Interface‐Engineered Bi0.5Sb1.5Te3/WSe2 Heterostructures for Enhanced Thermoelectric Performance
Karan Giri, Yen‐Ling Wang, Yi‐Ting Wu, Chun‐Hua Chen
TL;DR
This paper explores how engineering interfaces in BST/WSe2 heterostructures improves thermoelectric performance through structural and electronic effects.
Contribution
The study demonstrates that interface engineering in BST/WSe2 heterostructures enhances thermoelectric performance via lattice strain and moiré patterns.
Findings
BST/WSe2 heterostructures show high crystallinity and lattice compression with minor secondary phases.
Interfacial strain and moiré patterns enhance Seebeck coefficients and power factors above 50 µW cm−1 K−2.
Non-uniform WSe2 domains create p–n junction-like regions that reinforce energy filtering and carrier control.
Abstract
BST/WSe2 heterostructures are deposited via pulsed laser deposition to examine the influence of interfacial structure and compositional heterogeneity on thermoelectric performance. X‐ray diffraction reveals high crystallinity with uniform lattice compression along both in‐plane and out‐of‐plane directions, accompanied by minor secondary phase inclusions. Cs‐STEM resolves well‐defined quintuple layers and compressed van der Waals gaps. At the same time, lattice strain is evident in structurally distorted regions, and moiré patterns arising from lattice mismatch are specifically observed within the mixed region adjacent to the BST interfaces. Temperature‐dependent transport measurements show enhanced electrical conductivity, with S3 reaching the highest values and S2 closely following, driven by thermally activated carrier generation. Positive Seebeck coefficients indicate p‐type…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
FIGURE 1
FIGURE 2
FIGURE 3
FIGURE 4
FIGURE 5
FIGURE 6| Element | S1 (at. %) | S2 (at. %) | S3 (at. %) | S4 (at. %) |
|---|---|---|---|---|
| Bi M | 8.27 | 10.02 | 10.06 | 9.36 |
| Sb L | 25.08 | 30.52 | 29.37 | 30.84 |
| Te L | 60.73 | 58.11 | 53.8 | 54.9 |
| W M | 2.45 | 0.91 | 4.03 | 2.24 |
| Se L | 3.47 | 0.43 | 2.74 | 2.67 |
- —Taiwan's National Science and Technology Council (NSTC)
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAdvanced Thermoelectric Materials and Devices · 2D Materials and Applications · Topological Materials and Phenomena
Introduction
1
In the contemporary era, sustainable green energy solutions have attracted significant scientific interest, reflecting global efforts to reduce reliance on fossil fuels, address environmental challenges, and ensure long‐term energy security. Among these, thermoelectric power generation [1] offers a promising approach for energy harvesting and solid‐state refrigeration. Bi_0_.5_Sb_1.5_Te_3 (BST), a van der Waals (vdW) layered material composed of stacked quintuple layers (QLs), remains one of the most effective thermoelectric materials for near‐room‐temperature applications. The p‐type Bi_x_Sb_2_ − x_Te_3 alloys demonstrate figure‐of‐merit (ZT) values exceeding 1.2 within 300–400 K [2, 3], owing to their narrow bandgap (∼0.1 eV), high carrier mobility (µ), large density‐of‐states effective mass (m*), high carrier concentration (n), high electrical conductivity (σ), intrinsically high Seebeck coefficient (S), and low thermal conductivity (κ) [4, 5]. These properties have enabled applications in low‐grade waste heat recovery [6], solid‐state cooling [7], electronics [8], biomedical devices [9], aerospace power generation [10], microelectronic energy harvesting [11], sensors [12], and wearable technologies [13]. Performance improvements in BST have been demonstrated through doping with Cu [14, 15, 16], Ag [17, 18], C [19, 20], W [21], and other suitable dopants. Beyond conventional doping, integrating layered materials such as transition metal dichalcogenides (TMDs) offers a strategy to enhance thermoelectric performance through interface and structural modifications. Recent studies highlight the role of intralayer vdW dynamics in Bi_0_.4_Sb_1.6_Te_3, where atomistic diffusion and crystal‐amorphicity duality increase n while maintaining S, boosting power factor (PF) to ∼49 µW cm^−^ ^1^ K^−^ ^2^ [22]. Concurrent surface reconstruction reduces the lattice thermal conductivity (*κ_l_ *) to ∼0.97 W m^−^ ^1^ K^−^ ^1^, achieving a ZT of ∼1.54 at 373 K [23, 24].
