Introduction

Flexible thermoelectric technology can convert the temperature difference between the human body and the environment into electricity by the Seebeck effect, offering a potential solution to the challenge of providing a lightweight, safe, and sustainable power source for wearable electronics1,2,3. To improve the thermoelectric conversion efficiency, it is crucial to optimize the thermoelectric properties of the materials2. The thermoelectric performance is generally evaluated using the figure-of-merit (ZT), which can be expressed as \({ZT}=\frac{{S}^{2}\sigma }{\kappa }T=\frac{{S}^{2}\sigma }{{\kappa }_{e}+{\kappa }_{l}}T\), where S is the Seebeck coefficient, σ is the electrical conductivity, S2σ is the power factor, κ is the thermal conductivity, κe is the electronic thermal conductivity, κl is the lattice thermal conductivity, and T is the absolute temperature4. Since the S, σ, and κe are directly coupled with the carrier concentration (n), tuning the n is effective for optimizing the overall thermoelectric performance. To achieve this goal, historically, doping has been used to manipulate the band structure5, and introducing secondary phases can stimulate the energy filtering effect6, effectively decoupling the S, σ, and κe, thereby improving the ZT value. The κl is typically associated with phonons. Crystal and lattice defects introduce additional scattering centers into the phonon transport path, causing phonons with different wavelengths to encounter more obstacles during propagation and subsequently undergo scattering, thereby reducing the κl. However, these defects may also scatter charge carriers, thereby reducing the carrier mobility μ and lowering the σ, which makes the optimization of thermoelectric performance a challenging task.

Wearable thermoelectric devices operate based on the temperature different (ΔT) between the body surface and the surrounding environment7. Therefore, the materials in these devices need to exhibit high thermoelectric performance and flexibility within the near-room temperature range2. Bi2Te3 has been widely regarded as one of the best thermoelectric materials for near-room temperature applications8. However, limitations such as high costs and inherent brittleness hinder the assembly of flexible films using Bi2Te3-based materials7,9,10, making practical wearable applications challenging. As alternatives, silver chalcogenides Ag2X (where X is S, Se, and Te) show promising potential for near-room temperature applications11. Compared to Bi2Te3, Ag2S exhibits good ductility but lower thermoelectric performance12. Ag2Te is more brittle and has relatively high intrinsic n, resulting in insufficient thermoelectric performance optimization13. In this context, Ag2Se, after certain optimizations, can exhibit good thermoelectric performance in the near-room temperature range and can exhibit greatly improved ductility through doping, compositing, and hybridization strategies14,15,16. Currently, methods for preparing Ag2Se films include vacuum filtration17,18, cation exchange19, screen-printing20, co-evaporation21, magnetron sputtering22,23, and physical-vapor-deposition (PVD) methods24. As of today, the reported S2σ of Ag2Se-based films has sharply increased from 8 to 27 μW cm−1 K−217,18,19,20,21,22,24,25,26. Given its remarkable development pace, the potential application of Ag2Se in wearable/portable devices is evident.

Despite the flourishing research on Ag2Se films, significant controversies persist within this material system14. A major point of contention revolves around the relationship between the orientation of Ag2Se films and their thermoelectric performance. For example, compared to (013)-textured films, the films with strong (00 l) can yield high S and a high S2σ of 21.6 μW cm−1 K−227. This enhancement can be achieved in the oriented films by adjusting the Ag to Se ratio27. However, stronger (00 l) oriented Ag2Se films show significantly lower σ compared to those that are (013) oriented17. Other works reported that a strongly (201)-oriented Ag2Se thin film, prepared by PVD, exhibited a S2σ of 25.9 μW cm−1 K−2 within the near-room temperature range24. More recently, Te-doping-triggered strong (00 l)-oriented Ag2Se films, prepared by co-evaporation, showed a S2σ exceeding 24.8 μW cm−1 K−228. However, these studies, focusing on film preferred orientation, are still lacking clear experimental and theoretical validations14. Furthermore, most of these studies have not clarified the relationship between the preferred orientation observed through X-ray diffraction (XRD) and the orientation during performance measurement, as the direction of thermoelectric performance testing is typically in-plane, while XRD observation of film orientation is usually conducted in the out-of-plane direction. Therefore, those discussions regarding the thermoelectric performance of Ag2Se based on crystal orientation remain contentious14. Additionally, many previously reported strategies for enhancing the thermoelectric performance of Ag2Se thin films have focused on increasing σ, often targeting values above 2000 S cm−1 to achieve a higher S2σ29. However, a major drawback of this approach is that high σ also increases the κe of the film, a fact that is often overlooked. Therefore, effectively enhancing the S of Ag2Se thin films to improve overall thermoelectric performance should be the primary focus of research. Furthermore, there is still room for improving the practical application potential of the Ag2Se films, such as further improving their thermoelectric performance and stability, as well as simplifying the fabrication process14.

