Introduction

Low-grade heat below 373 K is about 2/3 of the total waste heat in daily life, such as industrial waste heat, body-released sensible heat, and electronics-dissipated heat, etc., the effective utilization of which is expected to relieve energy crisis1. Thermoelectric (TE) conversion system can directly convert heat into electricity, showing considerable promise for harvesting low-grade heat2. In comparison to traditional TE materials based on inorganic semiconductors and semi-metals, which suffer from low thermopower, rigid, and high cost, the emerging ionic thermoelectric (i-TE) materials in recent years, including TE ionogels, polyelectrolytes, and hydrogels, etc., have attracted much attention owing to their nature of eco-friendly property, flexibility, and high thermopower (>10 mV K−1)3,4.

Among the i-TE materials, TE ionogels - that is, polymeric networks swollen with salts, especially have emerged as a potential material for harvesting heat from complex geometrical heat source, owing to their quasi-solid state, excellent thermal stability, and good mechanical properties. Over the past few years, great effort has been made to extend the scope of TE ionogels material and optimize the TE performance. Polymeric systems including polyethylene oxide (PEO), polyvinyl alcohol (PVA), poly(vinylidene fluoride-co-hexafluoropropylene) (PVDF-HFP), and bacterial cellulose (BC) swollen with salts have been proven effective for constructing TE ionogels5,6,7,8. And thermal mobility difference between anions and cations inside these polymeric systems accounts for the generation of thermopower. To further enhance the thermopower, enlarging the thermal mobility difference between the anions and cations is required. The strategies including selective combination of polymeric networks and multiple ionic species, incorporation of inorganic additives into polymeric networks, and modulation of humidity conditions, have been confirmed effective to enlarge this thermal mobility difference. For example, Li et al. prepared a hybrid i-TE gel containing 1-ethyl-3-methylimidazolium tetrafluoroborate (EMIM BF4) ionic liquid (IL), polyethylene glycol (PEG), and poly(acrylamide)/ alginate (Pam-alginate) double-network. By introducing PEG9, the electrostatic attraction interaction between PEG and EMIM+, as well as the electrostatic repulsion interaction between PEG and BF4-, enabled a larger ionic mobility differential, thereby ensuring an enhanced thermopower of 19.3 mV K−1. Besides, as reported by Jang et al. the incorporation of SiO2 additives into polyaniline: poly(2-acrylamido-2-methyl-1-propanesulfonic acid): phytic acid (PANI:PAAMPSA:PA) ternary polymer10 not only promoted ionic dissociation within the system, but also weakened the binding between cations and polymers, enhancing the p-type thermopower from 14.9 to 17.9 mV K−1. In addition, Liu et al. incorporated EMIM Cl into the EMIM TFSI/PVDF-HFP ionogel to enable a thermopower of 9.5 mV K−1 at humidity of 40% RH, and the thermopower could be further improved to 19 mV K−1 under high humidity (70% RH)6. Despite the effectiveness of the forementioned strategies in enlarging thermopower, the complexity of multicomponent systems and high humidity conditions inevitably lead to trade-offs between thermopower, mechanical properties, homogeneity, and liquid leakage, limiting the practical application11,12. Thus, the exploration of advanced polymeric network, that can intrinsically enable high ionic thermal mobility difference to contribute a high thermopower, would be highly desirable.

To promote the TE ionogels toward practical application, beyond materials performance, rational-design device that can adapt to heat source with complex-shaped geometry, at the same time without compromising the TE performance, are equally important. However, the development of i-TE devices is still in infancy, whose advances fall far below that of conventional electronic TE device. The current i-TE devices are integrated by connecting i-TE films in series on a planar substrate13. However, the substrates unavoidably limit the practical application owing to four aspects: (1) The softness of the substrate is far inferior to quasi-solid TE ionogels itself, weakening the intrinsic advantages of ionogels in contacting complex-shaped thermal sources; (2) The substrate will inevitably cause thermal shunting, which will damage the final output performance of the device; (3) During long-term heat harvesting, the different thermal expansion of ionogle and the substrate will result in unreliably contacting low-grade (<373 K) heat source, causing device failure; (4) Planar two-dimensional devices cannot capture vertical heat flow in space. Therefore, how to establish an effective i-TE device, that can self-adapt to complex-geometric heat sources and capture vertical heat flow, remains particularly important yet challenging.

Taking the above challenges of TE materials and devices in mind, the integration of thermally actuated function and high thermoelectric function in TE ionogels come into being, which is expected to enable dynamically adaptive heat harvesting from low-grade heat source. A recently reported work developed an ionic thermoelectric actuator14, but thermopower is only 0.6 mV K-1, and the actuating feature originates from the physically integrated multilayers, which limited the application in efficient heat harvesting. How to achieve both high TE performance and actuated function in monolithic materials, at the same time ensure effectiveness at device level is challenging. A thermally actuating material - liquid crystal elastomer (LCE) may provide the entrance15. LCE possesses a tunable network composed of mesogenic units chemically incorporated into soft polymer chains, in which the soft chain with functional groups (including ether bonds C-O-C) could provide platform tailoring the ion-polymer interactions, at the same time the mesogen alignment could provide platform regulating ion transport. Thus, LCEs may not only serve as an effective polymer matrix for high thermopower via tuning ionic thermal mobility, but also construct a dynamically adaptive TE device attributing to the thermally responsive properties of LC, providing an innovative idea for future designing thermoelectric ionogels16,17.

