2.76 × 10−3 (350 K)	Fiber (μm)	[105]

Note: Thermoelectric properties not directly mentioned in the manuscript of each reference were estimated based on the plots of each parameter (α, σ, power factor, κ, zT); unless specified otherwise in parentheses, all values are reported at room temperature.

Abbreviations: FLG, few-layer graphene; PEI, polyethylenimine; rGO, reduced graphene oxide.

a

“nm” and “μm” categorize sample; nm: t < 1 μm and μm: 1 μm ≤ t < 1 mm, and mm: 1 mm ≤ t.

Figure 10. (A) High-resolution transmission electron microscopy images and selected area diffraction patterns of few-layer graphene (FLG) films before oxygen plasma treatment. Reproduced with permission: Copyright 2011, American Chemical Society.103 (B) High-resolution transmission electron microscopy images and selected area diffraction patterns of FLG films after oxygen plasma treatment. The small crystals of carbon are shown in yellow circles, and the disordered arrangement of carbon atoms is shown in red circles. Reproduced with permission: Copyright 2011, American Chemical Society.103 (C) Electrical conductivity and Seebeck coefficient of as-prepared rGO, annealed rGO under N2 atmosphere, and restored rGO. Reproduced with permission: Copyright 2017, Elsevier.104 (D) Thermal conductivity (left) and electrical conductivity (right) of graphene fibers in the presence of bromine doping as a function of temperature. Reproduced with permission: Copyright 2018, Springer Nature105

In 2016, Tu et al.¹⁰⁴ prepared reduced graphene oxide (rGO) film by obtaining graphene oxide (GO) using the modified Hummers' method from commercially available expandable graphite and chemically reducing the as-prepared GO using hydrazine. The rGO posed p-type character with Seebeck coefficient of 16.9 μV·K⁻¹, which was induced by hole doping derived from water and oxygen molecules adsorbed on the surface of rGO. The researchers further annealed the rGO film at 423 K under N₂ atmosphere for 1 h. As a result, water and oxygen molecules adsorbed unintentionally were removed, giving n-type character in which Seebeck coefficient is −4.87 μV·K⁻¹. Despite annealing rGO film under N₂ atmosphere, it exhibited p-type character when exposed to moisture environment for 5 days. The changes in conductivity and Seebeck coefficient for as-prepared rGO, annealed rGO under N₂ atmosphere, and restored rGO are shown in Figure 10C. Finally, the rGO film was doped with PEI by soaking rGO film into 1, 5, and 20 mg·mL⁻¹ PEI solution in ethanol. The electron-donating group of primary, secondary, and tertiary amines from PEI induced n-type doping as surface electrons are transferred from PEI to rGO. Optimal n-type doping of graphene with PEI was achieved at 5 mg·mL⁻¹ PEI solution, where α and σ are −16.3 μV·K⁻¹ and 2.83 S·cm⁻¹, respectively. Moreover, PEI-doped graphene maintained its n-type character for 3 weeks.

Another approach to enhance the TE performance of rGO was employed by Bark et al.¹⁰⁶ The researchers reduced bulk GO film at 773, 973, 1173, and 1373 K for 2 h under Ar atmosphere. It was found that Seebeck coefficient changed its sign from negative to positive as the reduction temperature increased. They have shown that the sign and the absolute value of Seebeck coefficient are determined from the difference between the energy level of charge neutrality point and the Fermi level; as reduction temperature increases, the energy level of charge neutrality point shifts toward Fermi level, which gives a perfect sp² network of graphene. Finally, the doping effect was found to be derived from the healing of structural defects and elimination of residual oxygen in graphene as temperature increases.

In 2017, Ma et al.¹⁰⁵ prepared bromine-doped graphene fibers by using a two-zone vapor transport method. The presence of bromine doping strengthened phonon scattering due to more defects, and thermal conductivity reduction was driven from higher concentration of defects. Furthermore, doping with bromine lowered the Fermi level, increasing electrical conductivity and Seebeck coefficient. The changes in thermal conductivity and electrical conductivity with the introduction of bromine doping as a function of temperature are shown in Figure 10D. Finally, four times higher zT value of 2.76 × 10⁻³ for bromine-doped graphene fibers as compared with undoped graphene fibers was obtained, followed by power factor of 624 μW·m⁻¹·K⁻² at room temperature.

Graphene nanocomposites for TE property enhancement Composites with inorganic materials

The TE performance of graphene/inorganic material composites is summarized in Table 6. In 2013, Liang et al.¹⁰⁷ prepared a nanostructured graphene/Bi₂Te₃ composite with 0, 0.1, 0.2, and 2 vol% content of graphene by hydrothermal synthesis and a spark plasma sintering method. Graphene/Bi₂Te₃ composite with 0.2 vol% of graphene content showed 31% enhanced zT value of 0.21 at 475 K as compared with pristine Bi₂Te₃ due to the improvement in Seebeck coefficient and reduction in thermal conductivity. Similarly, Agarwal et al.¹⁰⁸ prepared graphene/Bi₂Te₃ composite by mixing 0.05 wt% of graphene in Bi₂Te₃ powder using an agate mortar and pestle, followed by pressing into pellets. The zT value of graphene/Bi₂Te₃ with 0.05 wt% of graphene reached up to 0.92, compared with 0.68 at 402 K for pristine Bi₂Te₃. The enhanced zT value is due to the reduction in thermal conductivity and improvements in carrier mobilities and Seebeck coefficient. Presence of graphene in Bi₂TE₃ matrix induced filtration of low-energy charge carriers and enhancement in phonon scattering which further increased the Seebeck coefficient and reduced thermal conductivity. Furthermore, high carrier mobilities of graphene were attributed to the improvement in carrier mobilities, thereby enhancing the zT value.