In this study, guided by the first-principles density functional theory (DFT) calculations and by finely adjusting the parameters of the selenization reaction, the orientation of Ag2Se thin films can be favorably altered, leading to high μ (~1500 cm2 V−1 s−1) and σ (1387 S cm−1), and in turn, an S2σ of 30.8 μW cm−1 K−2 at 343 K, which ranks as a top value in the literature17,24,25,27,28,30,31,32,33. Through computational calculations and characterizations, the presence of slight Se vacancies (VSe) and elemental Ag nanoinclusions in the thin films can be found to contribute to this exceptional thermoelectric performance. Furthermore, by introducing ethanol to reduce the liquid surface tension of the Se precursor, the adhesion of Ag2Se thin films to the polyimide (PI) substrate can be improved, ensuring the overall flexibility and durability of the films. Building upon this improvement, a slotted thermoelectric device was assembled. At a temperature difference (ΔT) of 20 K, the device generated a maximum output power (P) of 0.58 μW, a power density (ω) of 807 μW cm−2, and a normalized power density (ωn) of 1.8 μW cm−2 K−2.

Results and discussion

Although there is currently considerable debate surrounding the preferred orientation of Ag2Se thin films, there are still certain pivotal criteria that can be followed. For example, it is crucial to identify suitable crystal orientations that can enhance the weighted mobility (μw), thereby increasing the S2σ (quality factor). Consequently, we followed the guideline of seeking higher μ to design Ag2Se thin films with optimal orientations. To achieve this objective, we first investigated the crystal structure of Ag2Se. As depicted in Fig. 1a, near-room-temperature Ag2Se features an orthorhombic structure (powder diffraction file PDF#010807685)34. Within the crystal structure, silver ions (Ag+) and selenium ions (Se2−) are arranged in the lattice in specific proportions. Consequently, the arrangement of Ag and Se atoms varies across different crystal planes. Figure 1b illustrates the calculated electron charge density distributions on the nearly parallel planes, specifically (002), (013), and (014). Here, “nearly parallel planes” indicates that (013) and (014) form a small angle (<15°) with (002), with only a 4° angle between (013) and (014). Supplementary Figs. 1 and 2 display the charge density distributions on planes perpendicular to (00 l) and (01 l). In comparison to the views of (100) and (010) planes with high charge accumulation, the (00 l) plane exhibits relatively ordered periodic and lower charge concentration17. High charge density distribution may trap the carriers, and prevent electrons gaining sufficient energy for transition, reducing the effective carrier concentration during thermoelectric testing. It should be noted that the observed out-of-plane (00 l) in XRD indicates that the thermoelectric performance test is conducted along the in-plane direction, which is perpendicular to the (00 l) plane. Similarly, the (01 l)-perpendicular planes show a non-uniform charge distribution with a larger periodic structure. This uneven distribution causes electron-electron and electron-atom collisions during transit, resulting in unnecessary carrier scattering and a subsequent decrease in μ. In contrast, planes like (h00)/(0k0), which are perpendicular to (01 l)/(00 l), exhibit a more uniform distribution, offering potential for higher μ. Therefore, to achieve Ag2Se thin films with enhanced μ, we focus on orientations perpendicular to the nearly parallel (00 l) planes.

Fig. 1: Introduction of highly orientated Ag2Se thin films.
figure 1

a Unit cell of orthorhombic Ag2Se (powder diffraction file PDF#010807685) with highlighted (013), (014) and (002) planes. b Electron charge density distribution for the nearly parallel planes (013), (014) and (002). c Illustrations of the fabrication process of the Ag2Se thin film and the comparison of the directions between the X-ray diffraction (XRD) detection and thermoelectric performance measurement. d Comparison of maximum power factor (S2σ) between this work and reported Ag2Se thin films in recent years17,24,25,27,28,30,31,32,33,36,37,38,39,40,41,42,43,44,45,46.

Full size image

In a thermoelectric thin film, the measured thermoelectric property is in-plane direction while the XRD scanning direction is out-of-plane of thin film, as depicted in Fig. 1c (bottom). This implies that when the obtained XRD patterns of the thin film show a strong nearly parallel planes of (00 l), it is possible for the strong anisotropy along the direction of thermoelectric performance measurement to be (100), (010) or other planes perpendicular to nearly parallel planes of (00 l). In other words, when the thin film shows strong nearly parallel planes of (00 l)-oriented, the measured macroscopic in-plane μ and σ will be higher. Therefore, it is better to obtain nearly parallel planes of (00 l)-oriented Ag2Se thin films, suggested by XRD.