In this study, we incorporate tungsten diselenide (WSe_2_), a vdW TMD, into the BST matrix via dual‐beam pulsed laser deposition (PLD) to improve thermoelectric performance through interface engineering. The ZT of a thermoelectric material is defined as
where S ^2^ σ is the PF reflecting electrical transport, T is absolute temperature, and κ = *κ_e_ *+*κ_l_
- represents total thermal conductivity. While reducing *κ_l_
- through interface engineering [25, 26], defect incorporation, and microstructural modifications is well‐established [22, 27, 28], PF enhancement is more challenging due to the intrinsic interdependence of S, σ, and n [29]. Here, this improvement is primarily achieved through tailored interface and defect engineering [30] in BST/WSe_2_ heterostructures, enabling controlled carrier transport and phonon scattering. Prior PLD‐based investigations [31] have demonstrated the efficacy of such approaches in enhancing thermoelectric performance, supporting the strategies adopted herein.
Interface engineering plays a crucial role in optimizing PF by enhancing S and modulating carrier transport while suppressing *κ_l_ *. For instance, LaFeSi/BST nanocomposites exploit vacancy‐induced phonon scattering and interfacial interactions to achieve ZT = 1.11 at 380 K [32]. In contrast, nano‐TiC/BST systems achieve a ZT of 1.3 at 400 K through the combined suppression of κ_l_ and energy filtering [33]. Nanocomposites with metallic nanoinclusions enhance S through interface potential barriers, and dense nanoinclusions (1–2 nm) further improve performance via energy‐selective carrier scattering [34]. Similarly, interface engineering through MAPbI_3_ incorporation into Sb_2_Te_3_ yields a maximum output power of 33.5 nW under a 10 K temperature gradient [13]. CuGaTe_2_/BST nanocomposites deliver ZT = 1.36 over 300–500 K via energy filtering and bipolar effect suppression [35], while Cu_3_SbSe_4_/BST heterostructures attain ZT = 1.6 at 476 K through heterojunction interfaces [36]. Flexible BST heterostructure films containing Te and Sb_2_Te_3_ nanoinclusions exhibit improved PF and enable high‐power flexible generators [37]. Mixed‐grain BST achieves PF = 4.2 × 10^−^ ^3^ W m^−^ ^1^ K^−^ ^2^ at 300 K and ZT = 1.16 at 350 K [38]. Collectively, these studies demonstrate that defect‐controlled, heterostructure‐based, and grain‐engineered interfaces can synergistically balance the intrinsic trade‐off between S and σ while optimizing their combined contribution, thereby mitigating phonon transport and enabling substantial PF enhancement.
WSe_2_ emerges as an effective interface modifier for BST‐based systems. As a vdW layered TMD with P6_3_/mmc symmetry, WSe_2_ provides favorable band alignment and high carrier mobility, enabling selective energy filtering and enhanced S [39]. Depending on thickness, it can exhibit n‐type, p‐type, or ambipolar conduction [40]. Incorporated into the BST matrix, nanoscale WSe_2_ phases act as phonon‐scattering centers to suppress *κ_l_
- and optimize carrier concentration through partial W^4^ ^+^ substitution of Sb^3^ ^+^ or Bi^3^ ^+^ [41]. The layered nature of both BST and WSe_2_ promotes c‐axis orientation on SiO_2_/Si substrates, facilitating coherent van der Waals interfaces [42]. Even small additions of WSe_2_ can induce atomic‐scale rearrangements across vdW gaps, influencing carrier transport and phonon dynamics. Although vdW heterostructures are widely explored in electronics and optoelectronics [43], their application in layered thermoelectrics remains underexplored. Here, we quantify the QL thickness and van der Waals gap variation, establishing a direct link between interfacial structure and thermoelectric performance.
Building on our previous work [44], this study investigates the incorporation of WSe_2_ into p‐type BST using dual‐beam PLD at four substrate temperatures, 573, 623, 673, and 723 K, designated S1–S4, under a deposition pressure of 0.1 Torr. The resulting heterostructured films consist of alternating BST and BST/WSe_2_ regions without sharply defined boundaries (Figure 1a), generating semicoherent van der Waals–coupled interfaces. Magnified schematic illustrations (Figure 1b) and crystal models (Figure 1c) highlight the layered nature of both materials, which facilitates structural integration.
Schematic representation and structural characterization of BST/WSe2 thin films. (a–c) Architecture of the heterostructures on SiO2/Si substrate, illustrating the distribution of BST and WSe2 phases together with their crystal structures. (d–f) XRD patterns of Samples S1–S4 compared with reference files, confirming the coexistence of BST, WSe2, and minor W‐related phases. High‐resolution scans highlight reflections from WSe2 and W phases, while the extracted lattice parameters (a and c) exhibit sample‐dependent variations, indicating subtle structural modifications across the series.