Table 6 Reported enhanced thermoelectric performance of graphene in various composite systems

Material	Seebeck coefficient (μV·K−1)	Electrical conductivity (S·cm−1)	Power factor (μW·m−1·K−2)	Thermal conductivity (W·m−1·K−1)	zT	Sample (thickness/diameter)a	References
Graphene/Bi2Te3	~−113 (475 K)	770 (475 K)	~983 (475 K)	~2.25 (475 K)	0.21 (475 K)	Pellet (mm)	[107]
Graphene/Bi2Te3	~−144 (402 K)	~970 (402 K)	2011.4 (402 K)	~0.88 (402 K)	0.92 (402 K)	Pellet (mm)	[108]
Bi2Te3/GQD nanosheet	−145 (425 K)	~420 (425 K)	890 (425 K)	~0.65 (425 K)	0.55 (425 K)	Pellet (mm)	[109]
CoSb3/graphene	~140	~400	790 (800 K)	~1.13 (800 K)	0.61 (800 K)	Pellet (mm)	[110]
BST/expanded graphene	~220	~650	~3200	~0.8	~1.18–1.24 (360 K)	Ingot (mm)	[111]
Graphene/BiSbTe	177.5		~4700	~1.1	1.29–1.54 (440 K)	Pellet (mm)	[112]
Graphene/BiSbTe	~168 (323 K)		3340 (323 K)	~0.95 (323 K)	1.26 (423 K)	Bulk (mm)	[113]
PANI/graphene	~32	~55	5.60			Pellet (mm)	[114]
PANI/graphene			2.6		1.95 × 10−3	Pellet (mm)	[115]
GNP/PANI	33.8	123	14			Pellet (mm)	[116]
PANI/graphene	26	~814	55			Film (μm)	[117]
PDG/PANI	~604	~14.5	~529	~0.22	0.74	Tablet (mm)	[118]
3D graphene/PANI	45.6	394.1	81.9			Film (μm)	[119]
GNPs/ethyl cellulose	~18 (332 K)	3.554 (332 K)	0.254 (332 K)			Film (μm)	[120]

Note: Thermoelectric properties not directly mentioned in the manuscript of each reference were estimated based on the plots of each parameter (α, σ, power factor, κ, zT); unless specified otherwise in parentheses, all values are reported at room temperature.

Abbreviations: GNP, graphene nanoplatelet; GQD, graphene quantum dot; PANI, polyaniline; PDG, p-phenediamino-modified graphene.

a

“nm” and “μm” categorize sample; nm: t < 1 μm and μm: 1 μm ≤ t < 1 mm, and mm: 1 mm ≤ t.

Further research has been conducted to improve the TE properties of Bi₂Te₃-based graphene nanocomposites by changing material dimensions from micrometer to nanometer. For example, Li et al.¹⁰⁹ demonstrated a simple method for fabricating graphene Bi₂Te₃/GQDs hybrid nanosheets in which a Bi precursor solution containing Bi(NO₃)₃·5H₂O, vitamin C, and GQDs solution was injected rapidly into the Te precursor solution containing TeO₂ and NaOH. The schematic drawing of Bi₂Te₃/GQDs formation by a solution synthesis process is shown in Figure 11A. As GQDs contain oxygen-containing functional groups of –OH and –COOH on the surface, the functional groups act as heterogeneous nucleation centers; the Bi ion is easily absorbed on the oxygen-containing functional groups on the surface of GQDs by tight chemical bonding, facilitating further selective growth of Bi₂Te₃ on surfaces of GQDs. The O-doping at the interface of Bi₂Te₃/GQDs increases the concentration of Te vacancies, giving more electrons. The excess charge carriers from the charged Bi₂Te₃/GQDs interface could be injected into the Bi₂Te₃ grain core, contributing to the carrier concentration increase. Consequently, increased Coulombic barriers from the charged Bi₂Te₃/GQDs interface lead to selective scattering of holes over electrons. Furthermore, maximum reduction of lattice thermal conductivity is induced by the enhanced phonon scattering, targeting the wide spectrum of phonons derived from atomic-scale defects and grain boundaries at high-density Bi₂Te₃/GQDs interface. The maximum power factor of 890 μW·m⁻¹·K⁻² and the maximum zT value of 0.55 were obtained from the hybrid nanosheet of Bi₂Te₃ with 4 mg·mL⁻¹ GQDs solution and 20-nm GQDs at 425 K, respectively. The TE properties of Bi₂Te₃/GQDs are shown in Figure 11B.

Figure 11. (A) Schematics of Bi2Te3/graphene quantum dot (GQD) hybrid nanosheet formation. Reproduced with permission: Copyright 2017, American Chemical Society.109 (B) Temperature dependence of Seebeck coefficient and total thermal conductivity of Bi2Te3/GQDs hybrid nanosheets with different GQD sizes. Reproduced with permission: Copyright 2017, American Chemical Society.109 (C) Schematic diagrams of polyaniline (PANI)//graphene (GP) formation by mixing and by in situ polymerization. Reproduced with permission: Copyright 2015, The Royal Society of Chemistry.117 (D) Seebeck coefficient and power factor of PANI/GP based on graphene contents. Reproduced with permission: Copyright 2015, The Royal Society of Chemistry117

Antimony-based composites also have attracted researchers' attention toward TE materials. Feng et al.¹¹⁰ fabricated p-type nanostructured CoSb₃/graphene composite with 1.5 wt% of graphene by solvothermal method and hot pressing method. Dispersion of graphene in CoSb₃ homogeneously gives more interfaces between CoSb₃ and graphene that enables phonons to be scattered and thermal conductivity to be reduced. CoSb₃/graphene reached the maximum power factor of 790 μW·m⁻¹·K⁻² and zT value of 0.61 at 800 K, whereas pristine CoSb₃ reached the zT value of 0.26 at 700 K.

