The 2D-TMDs have been investigated as good candidates for TE materials in past years. One advantage of 2D-TMDs was that their 2D crystal nature led to sharp features in DOS [30,31]. Another advantage was that the interdependent electrical and thermal conductivity can be optimized individually for maximizing energy conversion. Generally, 2D sheets of materials could be stacked to form weakly bonded layered structures. The intralayer atoms in each layer were conjugated by strong chemical bonds, while neighboring layers were combined by weak van der Waals (vdWs) interactions [32,33]. Thus, it was feasible to isolate, mix, and match dissimilar atomic layers to create a wide range of vdWs heterostructures, without the constraints of lattice matching and processing compatibility [33]. The crystal structure and chemical bonding of 2D-TMDs make themself have low thermal conductivity in the out-of-plane direction, and extremely high thermal conductivity in the in-plane direction. Therefore, these structures have a large anisotropy, which are promising for designing TE materials with high ZT. From the aspect of application, 2D-TMDs in the out-of-plane direction are good for TE or thermionic applications, while 2D-TMDs in the in-plane direction are more suitable for heat dissipation related applications.

Besides, it was also possible to tune the band gap and carrier density of 2D-TMDs by changing the number of layers [34,35], strain [36,37], electric field [34,38], and the structural composition of the materials (by hydrogenation [13,39] or oxidation [13,40]). A smooth transition from semi-metallic to the semi-conductive state has been demonstrated by varying the number of layers in 2D-TMDs [35]. Applying strain or electric field to 2D-TMDs can also influence the band structure. For instance, by applying 1% tension stress, direct band gap of mono-layer MoS₂ was turned to an indirect band gap [37]. The valley degeneracy in mono-layer MoS₂ could be observed under an external electric field [34]. The manipulation of electrical and thermal transport in 2D-TMDs will be discussed in Section 3 detailedly. Section 4 is devoted to the discussion of the recent advances in the study of electrical, thermal transport, TE properties of some typical 2D-TMDs, and TE-related applications.

The energetically degenerated valence/conduction valleys in monolayer TMDs leads to significant thickness effect on electrical properties and power factor in 2D materials (Hong et al. [34]). The calculated electrical properties and power factor in mono- and multi-layer MoS₂ are shown in Figures 2A–F. For both p-type and n-type MoS₂, the Seebeck coefficients are almost independent of layer number (Figures 2A,D), while the electrical conductivity (Figures 2B,E) and power factor PF/τ (Figures 2C,F) are dependent of layer number. It indicates that the Seebeck coefficient and electrical conductivity can be separately modulated by varying thickness, thus tuning valley degeneracy. The thickness dependence of valley degeneracy can be understood from the hybridization strength of different orbitals. The valence/conduction valleys at the K point are dominated by the localized Mo 4d orbital, thus the corresponding energy levels are not sensitive to the layer thickness. However, S 3p and Mo 4d orbitals are strongly hybridized, thus valence/conduction valleys at Γ and Σ _min points can be varied. Hence, the energy levels of valence/conduction valleys at Γ and Σ _min points are sensitive to thickness [43]. The so-called band convergence has also been proved to enhance the power factor of bulk TE materials. Similar effect of thickness on the power factors can be also observed in multi-layer MoSe₂, WS₂, and WSe₂ [37,41,42]. A general principle for these four TMDs was that the maximum ZT coincided with the greatest near-degeneracy within k _B T (k _B was the Boltzmann constant, and T was temperature) of the band edge, and the optimal thickness for maximum ZT was generally not the monolayer [43].

Apart from the electrical properties, the thickness effects on thermal conductivity have also been studied by different groups theoretically and experimentally [44–56]. However, the physical mechanisms are complex and the results are inconsistent. According to the non-equilibrium molecular dynamics (NEMD) method, the thermal conductivity of MoS₂ was insensitive to layer number. The finite energy gap of MoS₂ in the phonon spectrum made the phonon-phonon scattering channel almost unchanged with increased layer number [44]. Gu et al. [45] (Boltzmann transport equation method) and Xu et al. [46] (NEMD) reported a reduced thermal conductivity with increasing layer number from monolayer to trilayer. It could be attributed to the change of phonon dispersion and the thickness-induced anharmonicity, which was associated with the breakdown of a mirror symmetry in monolayer MoS₂. These conclusions qualitatively agreed with the experimental results performed on suspended single crystal MoS₂ [47], while it contradicted with those obtained from MoS₂ grown with chemical vapor deposition [48]. Such conflicts can be attributed to the difference in MoS₂ quality, the presence of dangling bonds at MoS₂/substrate interface, and other experimental variations. The relevant results are presented and visualized in Figure 2G.