Here, we report a LCE-based p-type TE ionogel fiber, composed of LCE network and 1-Ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide (EMIM TFSI), to address the above challenges in both i-TE materials and devices. The fiber molding contributes to aligned mesogens, which in turn induce a remarkable ~3-fold boom in thermopower (25.8 mV K−1) and a dramatic ~30-fold electrical conductivity boom (21.5 mS m−1). Meanwhile, combining the LC phase-induced axial actuation inside the i-TE fibers and the specific weaving patterns, a gripping-like and textiles-based actuatable i-TE device was successfully fabricated, which could achieve dynamically adaptive heat harvesting from complex-shaped heat source, and exhibit decoupled identification of size/shape and temperature of heat source. This design of LCE i-TE fibers not only provides an innovative approach for optimizing i-TE performance, but also unlocks the integration of actuator science with TE ionogels, offering a unique entry point for the application of TE ionogels.

Results

The thermal voltage of TE ionogels is produced by the difference in migration rate between cations and anions in a polymeric network under temperature difference, thus how to enlarge the advantageous migration of anion or cation as possible in polymer network is significant for achieving high thermopower. In order to achieve the above vision, a LCE network is synthetized and designed (Supplementary Fig. 1 and Fig. 1a), in which 2,2’-(ethylenedioxy) diethanethiol (EDDET) with ether bond C-O-C is served as the soft chain, aiming to provide the possibility of selective/preferential reactions with cations to produce ion migration difference, at the same time the orientation of the mesogen (1,4-Bis-[4-(3-acryloyloxypropyloxy) benzoyloxy]-2-methylbenzene-RM257) in LC phase is fine-tuned and aligned, aiming to further expand this migration difference. As a result, a strategy of synergies of oriented channel and intermolecular interaction is processed to enhance the thermopower via a “highway” for the rapid passage of ions. This synergistic effect is expected to promote a high intrinsic thermopower of the LCE ionogels.

Fig. 1: Design principle and mechanism analysis on synergistic effect of the intermolecular interaction and the mesogen orientation in a LCE network for a high intrinsic thermopower.
figure 1

a Design principle. b 1D intensity profiles of LCE fibers and films. Inset picture are 2D WAXD pattern of LCE films (down) and fibers (top). c Schematic of molecular chain inside LCE fibers and films. d–e Local graph of FTIR spectra for LCE fibers showing the stretching vibration band of S=O and bending vibration band of C-O-C. (f) Raman spectra for TFSI- in LCE fibers at wavenumber range from 800–600 cm−1. g–h MD snapshots of LCE i-TE fibers and films. i The binding energy between EMIM+/TFSI- and molecular chains in LCE i-TE fibers and films. j Diffusion coefficients (MSD) of EMIM+ and TFSI- in LCE i-TE fibers and films. Error bars denote mean ± standard deviation.

Full size image

According to the design principle, to optimize mesogen alignment for ion motion inside LCE materials, fiber molding process were carried out as schematically exhibited in Supplementary Fig. 1. During injection, the shear force generated at the nozzle aligns the mesogen, ultimately forming ordered channels with oriented molecular chains18. In order to confirm the effect of fiber molding on the molecular chain structure, the two-dimensional wide-angle x-ray scattering (2D WAXD) analysis is performed on both fiber and film samples. It can be observed from the 2D WAXD color map (Fig. 1b) that both fiber and film state exhibit isotropic diffraction loops. In addition, the scattering intensity (I) versus scattering vector (q) profiles (q–I curves) in Fig.1b reveals distinct differences in the main peak positions between the fiber and film forms. Then, the molecular chain spacing (d) is concluded via the equation [1] \(d=2\pi /q\), and the concluded result shows that d of LCE fibers (4.58 Å) is slightly larger than that of the films (4.33 Å). Meanwhile, the order parameter (S) is further calculated, whose results indicate a larger S of LCE fibers (S = 0.34) than that of LCE films (S = 0.28)19,20. These larger d and S indicate a wider molecular chain spacing and smoother diffusion channels of fiber compared with film due to the higher liquid crystal orientation, which is expected to accelerate the rapid passage of ions (Fig. 1c).

Then, the selective/preferential reactions of cations from EDDET soft chain segment inside LCE materials are tailored by immersing LCE fibers in EMIM TFSI ionic liquid. And the Fourier transform infrared spectroscopy (FTIR) and Raman spectra are further carried out to provide physicochemical insights into molecular interaction mechanisms between the soft chain and ions. By comparing the FTIR spectra of EMIM TFSI ionic liquid, pure LCE fiber and EMIM TFSI treated LCE fibers for 1, 3 and 6 hours (Supplementary Fig. 2 and Fig. 1d, e), it is discovered that the S=O stretching band of TFSI- shifts from 1350 cm−1 to 1355 cm−1 after being introduced into LCE fibers21. And the characteristic absorption peaks of C-O-C of soft chain in 1247 cm-1 shifts to the higher wavenumbers after immersing in ionic liquid. These results indicate a hydrogen-like non-covalent bonding interaction between the TFSI- and soft chain18,22. In addition, a tiny characteristic peaks at 1188 cm−1 from C-N-C of EMIM+ is observed in EMIM TFSI treated LCE fibers, but the peak position remains unchanged (Supplementary Fig. 3)23. Meanwhile, the C-O-C bending vibrational peak of LCE at 1071 cm−1 shows no peak shifts and no disordered phenomenon in EMIM TFSI treated LCE fibers, which means no obvious hydrogen bond between EMIM+ and LCE molecular chain (Supplementary Fig. 4)19. Furthermore, as shown in the Raman spectra (Fig. 1f), the characteristic peaks of TFSI- of EMIM TFSI appears in 747 cm−1, while this peek shifts to 733 cm−1 when introducing EMIM TFSI into LCE fibers, at the same time, the half-peak width of TFSI- peaks gradually increase with the extension of immersion time, suggesting a strong interaction between TFSI- and soft chain inside LCE i-TE fibers24,25.