Following that, Suh et al.¹¹¹ invested TE properties of p-type BST for potential TE materials and optimized TE properties of BST by the introduction of graphene. BST was prepared by spark plasma sintering of Bi₂Te₃ and Sb₂Te₃ nanoplates, and ball milling was used to get a uniform BST powder with 0.5–2.5-μm size. BST was mixed with 0–0.5 vol% of ball-milled EG, and further spark plasma sintering was performed to form BST/EG composites. The enhancement of TE properties by graphene introduction is achieved from the high intrinsic carrier concentration of p-type graphene, thereby increasing electrical conductivity and carrier concentration. Increased carrier concentration suppresses the bipolar conduction as temperature increases, leading to a slight decrease in Seebeck coefficient, increase in power factor, and a slight increase in bipolar thermal conductivity. The maximum zT value of 1.13 and power factor of ~2400 μW·m⁻¹·K⁻² were measured at BST–0.1 vol% EG at 360 K. The researchers further prepared ingot BST–EG by using ball-milled BST ingot powders. The ingot BST–0.1 vol% EG reached maximum zT value of 1.24 at 360 K and a power factor of ~3200 μW·m⁻¹·K⁻² at 300 K. Li et al.¹¹² and Zhang et al.¹¹³ further researched the composite of graphene and Bi_0.4Sb_1.6Te₃ (BiSbTe). Li et al.¹¹² prepared graphene/BiSbTe by blending (Bi₂Te₃)_0.2(Sb₂Te₃)_0.8 powders into the colloidal dispersions of graphene, whereas Zhang et al.¹¹³ prepared graphene-composited BiSbTe bulks by high-pressure and high-temperature method, followed by high-pressure sintering. Li et al.¹¹² reached the maximum zT value of 1.54 at 440 K for 0.3 vol% of embedded graphene, and Zhang et al.¹¹³ reached the maximum zT value of 1.26 and the maximum power factor of 3340 μW·m⁻¹·K⁻² for BiSbTe bulks with 0.05 wt% graphene at 423 and 323 K, respectively.

Composites with PANI

The TE performance of graphene/PANI composites is summarized in Table 6. PANI/graphene nanocomposite has been widely studied for organic TE materials. Du et al.¹¹⁴ investigated the TE properties of PANI/graphene pellets and films by varying PANI-to-graphene weight ratio from 4:1 to 1:1. PANI/graphene pellets were prepared by drying the mixed solution containing N-methyl-2-pyrrolidone (NMP), PANI powder, and graphene under vacuum and cold pressing the as-prepared powder. PANI/graphene films were prepared by drying the mixed solution on a glass substrate. The power factors of PANI/graphene pellets and films reached 5.60 and 1.47 μW·m⁻¹·K⁻² for 1:1 PANI-to-graphene weight ratio, respectively. Lu et al.¹¹⁵ prepared composites of protonated PANI and graphene by an in situ chemical polymerization method. Graphene and aniline monomer were added to HCl and oxidant ammonium peroxydisulfate and kept under room temperature to generate PANI/graphene composites. Appropriate amounts of graphene were added to get composites of graphene-to-PANI ratio of 0.05:1, 0.15:1, 0.2:1, and 0.3:1. The PANI/graphene composite shows degradation of electrical conductivities as temperature increases. Graphene adsorbed on PANI changes morphology of PANI from a twisted structure into extended structures, resulting in higher carrier mobility. Moreover, the increase in the electrical conductivity is caused by the interchain carrier transport bridged by graphene through a tunneling mechanism and the interaction with PANI. The maximum power factor of 2.6 μW·m⁻¹·K⁻² and the maximum zT of 1.95 × 10⁻³ were achieved by PANI with 30 wt% graphene. Soon after, Abad et al.¹¹⁶ investigated the preparation of exfoliated graphene nanoplatelets (GNPs/PANI) nanocomposite. Abad et al.¹¹⁶ prepared the GNPs/PANI nanocomposites by mechanical blending of synthesized PANI solution with commercially available GNPs, followed by cold pressing into a solid pellet. GNPs/PANI composite containing 50 wt% GNPs showed enhanced TE properties; the electrical conductivity increased from 0.48 to 123 S·cm⁻¹ and the Seebeck coefficient increased from −2.6 to +33.8 μV·K⁻¹ after the incorporation of 50 wt% GNPs. The enhancement in the electrical conductivity is due to the percolation law; the electrical conductivity largely increases when the critical volume fraction of filling material is established. Furthermore, the enhancement in Seebeck coefficient was due to energy filtering of carriers. The GNPs/PANI nanocomposites consisting of 50 wt% GNPs reached a maximum power factor of 14 μW·m⁻¹·K⁻².

Wang et al.¹²¹ discovered how structural defects and oxygen content affect the TE properties of the PANI/graphene composite. PANI was first doped with CSA at 2:1 mole ratio and dissolved in m-cresol for conformational change from a compacted to an expanded coil. Then, appropriate amounts of three types of graphene were added, respectively, and mixed by solid-state grinding. Next, they were dropped on glass substrates and dried to obtain PANI/graphene composite films. Three types of graphene are as follows: graphene 1 (GP1), graphene 2 (GP2), and graphene 3 (GP3), with ~10–50-μm diameter and 0.97 wt% oxygen content, ~0.5−20-μm diameter and 1.24 wt% oxygen content, and 0.5–2-μm diameter and 10.17 wt% oxygen content, respectively. Addition of GP1 resulted in tighter adsorption of PANI on the surface of GP1 due to the improved π–π conjugation interactions between GP1 and PANI, further expanding the PANI molecular chains. Moreover, ordered chain packing from expanded PANI molecular conformation results in less conjugation defects along the PANI molecular chain, increasing the carrier mobility. However, structural defects and oxygen contents on the surface of GP2 and GP3 result in a weaker interaction between graphene and PANI, lowering electrical conductivity. Therefore, higher electrical conductivity was obtained from PANI/GP1. In addition, PANI/GP1 composite shows a slightly lower Seebeck coefficient, compared with PANI/GP2 and PANI/GP3 composites, due to the poor dispersion of GP1 in PANI, which is caused by stronger van der Waals attraction of graphene, owing to low oxygen impurity and structural defect. As a result, the maximum power factor of 19 μW·m⁻¹·K⁻² and electrical conductivity of 856 S·cm⁻¹ were measured for the in-plane PANI/GP1 composite film at graphene content of ~45 wt%. Wang et al.¹¹⁷ further synthesized PANI/graphene composite powder by in situ polymerization of aniline with the addition of GP1 with the oxidant of ammonium peroxydisulfate. Also, the composite powder was doped with CSA at 2:1 mole ratio and dissolved in m-cresol, followed by dropping on glass substrates and drying, similar to their previous work. Compared with their previous work, PANI/graphene fabricated by in situ polymerization had enhanced dispersion of graphene in the PANI matrix, forming a more homogeneous structure. A schematic drawing of synthesis of PANI/GP by mixing and in situ polymerization and their corresponding TE properties are represented in Figure 11C,D. The homogeneous structure prevented graphene aggregation and weakened van der Waal interaction between graphene sheets. As a result, enhanced graphene dispersion led to more nano-interfaces and improved π–π conjugation interactions between graphene and PANI, forming more ordered and expanded molecular conformation, which improved the Seebeck coefficient. Consequently, the PANI/graphene at 48% graphene content exhibited enhanced in-plane power factor and Seebeck coefficient of 55 μW·m⁻¹·K⁻² and 26 μV·K⁻¹, respectively, due to molecular ordering.