In contrast to the previous reports, such as Te-doping-triggered orientation adjustment28, our objective is to design a relatively simple synthesis method that avoids unnecessary orientations while quantitatively demonstrating its application potential. With these goals, we proposed a combined approach using electron beam deposition and solution immersion to prepare Ag2Se thin films, as illustrated in Fig. 1c (top). More detailed fabrication processing and reaction principles is illustrated in chemical reaction principles and Supplementary Fig. 3 in the Supporting Information. The two-step fabrication method promotes the preferred orientation in Ag2Se thin films. To rigorously analyze the film orientation, we utilized two XRD scan modes. The 2θ scan (grazing incidence X-ray diffraction (GIXRD) scan modes) provides a broad view of crystallographic orientations in the sample, helping to identify the presence of any texture or preferred orientation. However, due to the limitations of the 2θ geometry, it may not reveal the specific planes with preferential alignment. In contrast, the symmetric θ-2θ scan emphasizes diffraction peaks associated with dominant texture features, enhancing preferred orientations while potentially suppressing other planes. This makes the θ-2θ scan particularly effective for identifying primary texture characteristics, albeit with reduced sensitivity to less prominent orientations. First, Ag thin films with a specific (111) preferred orientation were deposited on PI substrates using electron beam deposition, as shown in Supplementary Figs. 4 and 5. The (111) orientation of Ag, recognized for its close-packed spherical arrangement, demonstrates excellent adsorption with chalcogenides, which facilitates a thorough reaction with the Se precursor. This results in pre-heat-treatment samples with anisotropic properties, as shown in Supplementary Figs. 6 and 735. Please note that the Al peak in Supplementary Fig. 7 originates from the XRD holder and is not from the sample itself. Subsequently, through manipulation of thermal treatment, preferred orientation of (002), (013), and (014) were achieved. The significant orientation change may result from the recrystallization process occurring during heat treatment at 180 °C. With this strategy, we achieve an exceptionally high S2σ value of 30.8 μW cm−1 K−2, exceeding previously reported values, as shown in Fig. 1d17,24,25,27,28,30,31,32,33,36,37,38,39,40,41,42,43,44,45,46, Supplementary Fig. 817,18,21,24,25,27,28,30,31,32,33,37,38,39,40,47,48,49,50,51,52,53,54,55,56, and Supplementary Table 115,17,18,20,24,26,27,28,30,32,37,47,48,49,50,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67 in the Supporting Information. From the comparation, it should be noted that our S2σ is mainly contributed by S, which offers potential for achieving thin-film device application, as higher S leads to an increased output voltage (Voc).

To elucidate the influence of Se precursor on the macroscopic orientation of Ag2Se films, we varied the precursor concentration (x) during the selenization process, with x = 15, 10, 5, 2.5, and 1.25 mmol, for investigation and measured their GIXRD pattern, Fig. 2a is a 2θ mode pattern, displaying the lattice information of Ag2Se thin films in the 2θ range of 20 to 60°. All diffraction peaks match well with the standard near-room-temperature Ag2Se PDF card (#010807685), and no apparent impurities can be detected within the precision range of XRD examination. A rightward peak shift can be observed (Supplementary Figs. 9 and 10), possibly attributed to lattice contraction, caused by spontaneous selenium vacancies (VSe) in the films. As shown in Fig. 2a (2theta mode pattern), Ag2Se thin films prepared from different concentration precursors exhibit preferred orientation existing, which might be (002) and (013). However, as mentioned above, 2θ mode is not precise enough to judge the position of preferred orientation. Therefore, as shown in Fig. 2b, the XRD symmetric θ-2θ mode pattern reveals that the thin film exhibits preferred orientations along three main textures: (00 l), (013), and (014). These orientations can be considered nearly parallel planes relative to (00 l). The Lotgering factor of (00 l), (013), and (014) can be seen in Supplementary Fig. 11.

Fig. 2: Phase, structures, and compositions of Ag2Se thin films.
figure 2

a Grazing incidence diffraction GIXRD patterns (2θ mode) of Ag2Se thin films prepared at different Se precursor contents x (x = 15, 10, 5, and 2.5 mol). The 2θ ranges from 20° to 60°. b Symmetric mode (θ−2θ mode) of Ag2Se thin films prepared at different Se precursor contents x (x = 15, 10, 5, and 2.5 mol). The 2θ ranges from 20° to 60°. c X-ray photoelectron spectroscopy (XPS) spectra of Ag 3d3/2 and Ag 3d5/2 for the thin film with x = 10 mmol. Scanning electron microscopy (SEM) images of thin films with d x = 2.5 mmol and e x = 10 mmol from top views. The insets show corresponding SEM images from cross-sectional views. f SEM backscattered electron (BSE) image along with corresponding energy dispersive X-ray spectroscopy (EDS) maps of g Ag and h Se in the film with x = 10 mmol.

Full size image

According to Fig. 2b, when x = 2.5, 5 and 10 mmol, Ag2Se films show a strong (002) orientation. When x = 15 mmol, the (013)/ (014) orientation begins to strengthen. For the sample with x = 1.25 mmol, its uniqueness lies in the distinct presence of pure Ag peaks alongside the monoclinic crystal phase of Ag2Se in the diffraction pattern, attributed to an excess of Ag due to insufficient Se sources during the selenization reaction, as shown in Supplementary Fig. 12. Moreover, when the Se concentration is too low, there is a competition reaction between S and Se in the precursor during selenization (the precursor includes Na2S·9H2O and Se powder). Due to the much lower concentration of Se atoms compared to S atoms, a transition phase of Ag2SeδS1-δ is formed68, as shown in Supplementary Fig. 13. Specifically, the samples with x = 2.5, 5, and 10 mmol exhibit similar preferential orientations, predominantly along the (00 l) planes, with noticeable (013)/(014) orientations also present. Since no additional elements were introduced into the system and the heat treatment conditions remained consistent, we attribute the formation of these nearly parallel (00 l) planes to the recrystallization of the films during the heat treatment process. In contrast, the x = 15 mmol sample displays a stronger (013)/(014) orientation. This difference can be attributed to the fact that, as shown in Supplementary Figs. 9 and 10, its unit cell contraction is less pronounced compared to the other samples. This suggests that it may have fewer Se vacancies, potentially leading to more distinct (013)/(014) orientations during the recrystallization process.

