Introduction

Exploring novel, efficient and sustainable waste heat utilization technologies has emerged as a global research area for energy saving and realizing carbon neutrality1,2,3,4,5, as nearly 60% of the energy in the world is lost as waste heat6. Among them, a huge amount of waste heat is distributed near ambient temperature (classified as low-grade heat). Thermoelectric devices, such as traditional semiconductor thermoelectric generators (TEGs)7,8 and emerging ionic thermocells (characterized by the Soret effect9,10 or thermogalvanic effect11,12,13,14), which directly convert heat energy into electrical energy without any moving parts or vibration noise, offer great promise for directly powering wearable electronics. However, the effectiveness of traditional TEGs for low-grade heat conversion (i.e., body heat energy) is limited by their relatively low thermopower (Seebeck coefficient (Se), ~μV K−1), which results in a low energy conversion efficiency near ambient temperature15,16. Although recently developed liquid-based thermogalvanic cells (TGCs) have garnered substantial attention due to their large Se (~mV K−1) and stability, they are typically encapsulated using rigid devices and suffer from the inevitable risk of liquid leakage, which hinders their integration into wearable electronics.

Emerging flexible quasi-solid TGCs, which confine liquid electrolytes into three-dimensional gel networks, offer a possible solution to these challenges. As an important candidate to utilize body heat for powering wearable electronics directly, to realize high ηr and wearability, the thermocells must be highly efficient and mechanically robust to resist external shock. Nevertheless, state-of-the-art quasi-solid TGCs remain limited by a challenging trade-off between thermoelectric properties and mechanical strength due to the difficulty in decoupling their different requirements in material design. On the one hand, enhancing the Se of quasi-solid TGCs can be achieved by introducing functional additives such as guanidine hydrochloride (GdmCl) to enlarge the entropy difference of redox couples in thermocells. However, these additives typically deteriorate the mechanical toughness of quasi-solid TGCs substantially due to the Hofmeister effect-induced salting-in effect, which significantly weakens the interactions of polymer chains17,18. Despite this, existing quasi-solid TGCs still undergo a low ηr value and could not break the bottleneck of the commercialization threshold (5%)17,19,20,21. On the other hand, improving the mechanical strength of quasi-solid TGCs can be achieved through the reinforcement of polymer networks by the design of double networks19 or anisotropic structures via ice-template22 or mechanical training20,21 methods. However, these approaches often compromise the thermoelectric properties, exhibiting low output power densities and depressed ηr due to the low thermopower19,22 or slow ionic migration speed21 in thermocells. Therefore, effectively coupling mechanical and thermoelectric performance through a traditional single approach remains a significant challenge.

In this work, we present an ultrastrong yet flexible thermogalvanic armor (FTGA) with an ultrahigh ηr of 8.53% near room temperature through the synergy of molecular engineering and structural design, decoupling the trade-off between thermoelectric and mechanical properties in conventional quasi-solid TGCs. In contrast to conventional approaches, a novel small molecule, guanidine sulfate ((Gdm)2SO4), is first leveraged to induce thermosensitive crystallization for boosting thermopower, as well as reinforce the mechanical toughness via the Hofmeister effect23,24,25. In combination with molecular engineering, we also employ a directional freezing method to formulate the aligned channels as an “ion-transport highway” to further enhance the ion conductivity and thermopower. Our FTGA exhibits superior thermoelectric properties, including a high thermopower of 5.58 mV K−1, a boosted specific output power density of 11.90 mW m−2 K−2 and a high ηr of 8.53%, as well as delivers impressive mechanical properties, such as desirable toughness of 70.65 MJ m−3, high strength of 11 MPa and large elongation (~900%). To the best of our knowledge, this study pushes the boundary of the ηr of quasi-state TGCs to exceed 5%, which is much higher than the previous studies. As a proof of concept, our FTGA can be assembled into an integrated module for driving various wearable electronic devices, including an LED array and a smartwatch. We anticipate that this work will provide new insight into the design of highly efficient and flexible thermoelectric devices.