In 2018, Lin et al.¹¹⁸ synthesized the p-phenediamino-modified graphene (PDG)/PANI composite for high-performance TE material. PDG was synthesized by mixing equal volumes of p-phenylenediamine solution and graphene solution dissolved in ethanol, respectively, and extracting the solvent. Then, it was dried in an oven within the temperature range of 30–80°C for 24 h. The modified graphene was then dissolved in HCl, followed by the addition of aniline monomer, and further dried in an oven. PDG/PANI composite contains semi-interpenetrating networks of PANI that provide pathways for carriers to move through the composites. From the semi-interpenetrating networks, highest electrical conductivity of 68.5 S·cm⁻¹ and highest zT value of 0.74 were measured from PDG/PANI composites consisting of 15 wt% graphene that was dried under 80°C and 3 wt% graphene that was dried under 30°C, respectively.

Recently, Hsieh et al.¹¹⁹ fabricated flexible 3D graphene/PANI composite for p-type TE material. 3D graphene was fabricated by CVD of graphene and integrated with PANI by in situ electrochemical polymerization process. Maximum TE properties of 81.9 μW·m⁻¹·K⁻², 45.6 μV·K⁻¹, and 394.1 S·cm⁻¹ were obtained at 80 wt% PANI at room temperature. The high TE properties are achieved due to the interconnected graphene flakes in the monolithic structure of 3D graphene, which enable fast electron transport within the entire macrostructure. In addition, PANI polymer disperses into 3D graphene evenly due to the inherited porous structure of 3D graphene, inducing stronger π–π interaction between graphene and PANI. Moreover, the in situ polymerization of PANI on graphene enables intimate adhesion, which further enhances π–π interaction between 3D graphene and PANI. Due to the enhancement of π–π interaction between 3D graphene and PANI and the electron energy filtering effect driven by uniform PANI dispersion, high TE properties of 3D graphene/PANI composite have been obtained.

Studies on composites consisting of graphene and other conductive polymers such as polyvinyl acetate, polypyrrole, and polyvinylidene fluoride are still ongoing.^122–125 Recently, Mardi et al.¹²⁰ reported GNPs/ethyl cellulose nanocomposites for TE. Ethyl cellulose is a nonconducting polymer that is a cheap, abundant, green, and biocompatible material. The prepared GNP pastes consisted of ethyl cellulose as polymer matrices and GNPs as fillers, and GNPs/ethyl cellulose nanocomposites were formed by blade-coating GNP pastes on glass substrate. They have optimized the GNP amounts of GNP pastes and obtained maximum electrical conductivity of 3.554 S·cm⁻¹ and 0.254 μW·m⁻¹·K⁻² at 7:10 weight ratio of GNPs and ethyl cellulose at 330 K. Also, they prepared GNPs/ethyl cellulose nanocomposite pellets by dropping GNP pastes into a pellet maker and cold pressing into a pellet. Furthermore, the maximum output powers of 16.2 and 40.1 nW were generated from the pellets at temperature differences of 20.5 and 32.3 K, respectively. The authors have taken the first step of using easily processable, low-cost, and sustainable polymers in TE materials for commercialization in the future.¹²⁰ The TE properties of carbon allotropes (CNT, graphene) that were discussed above are summarized in Figure 12.

Figure 12. Thermoelectric properties of carbon allotropes (CNT, graphene) with various strategies. CNT, carbon nanotube; PANI, polyaniline; PEDOT, poly(3,4-ethylenedioxythiophene)

Graphene for flexible TE

Graphene is considered an attractive material for researchers, not only for its electrical conductivity but also for its flexibility. Graphene devices are widely researched for wearable TE devices and smart textiles due to its superior mechanical properties and high electrical conductivity. In addition to the studies of Tu et al.¹⁰⁴ mentioned above, they further fabricated TE device based on n-type rGO film (16.9 μV·K⁻¹) and p-type PEI–rGO film (−16.3 μV·K⁻¹). The flexible device contained electrically connected 24 pairs of n-type rGO film and p-type PEI–rGO film. The device exhibited output voltage of 11, 21, and 26 mV at the temperature difference of 23, 35, and 50 K, respectively. Furthermore, maximum power of 1.36 nW was obtained at the temperature difference of 50 K.

In 2018, Juntunen et al.¹²⁶ investigated the preparation of graphene films using large-area inkjet printing for further use in flexible TE devices. Graphene dispersions were obtained by exfoliation of bulk graphite by ultrasonic-assisted liquid-phase exfoliation in isopropyl alcohol (IPA) with polyvinylpyrrolidone (PVP). IPA helps fast drying, whereas PVP helps to stabilize the graphene dispersion. The exfoliated graphene flake dispersion has an estimated graphene concentration of 0.40–0.93 mg·mL⁻¹, and 50% of graphene flakes are perceived to have <10-mm thickness with an average lateral size of ≈200 nm. Graphene dispersion, 2 mL, was then spray-coated on glass. A maximum power factor of 18.7 μW·m⁻¹·K⁻² has been obtained with outstanding thermostability against 385-K temperature. The patterns of inkjet-printed graphene on PET substrate are shown in Figure 13A,B. The researchers further built a device comprising 20 inkjet-printed graphene legs, which are connected by inkjet-printed silver, and its thermovoltage response depending upon temperature gradient is represented in Figure 13C. Furthermore, the device showed durability against over 10,000 bending cycles of mechanical deformation. The study showed that inkjet-printed large-area graphene has high potential to be used in TE devices, not only due to the low cost and scalability, but also due to its promising TE properties, thermostability, and durability.