Defects in TMDs can be categorized into intrinsic defects such as vacancies and interstitials, substitution dopants, and alloy atoms, which are often applied to modify the charge carrier concentration [57,63,64]. Basically, point defect such as vacancies inevitably resides in samples during the preparation of 2D-TMDs. It is well known that defective states formed within band gap were able to affect electronic mobilities [58,59] and excitons [40,60]. In the presence of S vacancies (V_S), donors (acceptors) induce carrier accumulation (depletion) around the V_S center, which in turn strongly affect the screening and binding of excitons within MoS₂. Consequently, the HOMO and LUMO states for donor and acceptor are shifted downwards. Thanks to that, carrier density, polarity and Seebeck coefficient can be modulated by properly introducing dopants and V_S (Cai et al. [59]). Other types of defects, such as Mo vacancies, have not been studied in detail. Recently, Sharma et al. [18] studied four types of atomic defects, including V_S (one S vacancy), V_Mo (one Mo vacancy), V_MoS (one Mo and one S vacancy), and V_MoS2 (one Mo and two S vacancies). The effects of various atomic defects on structure, electronic, and TE properties of MoS₂ were explored by using density functional theory (DFT). It found that the electronic structure of mono-layer MoS₂ showed tunable band gap with the atomic defects. Moreover, the room temperature ZT of MoS₂ with V_S and V_MoS2 defects was as high as 6.24 and 1.30, respectively.

Vacancies can also offer an efficient way to tune the thermal conductivity and TE properties of materials, since the phonon scattering depends on the density of defects [61–63]. In experiment, Aiyiti et al. [47] used oxygen plasma treatment to controllably introduce point defects to suspended few-layer MoS₂, and then studied their effects on thermal conductivity. Experiments show that, as expected, with the increase of defect concentration, the thermal conductivity decreased. Apart from vacancy defects, isotopic doping also had a considerable effect on thermal transport. For example, in MoS₂, Mo isotopes had a larger impact on thermal conductivity than that of S isotopes. The reason was that S isotopes influenced high-frequency phonon modes, while Mo (heavier atomic mass) isotopes influenced low-frequency range. In thermal transport, low-frequency phonons played a more significant role [64].

In general, chemical functionalization for 2D materials is realized by two strategies: (1) substitutional doping, which replaces the structure composition with other elemental atoms; and (2) charge transfer doping, which exchanges charge with physical adsorbents, acids, and specific solutions. Both strategies can modify the TE performance of materials by modulating the local DOS, carrier concentration, and carrier mobility [59,65–70]. For substitutional doping, Jenisha et al. [67] recently reported the TE properties in Ni-doped MoS₂. At a 7.5% Ni-doping concentration, interlayer distance was reduced and a lower thermal conductivity was confirmed. However, because of the formation of NiS phase, the electrical conductivity in Ni-doped MoS₂ was increased, which resulted in a decrease in ZT value.