Results

Design principle

Our proposed FTGA, consisting of poly(vinyl alcohol) (PVA), K3Fe(CN)6/K4Fe(CN)6 and (Gdm)2SO4, is constructed by coupling molecular engineering and structural design through directional freeze-casting and subsequent salting-out processes (Fig. 1a, Supplementary Fig. 1). First, from the perspective of molecular engineering, we selected biocompatible and cost-effective PVA as a strong gel skeleton, owing to its hydrophilic and biocompatible nature26,27, and K3Fe(CN)6/K4Fe(CN)6 as the redox couples because of its high intrinsic thermopower (1.4 mV K−1). Distinct from most K3Fe(CN)6/K4Fe(CN)6-based thermocells, the biggest merit of our design is to exploit (Gdm)2SO4 as a dual molecule mediator to replace the commonly used GdmCl. In previous studies using GdmCl, Gdm+ selectively induced [Fe(CN)6]4− crystallization to enlarge the solvent-dependent entropy difference (ΔS) of [Fe(CN)6]3−/[Fe(CN)6]4− (~1.7 times)11,12 (Supplementary Fig. 2), and further enhance the thermopower. However, Gdm+ is also a strong chaotropic cation in the Hofmeister series (Supplementary Fig. 3)11 that can dissolve polymer chains due to the “salting-in” effect28, resulting in poor mechanical properties of quasi-solid thermocells17. In striking contrast, (Gdm)2SO4 offers exciting ability to resolve these trade-off issues because of its unique molecular structure (Fig. 1b). For (Gdm)2SO4, besides the superiority of Gdm+ in enhancing thermoelectric performance, its counterion, SO42−, is a strong salting-out anion in the Hofmeister series for enhancing the mechanical properties (~2.7 times for toughness, ~1.5 times for Young’s modulus, and ~1.8 times for tensile strength) (Supplementary Figs. 3 and 4). As a result, the molecular design of (Gdm)2SO4 allows the FTGA to simultaneously improve the thermopower and mechanical strength. Second, considering the structural design, the FTGA was equipped with honeycomb-like aligned channels by employing a directional freezing process (Fig. 1c1–c4). The aligned channels of FTGA possess intact cellular structures with compartment dimensions approximately between 1–5 μm (Fig. 1c1, c2), as well as highly ordered cellular arrays aligned with the ice-template direction (Fig. 1c3, c4). Unlike conventional isotropic gel-based thermocells (Supplementary Fig. 5), the unique aligned channels of FTGA could not only store a large number of ions (Fig. 1c5, c6), but also act as a “highway” for accelerating crystals sedimentation as well as ion transport, and further enhancing the thermopower (~1.4 times) and electrical conductivity (~1.3 times) (Supplementary Fig. 6). Taken together, by optimizing the synergistic effect of molecular engineering and structural design, our FTGA holds tremendous potential in attaining enhanced thermopower, improved toughness, and accelerated ion transportation, aiming at harvesting preferential energy conversion efficiency and mechanical strength.

Fig. 1: Design principle of the FTGA.
figure 1

a Schematic illustration of the fabrication of FTGA through the directional freezing-assisted salting-out method. b Mechanism illustration of guanidine sulfate to the performance of FTGA by molecular engineering. c SEM images showing the anisotropic structures of the FTGA. (c1, c2): top view, (c3, c4) side view, (c5, c6): energy dispersive spectrum of c4. The experiment was repeated three times using different samples. d Thermoelectric device for module electric performance measurement.

Full size image

Thermoelectric performance

We first elaborated an experimental device to evaluate the thermoelectric performance of the FTGA (Fig. 1d). In brief, the FTGA was sandwiched between two commercial platinum (Pt) wires as working electrodes, in which Pt wires are chemically inert materials that could eliminate the impact of electrode corrosion on thermopower. To measure the thermoelectric properties of the FTGA precisely, we employed a thermal stage (on the bottom) and a Peltier cooler stage (on the top) to manipulate the temperature gradient in the FTGA. When a temperature difference exists across the FTGA, the reversible reaction of the redox couples occurs on both sides, causing a potential difference in the FTGA at the two electrodes. Specifically, the oxidation reaction of [Fe(CN)6]4− to [Fe(CN)6]3− occurs on the hot side, releasing an electron to the electrode. This released electron is then transferred to the cold side through the external circuit and is ultimately consumed by the reduction reaction of [Fe(CN)6]3− to [Fe(CN)6]4−. During the whole process, due to the introduction of Gdm+ into the thermocells, the entropy difference (ΔS) of the redox couples could be enlarged, thus leading to an obvious improvement in the thermopower29.