Importantly, this work demonstrates that non‐epitaxial PLD growth can yield localized moiré‐type registry at the BST/WSe_2_ interfaces, an interfacial feature not previously reported. Comparison with our earlier BST/WSe_2_ films grown at the same substrate temperature but higher pressure confirms that the nearly temperature‐invariant PF in the present films cannot be attributed to compositional variations, as the WSe_2_ fraction at 623 K is comparable to other temperatures and does not dominate the microstructure. Instead, the stabilized PF, combined with a Seebeck coefficient of 400–430 µV K^−^ ^1^, carrier concentration of (2–4) × 10^1^ ^9^ cm^−^ ^3^, modest mobility reduction (100→47 cm^2^ V^−^ ^1^ s^−^ ^1^), and EDS evidence of WSe_2_‐rich nanoscale domains, indicates that interface‐mediated energy filtering and band modulation primarily govern the enhanced thermoelectric response.
Results and Discussion
2
Morphology and Structural Characterization
2.1
Figure 1d presents the θ–2θ X‐ray diffraction (XRD) patterns of the four samples in the 10°–70° Bragg angle range. All samples exhibit pronounced (001) and (015) reflections, indicative of a dual‐textured growth mode. The dominant (001) peak suggests strong c‐axis orientation, likely reflecting an initial epitaxial alignment with the SiO_2_/Si substrate. The emergence of the (015) reflection implies the development of a secondary crystallographic orientation, potentially arising from thickness‐dependent structural relaxation during film growth [45]. This evolution is consistent with our prior reports on preferential texture transitions in layered thermoelectric thin films.
The diffraction peaks are well indexed to rhombohedral BST (PDF #49‐1713), with additional minor peaks corresponding to hexagonal WSe_2_ (PDF #38‐1388) and cubic W (PDF #88‐2339), confirming BST as the predominant phase with limited incorporation of layered WSe_2_ and elemental W. Consistent 2θ positions across samples confirm structural stability, while sharp, well‐defined peaks indicate high crystallinity. Selenium volatility at elevated temperatures likely contributes to the elemental W peaks observed. The weak or absent WSe_2_ reflections correspond to its low concentration, resulting from periodic ablation and reduced laser intensity. Two peaks nearly at 33° and 61° marked with asterisks are attributed to WO_3_ (PDF #89‐8053), likely originating from partial pre‐oxidation of tungsten in the WSe_2_ ingot (Figure 1e). The absence of this oxide phase in S2 and S4 suggests that tungsten oxidation is influenced by sample‐specific growth or handling conditions.
The partial overlap of W reflections near 38°, 44°, and 64° with BST peaks contributes to the broadening of the (*1010 *), (*0015 *), and (*0021 *) reflections. This broadening is indicative of nanoscale phase coexistence, reduced grain sizes, and interfacial strain. Crystallite sizes, estimated via Scherrer's equation, range from 13 to 24 nm (S1), 18 to 34 nm (S2), 23 to 35 nm (S3), and 24 to 36 nm (S4), which are notably smaller than those of our earlier samples grown at 0.5 Torr and pristine BST (32–46 nm). Corresponding dislocation densities of 3.36 × 10^−^ ^3^, 1.44 × 10^−^ ^3^, 1.90 × 10^−^ ^3^, and 1.91 × 10^−^ ^3^ nm^−^ ^2^ for S1–S4 indicate moderate defect levels, with the higher density in S1 potentially stemming from localized growth variations such as nucleation dynamics or substrate‐film interactions.
Lattice parameter analysis reveals a uniform contraction relative to standard BST (a = 4.28 Å, c = 30.52 Å), with measured values of a = 4.18–4.19 Å (≈2.2%–2.4% reduction) and c ≈ 29.68–29.75 Å (≈2.5%–2.8% reduction). This contraction is primarily attributed to Te deficiency, particularly at the Te1 sites, which induces isotropic lattice compression; Te vacancies are well known to reduce both a and c lattice constants in Bi_2_Te_3_‐type structures. Grain‐size effects may also contribute, as studies on highly oriented nanocrystalline BiSbTe thin films have demonstrated a clear correlation between reduced grain size and lattice strain, consistent with grain‐boundary‐mediated perturbations of the crystal lattice [46]. In addition, interfacial strain arising from lattice mismatch with WSe_2_ (c = 12.96 Å) perturbs epitaxial alignment and vertical growth, since lattice mismatch is a well‐established source of residual stress that influences thin‐film orientation and strain relaxation behavior [47].