For 2D materials, charge transfer doping is more efficient than substitutional doping. To date, charge transfer doping for 2D materials is realized through various approaches by introducing different groups. Cai et al. [59] investigated the effect of chemical molecular doping on the TE properties of monolayer MoS₂. Tetrathiafulvalene (TTF) and dimethyl-p-phenylenediamine molecules (DMPD) are effective donors (Figures 3A,B), whereas tetracyanoethylene (TCNE) and tetracyanoquinodimethane (TCNQ) are effective acceptors (Figures 3C,D). The molecular doping is found to be able to change the energetic positions of defective states in the band gap. Although the maximal Seebeck coefficient is almost insensitive to molecular doping, it can be adjusted by the molecular doping induced chemical potential variation. Thus, it is possible to switch the carrier type via doping different organic molecules. Guo et al. [68] reported in-situ grown Au-decorated MoS₂-assembled heterojunction (Figure 3E), as well as the TE performance of this material. The Seebeck coefficient and electrical conductivity were increased simultaneously. The first reason was the occurrence of p-type doping of the MoS₂ 2H phase. The second reason was the injection energy filtering of dopant-originated carriers around the local band bending at the interface. In addition, doping also can reduce the rate of thermally activated hopping, thus lowering the thermal conductivity. Kong et al. [69] prepared oxygen doped MoS₂ by heating the sample in air. A low thermal conductivity value of 3.1 Wm⁻¹·K⁻¹ was obtained at 327K. Abinaya et al. [70] presented an ultralow thermal conductivity of 0.248 Wm⁻¹·K⁻¹ in MoS₂/polyaniline nanocomposites, with variable range hopping property. Nevertheless, the available experimental data on TE transport properties of doped MoS₂ have been quite scarce. Most experiments have focused on the electrical transport properties and failed to measure the thermal properties. Some merely worked on the thermal conductivity. Based on those incomplete experimental data, it remains a big challenge to experimentally find a proper dopant and evaluate the TE efficiency after doping in a wide range. Thus, the ZT cannot be optimized.

Strain engineering is a common method that has been widely applied to tune the physical properties of 2D-TMDs (Figure 4). In experiment, in-plane strain can be applied by several approaches, such as bending the TMDs prepared on a flexible substrate, introducing winkles and puddles to 2D-TMDs, indending the layer with AFM, substrate thermal expansion, and so on [71–81]. The out-of-plane strain could be applied using diamond anvil cells set up [72,73]. One of the most important phenomena is direct-to-indirect band gap transition induced by tensile strain [82–88] (in the case of WSe₂, tensile strain induces an indirect-to-direct band gap transition [87,88]). In the meantime, strain also modulated the band gap, effective mass, and valley degeneracy at conduction/valance band edges of TMDs, leading to strain-tunable electrical conductivity and correlated TE properties [89,90]. For n-type WS₂, the relaxation time scaled power factor (S ² σ/τ) was increased by the application of compressive strain, whereas for p-type WS₂, it was increased with the application of tensile strain due to valley degeneracy [91].

Apart from electrical properties, the effect of strain on phonon transport properties also assists in tuning the thermal conductivity of 2D-TMDs. In general, while the tensile strain becomes large, the appearance of phonon softening lowered the phonon group velocity, leading to a reduced thermal conductivity of 2D planer nanostructures [92]. The thermal conductivity of MoS₂ was monotonically decreased with the increase of tensile strain (Jiang et al. [93]). Additionally, the analysis of lattice dynamics and phonon spectroscopy revealed that the thermal conductivity was dramatically increased with pressure in the out-of-plane direction. It was mainly due to the interlayer force strengthening and the group velocity enhancing of longitudinal acoustic phonons. When the pressure was over 15 Gpa, the thermal conductivity tended to be saturated. This was due to the combined effect of increased group velocity and reduced phonon lifetime. The phenomena were observed in most vdWs heterostructures. A 9% compressive strain can cause a 7-fold increase of the thermal conductivity in the out-of-plane direction for the multilayer MoS₂ (Meng et al. [94]).