Finaly, molecular dynamics simulations (MD) are conducted to further validate the molecular chain ordering and ion/molecule interactions in both LCE i-TE films and i-TE fibers. From the MD snapshots (Fig. 1g, h), it can be clearly observed that LCE i-TE fibers exhibit more ordered molecular chain alignment compared to the disordered arrangement in LCE i-TE films, potentially providing effective channels for ionic diffusion. Then the binding energy between cations/anions (EMIM+/TFSI-) and molecular chains, as well as ion diffusion coefficients in LCE i-TE fibers/films are calculated. As displayed in Fig. 1i, molecular chains in LCE i-TE films, owing to their complex disordered structure, show higher binding energy with ions compared to those in LCE i-TE fibers, and TFSI- exhibits stronger binding energy with molecular chains than EMIM+ in both systems. This result suggests that ionic migration in LCE i-TE films may encounter greater hindrance than fibers. Meanwhile, ionic diffusion coefficients in LCE i-TE fibers are significantly higher than those in LCE i-TE films, with cation migration being dominant in both cases (Fig. 1j). Thereby, as can be concluded from the above experimental and computational results, owing to the synergistic effect of the intermolecular interaction and the mesogen orientation, ion migrates faster inside LCE i-TE fibers than LCE i-TE films, and EMIM+ migrates faster than TFSI- in LCE network, which is more obvious inside LCE i-TE fibers than LCE i-TE films. These effects are predicted to contribute to a high intrinsic thermopower in LCE i-TE fibers.

To verify the synergistic effect of the intermolecular interaction and the orientation pathway on the intrinsic thermopower, the thermoelectric properties are further investigated in detail. Since the thermopower of i-TE materials are still measured by homemade experimental setup in the field of i-TE, we conducted quantitative benchmarking correction with previously reported i-TE materials, confirming the testing apparatus accuracy, as shown in Supplementary Fig. 53,8. The LCE i-TE fibers prepared by impregnation method are firstly compared with that prepared by a one-step synthesis method, where ionic liquid was directly added during the LCE synthesis. It can be discovered that the LCE i-TE fibers synthesized by one-step method appear as white gelatinous, and show low thermopower below 5 mV K−1 ((Supplementary Figs. 6 and 7)). Meanwhile, neither the modulation of synthesis parameters nor ion species can significantly increase the thermopower, which may be attributed to the selective reactions of Michael addition26. Unexpectedly, the simple impregnation method in this paper addresses the above questions. As shown in Fig. 2a, the thermopower of LCE i-TE fibers fabricated by impregnation method with the identical ion species are superior to that by one-step synthesis, confirming the effectiveness and applicability of impregnation method in improving the thermopower of LCE i-TE materials. In addition, the ion species can alter the polarity of the thermopower, where impregnation of BMIM PF6 or LiBF4 endows n-type characteristics, while impregnation of BMIM OAC, EMIM TFSI, AMIM TFSI or EMIM Cl presents the opposite case. And the value of thermopower reaches its maximum (9.6 mV K-1) in the case of EMIM TFSI (Fig. 2a).

Fig. 2: Thermoelectric properties and reliability of LCE i-TE fibers.
figure 2

a Thermopower of LCE fibers immersed in different types of ionic liquids. b Thermopower of LCE i-TE films and fibers immersed in different concentrations of EMIM TFSI. c Thermopower of LCE i-TE fibers synthesized with various crosslinker contents at different immersion time. d Exploded schematic of the layer at different depths in LCE i-TE fibers. e–f Two-dimensional statistical kernel density estimates of the N and F elements contents along the radial direction of the LCE i-TE fibers in the case of different soaking time, respectively. g Comparison in thermopower of ionic thermoelectric materials measured at humidity less than 30% RH between the reported literatures and this work, inner graph shows a comparison in the thermopower of TE fibers between reported literatures and this work6,8,9,12,13,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58. h Ionic electrical conductivity of LCE i-TE fibers and films synthesized with different concentrations of EDDET and PETMP. i Voltage response curves of LCE i-TE fibers impregnated with different concentrations of ionic liquids over 10,000 s. j Cyclic voltage-temperature curves of LCE i-TE fibers. k Capacitor mode working mechanism diagram. All error bars denote mean ± standard deviation.

Full size image

Then, the thermoelectric performance of LCE i-TE fibers and films treated with EMIM TFSI is investigated and compared (Note that LCE i-TE materials immersing in different EMIM TFSI concentration are named as LCE-x EMIM TFSI with x = 0.5, 1.0, 1.5, 2.0, 2.5 and 3.0 M, respectively). As shown in Fig. 2b, the thermopower of both the LCE i-TE films and i-TE fibers increases at low EMIM TFSI concentrations and then decreases at high EMIM TFSI concentrations, saturating at 1 M. And the thermopower of LCE i-TE films can reach 10.4 mV K−1 at 1 M EMIM TFSI concentrations, while that of LCE i-TE fibers reach 17.4 mV K−1. This pronounced increase in the thermopower of LCE i-TE fibers in contrast to films confirms the important effect of mesogen orientation on the intrinsic thermopower, as revealed in Fig. 1b–h, which is attributed to the more favorable thermal diffusion of free cation and thus the larger difference of thermal mobility between anions and cations. To further prove that the above higher thermopower of LCE i-TE fibers compared with films indeed stems from different mesogen orientation, other than different shape between fibers and films, we fabricated LCE i-TE fibers with different diameters (1 mm, 2 mm, 3 mm) that suffered from different shear force, and studied the effect of mesogen orientation on the thermopower of these fibers. From the WAXD results of these LCE i-TE fibers (Supplementary Fig. 8), it can be seen that diameters alter the mesogen orientation, and the LCE i-TE fiber with diameter of 2 mm exhibits superior molecular chain spacing (d = 4.58 Å) and order parameter (S = 0.34) compared to its 1 mm (d = 4.47 Å, S = 0.30) and 3 mm (d = 4.39 Å, S = 0.29) counterparts. At the same time, the thermopower exhibits closely positive correlation with the order parameter, where the highest order parameter in the case of 2 mm diameter enable the highest thermoelectric performance, and the smallest order parameter (3 mm diameter) lead to the smallest thermopower, confirming the crucial role of molecular orientation in improving thermoelectric properties. More importantly, it is found that LCE i-TE fibers with diameter of 3 mm (S = 0.29) and LCE i-TE films (S = 0.28, as shown in Fig. 1b) with the similar order parameter contributes a similar thermopower (10 mV K−1, the thermopower of LCE i-TE films is shown in Supplementary Fig. 9). It means that the same molecular orientation whatever appearing in fiber-format or film-format will cause the same thermopower, further confirming the impact of molecular orientation on TE performance.