Figure 13. (A) Photographic image of graphene patterns produced by inkjet printing. Reproduced with permission: Copyright 2018, Wiley-VCH.126 (B) Photographic images of a flexible device comprising of 20 inkjet-printed graphene legs connected by inkjet-printed silver. Reproduced with permission: Copyright 2018, Wiley-VCH.126 (C) Thermovoltage response of flexible thermoelectric (TE) device depending on temperature gradient. Reproduced with permission: Copyright 2018, Wiley-VCH.126 (D) Power output of a device consisting of 10 couples of legs of n-type rGO/Bi2Te3 and p-type SWCNT/Sb2Te3 on polyimide substrate. Reproduced with permission: Copyright 2019, Wiley-VCH.127 (E) Photographic image of a potential use of a flexible, wearable TE device. Reproduced with permission: Copyright 2019, Wiley-VCH.127 (F) Photographic image of a photovoltaic (PV)–TE device attached to a T-shirt. Reproduced with permission: Copyright 2019, Wiley-VCH127

A 3D graphene/PANI composite fabricated by Hsieh et al.¹¹⁹ showed high potential to be used as a wearable device that harvests energy. For wearable device applications, thermoelements of millimeter-level thickness are required in order to generate a power output exceeding 100 uW. 3D graphene is easily fabricated with thickness larger than 100 μm, and by stacking up, a sufficiently thick film for TE devices can be obtained. Tensile strength of 3D graphene reached 44 Mpa and elongation at break of 1.9. Mechanical strength of composite decreased with greater amounts of PANI, due to the weaker mechanical strength of PANI. The researchers further tested TE properties of 3D graphene/PANI after bending to a 2-mm radius and releasing to flat up to 1500. The power factor saturation was achieved at ~57 μW·m⁻¹·K⁻² after the bending test. 3D graphene/PANI has high potential to be used as a wearable device, not only due to the scalable production of 3D graphene/PANI and sufficient thickness, but also due to its flexibility and mechanical properties.

In 2019, all-inorganic hybrid TE device with high TE performance was developed by Wu et al.¹²⁷ They fabricated n-type rGO and Bi₂Te₃ composite by adding appropriate Bi₂Te₃ nanoplates into graphene dispersed in deionized water, followed by annealing at 773 K under NH₃ atmosphere. The n-type rGO/Bi₂Te₃ showed optimal TE properties at 6 wt% of rGO content: 108 μW·m⁻¹·K⁻² for power factor and 3.5 × 10⁻³ for zT value. In addition, the researchers fabricated p-type SWCNTs/Sb₂Te₃ composite. SWCNTs and sodium 4-dodecylbenzenesulfonate were dissolved into ultrapure water, followed by sonication for 3 h in an ice water bath and addition of appropriate amount of Sb₂Te₃ nanoplates into the solution. The solution was finally annealed at 673 K under Ar atmosphere to prepare p-type SWCNTs/Sb₂Te₃. The p-type SWCNTS/Sb₂Te₃ exhibited optimal TE properties at around 16 wt% of SWCNTs content: 55 μW·m⁻¹·K⁻² for power factor and 1.6 × 10⁻³ for zT value. They finally fabricated a wearable TE device with flexibility by connecting 10 couples of legs of n-type rGO/Bi₂Te₃ and p-type SWCNTS/Sb₂Te₃ on polyimide substrate. The flexible device showed around 135 mV of voltage and 23.6 μW of power output with the external load of 210 Ω when the temperature gradient was 70 K (Figure 13D). The device showed a rapid response to the temperature with uniform voltage and current responses, which implies that the device is stable for practical usage. Moreover, flexible photovoltaics and TE (PV–TE) integrated device was fabricated by attaching four pairs of n-type rGO/Bi₂Te₃ and p-type SWCNTS/Sb₂Te₃ on the back of photovoltaic cell to the hot end and binding an insulating polyethylene foam to the cold end, followed by placement of aluminum foil on the polyethylene foam to minimize the cold-end temperature. The PV–TE device showed a rapid response to heat and maximum thermally induced voltages of 8.0 and 13.7 mV when irradiated under AM 1.5 and AM 0, respectively. Furthermore, the PV–PE device showed ~2% higher photoelectric conversion efficiency as compared with the PV cells from the increased fill factor derived by heat removal. Moreover, the PV–TE device showed a stable current density after 1200 s; however, PV cells showed a slight decrease in current density after 600 s. This study showed great commercial potential of all-inorganic hybrid film TE devices for capturing low-grade heat, as well as incorporation with PV cells. A photographic image of the potential usage of a PV–PE device is shown in Figure 13F.

Very recently, Novak et al.¹²⁸ prepared a TE device consisting of all-inorganic graphene with high power factors. First, they prepared non-oxidized graphene flakes (NOGF) by mixing raw graphite in a potassium naphthalenide solution with 30 min of bath sonication in DMSO. Furthermore, PVP or pyrenebutyric acid (PBA) were added for the functionalization of NOGF from adsorbed organic molecules, thereby posing n-type and p-type NOGF. PBA–NOGF poses p-type characteristics due to the electron-withdrawing group of the carboxylic group, and PVP–NOGF poses n-type characteristics due to the electron-donating group of amine group. The high HOMO (highest occupied molecule orbital) level of PVP further results in electron donation in graphene. At 313 K, PBA–NOGF exhibited electrical conductivity of 2330 S·cm⁻¹, Seebeck coefficient of 53.1 μV·K⁻¹, and power factor of 655 μW·m⁻¹·K⁻², and PVP–NOGF exhibited electrical conductivity of 3010 S·cm⁻¹, Seebeck coefficient of 45.5 μV·K⁻¹, and power factor of 621 μW·m⁻¹·K⁻². Finally, p-type PBA–NOGF and n-type PVP–NOGF were combined to construct an all-graphene TE device. At an approximate temperature gradient of 50 K, the device exhibited the maximum power output of 5.0 nW at 2.7 mV.