The most common 2D-TMDs (such as MoS₂, MoSe₂, WS₂, and WSe₂) exhibit very similar electronic properties. The band structures of the 2D-TMDs show a similar transition from the direct band gap to the indirect band gap when the atomic layer increased from a mono-layer to a bilayer [95,96]. The theoretically predicted ZT values of MoS₂, MoSe₂, WS₂, and WSe₂ along the in-plane direction are displayed in Figures 5A–D. Few-layer TMDs (bilayer or trilayer) show larger increase in ZT values compared to that of bulk materials [42]. Many theoretical works on the TE properties of 2D-TMDs have also been reported. For example, Chen et al. [97] studied the TE properties of mono-layer MoS₂, MoSe₂, WS₂, and WSe₂ by the first principles calculations. The ZT value of small nanotubes was lower than that of monolayers due to the lower Seebeck coefficient. It was predicted that the highest ZT of mono-layer WSe₂ was 0.91 at room temperature, which was much higher than that of zigzag WSe₂ nanotube. Kumar et al. [98] explored the TE properties of bulk and monolayer MoSe₂ and WSe₂ by using the first principles calculations and the semiclassical Boltzmann transport theory. It proved that the electrical conductivity, the Seebeck coefficient, and the thermal conductivity of the mono-layer materials were extremely different from those of bulk materials. It indicated that much effort could be done to optimize the TE properties of unit material. Jia et al. [99] predicted the TE performance of pure MoS₂, MoS₂/MoSe₂ lateral heterostructures (LHS), and MoS₂/MoSe₂ vdWs by using DFT calculations and nonequilibrium Green function (NEGF) method. Compared to pure MoS₂, the ZT value of MoS₂/MoSe₂ LHS has been improved by the introduction of lateral interface. Moreover, the room-temperature ZT value of MoS₂/MoSe₂ vdWs stacking reached 3.5, which was ∼3 and ∼6 times higher than MoS₂/MoSe₂ LHS and pure MoS₂, respectively. The results suggested that MoS₂/MoSe₂ vdWs can be used to construct efficient TE devices.

Apart from these theoretical studies, the experimental results of 2D-TMDs have also been implemented. The typical structure of the thermoelectric micro-nano devices is shown in Figure 5E, which is used to measure the TE performance of a mono-layer or multi-layer 2D material in the in-plane direction. The temperature gradient through the nanosample one is generated by building one or two heaters at both sides. The metal electrodes are deposited onto the sample via an electron beam lithography. The Au/Cr metal contacted with the sample serves as an electrode as well as a thermometer. The TE properties of an exfoliated MoS₂ fake with different thicknesses were studied by Hippalgaonkar [35]. Because of the high electron concentration n = 1.06 × 10¹³ cm⁻² at a high external electric field, the bilayer MoS₂ exhibited a high power factor of 8.5 mWmK⁻². Although many theoretical researchers predicted that the 2D-TMDs should exhibit an outstanding TE performance, these results have not been proven by experimental data.

TiS₂ is another typical and widely studied 2D-TMD material. In 2001, it was reported to provide a large power factor of 37.1 mW K⁻²·cm⁻¹ and a low resistivity of 1.7 mΩ cm. In addition, the Seebeck coefficient was -251 mV K⁻¹ at 300 K in the in-plane direction [100]. These values of TiS₂ were comparable to that of the best TE material, i.e., Bi₂Te₃ alloy. In spite of the large power factor, TiS₂ exhibited a relatively large thermal conductivity of 68 mW K⁻¹·cm⁻¹, which comes from the reducible lattice thermal conductivity. Thus, the TE performance of TiS₂ can be further improved by decreasing the lattice thermal conductivity. The main strategies to maximize the TE performance of TiS₂ are nonstoichiometry, intercalation, substitution, and the block layer concept [101]. Instead of stoichiometric TiS₂, it was prone to obtain Ti-rich Ti_1+x S₂, where excessive Ti atoms intercalated between the vdWs gap [101,102]. According to a study of the effect of Ti/S stoichiometry on the TE properties of Ti_1+x S_2, the S was decreased with increasing x in the system (Beaumale et al. [102]). A maximum power factor of 1.7 mW m⁻¹·K⁻² in x = 0.015 was obtained at 300 K by manipulating Ti/S stoichiometry. For optimizing the TE performance of TiS₂, various heavy atoms (Au, Cu) were intercalated in the vdWs gap between the S-Ti-S layers [103,104]. The methods to minimize lattice thermal conductivity have been performed by substituting Ti atom with Ta and Nb atoms in solid solution alloy [105,106]. Recently, Zhou et al. [107] employed a novel chemical welding method, in which chemically exfoliated TiS₂ nanosheets bridged with multivalent cationic metal Al³⁺ to crosslink the nearby sheets, thus achieving a high TE performance. It is worth mentioning that all the strategies which have been employed to optimize the TE performance of TiS₂ can also be applicable to TiSe₂.