In addition, the thermopower of LCE i-TE fibers is further optimized via regulating synthesis parameter and post-immersion time. From Fig. 2c and Supplementary Fig. 10, it can be found that LCE i-TE fibers show best thermopower when the crosslinker Pentaerythritol Tetra (3-mercaptopropionate) (PETMP) is 0.11 M and soft chain EDDET is 1.7 M. The thermopower versus crosslinker and soft chain can be attributed to the change in the viscosity and crosslinking degree of LCE materials, where excessively low viscosity and crosslinking at low crosslinker content decrease the thermopower by preventing the binding between ions and soft chain, unduly high viscosity and crosslinking degree also deteriorate the thermopower via restricting the thermal diffusion behavior of cations and anions. Besides, the thermopower of LCE i-TE fibers obviously increase and then decreases with the elongation of immersion time, saturating at 6 h in the case of 0.11 M crosslinker. This change trend in thermopower has been also reported in the profile of thermopower versus ion concentration6,12. To figure out the reason why the highest thermopower appears at a moderate immersion time, we study the ionic liquid distribution using energy dispersive x-ray spectroscopy (EDS) and layer-by-layer x-ray photoelectron spectroscopy (XPS). From the result of EDS, four elements of IL (C, N, O, and S) are uniformly distributed on both the fiber surface and cross-section after 6 h immersion, directly demonstrating that the IL can effectively and uniformly penetrate into the fiber interior (Supplementary Fig. 11). However, it remains unclear whether the concentration of cations and anions at different depths affects the performance. Therefore, as shown in Fig. 2d–f, the two-dimensional statistical kernel density estimation, which is plotted using the R language, ggplot2 based on the data of layer-by-layer XPS, is further carried out, demonstrating a non-positive correlation between immersion time and the content of ionic species (N elements for cations and F elements for anions). During the initial immersion, ions diffuse from the liquid into both the surface and interior of the fibers, leading to increased ion content. As immersion time extends, fibers swelling reaches an optimal state. However, excessive immersion causes over-swelling, leading to relaxation and deterioration of the internal network structure. Ultimately, the highest ion content at each layer (excluding the surface layer) and the most uniform ion distribution are observed at immersion time of 6 h. This phenomenon reveals that the ionic content and distribution accounts for the immersing time-controlled thermopower.

In brief, the thermopower of LCE i-TE fibers consistently exceed that of LCE i-TE films, and this superior performance can be attributed to the crucial role of mesogen orientation induced ion migration difference27,28. Ultimately, the optimized thermopower of the LCE i-TE fiber reach 25.8 mV K−1 (Fig. 2c and Supplementary Fig. 12), while that of the i-TE film is just 10 mV K−1. Particularly, the thermopower of this LCE i-TE fiber is higher than that of the reported ionic thermoelectric materials measured at less than 30% RH, and is the largest one among all reported thermoelectric fibers (Fig. 2(g))6,8,9,12,13,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58.

Then, the ionic electrical conductivity (σi) of LCE i-TE films and fibers are studied. As shown in Fig. 2(h), the σi of the LCE i-TE films remain relatively stable regardless of synthesis parameter variations, with a maximum value of only 0.7 mS m−1. However, a breakthrough increase of the σi can be found in the LCE i-TE fibers, whose highest value reaches 21.5 mS m−1 at the same condition (1.7 M EDDET and 0.11 M PETMP). The enormous differences in σi between LCE i-TE films and fibers is mainly related to differences in ionic diffusion channels arising from mesogen orientation. Furthermore, the conductivity of the LCE i-TE fibers gradually increases with both ionic liquid concentration and immersion time, reaching a maximum value of 32.7 mS m−1 (Supplementary Fig. 13). Meanwhile, thermal conductivity tests reveal that the LCE i-TE fibers exhibit notably low thermal conductivity of 0.11 ~ 0.12 W m−1 K−1 (Supplementary Fig. 14). And this low thermal conductivity can contribute an effective utilization of the environmental temperature gradient. To prove this high utilization, an external temperature difference of 6 K is imposed on the fiber, as the three-dimensional waterfall plot of the temperature shown (Supplementary Fig. 15), the temperature difference along the fiber maintained 5.6 K, achieving a utilization of 93%. Finally, the good thermopower, electrical conductivity, and low thermal conductivity result in a maximum ZTi value of 0.035 (Supplementary Fig. 16).