Organic–inorganic hybrid halide perovskite Perovskite doping and treatment for TE property enhancement

Organic–inorganic hybrid halide perovskite of CH₃NH₃PbX₃ and CH₃NH₃SnX₃, where X = I, Br, Cl, started to be considered as a promising TE material due to its high Seebeck coefficient and low thermal conductivity.¹²⁹ The TE performance of organic–inorganic hybrid halide perovskites is summarized in Table 7. Mettan et al.¹³⁰ synthesized CH₃NH₃PbI₃ (MAPbI₃) and CH₃NH₃SnI₃ (MASnI₃) and enhanced the TE properties of MAPbI₃ by photoinduced doping and MASnI₃ by chemical doping. MAPbI₃ crystals were obtained by the precipitation of the solution of hydriodic acid and lead(II) acetate trihydrate. CH₃NH₂ and MASnI₃ crystals were obtained by the solution containing hydriodic acid, tin(IV) iodide, and different amounts of CH₃NH₂. The study showed a dramatic decrease in electrical resistivity under the higher white-light illumination of 80, 165, and 220 mW·cm⁻² for MAPbI₃, because the white-light illumination induces an increase in the photoexcited carriers' number. However, Seebeck coefficient decreased from 0.82 to 0.54 mV·K⁻¹ under 220 mW·cm⁻² of the white-light illumination, as the photoinduction of electrons in the conduction band decreases the value of Seebeck coefficient. At room temperature, the zT value of MAPbI₃ increased from ~10⁻⁹ to ~10⁻⁶ under 220 mW·cm⁻² of the white-light illumination. For MASnI₃, the electrical resistivity largely decreased when MASnI₃ was unintentionally doped. MASnI₃ with more unintentional doping induced large reduction of electrical resistivity from 1 × 10⁻³ to 3 × 10⁻⁵ Ω·cm. Thus, the zT values of MAPbI₃ and MASnI₃ were also largely enhanced under the higher temperatures due to the enhancement of the charge mobility or the carrier density.

Table 7 Reported thermoelectric performance of organic–inorganic hybrid halide perovskites

Material	Seebeck coefficient (μV·K−1)	Electrical conductivity (S·cm−1)	Power factor (μW·m−1·K−2)	Thermal conductivity (W·m−1·K−1)	zT	Sample (thickness/diameter)a	References
MAPbI3	540	10		~0.5	~10−6	Crystal (mm)	[130]
MASnI3	−720			~0.09		Crystal (mm)	[130]
Bi-doped MAPbI3	77 (343 K)	6.24 × 10−6 (343 K)	3.8 × 10−6 (343 K)	~0.48 (343 K)		Film (nm)	[131]
MASnI3	66.03 (323 K)	3.56 (323 K)	1.55 (323 K)			Film (nm)	[132]
PCBM, PFO-MAPbBr3	~1350 (313 K)					Crystal (mm)	[133]

Note: Thermoelectric properties not directly mentioned in the manuscript of each reference were estimated based on the plots of each parameter (α, σ, power factor, κ, zT); unless specified otherwise in parentheses, all values are reported at room temperature.

Abbreviations: PCBM, [6,6]-phenyl-C61 butyric acid methyl ester; PFO, poly(9,9ʹ-dihexyl fluorenyl-2,7-diyl).

a

“nm” and “μm” categorize sample; nm: t < 1 μm and μm: 1 μm ≤ t < 1 mm, and mm: 1 mm ≤ t.

In 2019, Xiong et al.¹³¹ fabricated Bi-doped MAPbI₃ thin films by a one-step method; some PbI₂ was replaced with BiI₃ to prepare 1%, 3%, and 5% Bi-doped MAPbI₃ solution and the MAPbI₃ solution was spin-coated on ITO, followed by vacuum thermal evaporation of Au to fabricate ITO/Bi-doped MAPbI₃/Au structure. Researchers found that doping of MAPbI₃ with higher contents of Bi ions increases the TE properties: electrical conductivity and Seebeck coefficient were enhanced from 0.025 to 0.077 mV·K⁻¹ and 7.7 × 10⁻⁶ S·m⁻¹ to 6.24 × 10⁻⁴ S·m⁻¹ with 5% bismuth doping of MAPbI₃, respectively (Figure 14A,B). Thus, 3.8 × 10⁻⁶ μW·m⁻¹·K⁻² of power factor was obtained for 5% Bi-doped MAPbI₃. The thermal conductivity showed a gradual increase with greater Bi ion doping, but was maintained to a relatively low level, as seen in Figure 14C. Bi ions at the grain boundary modify carrier channels, reduce ion migration, and help in charge transportation. Bi ions also passivate defects in MAPbI₃, which causes the polarization effects and the mobility to be increased. Moreover, Bi ions reduce grain sizes, hindering phase transition of tetragonal to cubic phase and making MAPbI₃ stable. Therefore, Bi ions contribute to the enhancement of the TE properties of MAPbI₃.

Figure 14. (A) Seebeck coefficient of MAPbI3 with different bismuth dopings. Reproduced with permission: Copyright 2019, Wiley-VCH.131 (B) Electrical conductivity of MAPbI3 with different bismuth dopings. Reproduced with permission: Copyright 2019, Wiley-VCH.131 (C) Thermal conductivity of MAPbI3 with different bismuth dopings. Reproduced with permission: Copyright 2019, Wiley-VCH.131 (D) Electrical conductivity of MASnI3 thin film annealed at different times: 5 min (black), 10 min (red), and 15 min (blue). Reproduced with permission: Copyright 2020, Springer Nature.132 (E) Seebeck coefficient of MASnI3 thin film annealed at different times: 5 min (black), 10 min (red), and 15 min (blue). Reproduced with permission: Copyright 2020, Springer Nature.132 (F) Power factor of MASnI3 thin film annealed at different times: 5 min (black), 10 min (red), and 15 min (blue). Reproduced with permission: Copyright 2020, Springer Nature132

Very recently, annealing time of MASnI₃ thin films was optimized by Saini et al.¹³² MASnI₃ precursor was spin-coated under N₂ atmosphere and annealed at 373 K for 5, 10, and 15 min. It was found that optimal annealing time was 5 min for MASnI₃ thin films. The scanning electron microscopy images revealed that annealing time of 5 min gives larger grain sizes as compared with 10 and 15 min, forming less nucleation sites. The Sn⁴⁺ ion formations in the MASnI₃ thin films might have led to an increase in the electrical conductivity with the increase in the carrier concentration or mobility. The power factor of 1.55 μW·m⁻¹·K⁻² was measured at 323 K for the MASnI₃ thin films annealed for 5 min with corresponding σ and α of ~3.5 S·cm⁻¹ and ~66.03 μV·K⁻¹ (Figure 14D–F).