To investigate the extent of residual elemental Ag in the films after solution-phase selenization, we examined the valence state of Ag in the films at x = 10 mmol by X-ray photoelectron spectroscopy (XPS). As shown in Fig. 2c, the XPS spectrum primarily displays Ag+ valence states associated with the Ag2Se matrix with a small amount of Ag0, and the peak positions are consistent with values reported in the literature69,70. However, at x = 10 mmol, we did not observe crystalline signals of pure Ag in the GIXRD diffraction pattern. One potential reason may be due to the extremely small amount of residual crystalline Ag that is below the detection limit of XRD when the Se precursor concentration is sufficiently large, while another reason should be that the sample at x ≥ 2.5 mmol may contain a small amount of nano-crystalline Ag, which is difficult to be examined by XRD due to its board humps.

To better understand the structure and composition of Ag2Se films prepared from solutions with different concentrations of Se precursor, we investigated sample morphology using scanning electron microscopy (SEM) and conducted energy-dispersive X-ray spectroscopy (EDS) analysis to analyze composition. Figures 2d and 2e display top-view SEM images of Ag2Se films with x = 2.5 and 10 mmol, revealing significant differences in morphology. This observation reinforces the notion that, despite the similar preferred orientations indicated by the XRD results, variations in precursor concentration still influence the morphology and crystallization of the thin films. Consequently, these differences in morphology can potentially lead to variations in the performance of the thin films. Insets in Fig. 2d, e show cross-sectional SEM images, indicating an average film thickness of approximately 800 nm. To investigate the uniformity of film composition, the backscattered electron (BSE) image and corresponding EDS maps of the sample with x = 10 mmol are provided in Fig. 2f–h. As can be seen, the distributions of Ag and Se are uniform without significant elemental segregation. Additional SEM and corresponding EDS information on samples prepared using different Se precursor contents are shown in Supplementary Figs. 14–24 in the Supporting Information.

To obtain insights into the microstructure and detailed composition of Ag2Se thin film, we prepared transmission electron microscopy (TEM) specimens using focused ion beam (FIB) technology for characterization using aberration-corrected scanning TEM (Cs-STEM). Figure 3a depicts a low-magnification high-angle annular dark-field (HAADF) image of the FIB-prepared sample of Ag2Se film with x = 10 mmol. The image reveals a relatively dense film without significant porosity, exhibiting good continuity at the microscopic scale. However, some defect-like structures causing local variations in image contrast are observed. Additionally, insets depict corresponding selected-area electron diffraction (SAED) patterns, confirming the view direction of the FIB sample along [100]. Since the observation plane of the samples prepared by FIB is perpendicular to the film (owing to the FIB preparation method), this confirms a high probability of subsequent film thermoelectric performance measurement along [100] (parallel to the (100) orientation), aligning with our experimental design concept. Figure 3b is a high-resolution Cs-STEM HAADF image from a normal area selected from Fig. 3a. The lattice arrangement in the selected area appears ordered, confirming that the normal matrix of the film is highly crystalline, validating our film preparation process to achieve high crystalline quality. The viewed zone axis is along the [100] direction, indexed by the corresponding unit cell shown in Supplementary Fig. 25 for reference. However, the image presents lattice distortion along a certain direction. Figure 3c, obtained from an area near the contrast difference in Fig. 3a, highlights potential point defect regions within the matrix, likely originating from VSe, which aligns with the rightward peak shift observed in the XRD results. Figure 3d presents a Cs-STEM HAADF image from a typical area selected from Fig. 3a, accompanied by corresponding strain maps in different directions, which reveal significant strain in the x-direction, indicating potential lattice distortion. Figure 3e, from another region near the defect, exhibits a stronger contrast difference, indicating more pronounced lattice distortion. The inset shows the corresponding fast Fourier transform (FFT) pattern. Its slight pattern overlap confirms subtle variations in the zone axis due to severe lattice distortion in that area. Figure 3f provides a filtered image corresponding to Fig. 3e, exhibiting a typical interplanar spacing of 3.81 Å in the (002) orientation, perpendicular to the [100], which can be corroborated with the previous XRD results, and potential edge-like dislocations within the selected range (marked by red squares). The inset image shows an enlarged view of the red square-marked area, featuring a suspected edge-like dislocation possibly triggered by the formation of VSe. Figure 3g is an enlarged image of the highly distorted area in Fig. 3a, revealing ripple-like lattice contrast differences, indicating more pronounced lattice distortion, supported by the observed overlapping FFT pattern shown in Fig. 3h as inset. Figure 3i displays a low-magnification Cs-STEM HAADF image, corresponding EDS maps, and determined atomic ratios of Ag and Se. Overall, the distribution of elements appears relatively uniform, but slight Ag enrichment regions can be observed from the EDS and HAADF images (as indicated by arrows). Combining the micro-area EDS results from Fig. 3i with our previous XRD and XPS results, we infer the presence of minor amorphous pure Ag inclusions in the sample. For comprehensive lattice calibration information, refer to the SAED images in Supplementary Fig. 26.