In‐plane lattice constants remain consistent (SD ≤ 0.0077 Å) (Figure 1f), with slightly higher a‐values in S2 and S3, likely reflecting localized strain or compositional heterogeneity. The out‐of‐plane c‐parameter exhibits minimal variation, with S1 exhibiting a marginally larger value and standard deviation, possibly arising from stacking faults or anisotropic lattice relaxation. Overall, these observations confirm that the films are structurally robust, phase‐pure, and highly crystalline, with reproducible lattice characteristics across all growth temperatures.
Scanning electron microscopy (SEM) surface and cross‐sectional analyses (Figure 2) reveal pronounced microstructural variations among the samples, primarily governed by substrate temperature. While all films exhibit uniform coverage, their internal architectures differ slightly. Sample S1 shows a dense and compact surface morphology composed of vertically aligned flakes. In contrast, S2 exhibits less compact structure, lacking the pronounced vertical flake alignment observed in S3 and S4, and displays smaller vertical flake dimensions compared to previous films deposited at the same temperature under 0.5 Torr.
Top‐view and cross‐sectional SEM micrographs of BST/WSe2 thin films (S1–S4), highlighting the evolution of surface morphology and cross‐sectional architecture with varying deposition conditions.
In both the current and earlier studies, most films display similar cross‐sectional features, with the notable exception of S2 deposited at 623 K, which exhibits a distinct morphology. At the reduced deposition pressure of 0.1 Torr, the lower frequency of gas‐phase collisions increases the kinetic energy of adatoms, thereby promoting greater surface mobility and longer migration distances before nucleation [48]. As a result, except for S2, samples tend to develop larger flake dimensions and lower nucleation densities with increasing substrate temperature. In particular, S3 and S4 form well‐defined, flake‐like architectures oriented perpendicular to the substrate, with lateral flake dimensions and porosity both increasing at higher substrate temperatures, likely due to shadowing effects during growth. Such porous architectures are expected to hinder carrier transport by introducing additional scattering centers.
Elemental analysis of Samples S1–S4 (Table 1) highlights S2's compositional advantages. It exhibits the most favorable stoichiometry for thermoelectric performance, with a well‐balanced distribution of Te (58.11 at.%), Sb (30.52 at.%), and Bi (10.02 at.%), accompanied by trace amounts of W (0.91 at.%) and minimal Se (0.43 at.%). These values closely correspond to the nominal composition of Bi_0_.5_Sb_1.5_Te_3 with a minor WSe_2_ fraction, reflecting improved dopant control and compositional uniformity relative to the other samples. This optimized composition likely underpins S2's enhanced microstructural stability, which in turn is expected to contribute to its superior thermoelectric properties. To further elucidate these correlations and probe the defect landscape at the atomic scale, a comprehensive STEM investigation of S2 is performed.
Structural Analysis Using Transmission Electron Microscopy
2.2
High‐resolution Aberration‐Corrected Scanning Transmission Electron Microscopy (Cs‐STEM) imaging of the pure BST region (Figure 3a) reveals well‐defined QLs. The corresponding fast Fourier transform (FFT) from the boxed area confirms the crystalline nature of BST through distinct diffraction spots, and the inverse FFT (IFFT) reconstruction (Figure 3b) enhances visualization of periodic QLs and vdW gaps. Statistical analysis across multiple regions (Figure 3c,d) reveals QL thicknesses ranging from 6.9 to 10.9 Å, predominantly clustering between 9.7 and 10.3 Å (mean: 9.54 Å; median: 9.97 Å). The vdW gaps exhibit a mean of 1.06 Å and a median of 0.94 Å, reflecting contractions of ∼59%, relative to bulk Bi_2_Te_3_ (2.582 Å) [49], and of ∼65%–75% relative to Sb_2_Te_3_ (∼3.0 Å) [50]; (3.71 Å) [27]. These contractions contribute to the variation in QL thickness, suggesting a structural compensation mechanism that accommodates local strain.
Structural and statistical analysis of the BST phase. (a) Cs‐STEM image revealing the layered structure of the BST matrix. (b) The IFFT image emphasizes the QLs and vdW gaps within the BST phase. (c) Histogram of QL thicknesses, with a mean of 9.54 Å and a median of 9.97 Å. (d) Histogram of vdW gap widths, exhibiting a mean of 1.06 Å and a median of 0.94 Å. (e) Correlation plot illustrating a strong inverse relationship between QL thickness and vdW gap, with a linear fit (y = −0.35x + 5.79; R 2 = 0.9316; r = −0.957).