Recently, a series of synthetic misfit compounds have been intensively studied. These misfit compounds involve two sub-lattices with different lattice constants, which hold together in an alternating fashion with weak vdWs forces. The structural summary of misfit layer compounds is shown in Figure 6A. The empirical formula can be denoted as (MX)_1+x (TX₂) _m , where M = Pb, Bi, Sn, Sb; T = Cr, Ta, Ti, V, Nb; X = Se, S; m = l-3 and 0.08 < x < 0.28 [108–111]. The TX₂ layer in the misfit compounds allows high mobility of charge carriers, and the intercalated MS layer can suppress the transport of phonons by the interaction between the TX₂ layers. The misfit compounds show a phonon-glass like behavior due to the lattice mismatch between the MX and TX₂ layers. These attractive properties make (MX)_1+x (TX₂) _m an ideal class of compounds for TE energy conversion applications [112–115]. The powder of misfit compounds is often prepared by using a solid-liquid-vapor reaction method. However, although many known misfit compounds are in the sulfide space, the fabrication of kinetically controlled sulfide superlattices is still challenging due to the experimental challenges during sulfur deposition. Very recently, Roberts et al. [116] demonstrated the preparation of nanoscale SnS-TaS₂ superlattices with independent sulfur source from designed amorphous precursors, which were deposited by using a radio-frequency (RF) co-sputtering technique. It is the first report of artificial sulfide misfit superlattices with kinetically controlled architectures (Figure 6B).

Many researchers have focused on the TE properties of (MX)_1+x (TX₂) _m materials. Since 2004, a very low total thermal conductivity of 0.80 Wm⁻¹ K⁻¹ at 300 K was reported in misfit layered chalcogenide (Yb_0.95S)_1.24NbS₂ along the out-of-plane direction [117]. A even lower thermal conductivity of 0.17 Wm⁻¹·K⁻¹ was reported in the out-of-plane direction of misfit (SnSe)_m (TiSe₂)_n [118]. However, the electronic transport of this material was poor, manifested with low electrical conductivity and Seebeck coefficient. The reason was the sequential positioning of n-type and p-type layers. Wan et al. [110] investigated the (MS)_1+x (TiS₂)₂ TE materials by intercalating one layer of MS (M = Pb, Sn, Bi) into the vdWs gap of the TiS₂ layers, which established a natural superlattice structure in a bulk material. Compared to pristine TiS₂, these stacking faults significantly reduce the lattice thermal conductivity along the layers, without deteriorating the electron mobility (Figures 6C–E).

To improve the TE performance in (MX)_1+x (TX₂) _m materials, various strategies have been proposed to promote their electrical transport properties [114,119–121], including improving carrier mobility, optimizing carrier concentration, and enhancing DOS distortion. Alternatively, thermal conductivity could be minimized by softening lattice or introducing planar defects (such as translational displacement and stacking faults) [110,115,122]. Among all the misfit compounds (SnS)_1.2(TiS₂)₂ showed the highest ZT of 0.5 at 625 K [123]. However, the above-mentioned highest possible ZT mainly existed in the in-plane direction while the out-of-plane properties have gathered less attention. More importantly, it is unknown to what extent anisotropic tuning can improve TE performance in the misfit-layered chalcogenide, which hinders us from maximizing the ZT value.

TE materials could be applied to realize thermionic coolers and power generators in solid-state (proposed by Mahan et al. [124] and Shakouri et al. [125]). In solid-state thermionic power generators, a semiconductive layer was sandwiched between metallic electrodes as an potential barrier for carrier transport, which lowered the work function from a few eVs to less than 1 eV [126]. The working principle was described as follows: when electrons in the hot side were heated to gain enough velocity, they would overcome the energy barrier (material’s work function) and move along the out-of-plane direction. Eventually, some of these electrons were collected at another side, and then flowed through the outer circuit [125–127]. The current flow was referred to as thermionic current. It should be noted that the efficiency of a thermionic device was almost always smaller than the efficiency of a TE device with similar parameters. It was due to the size effects and Kapitza resistances at the interface [128]. The biggest challenge for a solid-state thermionic power generator is the back flow of the heat from the hot side to the cold side, due to the significant thermal conductance of the semiconductor layer [126]. The overcome solution is to reduce the thermal conductivity to a very low value, so that a large temperature drop can be established in a very small length range.