Fig. 3: Micro/nanostructural characterizations of Ag2Se thin film with x = 10 mmol.
figure 3

a Transmission electron microscopy (TEM) high-angle annular dark-field (HAADF) image of the specimen prepared by focused ion beam (FIB) technique. The inset is the corresponding selected area electron diffraction (SAED) pattern with indexed lattice information. b High-resolution spherical aberration-corrected scanning TEM (Cs-STEM) HAADF image of the matrix taken from a normal area in (a). c Cs-STEM HAADF image with areas of different contrast caused by intensive point defects taken from (a). d Cs-STEM HAADF image with corresponding strain maps along different directions. e Another Cs-STEM HAADF image taken from (a). The inset is the corresponding fast Fourier transform (FFT) pattern with indexed lattice information. f Filtered image taken from e suggesting the potential presence of edge-like dislocation. The inset shows the magnified edge-like dislocation area. g Cs-STEM HAADF image taken from the highly distorted area in (a). h Cs-STEM HAADF image taken from (g) with highly distorted area. The inset shows the corresponding FFT pattern with indexed lattice information. i HAADF image of a particle in the matrix of Ag2Se with corresponding EDS maps and determined atomic ratio between Ag and Se.

Full size image

To illustrate the potential impact of VSe and elemental Ag formation on the thermoelectric properties of Ag2Se thin films, we conducted first-principles calculations using DFT. Initially, we computed the formation energies of Ag vacancies (VAg), VSe, and point defects by Se replacing Ag (SeAg), which were found to be 2.57, 1.15, and 5.07 eV, respectively. The formation energy of VSe at only 1.15 eV suggests their likelihood to exist during the solution-phase selenization process of Ag films, leading to lattice contraction, consistent with the overall right shift observed in XRD characteristic peaks. Consequently, the actual film should exhibit n-type semiconductor transport characteristics. Furthermore, we calculated the band structures of intrinsic Ag2Se and Ag2Se with Ag and Se vacancies, as well as the band structure of pure Ag. Figure 4a illustrates the band structure of intrinsic Ag2Se, where it is evident that the semiconductor properties of pristine Ag2Se are contributed by both the valence band and conduction band. Figure 4b depicts an enlarged view of the Fermi level position in the band structure of intrinsic Ag2Se. The conduction band is seen to intersect the Fermi level, suggesting that the material may exhibit semi-metallic characteristics. However, theoretical calculations are generally performed under 0 K conditions, which can introduce discrepancies compared to experimental results. Consequently, we speculate that pristine Ag2Se may have an extremely narrow bandgap, allowing for higher excitation energies of electrons that facilitate electron transitions. This characteristic could significantly contribute to its high σ. In the presence of VAg (Fig. 4c), higher contributions from the valence band lead to a material transition towards p-type conductivity. This transition is further evidenced by the enlarged electronic structure (Fig. 4d). However, as mentioned earlier, the formation of VAg in Ag2Se is challenging, which is why practical Ag2Se materials exhibit n-type thermoelectric properties. Conversely, in the presence of VSe, the conduction band crosses the Fermi level, with the conduction band contributions prominent in the band structure, resulting in typical n-type thermoelectric characteristics of the material, as shown in Fig. 4e. Figure 4f illustrates the band structure of pure Ag, where conduction and valence bands overlap, exhibiting typical metal conductivity. However, the significant electronic structural differences between Ag and Ag2Se may lead to the presence of Ag nanoscale inclusions in the Ag2Se samples potentially enhancing both κl and S.

Fig. 4: Electronic structures of Ag2Se1−δ/Ag.
figure 4

Calculated band structures of a pristine Ag2Se (Ag32Se16), b magnified Ag32Se16, c Ag2Se with Ag vacancies (Ag31Se16), d, magnified Ag31Se16, e Ag2Se with Se vacancies (Ag32Se15), and f pure Ag (Ag48).

Full size image

To determine the specific impact of precursor Se content on the thermoelectric properties of Ag2Se thin films, we measured the thermoelectric properties of Ag2Se films corresponding to x = 15, 10, 5, and 2.5 mmol. The performance of the sample with x = 1.25 mmol can be found in Supplementary Fig. 27, where severe inadequacy in selenization significantly affects the overall thermoelectric properties of the film due to the decisive influence of Ag on the thermoelectric properties. Figure 5a–c display the temperature-dependent σ, S, and S2σ of Ag2Se. As can be seen, Ag2Se films with x = 10 and 5 mmol exhibit higher σ, leading to S2σ values as high as 30.8 and 27.5 μW cm−1 K−2 at 343 K, respectively. The result of S2σ > 30 μW cm−1 K−2 can be repeated when the material was synthesized by selenization process with a selenium precursor at x = 10 mmol under E-beam deposition conditions with voltages (V) and currents (I) of 10 kV and 31 mA, respectively, which are depicted in Supplementary Figs. 28 and 29. The measurement method using ZEM-3 is illustrated in Supplementary Fig. 30. When the x is sufficiently large (x greater than or equal to 5 mmol), the S are essentially the same, indicating that the change in thin-film orientation has a weak effect on the S, as hypothesized earlier. When x is small (such as x = 2.5 mmol), the presence of more residual Ag impurities significantly reduces the S. Compared with other literature17,24,25,27,28,30,31,32,33,36,37,38,39,40,41,42,43,44,45,46, our as-fabricated thin-film sample with precursor Se content x = 10 mmol exhibit relatively good value in both σ and S, which results in overall optimized S2σ. This phenomenon may be attributed to the energy filtering effect71.