In addition, a lateral gradient in vdW gap width is observed along the in‐plane direction (as indicated by the arrow), with out‐of‐plane gaps being wider at one end and gradually narrowing toward the opposite side. This trend reflects a spatial variation in c‐axis contraction, as suggested by lattice parameter analysis, likely driven by in‐plane strain relaxation. The measured vdW gaps are narrower than the atomic diameter of W (∼2.7 Å), making static intercalation geometrically unfavorable. Consistent with this, and unlike our earlier study at 0.5 Torr [44], where W diffusion into vdW gaps was evident, no such incorporation is detected at the lower pressure of 0.1 Torr, suggesting that reduced interlayer spacing and deposition pressure jointly suppress W intercalation. While XRD patterns show minor reflections from elemental W and faint WSe_2_ signatures, Cs‐STEM and SEM‐EDS reveal no discrete W‐rich phases but do confirm localized WSe_2_ domains. This apparent discrepancy is attributed to the nanoscale, spatially heterogeneous distribution and extremely low volume fraction (<∼1%) of these phases, which are detectable in bulk‐averaged XRD. Moreover, localized reductions in QL thickness and vdW gap width, accompanied by subtle contrast variations, may indicate the presence of Te vacancies; as Te atoms occupy the outer layers of each QL, their absence could lead to compaction along the c‐axis. While direct confirmation requires chemical analysis, the structural signatures observed here are consistent with partial Te deficiency.
As shown in Figure 3e, the measured QL thickness and vdW gap width exhibit a strong inverse correlation (Pearson correlation coefficient [51]: r = –0.957; R ^2^ = 0.916), indicating that most of the apparent variation in vdW gap width is accommodated by subtle adjustments in QL thickness, despite the latter being generally considered structurally rigid. This compensatory response suggests that the more flexible vdW gaps primarily accommodate strain, preserving the overall structural coherence [52]. These TEM‐based observations are corroborated by XRD analysis, which reveals lattice compression across all samples. Overall, these findings highlight the structural adaptability of layered thermoelectric heterostructures and the critical role of interlayer modulation in accommodating mechanical and compositional perturbations.
Figure 4a presents a high‐resolution Cs‐STEM image of the WSe_2_ phase, showing densely packed atomic planes with interplanar spacings of ∼2.46, 1.78, and 1.77 Å. These spacings are consistent with characteristic WSe_2_ lattice planes and confirm its structural identity; the corresponding FFT from Box a1 is shown in Figure 4d‐(a1).
Cs‐STEM micrographs and corresponding diffraction patterns illustrating interfacial structures within the heterostructure. (a) High‐resolution image of the WSe2 phase, showing well‐resolved lattice fringes and corresponding interplanar spacings. (b) Image highlighting the pristine BST region and the interface with the adjacent mixed‐phase zone, with lattice spacings identified. (c) Interface between the mixed‐phase region and a highly distorted domain, exhibiting pronounced lattice distortion. (d) FFT patterns corresponding to areas a1, b1, b2, and c1, highlighting differences in crystallographic orientations and periodicities. (e–f) IFFT images derived from the FFTs of selected regions, revealing local lattice modulations and dislocations. (g) Schematic representation of possible atomic stacking configurations (AB, AA, and BA) within the Se–W–Se tri‐layer structure of WSe2.
Figure 4b,c further highlights two distinct interface types within the heterostructure. In Figure 4b, a sharp boundary separates a pure BST region from a mixed‐phase zone containing interspersed BST and WSe_2_ subdomains without sharply defined phase boundaries. The BST region (Box b1) displays (006) lattice fringes, as corroborated by the FFT in Figure 4d‐(b1). In the adjacent mixed region, moiré fringes arise from the superposition of the (014) and (103) planes of WSe_2_ subdomains embedded within the BST matrix, producing the characteristic four‐lobed diffraction pattern in Figure 4d‐(b2).
The periodic array of bright spots in Figure 4e, arranged in a square or rectangular lattice, corresponds to a localized moiré pattern originating from the interference between the nearly orthogonal WSe_2_ lattice planes (014) and (103). This contrast arises primarily from an interplanar spacing mismatch of 3.04 and 3.26 Å, respectively, rather than rotational misorientation, as confirmed by the absence of angular displacement in the FFT. These moiré features are highly localized, extending only ∼5–7 nm, which is substantially shorter than the intrinsic moiré periodicity typically reported for WSe_2_ systems [52, 53]. Such short‐range modulation indicates that local lattice mismatch, rather than long‐range twist, governs the observed pattern. Analysis of additional WSe_2_ planes (2.46 Å vs 1.77 Å) reveals a larger mismatch of 28%–39%, which further supports that the interface accommodates strain through localized registry and semicoherent coupling instead of forming extended coherent moiré superlattices.