It is well known that 2D-TMDs exhibit a high anisotropy along the in-plane and out-of-plane directions. Van der Waals heterostructures based on 2D-TMDs are good candidates for the barrier layers in solid-state thermionic power generator. Their low thermal conductivity in the out-of-plane direction is originated from the weak vdWs bonds. In addition, the band gap, which plays a key role of a barrier, can be easily tuned via changing the layer number of the 2D-TMDs. It provides the opportunity to optimize the barrier heights for thermionic transport. Furthermore, Zebarjadi [127] developed an analytical model to investigate the performance of the device and suggested that the ideal thermal conductance values should be as low as 1 mW m⁻²·K⁻¹. Thus, vdWs heterostructures meet both needs for designing high performance solid-state thermionic power generators. Sadeghi et al. studied the TE performance of mono-layer MoS₂ sandwiched between two graphene monolayers. The out-of-plane ZT of graphene/MoS₂/graphene heterostructure was 2.8, much higher than that of mono-layer MoS₂ (0.3) and mono-layer graphene (negligible) [129]. Liang et al. [130] proposed a solid-state thermionic power generator where vdWs heterostructure sandwiched between two graphene electrodes. The vdWs heterostructure was composed of suitable multi-layer TMDs, such as MoS₂, MoSe₂, WS₂, and WSe₂. The device was expected to harvest waste heat at 400 K with efficiency of about 7–8%.

Although several theoretical studies have highlighted the potential of 2D-TMDs in thermionic applications and predict large ZT values [129,131], there is still no successful experimental result to confirm the high TE performance in solid-state thermionic power generators. In fact, it is difficult to measure the thermal conductivity, the Seebeck coefficient, and the electrical conductivity of single-layer and multi-layer 2D materials along the out-of-plane direction. To overcome this technical challenge, Li et al. [118] developed a specific platform. Figure 7 shows the schematic diagram of device fabrication process and an optical photo of completed device. Using the device, the out-of-plane TE properties of SnSe₂ and (SnSe) _n (TiSe₂) _n (n = 1, 3, 4, 5) layered film are measured. The out-of-plane Seebeck coefficient was decreased from -31 to -2.5 μV K⁻¹, while the out-of-plane effective thermal conductivity was reduced by a factor of two when n decreased from five to 1. It was due to the enhanced interfacial phonon scattering.

With the increase in the demand for wearable and portable flexible devices, thin-film TE materials have received wide attention. Besides of high TE performance, there are many other criteria for flexible TE materials, like low cost, light weight, flexibility, stability, scalable, and easy preparation [132]. Inorganic materials have high TE performance, but they are inflexible [133,134]. In contrast, organic TE materials (mainly polymers) have several advantages, such as intrinsically low thermal conductivity, mechanical flexibility, and easy processability over a large scale [135,136]. Besides of the investigations on polymer-based thin-film TE materials, recent researchers have mainly focused on the integration of 2D-TMDs such as MoS₂, TiS₂, WS₂, and NbSe₂, which exhibit promising TE properties and facile preparation. Huang et al. [137] studied the TE properties of metallic 1T-phase MoS₂ nanosheets. The MoS₂ nanosheets were synthesized via a chemically exfoliated process, and the thin films were fabricated via a simple vacuum-assisted filtration process. The power factor of the MoS₂ film reached 73.1 μW mK⁻², which was much higher than that of pristine graphene or single wall carbon nanotubes. Moreover, the TE performance of MoS₂ can be further enhanced by various functionalization [65,68]. Decorating the MoS₂ surface with Au nanoparticles allowed composite TE films to achieve a power factor value up to 166.3 μW mK⁻² at room temperature [68]. It had great potential for harvesting heat from human body. Flexible n-type TiS₂ films have provided a power factor of ∼216.7 μW mK⁻² at room temperature. This value was the highest among chemically exfoliated 2D-TMDs based films, and it was comparable to the best flexible n-type TE films. Specifically, the neighboring TiS₂ sheets were bridged with multivalent cationic metal Al³⁺, which highlighted the importance of chemical functionalization [107]. Chemically exfoliated 1T-phase WS₂ and NbSe₂ nanosheets are used as TE based self-powered wearable electronics (Figures 8A,B), which can constantly produce an output power up to 38 nW at a temperature gradient of 60 K [138]. Near room temperature, the power factors of WS₂ (n-type) and NbSe₂ (p-type) nanosheets films were evaluated to be 5–7 mW mK⁻² and 26–34 mW mK⁻², respectively.