Fig. 5: Thermoelectric performance and charge distribution of Ag2Se thin films at different Se precursor content x (x = 15, 10, 5, 2.5 mmol).
figure 5

Temperature-dependent a electrical conductivity (σ), b Seebeck coefficient (S), and c S2σ. Here, 5% error bars are employed. d Room-temperature carrier concentration (n) and mobility (μ) as a function of x. Here, 5% error bar is employed e Room-temperature effective mass (m*) and deformation potential (Edef) as a function of x calculated by a single parabolic band (SPB) model. f Comparison of the charge distribution maps for Ag2Se with Se vacancies along different orientations. x-dependent room-temperature g thermal conductivity (κ) and h lattice thermal conductivity (κl) and electrical thermal conductivity (κe). Here 10% error bar is employed for (g), and 5% error bar is employed for (h). i Estimated temperature-dependent ZT. Here 5% error bar is employed.

Full size image

To explore deeper into the nature of the variations in σ and S, we explored the relationship between μ and n at room temperature for different films, as shown in Fig. 5d. Ag2Se films with x = 10 and 5 mmol exhibit high μ values of 1500 cm2 V−1 s−1, validating the correctness of our orientation design concept. Additionally, the values of n fall within the range of 4.94 × 1018 cm−3, which is within a reasonable range compared to the reported optimal n of 1–2 × 1018 cm−3 68. In this regard, the n of Ag2Se films synthesized under x = 15 mmol precursor is closer to the optimal n. Moreover, the μ of the samples with x = 10 mmol and 5 mmol is similar. However, the overall electrical performance of the x = 15 mmol sample is lower. This can be explained by thin film anisotropy. Considering the XRD analysis results provided earlier, Ag2Se films prepared with x = 10 and 5 mmol exhibit a stronger preferred orientation along the (00 l) direction, where the in-plane direction (i.e., the direction of properties testing) results from planes perpendicular to nearly parallel planes of (00 l) and (01 l) such as (100) and (010), thus facilitating excellent charge transport. However, samples with precursor x = 15 mmol exhibit strong orientation in (002), (013), and (014) directions, complicating the crystal orientation for in-plane performance testing. This complexity might lead to significant carrier scattering at grain boundaries during the transport, resulting in a decrease in overall conductivity. For Ag2Se films synthesized with precursor concentration x = 2.5 mmol, the relatively pronounced unit cell contraction observed in the GIXRD results, combined with the increased in-plane nanoinclusions resulting from the lower precursor concentration, leads to higher n and lower μ. This combination ultimately results in a simultaneous degradation of both the S and σ.

To explain the variation in n and μ of the thin films, the effective mass m* and deformation potential Edef of films with different x at room temperature were calculated by using the single parabolic band (SPB) model72, as shown in Fig. 5e. The variation in m* is not particularly significant, mainly due to the insignificant change in n with varying precursor concentration x. For the sample with x = 15 mmol, the change in m* might stem from the lower n value. Additionally, with decreasing x, a slight increase in m* can be seen, attributed to the reduced completion degree of the selenization reaction, implying more residual nanoscale Ag inclusion phases. Since more residual inclusion phases increase the average intensity of interfaces between the nanophase and the Ag2Se matrix, the energy filtering effect becomes more pronounced, thus leading to more scattering of low-energy carriers and in turn, an increase in m*. Ag2Se films synthesized with precursor concentrations x = 10 and 5 mmol exhibit relatively low Edef, indicating relatively easy lattice deformation capabilities. This indirectly confirms the enhanced orientation of the films, leading to relatively higher μ. Therefore, we demonstrate the consistency between the experimental and calculated results of the electrical transport properties.

Furthermore, considering the presence of VSe in actual thin-film materials, we calculated the charge distribution of Ag2Se incorporating VSe, as shown in Fig. 5f. Compared to the (002), (013), and (014) planes in Ag2Se without VSe, clear atomic vacancies are observed in the charge distribution of Ag2Se with VSe, resulting in a more uneven charge density distribution. Consequently, from the calculated charge distribution results, we can infer the significance of achieving a strong orientation.