The presence of moiré fringes and subtle contrast variations implies partial atomic reconstruction within the mixed‐phase region. Local lattice strain likely facilitates the formation of energetically favorable stacking configurations, such as AB‐ or BA‐like domains (Figure 4g), consistent with prior reports [54]. Unlike ideal twisted bilayer systems, the domain boundaries are not sharply defined, and reconstruction is most evident within the mixed‐phase region, with limited contributions in interfacial zones, where more diffuse structural modulations are observed. Figure 4c additionally presents the interface between the mixed region and a highly distorted domain. The FFT of Box C1 (Figure 4d‐(c1)) and its corresponding IFFT (Figure 4f) reveal pronounced lattice distortions, edge dislocations, and disrupted periodicity, indicative of localized strain caused by structural mismatch or interfacial stress.
These TEM observations quantitatively demonstrate that non‐epitaxial PLD growth produces semicoherent vdW‐coupled interfaces with localized moiré patterns. Defects such as vacancies, edge dislocations, and isolated W atoms are distributed at the interface and within the BST matrix, providing a structural basis for the enhanced power factor and thermoelectric performance observed experimentally.
Transport Properties
2.3
Figure 5 presents the temperature‐dependent electrical transport behavior of the BST/WSe_2_ heterostructures (S1–S4), highlighting the variations in electrical conductivity and Seebeck coefficient arising from differences in carrier generation, interfacial effects, and secondary phase contributions. BST/WSe_2_ heterostructures exhibit pronounced temperature‐driven electrical conductivity enhancements across all samples (Figure 5a), with S3 attaining the highest values and S2 closely following. This enhancement arises primarily from thermally activated carrier generation (Figure 5b) that surpasses phonon and charge scattering losses (Figure 5c), whereas S1 and S4 remain limited by lower carrier concentrations. Positive Seebeck coefficients confirm p‐type transport, with S2 achieving >430 µV K^−1^ over 310–440 K (Figure 5d) due to robust interfacial energy filtering, followed by a high‐temperature drop attributed to bipolar conduction. S3 exhibits a steady thermopower increase; however, its slightly lower performance relative to S2, along with a more pronounced peak, may partly result from WO_3_ phases that modify carrier scattering [55]. A similar WO_3_ contribution in S1 likely explains its weaker and earlier Seebeck peak, whereas S4, lacking WO_3_, exhibits a peak at ∼335 K, indicating that intrinsic interface effects can also induce temperature‐specific enhancements.
Temperature‐dependent transport properties of BST/WSe2 heterostructures: (a) electrical conductivity, (b) carrier concentration, (c) mobility, (d) Seebeck coefficient, (e) effective mass, and (f) power factor. Data show the comparative performance of S1–S4, with S2 exhibiting the highest S and PF over a broad temperature range. The measurement reproducibility and corresponding error bars are provided in the Supporting Information (Figures S1 and S2).
While elevated substrate temperatures promote Te volatility and vacancy formation, these effects alone cannot fully account for the observed carrier distributions, particularly the low n in S4. Compositional effects, specifically WSe_2_ incorporation, further modulate transport. Cs‐STEM reveals non‐uniform WSe_2_ domains (∼65 nm) embedded in the BST matrix, likely exceeding the threshold for local n‐type behavior [40]. These domains form internal p–n junction‐like regions that locally adjust carrier concentration, facilitate electronic compensation, and enhance interfacial energy filtering, thereby contributing to the superior thermoelectric response of S2 and S3, including enhanced Seebeck coefficients.
Effective mass analysis indicates substantial band flattening in S2 (1.36 to 1.61 m_0_) and S3 (0.79 to 1.24 m_0_) (Figure 5e), which can be directly linked to interfacial strain and moiré‐induced band reconstruction. High‐resolution Cs‐STEM reveals short‐range moiré patterns, lattice distortions, and vacancy clusters, which generate nanoscale potential fluctuations and local band offsets. Such structural features act as selective scattering centers for low‐energy carriers, thereby intensifying interfacial energy‐filtering effects and enabling the unusually high thermopower observed in the heterostructures [56].