Apart from the high TE performance, reliability and stability of this high TE property is also essential to practical application. Therefore, long-term voltage stability and cycling performance tests are further conducted to evaluate the practicality of LCE i-TE fibers. As observed in Fig. 2i, the voltage profile of LCE i-TE fibers immersed in different IL concentration exhibits remarkable stability under a long measuring time up to 10,000 s, without fluctuations. Meanwhile, Fig. 2j shows that LCE i-TE fibers display stable voltage response to cyclic stimuli of temperature difference. And the air stability of LCE i-TE fibers is further studied. As shown in Supplementary Fig. 17, the thermopower of LCE i-TE fibers remains above 80% after a week-long performance monitoring, indicating excellent air stability. Moreover, thanks to the advantages of the post-immersion method, the LCE i-TE fibers do not suffer from ionic liquid leakage issues. When wiping the LCE i-TE fibers with lint-free wipes, no overflow or leakage of ionic liquid is observed (Supplementary Movie S1). Therefore, the above results of long-term voltage stability, cycling performance tests, air stability, and leakage measurement all demonstrated the excellent reliability of LCE i-TE fibers.

Accordingly, compared to currently reported ionic thermoelectric materials, the LCE i-TE fibers possess inherently ordered network structures for high TE performance, operability under low humidity conditions, simple components, no leakage risk, long-term stability, cyclability, and air stability, all of which demonstrates exceptional practicality and application potential in low-grade heat harvesting.

Lastly, we investigate the capacitive characteristics of LCE i-TE fibers, as shown in Fig. 2k. A complete working model is divided into four stages, an open-circuit voltage of ~41 mV is generated by LCE i-TE fibers, due to the diffusion and accumulation of ions on the hot/cold side when a temperature gradient of 1.8 K is applied at stage I. At stage II, electrons flow through the external circuit and neutralize the unbalanced charge on the electrodes, causing a thermal voltage decay to zero. Nevertheless, as the temperature gradient disappears and the external load is removed (stage III), the ions accumulate on the electrodes gradually return to their original random state, while the electrons remaining on the cold side generate a reverse voltage. Eventually, the external load is reconnected, dropping the reverse voltage of the stage III to zero (stage IV)39. Meanwhile, by calculating the energy obtained in the second and fourth stages using the equation [2] \({Energy}=\int {U}^{2}/{Rdt}\), the LCE i-TE fibers exhibit promising output energy of 69.3 nJ when being connected to an external load of 20 MΩ (Supplementary Fig. 18)5.

Beyond superior TE performance, the excellent mechanical property and scalable fabrication ability are crucial for i-TE materials and devices. As shown in Fig. 3a, b, large-scale fabrication of continuous LCE i-TE fibers longer than two meters, with inherently deformable, stretchable, homogeneous, and unfolding properties, can be easily achieved. This demonstrates a promising commercial production.

Fig. 3: Optical property, mechanical performance, and weavability of the LCE i-TE fibers.
figure 3

a–b Physical drawings of large-scale produced LCE i-TE fibers. SEM and POM images of LCE i-TE fibers before (c–f) and after (g–j) impregnation with ionic liquid solution. k Tensile stress-strain curves of LCE i-TE fibers with different impregnation time at stretching rate of 50 mm min−1. l Storage modulus (G′) and loss modulus (G″) of LCE i-TE fibers with different impregnation time. m Physical images of LCE i-TE fibers puncture test. n Displacement-stress curve corresponding to the LCE i-TE fibers puncture test in (m). o, p Physical diagram of Chinese knots suffering from horizontal and vertical tensile. q LCE i-TE fibers-based bracelet suffering from tensile and twist testing.

Full size image

The polarized optical micrographs (POM) and scanning electron microscopy (SEM) are employed to examine the polarization and surface morphology of the fabricated LCE i-TE fibers. The LCE i-TE fibers exhibit smooth surfaces without obvious impurities and defects before (Fig. 3c, d) and after immersion process (Fig. 3g, h), the diameter of fibers remains 2 mm regardless of impregnation times (Supplementary Fig. 19). Meanwhile, before immersion (Fig. 3e, f), POM observations reveal colorful birefringence when the LCE fiber is oriented at a 45° angle to the polarizing filter, while turning to completely dark at 0°. This phenomenon remains unchanged after immersion (Fig. 3i, j), indicating a characteristic of monodomain aligned and anisotropic of both the LCE fibers before and after immersion59,60. Furthermore, cross-sectional POM images of the LCE fibers before and after immersion demonstrate fully dark fields during heating and cooling cycles, attributed to the fact that molecular chains aligned parallel to the fiber axis resulting in disappearance of birefringence, which directly prove the highly ordered arrangement inside the LCE i-TE fibers (Supplementary Fig. 20).

In addition, the mechanical properties of LCE i-TE fibers do not show significant changes with varying IL concentrations (Supplementary Fig. 21), but displays obvious changes with IL immersion time. As the immersion time increased, the strain of LCE i-TE fibers gradually decreases from 255.2% to 137%, combined with tensile stress reducing from the 524.5 KPa to 308.9 KPa (Fig. 3k). The same tendency of changes in the strain and stress can be seen for LCE i-TE films (Supplementary Fig. 22). The main reason for these decays in mechanical properties may be attributed to the decrease of interchain force and mechanical properties induced by the LCE swelling61. Nonetheless, the LCE i-TE fibers exhibit superior mechanical properties compared to LCE i-TE films after the immersion process. Specifically, the LCE i-TE fibers still possess stress and strain values of 444.5 KPa and 202.6% following 6 h immersion, surpassing those of LCE i-TE films by 54.4% and 50.9%, respectively. And the excellent mechanical properties of the LCE i-TE fibers enable readily lifting 500 g weights without breaking. The LCE i-TE fibers also have good stretchability, which can return to their original state after being stretched to a strain of 100%, as shown in Supplementary Fig. 23. Dynamic mechanical measurements in Fig. 3l display higher storage modulus G′ than the loss modulus G″ of the LCE i-TE fibers impregnated for 1, 3, 6 h, indicating the quasi-solid gel behavior of LCE i-TE fibers46. Additionally, LCE i-TE fibers and films demonstrate excellent puncture resistance, as shown in Fig. 3m and Supplementary Fig. 24. When applying stress with a 1 mm sharp needle, the fiber exhibits a maximum puncture displacement of 11.4 mm and a stress of 3.1 N (Fig. 3n).