Besides the electrical transport properties of the thin-film materials, we utilized the laser photothermal intensity technique (PIT) alternating current (AC) method for thermal diffusivity D measurements. The methodology and principles for testing D of thin films are described in Supplementary Figs. 31 to 35. To calibrate the κ of the films, we determined the Lorentz parameters L of Ag2Se films by the SPB model (refer to Supplementary Fig. 36) and calibrated the porosity (refer to Supplementary Fig. 37). Porosity is inversely proportional to material density (ρ); therefore, as porosity increases, the material density decreases. Since the results obtained from the equipment are thermal diffusivity (D), and according to the formula D = \(\frac{\kappa }{\rho c}\) (here c refers to specific heat capacity), the ρ directly affects the calculation of κ. Therefore, we used ImageJ software to calibrate porosity, aiming to obtain relatively accurate values of κ. In this context, we obtained convincible κ data for Ag2Se films, as shown in Fig. 5g. Overall, the κ of the films are relatively low. Figure 5h compares the κe and κl of the Ag2Se films. κe can be calculated using the formula κe = LσT73, and κl can be calculated by subtracting κe from κ. It can be observed that for Ag2Se film samples with x = 10 and 5 mmol, the κl are relatively low due to the presence of defects in various dimensions observed in TEM characterizations. However, the κe are relatively high, which is mainly attributed to the excellent σ. Based on reliable measurements of σ, S, and κ, we obtained the overall room-temperature ZT for Ag2Se films with different x, reaching a maximum value of 0.9, as shown in Fig. 5i. Theoretical and experimental ZT as a function of n calculated from the SPB model are presented in Supplementary Fig. 38 for reference.

To quantify the practicality of the films, we conducted durability and flexibility tests on Ag2Se films. Figure 6a displays the performance changes of the films before and after exposure to the air for six months. There is no significant performance change, demonstrating that Ag2Se films prepared using electron beam and solution methods exhibit high stability. The slight increase in S2σ may be attributed to the effect of humidity on S74. This high longevity may stem from the well-crystallized nature of the thin films. Furthermore, we subjected samples prepared with different x to 1000 cycles of bending with a r of 5 mm (Fig. 6b and Supplementary Fig. 39), as well as 200 cycles of bending for the Ag2Se thin film with x = 10 mmol at different r values (Fig. 6c). The inset in Fig. 6c illustrates the process and method of the bending test. The results indicate that the normalized resistance change (ΔR/R0) of the film consistently stays within 10%, reflecting the high stability and flexibility of Ag2Se films. The high flexibility and stability are derived from the optimized processes of electron beam evaporation and subsequent selenization reactions, controlling the thickness of the film, and ensuring excellent contact between the film and the flexible PI substrate.

Fig. 6: Durability and flexibility of Ag2Se thin films and their device performance.
figure 6

a Durability testing results of thin films with different x (comparison of S2σ values before and after 6 months). Here, 5% error bar is employed. b Normalized resistance change (ΔR/R0) of thin films with different x as a function of bending cycle with a fixed bending radius r of 5 mm. c ΔR/R0 of Ag2Se thin film with x = 10 mmol as a function of r under 200 bending cycles. The inset photos exhibit the high flexibility of thin films. Here, 5% error bar is employed. d Three views of slotted device including front view at the top, side view at the middle, and the top view at the bottom. e Relations between voltage (V)and current (I) at different ΔTs, as well as f relations between I and output power (P) at different ΔTs. g Determined power density (ω) at different ΔTs. The inset is the wearability demonstration of the slotted thin film TED. h Photograph demonstrating the wearability of the four-leg conventional horizontal TED. i Measured V of the four-leg device while being worn during sitting, walking, and running as a function of duration.

Full size image

To validate the practical application potential of the prepared thin film, we designed a novel slotted thin-film device, as illustrated in Fig. 6d (with dimensions detailed in Supplementary Fig. 40). The device consists of three sets of three-leg modules connected in series. This configuration enables the stacking of multiple modules to achieve high-performance output. As shown in Fig. 6e, f, under a ΔT of 20 K, the Voc can reach up to 27.1 mV, and the output power (P) reaches 0.58 μW. The power density (ω), as depicted in Fig. 6g, is 807 μW cm−2, corresponding to a normalized power density (ωn) of 1.82 μW cm−2 K−2. The inset image in Fig. 6g is a wearable display of the slot device. Also, a Voc of 13.8 mV while wearing is shown in Supplementary Fig. 41. To verifying the accuracy of the performance test, we employ the formula ρ = RA/l, where ρ denotes the resistivity, A denotes the cross-sectional area, and l denotes the length, the σ of the device (σd) can be calculated to be ~925 S cm−1, exhibiting 77% of the thin-film material, indicating a reasonable compatibility between our device and material performance. However, the internal resistance of silver paste and the interface resistance between the material and silver paste are the main factors affecting the σd. To verify the difference between the P of our device and the theoretical power in the structure, we employ a matched load power formula: Pmax = \(({\frac{{V}_{{oc}}}{{R}_{{{\rm{in}}}}+{R}_{{{\rm{load}}}}}})^{2}\)× \({R}_{{\mbox{load}}}\), where Voc is calculated by Voc = n × S × ΔT (n refers to the number of legs), where Rin refers to device internal resistance of the slotted device, and Rload refers to external resistance in the test circuit. The tested values represent 90% of the calculated values, indicating the maximum P of the device under this structure. Additionally, the test results, along with the theoretical predictions, are presented in Supplementary Fig. 42.