These structural and compositional features underpin the enhanced Seebeck coefficients and nearly constant power factor of S2 (>50 µW cm^−1^ K^−2^) and S3 (∼45 µW cm^−1^ K^−2^). At the same time, S1 and S4 remain below 8 µW cm^−1^ K^−2^ due to weaker interfacial transport and bulk‐limited conduction (Figure 5f). The results highlight that optimal thermoelectric performance in BST/WSe_2_ heterostructures arises from deliberate control of microstructure, interface quality, strain, and compositional heterogeneity, which collectively balance carrier concentration, mobility, and energy filtering. Notably, the PF of S2 surpasses the 43.6 µW cm^−1^ K^−2^ reported for Cu_2_GeSe_3_/Se‐doped Bi_0_.5_Sb_1.5_Te_3 [57] and approaches the ∼49 µW cm^−1^ K^−2^ achieved in layered Bi_0_.4_Sb_1.6_Te_3 via intralayer vdW engineering [58], while WSe_2_‐doped Bi_0_.48_Sb_1.52_Te_3 and liquid‐Te‐assisted planar Bi_2_Te_3_ films also show comparable PF through carrier‐concentration optimization and interface‐mediated scattering [41, 59]. These results demonstrate that interfacial strain, moiré reconstruction, and compositional design in BST/WSe_2_ films can yield high power factors without the extrinsic dopants, highlighting the pivotal role of interface engineering in layered thermoelectric systems.
Figure 6 schematically illustrates the charge‐carrier dynamics in BST/WSe_2_ heterostructures. Holes primarily originate from BST and electrons from WSe_2_; at the interface, partial compensation optimizes net hole concentration for balanced transport. Carriers traverse two key interfaces: (i) BST–mixed region, where band offsets preferentially scatter low‐energy holes, enhancing average carrier energy, and (ii) mixed region–distorted WSe_2_ domain, where structural strain induces incoherent scattering that selectively filters carriers without severely reducing mobility. This hierarchical energy‐filtering landscape underpins the high Seebeck coefficient and moderate conductivity observed in S2, consistent with studies on oxide inclusions and graphene‐modified Bi–Sb–Te interfaces [60, 61].
Schematic illustration of charge optimization and its role in balancing electrical conductivity and the Seebeck coefficient to achieve a higher power factor in a BST/WSe2 heterostructure. Holes from p‐type BST and electrons from n‐type WSe2 converge at an interfacial region, where carrier balancing leads to an optimized net carrier concentration. This optimized concentration contributes to electrical conductivity and facilitates energy filtering at the interface. The horizontal arrows preceding the zigzag regions represent the directional flow of charge carriers encountering an interface‐modulated region, which acts analogously to a gate, influencing scattering or filtering behavior. The adjacent potential barrier symbolizes the energy filtering mechanism, which selectively allows high‐energy carriers to dominate transport, thereby enhancing the Seebeck coefficient. Both σ and the improved S contribute to the overall enhancement of the PF.
Conclusions
3
Comprehensive structural and compositional analyses using XRD, SEM‐EDS, and Cs‐STEM reveal that BST/WSe_2_ thin films possess well‐defined BST quintuple layers with contracted vdW gaps, lateral gradients in interlayer spacing, and semicoherent interfaces. Cs‐STEM further identifies localized WSe_2_ domains, moiré patterns, lattice distortions, and potential Te vacancies, reflecting nanoscale structural reconstruction and strain accommodation. SEM‐EDS confirms the heterogeneous spatial distribution of minor WSe_2_ phases, while XRD captures bulk lattice compression and minor W reflections. Electrical transport measurements demonstrate that interfacial engineering and compositional heterogeneity effectively optimize carrier concentration and energy‐filtering effects, yielding p‐type conduction with Seebeck coefficients exceeding 430 µV K^−1^ (S2) and a power factor of up to 50 µW cm^−1^ K^−2^. These results establish a direct correlation between nanoscale structural features and macroscopic thermoelectric performance, highlighting that precise control of interlayer spacing, interface quality, and secondary‐phase distribution can substantially enhance thermoelectric efficiency.
Experimental Section
4
Composite thin films were deposited on (001) SiO_2_/Si substrates via dual‐beam PLD. High‐purity BST (99.99%) and WSe_2_ (99.80%) powders were pressed into disc‐shaped targets. Ablation was performed in an Ar atmosphere at base pressure 2×10^−^ ^5^ Torr and a deposition pressure of 0.1 Torr, using a Q‐switched Nd: YAG laser (355 nm wavelength, 10 Hz repetition rate, 5 ns pulse duration, and ∼ fluence 8.84 J cm^−2^) with beam splitting set to 30% reflected and 70% transmitted. The primary BST target was ablated continuously for 55 min, while the secondary WSe_2_ target was ablated periodically in 5 min intervals (except at the first and last 5 min) to control WSe_2_ incorporation, thereby facilitating precise heterostructure growth.