Perovskite composites for TE property enhancement

Sun et al.¹³³ synthesized MAPbBr₃ single crystal by the inverse temperature crystallization method and measured the TE properties of MAPbBr₃ in the vertical and horizontal directions. The positive Seebeck coefficient was measured in the vertical direction, and the negative Seebeck coefficient was found in the horizontal direction. MAPbBr₃ single crystal contains p-type defects in the body and has higher hole mobilities. As a result, it showed p-type semiconductor characteristics with positive Seebeck coefficient in the vertical direction. The reason why it poses negative Seebeck coefficient is that MAPbBr₃ has a large number of defects on the surface with mainly n-type states that impede hole transportation. The maximum Seebeck coefficients of ~2.3 and ~−0.15 mV·K⁻¹ were obtained at 70°C and ~47°C in the vertical and horizontal directions, respectively. The researchers further investigated enhancement of the Seebeck coefficient of MAPbBr₃ by reducing the surface defects through the addition of organic semiconductor layers on the surface of MAPbBr₃. [6,6]-Phenyl-C61 butyric acid methyl ester (PCBM) and poly(9,9ʹ-dihexyl fluorenyl-2,7-diyl) (PFO) were used for the n-type and p-type organic semiconductors. The maximum Seebeck coefficients of ~-0.03 and ~0.45 mV·K⁻¹ at ~40°C and ~75°C were measured when PCBM and PFO were added on the surface of MAPbBr₃, respectively (Figure 15A). When PCBM is added, electrons are bound by defects, the surface charge distribution is changed, internal charge moves from the interior to the surface, charge concentration on the surface is increased, and finally the Seebeck coefficient is increased. For PFO, the carriers from PFO combine with the n-type surface defects of MAPbBr₃ and they decrease the surface defect concentration, producing the positive Seebeck coefficient. The decreased surface defect concentration reduces the carrier hinderance in the crystal body. Furthermore, metal-modified layers of silver and gold were also added to the surface of MAPbBr₃, and reduced surface defects. Silver penetrates into MAPbBr₃ crystal, reacts with Br ions, and reduces Br ions in the MAPbBr₃ crystal, increasing the Seebeck coefficient. When gold is introduced, gold penetrates into MAPbBr₃ crystal, impedes the lattice expansion, and increases the carrier scattering. As a result, maximum Seebeck coefficients of ~2.4 mV·K⁻¹ at ~60°C and 11.2 mV·K⁻¹ at ~45°C were obtained with the introduction of silver and gold layers (Figure 15B).

Figure 15. (A) The Seebeck coefficients of MAPbBr3 single crystal with the introduction of organic semiconductor layers of PCBM (left) and PFO (right). Reproduced with permission: Copyright 2019, Wiley-VCH.133 (B) Seebeck coefficient of MAPbBr3 single crystal with the introduction of metal-modified layer of Au (left) and Ag (right) in the vertical direction. Reproduced with permission: Copyright 2019, Wiley-VCH.133 PCBM, [6,6]-phenyl-C61 butyric acid methyl ester; PFO, poly(9,9ʹ-dihexyl fluorenyl-2,7-diyl)

Evaluation of the TE properties of carbon-based materials

To properly evaluate a TE material, it is critical to characterize all of the TE properties (α, σ, and κ), leading to the overall zT value of the material. So far, commercially available TE devices are composed of a matrix-like configuration of sintered cubes of inorganic semimetals. The comprising thermoelements are of millimeter scale, which makes the measurement of TE properties rather straightforward, allowing the use of commercial equipment or even home-built systems.

Although there have been reports of carbon-based TE materials that have successfully evaluated the zT value using commercial devices, reports often estimate or fail to present the zT value due to experimental limitations. This issue mainly stems from the fact that many carbon-based TE materials are reported as films or fibers. The significantly small thickness and unique shape of these materials make TE property evaluation (especially thermal conductivity) challenging. Another important aspect is reporting the TE properties in the direction equal to that of actual device operation. For instance, as thin films and fibers are more likely to be operated in the in-plane direction, it is ideal that the TE properties are characterized accordingly, especially for materials with anisotropic nature. As an alternative, some reports have predicted the zT value by calculating the property of interest utilizing the anisotropy ratio, but such endeavors are limited in terms of reliability.⁷

Characterization of the Seebeck coefficient and electrical conductivity is relatively easier as compared with the thermal conductivity. In most cases, the electrical conductivity can be precisely determined using the classical two-point method, colinear four-point probe method, or the van der Pauw method on a temperature-controlled stage, on the premise that the sample thickness is determined separately. The Seebeck coefficient is acquired using a potentiostat and temperature control system, with its reliability confirmed using reference TE materials. Thermal conductivity, however, is often determined using the steady-state setup or the laser flash technique combined with the measurement of the specific heat capacity (C_p) and sample density (ρ). Unfortunately, the above methods are inapplicable to thin-film or fiber samples, which evokes the necessity of new reliable and universal measurement techniques for the development of carbon-based materials.⁵