Apart from the purely inorganic materials, organic/2D-TMDs hybrids have emerged as a new class of flexible TE materials [139–145]. A novel electrochemical intercalation and solvent exchange process to synthesize n-type hybrid organic/inorganic materials (Figures 8C,D), where TiS₂ is intercalated by organic cations and polar molecules (such as DMSO, Hexylammonium, DEG, PEG1000, H₂O) were reported by Wan et al. [141,145]. The organic cations were distributed on two sides of each TiS₂ layer. As a result, an electrical conductivity of 790 S cm⁻¹ and a very high power factor of 0.45 mW mK⁻¹ were obtained in the case of the TiS₂/[(Hexylammonium) _x (H₂O) _y (DMSO) _z ] hybrid superlattice [141]. Moreover, the measured in-plane lattice thermal conductivity was about 0.12 W mK⁻¹, which was two orders of magnitude lower than that of single-layer and bulk TiS₂. Recently, a solution-printable and flexible n-type TE materials C₆₀/TiS₂ hybrid film is demonstrated [144]. A ZT of 0.3 at 400 K has been obtained. Most importantly, the solution of C₆₀/TiS₂ hybrid can be used as an ink for printing large-area flexible and spatial TE devices with an outstanding output power of 1.68 W m⁻² at a temperature gradient of 20 K, suggesting a facile fabrication process for manufacture.

Solar energy is clean and abundant, which is considered as an ideal energy source. However, 80% of solar radiation is unexploited, which can be converted into thermal energy. Hence, capturing solar energy has become an attractive research hotspot. The concept of solar TE generators was first proposed by Telkes in 1954 [146]. Different from current solar photovoltaic and solar thermal power generation, the photo-thermoelectric generator (Photo-TEG) has been developed for converting photo-energy into electricity based on the photo-thermal effect and the Seebeck effect [147–152]. The Photo-TEG enables generating a photocurrent due to the temperature gradient induced by the absorbed light on an electric voltage (Figure 9A). A fascinating advantage of Photo-TEG was that the device could convert the whole solar spectrum into electricity including infrared (IR) light, while conventional solar photovoltaic device only utilized the ultraviolet-visible light [151]. Photo-TEG is a promising technique for energy harvesting, however, the research in this field is limited so far, compared to photovoltaics and thermoelectrics. The key challenge was to create a significant temperature difference across the device that acted as a driving force of TE generators and improved the conversion efficiency of materials and devices.

Atomic thin semiconductors, such as 2D-TMDs, have excellent electronic and optoelectronic applications attributed to the large band gap. MoS₂ was an excellent photothermal material with higher absorbance in the IR region than graphene oxide and gold nanorods [153]. Upon the absorption of light (even at energies below the bandgap of MoS₂), a temperature gradient is produced in a mono-layer MoS₂ device, which generates a thermovoltage (Seebeck) to drive the photocurrent through the circuit [152] (Figure 9B). Very recently, Li et al. [154] reported a stable Photo-TEG by integrating a new 2D Cu₂S-phenol superlattice (CP-SL) and carbon nanotubes composites. It had a maximum output power of 534.7 nW at a temperature gradient of 39.7 K. The maximum open-circuit voltage and short-circuit current measured in the experiments are 6.86 mV and 0.31 mA, respectively.

The conventional Photo-TEG device (Figure 9C) contains many bulky components, such as a vacuum enclosure, condenser lens, and heat sink, which increase the weight and size of the device [154,155]. The huge size of the device limits its application environment. However, 2D materials based flexible Photo-TEGs can be easily mounted on various surfaces exposed to sunlight, such as clothes, windows, and the outer walls of a building [156–158]. A flexible Photo-TEG based on MoS₂/polyurethane photothermal film and Te/PEDOT thermoelectric layer has been demonstrated for harvesting environmental IR light [158]. When the infrared lamp is turned on and lasting for 50 s, the open-circuit voltage of Photo-TEG is rapidly increased and plateaued at 1.2 mV. The device is expected to be useful in applications such as self-powered wearable electronic devices.