Ultimately, the dyed LCE i-TE fibers are woven into Chinese knots, which show good endurance to a certain amount of tensile force applied in the transverse and longitudinal directions, with no inter-fiber fracture phenomenon (Fig. 3o, p). Similarly, no fracture or damage occurs when applying tensile and torsional stresses to the bracelet woven by LCE i-TE fibers, as shown in Fig. 3q. The above results confirm the excellent mechanical properties and good weavability of the LCE i-TE fibers, highlighting their significant potential for applications in wearable electronics and smart textile technologies.

Besides the above TE function, the liquid crystal primitive inside the LCE i-TE fibers can contribute the thermal-induced reversible phase transition, which will drive reversible shape change at the macro level62. Here differential scanning calorimetry (DSC) is employed to characterize the phase transition behavior of the LCE i-TE fibers. As shown in Supplementary Fig. 25, heat capacity increases rapidly at 38.9 °C, corresponding to the glass transition. Meanwhile, an additional slight change of the heat capacity at the 58.7 °C can be observed, which accounts for the nematic—isotropic transition point in monodomain LCE (clearing point)63. Accordingly, the liquid crystalline range of the fabricated LCE i-TE fiber is from 38.9 °C to 58.7 °C. Thus, temperature higher than 38.9 °C will cause entangle and curl behavior of LCE molecular chain, while gradually lowing temperature will lead to stretching and flattening behavior, achieving a reversible deformation, as the mechanism diagram shown in Fig. 4a. Based on this mechanism, the shrinkage in the LCE i-TE fibers driven by 60 °C heating temperature can reach 14.5%, and gradually return to original state as the temperature decreases (Fig. 4b). Meanwhile, as the physical exhibit in Fig. 4c, LCE i-TE fibers attach with a red sphere undergo obvious stretching (from 40 °C to 60 °C) and recovery (from 60 °C to 30 °C) phenomena. These results indicate the favorable thermally actuating responsiveness of the prepared LCE i-TE fibers, and this controlled deformation is especially important for the application of i-TE materials in the field of wearable and smart textiles.

Fig. 4: Thermal actuation properties of the LCE i-TE fibers.
figure 4

a Mechanism diagram of reversible thermal braking of LCE molecular chains. b The shrinkage variation ratio of LCE i-TE fibers at different temperatures. c Thermal actuation diagram of LCE i-TE fibers attached with a red sphere at different temperatures. d Diagram of plain stitch structure and deformation image. e Diagram of rib knit pattern structure and deformation image. f The deformation of plain stitch- and rib knit pattern-structured fabrics at different temperatures. g Radargram of LCE i-TE fibers in contrast with the areas of thermopower, tensile stress, fiber format i-TE materials, and actuation8,9,29,43,46,48,64.

Full size image

Afterwards, two special weaving structures, plain stitch and rib knit pattern, are utilized to fabricate textiles, aiming to convert the one-dimensional thermally actuating deformation along the fibers into three-dimensional deformation inside the textiles. As shown in Fig. 4d, it can be found that the plain stitch-structured LCE textile is composed of a single knitting ring arrangement, and the internal torque is accumulated in each knitting coil. When heating the sample, the LCE i-TE fibers shrink, so that the braided ring shrinks and thus the fabric is bent. In contrast, as shown in Fig. 4e, the LCE textile woven with rib knit pattern shows non-deformation, due to the mutual cancellation of the torque between the adjacent knitting rings. Then, both of the LCE fabrics woven by plain stitch and rib knit are put into the oven, and the deformability of these fabric are examined at different temperatures (Fig. 4f). The LCE fabric of these two structures has no obvious change when the oven temperature is 30 °C, but with the temperature over 40 °C, fabric woven by plain stitch shows gradually bending deformation, and when the temperature reaches 70 °C, the fabric realizes complete curling deformation. In contrast, the fabric woven by rib knit pattern exhibits no obvious deformation over temperature ranging from 40 °C to 70 °C. The braided structure-induced thermal actuating properties provides the basis for the subsequent construction of four-dimensional dynamically adaptive devices.

Based on the above results, as shown in Fig. 4g, the proposed LCE i-TE fibers not only exhibit the highest value of thermopower at low humidity, but also demonstrate excellent mechanical and actuation properties, providing a unique actuating TE material for smart wearables, soft robots, and flexible electronics8,9,29,43,46,48,64.

Based on the thermal response characteristics of fabrics, we developed self-adaptive thermoelectric devices with gripper-like configurations using plain stitch and rib knit patterns. These all-fiber devices integrate thermoelectric, actuating, and sensing capabilities. As illustrated in Fig. 5a, b and Supplementary Fig. 26, the top part of the device is made of rib knit pattern, whose deformation is not affected by temperature, while the four gripping arms were made of plain stitch-structured fabrics with excellent thermally actuating ability. The thermoelectric device demonstrates remarkable self-deformation ability. As evidenced in Supplementary Movie S2 and Supplementary Movie S3, it exhibits autonomous deformation at elevated temperatures. When being placed near a heat source, the device shows excellent grasping capabilities (Supplementary Movie S4). Importantly, the device returns to its original state as temperatures decrease (Fig. 5c and Supplementary Movie S5), enabling dynamic and adaptive heat harvesting capabilities. And when being exposed to an environmental temperature of 60 °C, the prepared TE device can easily grasp a red plastic ball within 90 s, as shown in Fig. 5d. Notably, this thermoelectric device demonstrates remarkable grasping capability by lifting a steel ball, whose weight is 17 times higher than the device’s (as shown in Supplementary Fig. 27). Additionally, the device exhibits reliable cyclic performance in its grasping functionality, as demonstrated in Supplementary Movie S6 and Fig. 5f, it can perform repeated cycles of gripping and releasing through temperature elevation and reduction processes.