In an attempt to overcome these limitations, there have been reports that have characterized the thermal conductivity of carbon-based materials through delicate measurement setup design and calculations. The in-plane TE properties of PEDOT thin film were characterized using the same sample for the first time using a suspended microdevice, as shown in Figure 16A.¹³⁴ PEDOT:PSS and PEDOT:Tos thin films were synthesized using vacuum vapor phase polymerization or chemical polymerization. The PEDOT samples were separated from their substrate and transferred on polymethylmethacrylate (PMMA) or polyvinylacid (PVA) sacrificial substrates, enabling the PEDOT samples to remain aligned to the microdevice. The TE properties were derived from four-probe measurements with the two electrodes and SiN_x membranes patterned with Pt resistance thermometers. The reliability of TE properties was enhanced by measuring the thermal and electrical contact resistances using six supporting SiN_x beams. The principle behind the derivation of κ is based on a separate report.¹³⁶ Utilizing the microdevice, it was demonstrated that although some κ values of PEDOT:PSS and PEDOT:Tos are comparable to that reported previously for supported PEDOT samples, the suspended PEDOT samples generally show higher κ values, showing the importance of the effect of substrates for κ characterization. It was also found that the increase of thermal conductivity with the increase of electrical conductivity deviates from that predicted from the Wiedemann–Franz law, emphasizing the importance of the experimental validation of the thermal conductivity.

Figure 16. (A) Scanning electron microscopy image of the poly(3,4-ethylenedioxythiophene) thin films suspended on the microdevice. Electrodes (red), sample (blue), and Au shadow-mask deposition (yellow) are false colored. Reproduced with permission: Copyright 2015, Wiley-VCH.134 (B) Schematic illustration of the graphene film thermal conductivity measurement setup using Raman spectroscopy. Reproduced with permission: Copyright 2017, IOP Publishing135

The applicability of Wiedemann–Franz law was also investigated for CNT thin films using a suspended Si–N membrane thermal isolation platform.¹³⁷ The characterization consists of two Si–N membranes suspended on four Si–N legs. The two membranes are connected with a narrow Si–N platform. CNT films are deposited on the Si–N platform through ultrasonic spray. Joule heating of one of the Si–N membranes induces a voltage drop in the heater to derive the Joule heat. The Joule heat and the temperatures of both Si–N membranes are used to calculate the thermal conductance (κ).

Raman spectroscopy (Figure 16B) was also used to determine the in-plane thermal conductivity of graphene monolayers to demonstrate the increase in zT value of monolayer graphene sheets through defect formation.¹³⁵ Thermal conductivity values of 543 and 36 W·m⁻¹·K⁻¹ were successfully measured for as-prepared and defect-induced graphene, respectively. The methodology behind the thermal conductivity measurement using Raman spectroscopy was based on a previous work.¹³⁸

The zT values of carbon-based TE fibers have also been evaluated. Wet-spun PEDOT:PSS fibers were characterized to exhibit a zT value of ~0.003 at 300 K.¹³⁹ Here, the Seebeck coefficient, electrical conductivity, and thermal conductivity were measured in separate setups. The electrical conductivity was measured using a commerical microvoltmeter in a two-probe method, and the Seebeck coefficient was acquired using a typical setup consisting of two Peltier devices and two thermocouples to read the temperature of each sides. The thermal conductivity was measured based on a self-heating technique by monitoring the sample resistance as a function of applied current. Although thermal conductivity was acquired at liquid nitrogen temperature to minimize the measurement error due to thermal radiation, self-heating can be a valid method for the zT evaluation of TE fibers.

Other examples that have also successively measured the thermal conductivity to evaluate the zT values, combined with the abovementioned reports, are summed up to give a significant variety of options for the zT evaluation of carbon-based TE materials.^140–142 With the present foundations, the evaluation of zT values for carbon-based TE materials is difficult but not impossible. However, for carbon-based TE materials to be used on the actual sites of the industry, a standardized method for the thermal conductivity measurement is still necessary, and ideally, the three TE properties (α, σ, and κ) should be determined using a single sample.

CONCLUSION AND OUTLOOK

Carbon materials have been proposed as next-generation TE materials for energy harvesting applications. Although the TE performance of carbon materials is not enough to replace inorganic semiconductors in their pristine form, various strategies have been demonstrated in an effort to produce high-performance carbon-based TE materials. PEDOT:PSS, CNTs, and graphene have proven to be potential high-performance TE materials by adopting strategies such as chemical doping, nanostructuring, and compositing. Moreover, the abovementioned materials have been successfully used for flexible TE devices, indicating a great leap toward commercial energy harvesting solutions. From both theroretical calcutions and experimentations, organic–inorganic hybrid halide perovskites also have been proved to be used as high-performance TE materials.

Despite the accomplishments of carbon-based materials in TEs, the TE performance is still unsatisfactory to be able to replace inorganic semiconductors. Therefore, this article will be concluded highlighting the key challenges and future research directions for the discussed materials in the viewpoint of TE performance enhancement.

• 1.

  Establishment of theoretical understanding of the TE parameters for carbon-based TE materials.

• 2.

  Understanding of the dependence of the apparent carrier transport properties (carrier density, carrier mobility, etc.) on the Seebeck coefficient for carbon-based TE materials.

• 3.

  Understanding of the mechanism to form well-correlated bulk 3D networks of PEDOT:PSS, CNT, and graphene to fabricate flexible TE devices that can form larger temperature gradients.

• 4.

  Pore structuring and pore size engineering strategies to effectively tune the thermal conductivity to the lowest value while minimizing the loss of carrier conducting properties, according to the material of interest.

• 5.

  Exploitation and TE characterization of new combination of carbon-based materials and dopants.

• 6.

  Synthesis and TE characterization of halide perovskites with modified atom sites and compositions.

Nevertheless, we believe that comprehensive understanding of carbon-based TEs provided in this review will be able to guide future TE researchers to design commercial high-performance carbon-based TE devices.

ACKNOWLEDGMENTS

This study was supported by the National Research Foundation of Korea (NRF) under the Ministry of Science, ICT & Future Planning (Basic Science Research Program (No. 2021R1A5A6002853) and Nano-Material Technology Development Program (No. 2017M3A7B4041696).

CONFLICT OF INTEREST

The authors declare no conflict of interest.