To demonstrate the utility of (Gdm)2SO4, we examined the influence of the concentration of (Gdm)2SO4 on the thermopower of the FTGA (Supplementary Fig. 7). With (Gdm)2SO4 concentration increasing, the Se of the FTGA gradually increases and then drops down. When the (Gdm)2SO4 concentration is 1 M, the Se of the FTGA reaches a peak value of 5.58 mV K−1, which is 3.2 times greater than that without (Gdm)2SO4. Based on these results, we selected 1 M (Gdm)2SO4 as the optimal concentration for further characterization of the FTGA. Next, we distinguished the contributions of (Gdm)2SO4 and the aligned channels to the high thermopower of the FTGA (Fig. 2a). When only possessing aligned channels without (Gdm)2SO4, the Se of the FTGA is significantly lower at 1.77 mV K−1. For the FTGA consisting of (Gdm)2SO4 but lacking aligned channels, the Se value is 4.07 mV K−1. These results imply that the high thermopower of the FTGA originates from the synergy between (Gdm)2SO4 and the aligned channels. This synergistic effect can be further revealed by inspecting the electrical conductivity of the FTGA and control samples. Figure 2b shows that the FTGA possesses a high effective electrical conductivity of 12.98 S m−1, which is much higher than the other two control samples, attributing to its aligned channels for providing a “highway” for ion transport. Furthermore, the FTGA exhibits a remarkable increase in thermopower and electrical conductivity compared to previously reported quasi-solid thermocells17,19,20,21,22,30,31,32,33,34,35,36,37 (Fig. 2c and Supplementary Table 1), owing to the advantage of our synergetic design.

Fig. 2: Thermoelectric output performance of the FTGA.
figure 2

a, b Optimized thermopower (Se) (a) and electrical conductivity (b) for the FTGA induced by different structures and additive materials. Data are presented as mean values ± s.d. For the measurements of Se and electrical conductivity, n = 3. c Property comparison of the electrical conductivity versus thermopower (Se) for various quasi-solid thermogalvanic thermocells. d–f Output voltage-current curves and their corresponding power densities at various temperature difference (ΔT) (d), corresponding maximum power density (Pmax) and specific output power density (Pmax/ΔT2) (e), and thermal energy conversion efficiency (η) and the Carnot-relative efficiency (ηr) (f). The cold side was fixed at 298 K. g Working mechanism of the enhanced thermopower in FTGA system through the selective crystallization of Gdm+ with [Fe(CN)6]4−. h Schematic solvation shell structures of [Fe(CN)6]3− and [Fe(CN)6]4− in FTGA.

Full size image

The electrical performance of the FTGA is further reflected by its output thermoelectric performance under various temperature gradients. The output voltage-current density curves and corresponding power densities of the FTGA were tested and presented in Fig. 2d, where the temperature of the cold side (Tc) was fixed at 298 K. The FTGA exhibits nearly linear voltage-current density and parabolic power density-current density relationships. For a sample with a cross-sectional area of 0.04 cm2 and an electrode distance of 10 mm, the open-circuit voltage (Voc) increases from 27 mV to 165 mV while the short-circuit current density (Isc) rises from 36.73 A m−2 to 241.35 A m−2, as ΔT increases from 5 K to 30 K. Based on these output electrical performances, the maximum output power density (Pmax) of the FTGA is calculated. As shown in Fig. 2e, the Pmax of the FTGA is nearly proportional to the square of ΔT, which is consistent with the theoretical result (Supplementary Note 1). Surprisingly, the normalized power density (Pmax/(ΔT)2) of the FTGA reaches an ultrahigh level of 11.90 mW m−2 K−2, which is 6 and 1.4 times higher than the reported state-of-the-art stretchable quasi-solid thermocells21 and photocatalytically-enhanced aqueous thermocells38, which have the values of 1.97 mW m−2 K−2 and 8.5 mW m−2 K−2, respectively.