Fig. 5: Preparation of all fiber-based self-adaptive thermoelectric devices and performance test.
figure 5

a, b Physical drawing for all fiber-based self-adaptive thermoelectric devices of top view and front view. c Recovery process of bending angle of devices after temperature reduction. d Flowchart of the system’s capture of a plastic red ball. e The deformation of plain stitch-structured TE device at different temperatures. f Flowchart for cyclic gripping properties of TE devices. g, h Gripping cylinders heat sources with different pipe diameters (Φthick: 16 mm, Φthin: 10 mm). i–j ΔR/R0 curves for cyclic bending (top) and stretching (down) of TE device/LCE i-TE fibers at different degrees of deformation. k Output voltage and ΔR/R0 of the device at different temperatures under a bending angle of 45°. l Output voltage and ΔR/R0 of the device at different bending angles under a heat source of 150 °C. m, n The current-voltage curves and localized plots of the TE device simultaneously enduring bending angles and temperature. o, p The ΔR/R0 of system captures the different-sized spheres and various shapes. All error bars denote mean ± standard deviation.

Full size image

To enable precise heat harvesting and deformation sensing, we embedded an LCE i-TE fiber into the grasping arm as its active part. As demonstrated in Fig. 5e, integrating fibers will not compromise the overall device’s thermally deformation properties. Then, the ability to dynamically adapt to low-grade heat source with complex-shaped geometry is studied, by employing a double-armed gripper-like TE device. It can be seen from Fig. 5g, h that the TE device can tightly self-wrap cylinder heat sources with different tube diameters and generate a voltage of 1.01 V at 160 °C, regardless of 10 mm or 16 mm (Supplementary Movie S7 and Supplementary Movie S8). Accordingly, the thermoelectric device can keep closely attaching to the heat source and output a stable voltage over time regardless of dynamically changing size or increasing temperature of the heat source, without extra treatment between the device and the heat source including sticking, pressing, etc. The dynamically adaptive TE device in four-dimensional space promotes their practical applications in harvesting low-grade heat.

Besides, the resistance changing rate (ΔR/R0) of LCE i-TE fibers embedded in gripping arms and the ΔR/R0 of single i-TE fibers show stable and reversible cyclically response to different bending angles and strain, respectively (Fig. 5i, j). Thus, the TE device can sense temperature and deformation via voltage signal and resistance signal, respectively. Then, the feasibility of decoupled sensing of temperature and deformation is verified. From the Fig. 5k, the corresponding output voltage of TE devices shows proportional relationship to the temperature difference when the heat sources are heated from 120 °C to 160 °C under a bending angle of 45°, while the ΔR/R0 remains constant as the TE devices tightly grabbed the heat source all the time. Simultaneously, we also measure the changes in output voltage and ΔR/R0 at different bending angles. As shown in Fig. 5l and Supplementary Fig. 28, the variation of output voltage and thermopower at different bending angles keep stable, which are only 3.5% and 5.2%, respectively, while the ΔR/R0 continuously increase with increasing angle. Therefore, the TE devices can directly and accurately reflect the bending angle through ΔR/R0, while the temperature is reflected by the output voltage, achieving decoupled recognition. To further display the decoupled sensing capability, current-voltage (I–V) curves are tested under simultaneous thermal and bending stimuli (Fig. 5m), it can be revealed that the I–V curves maintain parallel relationships with consistent slopes (resistance values) under fixed bending angles, while their Y-intercepts increase proportionally with temperature (Fig. 5n). With constant temperature differentials, the I–V curves demonstrate increasing slopes corresponding to larger bending angles. This behavior confirms the decoupled sensing of temperature and deformation, where temperature is detected through voltage response (Y-intercepts) and deformation is detected through resistance changes (slopes).

Finally, a TE device with four griping arms is integrated, whose two adjacent arms (named x arm and y arm, respectively) are the active sensing parts. Inside this device, the ΔR/R0 of the inlaying LCE i-TE fibers on the two adjacent arms (x arm and y arm) of the devices is recorded to detect the sizes and shapes of the captured objects. In the case of capturing a sphere with different size, it can be found in Fig. 5o, the bending angle of the grasping arm gradually decreases with the gradual increase of the sphere size, and the corresponding ΔR/R0 show a certain linear relationship with the size of the grasping object. Meanwhile, in the case of capturing a sphere with different shape (Fig. 5p), the ΔR/R0 shows significant variations capturing five objects (sphere (small), sphere (big), triangular, cuboid, cube), due to different bending angle when grasping arm touches objects of different shapes. Thus, the combined data of the ΔR/R0 from the two arms allows for accurate identification of object size and shape.

Discussion

In conclusion, we have investigated a p-type LCE ionic thermoelectric fiber with a high thermopower (25.8 mV K−1) via the synergistic effect of the intermolecular interaction originating from the soft chain and the orientation pathway originating from the LC phase, achieving the highest reported thermopower at low humidity in the field of ionic thermoelectrics and thermoelectric fibers. Based on the specific thermal actuated property, high TE performance, and excellent mechanical properties, an all fiber-based, self-adaptive, and griper-liked thermoelectric devices was successfully fabricated. And this griper-liked device can accurately recognize the shape/size and temperature of the capturing objects in a decoupled mode. The proposal of this high-performance ionic thermoelectric fiber and the dynamically self-adaptive thermoelectric devices provide an innovative idea for the future research of ionic thermoelectrics and have a highly promising application in the fields of smart wearables and soft robots.