Apart from the slotted thin-film device, we also assembled a conventional horizontal four-leg TED for wearing as shown in Fig. 6h and Supplementary Fig. 43. Supplementary Figs. 44 and 45 exhibits that the P and Voc closely match the theoretical values calculated for the design, with the maximum P of 115 nW at a ΔT of 28 °C. To verify the usability of the four-leg TED, we also adhered the device to the body surface and evaluated its voltage output under different motion states (Fig. 6i). Through these series of tests, the Ag2Se thin-film-based device manufactured using the electron beam/solution method has been proven to have excellent durability and practical prospects. However, to achieve better performance output, reasonable optimization in design is still inevitable.

In this work, we designed highly oriented Ag2Se thin films using the electron beam and solution immersion method, introducing a novel approach to controlling film anisotropy by adjusting the concentration of Se precursor to optimize its thermoelectric performance. Under selenization with a selenium content of 10 mmol, the electrical transport direction was steered away from the uneven charge distribution associated with the nearly parallel planes of (00 l) orientation. The combined effects of thin film anisotropy optimization, a small quantity of VSe, and Ag nanoinclusions in the thin film resulted in an exceptionally high S2σ of 30.8 μW cm−1 K−2 at 343 K. The precisely controlled film thickness, crystallinity, and strong adhesion to the flexible polyimide substrate led to excellent flexibility and stability, with performance variation staying within 5% after 2000 bending cycles at a radius of just 5 mm. The films also exhibited high durability, maintaining over >90% of with power factor after six months. A slotted, wearable thermoelectric device was fabricated using the optimized thin films, achieving a Voc of 27.1 mV, a P of 0.58 μW, and a ω of 807 μW cm−2 at a ΔT = 20 °C with a corresponding ωn of 1.8 μWcm−2 K−2. This research introduces a new method for controlling film orientation, with significant implications for high-performance, thin-film-based thermoelectrics.

Methods

Materials synthesis

Pure Ag thin films were prepared on PI film using electron beam evaporation (PVD 75 E-beam, Kurt J. Lesker). A carbon crucible (1.167 inch top × 0.563 inch height × 0.093 inch wall thickness and 15-degree wall angle) was employed to contain Ag particles (purity 99.99%, Kurt J. Lesker), which were placed into the deposition chamber. The PI film (30 mm × 30 mm × 0.12 mm, Cole-Parmer) underwent a 15-min sonication in ethanol before being adhered to the deposition plate inside the chamber. The background pressure was maintained below 5 × 10−6 Torr. Ag deposition source parameters were set to I of 31 mA and V of 10 kV, with thickness controlled by a detector to approximately 500 nm. Subsequently, the Ag thin film, prepared via electron beam deposition, underwent selenization using a solution method. The solvent portion of the precursor, comprising 20 ml of water and 20 ml of ethanol, was precisely measured using a graduated cylinder and thoroughly mixed in a centrifuge tube. Se powder (100 mesh, 99.99%, Sigma Aldrich) was dissolved in a mixed solution of ethanol and water, in varying amounts: 15 mmol, 10 mmol, 5 mmol, 2.5 mmol, and 1.25 mmol. The mixed solution also contained Na2S·9H2O (98%, Sigma Aldrich) in amounts of 60 mmol, 50 mmol, 20 mmol, 10 mmol, and 5 mmol, respectively. After soaking in the reaction for 20 min, the thin films were removed, and surface residue was washed with isopropanol and a small amount of water. After cleaning, the samples were placed into an oven (ACROSS international) and annealed at 180 °C for 10 h to obtain the Ag2Se thin film samples for testing.

Characterizations

Grazing incident XRD analysis was conducted using a Rigaku Smart Lab instrument with CuKα radiation across an angular range of 20° to 60° in 0.02° increments to ascertain the crystal orientation of the Ag2Se thin-film samples. XRD results were refined using TOPAS Pawley refinement, achieving a weighted profile p-factor (RWP) of 9.90 and a goodness of fit (GOF) of 1.17. XPS analysis was conducted using Kratos Axis Supra Photoelectron Spectroscopy. Morphological analysis, mapping, and composition assessment of the samples were performed using a JOEL 7001 F SEM. An EDS detector from Bruker, the EDS QUANTAX, was utilized for EDS analysis. Furthermore, lamina samples of Ag2Se thin films were prepared using the FEI Scios FIB. The Cs-STEM (Hitachi HF5000) was employed for HAADF in STEM mode to conduct microanalysis of Ag2Se thin film FIB samples.

Thermoelectric performance evaluation

The σ and S were measured using a Seebeck coefficient and electrical conductivity apparatus (ZEM-3). The D was determined using the alternative current method measurement system (RIKO Laser-PIT). The n and μ were investigated using a Van der Pauw Hall measuring instrument (CH-70, CH-magnetoelectricity Technology Co., Ltd., China) under a magnetic field up to 500 mT. The n and μ were determined by n = 1/eR and μ = σR, respectively.