Surface morphology and composition were analyzed using field‐emission scanning electron microscopy (FESEM, JEOL JSM‐7800F) coupled with energy‐dispersive X‐ray spectroscopy (EDS, OXFORD MAX150). Phase identification and crystal structure were analyzed using X‐ray diffraction (XRD, Bruker AXS D8 Discover) with Cu Kα radiation (λ = 1.54 Å). Cross‐sectional specimens for spherical aberration‐corrected scanning transmission electron microscopy (Cs‐STEM) were prepared using a focused ion beam system (FIB; Hitachi NX2000). High‐resolution imaging was performed on a Cs‐corrected TEM (STEM023300). In‐plane thermoelectric transport properties, including electrical conductivity and Seebeck coefficient, were measured using a LINSEIS L79 HCS 1 system under a 0.7 T magnetic field over the temperature range 300–475 K.
Author Contributions
K.G. designed the experiments, deposited the films, carried out structural and transport characterizations, performed calculations, and drafted the manuscript. Y.L.W. and Y.T.W. contributed to data curation, project administration, and manuscript editing. C.H.C. supervised the project, contributed to experimental design, edited the manuscript, and secured funding. All authors reviewed and approved the final version of the manuscript.
Conflicts of Interest
There are no conflicts of interest to declare.
Supporting information
Supporting Information File 1: smll72287‐sup‐0001‐SuppMat.docx
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1M. Wang , Q. Zhang , K. Pang , et al., “Robust Homogeneous Segmented Power Generator Driven by Sb 2Te 3 ‐Based Thermoelectrics,” Advanced Materials 37 (2025): 2503128, 10.1002/adma.202503128.40177872 · doi ↗ · pubmed ↗
- 2T. Cao , X.‐L. Shi , M. Li , et al., “Advances in Bismuth‐Telluride‐Based Thermoelectric Devices: Progress and Challenges,” e Science 3 (2023): 100122, 10.1016/j.esci.2023.100122. · doi ↗
- 3X. Tang , Z. Li , W. Liu , Q. Zhang , and C. Uher , “A Comprehensive Review on Bi 2Te 3‐Based Thin Films: Thermoelectrics and Beyond,” Interdisciplinary Materials 1, no. 1 (2022): 88–115, 10.1002/idm 2.12009. · doi ↗
- 4M. Tan , X.‐L. Shi , W.‐D. Liu , et al., “Enabling Ultra‐flexible Inorganic Thin‐Film‐Based Thermoelectric Devices by Introducing Nanoscale Titanium Layers,” Nature Communications 16 (2025): 633, 10.1038/s 41467-025-56015-5.PMC 1173066039805848 · doi ↗ · pubmed ↗
- 5C. Zhao , S. Cui , T. Li , et al., “Unexpected 18‐Fold Overlapped Feathery Fermi Pockets in Typical Thermoelectric Bi 0.5Sb 1.5Te 3 ,” Physical Review Letters 134 (2025): 086401, 10.1103/Phys Rev Lett.134.086401.40085913 · doi ↗ · pubmed ↗
- 6A. Nozariasbmarz , B. Poudel , W. Li , H. B. Kang , H. Zhu , and S. Priya , “Bismuth Telluride Thermoelectrics With 8% Module Efficiency for Waste Heat Recovery Application,” Iscience 23 (2020): 101340, 10.1016/j.isci.2020.101340.32688286 PMC 7369584 · doi ↗ · pubmed ↗
- 7J. Ballard , M. Hubbard , S.‐J. Jung , et al., “Nano‐Engineered Thin‐Film Thermoelectric Materials Enable Practical Solid‐State Refrigeration,” Nature Communications 16 (2025): 4421, 10.1038/s 41467-025-59698-y.PMC 1209556540399263 · doi ↗ · pubmed ↗
- 8T. Gong , C. Ma , L. Li , et al., “High‐Performance Planar Thin‐film Thermoelectric Cooler Based on Sputtered Nanocrystalline Bi 2Te 3/Bi 0.5Sb 1.5Te 3 Thin Films for On‐Chip Cooling,” ACS Applied Materials and Interfaces 17 (2025): 17008–17017, 10.1021/acsami.4c 19653.40045166 · doi ↗ · pubmed ↗