Methods

Materials

RM257 (98%) was purchased from Shanghai Bide Pharmaceutical Technology Co., Ltd., EDDET (95%), PETMP (95%) and ethyl acetate were supplied by Shanghai Titan Scientific Co., Ltd. DPA (99%) was bought from Meryer (Shanghai) Chemical Technology Co., Ltd. The EMIM Cl/TFSI, BMIM OAC/PF6, AMIM TFSI and LiBF4 were purchased from Lanzhou Greenchem ILS, LIPC, ACS (Lanzhou China). All reagents were used as received without any further purification.

Preparation of LCE i-TE films and fibers

The preparation of LCE i-TE could be described in the following two step: the first step was the mold forming of LCE fibers, where the 2.4 g RM257, 0.7 g EDDET and 0.1 g PETMP were dissolved in 2.4 ml of ethyl acetate solution and stirred at 70 °C for 1 h until complete dissolution, meanwhile, the LCE solution was obtained after the 9 μl of DPA was added and stirred for 1 min. LCE films were obtained by pouring the LCE solution into culture dishes and letting it stand at room temperature for 12 h. For LCE fibers, the LCE solution was transferred to a syringe and injected into a polytetrafluoroethylene tube by the mold method, and horizontally placed at room temperature for 12 h. Finally, the LCE fibers were obtained via stripping from the tube. The final step in the preparation of LCE i-TE fibers/films was the impregnation of LCE fibers/films in ionic liquid solution. Firstly, the ionic liquid solutions with a concentration of 0.5, 1.0, 1.5, 2.0, 2.5, 3.0 M were prepared, and then, the prepared LCE fibers/films were immersed in this solution for 1, 3, 6, 9, 12, 24 h to get the LCE i-TE fibers/films impregnation with different time.

Preparation of all fiber-based self-adaptive thermoelectric devices

The all fiber-based self-adaptive thermoelectric devices consisted of four fabric arms (double-armed gripping system was two fabric arms) and a top fabric, where the fabric arms as bending actuator modules were prepared from LCE i-TE fibers by plain stitch weave, while the top fabric was fabricated via rib knit pattern. Then the LCE i-TE fibers were used to sew the top fabric and the curved arms together, where the curved sides of the fabric arms faced each other and sewn to the four edges of the top fabric to create the gripping deformation. And the LCE i-TE fibers, which serves as a temperature recognition and sensing module, was directly embedded in one of the four curved arms to enable all fiber-based self-adaptive thermoelectric devices preparation.

Characterization

The micro morphology and optical properties of LCE i-TE fibers were conducted using FE-SEM (SU8010, Hitachi) and POM (DM750P, Leica). The FTIR spectroscopy (Nicolet6700, Thermo Fisher) with the attenuated total reflection accessory and Raman spectra (inVia-Reflex, Renishaw) with a 633 nm laser were recorded to analyze the ion and molecular interchain interaction. Thermopower was measured by a homemade equipment based on the equation [3] \({Thermopower}=-\varDelta V/\varDelta T\), here potential differences ΔV arising from eight temperature differences ΔT was recorded by Keithley 2182 A. The linear correlation (R2) between ΔV and ΔT should be > 0.999. Test items were tested at 3 times of each sample for an average value. The ionic conductivity σi was calculated as follows: \({\sigma }_{i}=d/(A * R)\), where the d, A, and R in the formula represent the thickness, area, and ionic resistance, respectively. The ionic resistance was tested by electrochemical impedance spectroscopy on an electrochemical workstation (DH7006,) with the frequency ranging from 0.1 to100000 Hz. Stress-strain curves were performed on an Instron (5969) testing instrument with a speed of 50 mm min−1 at room temperature. And the gelatin properties of LCE i-TE fibers, which the angular frequency dependencies of the storage modulus (G′) and loss modulus (G″) were performed by DMA (Q800, TA).

Computational details

Modeling and Simulation Methods: All ionic species and small molecules were parameterized using the next-generation general AMBER force field (GAFF2), with specific force field parameters generated using the sobtop software. The initial configurations were constructed using the packing optimization for molecular dynamics simulations (Packmol) program with a periodic simulation box of 40 × 40 × 40 nm. All molecular dynamics (MD) simulations were performed using the groningen machine for chemical simulations (GROMACS) 2022.5 package. The simulation protocol consisted of three main stages:

  1. 1.

    Energy Minimization

    The system was initially minimized using a combination of 5000 steps of steepest descent followed by 5000 steps of conjugate gradient algorithms to eliminate unfavorable contacts.

  2. 2.

    Constant number of particles, pressure, and temperature ensemble (NPT) Pre-equilibration

    The system was pre-equilibrated in the NPT ensemble using the V-rescale thermostat at 298 K and the Parrinello-Rahman barostat at 1 atm. Non-bonded interactions were treated with a cutoff radius of 1.2 nm, and the integration time step was set to 1 fs.

  3. 3.

    MD Simulation

Following equilibration, the temperature coupling was switched to the Berendsen thermostat. Bond lengths and angles were constrained using the linear constraint solver (LINCS) algorithm. A twin-range cutoff of 1.2 nm was employed for van der Waals interactions, while long-range electrostatic interactions were handled using the particle-mesh Ewald (PME) method. Trajectory frames were saved every 10.0 ps for subsequent analysis.