We continue to assess the output thermoelectric performance of the FTGA by figuring out the energy conversion efficiency (η) and the Carnot-relative efficiency (ηr) (Supplementary Note 2). Unlike conventional quasi-solid thermocells, our FTGA can also suppress thermal conductivity and achieve high η and ηr values. Stemming from the inhibited thermal convection caused by the crystallization of [Fe(CN)6]4− and Gdm+39, the FTGA delivers a low effective thermal conductivity (κeff) at both parallel and perpendicular direction (relative to the aligned channels) (Supplementary Table 2). Especially, the κeff at the parallel direction is as low as 0.4434 W m−1 K−1. Combining the merits of high thermopower, effective electrical conductivity, and low thermal conductivity, our FTGA achieves maximum η and ηr values of 0.79% and 8.53%, respectively (Fig. 2f). Notably, the ηr of FTGA largely exceeds the commercialization threshold (5%). Moreover, the ηr and (Pmax/(ΔT)2) values of our FTGA are much higher than those reported quasi-solid thermocells (Supplementary Fig. 9 and Table 3). These encouraging results strongly demonstrate the great potential for commercializing our FTGA in wearable electronics.

The high-output thermoelectric performance of the FTGA can be further understood from the molecular-level perspective. The thermopower of the FTGA is strongly related to the solvent-dependent entropy difference (ΔS)29 of [Fe(CN)6]3−/[Fe(CN)6]4− at the system. Owing to the higher charge density of [Fe(CN)6]4− than [Fe(CN)6]3−, Gdm+ could selectively induce [Fe(CN)6]4− crystallization based on chaotropic-chaotropic interaction11, so as to enlarge the ΔS of the redox ions and further enhance the Se. This conclusion is also supported by the density functional theory (DFT) calculations (Supplementary Note 3, Supplementary Fig. 10). Before adding (Gdm)2SO4, the hydration energy of [Fe(CN)6]4− and [Fe(CN)6]3− is −18.71 eV and −13.24 eV, respectively, indicating that a higher charge of [Fe(CN)6]4− causes a more closely packed solvation shell. After adding (Gdm)2SO4, the hydration energy of [Fe(CN)6]4− is −8.86 eV, while the hydration energy of [Fe(CN)6]3− is −10.45 eV. This demonstrates that more Gdm+ incline to bond with [Fe(CN)6]4− due to its higher charge, causing greater destruction of the hydration shell than [Fe(CN)6]3−. In contrast, (Gdm)2SO4 exhibits a slight impact on the solvation shells of [Fe(CN)6]3− (Fig. 2h and Supplementary Fig. 11). This distinct change in solvation shell structure significantly increases the ΔS of [Fe(CN)6]3−/[Fe(CN)6]4−. In addition to the change in solvation shell, Gdm+ also induces the crystallization of [Fe(CN)6]4− on the cold side11,12 of the FTGA. As this crystallization process occurs, the concentration of [Fe(CN)6]4− near the cold side decreases, promoting the reduction reaction of [Fe(CN)6]3− to [Fe(CN)6]4−. Moreover, the well-designed aligned channels can promote the formed crystals sedimenting quickly into the hot side and redissolve to generate a high concentration of [Fe(CN)6]4− near the hot side, which also accelerates the oxidation reaction of [Fe(CN)6]4− to [Fe(CN)6]3− at the hot side and contributes to a high thermopower value.

Flexibility and mechanical toughness

In parallel with desirable thermoelectric properties, the FTGA also imparts impressive mechanical properties, which are typically not achievable with conventional quasi-solid TGCs. To highlight the significance of the synergy of molecular engineering and structural design on mechanical properties, we conducted uniaxial tensile and compressive tests to compare the mechanical behavior of different FTGA series with control samples. As depicted in Fig. 3a, our FTGA exhibits superior overall mechanical performance, with an elongation capability of ~900%, as well as high tensile strength, Young’s modulus, and toughness of 11 MPa, 5.28 MPa, and 70.65 MJ m−3, respectively. In contrast, the tensile strength, Young’s modulus, and toughness of the sample without aligned channels are only 3.76 MPa, 2.82 MPa, and 11.4 MJ m−3, respectively, while the sample without (Gdm)2SO4 has a tensile strength, Young’s modulus, and toughness of 6.15 MPa, 3.56 MPa and 26.8 MJ m−3, respectively (Supplementary Fig. 12). Furthermore, compared to other reported flexible TGCs, the FTGA exhibits high tensile strain and strength (Supplementary Fig. 13, Supplementary Table 4).

Fig. 3: Quantitative mechanical properties and durability of FTGA.
figure 3

a, b Stress-strain curves for FTGA with different structures and additive materials from uniaxial tensile (a) and compressive (b) tests. c Schematic illustration of the SHPB test system (inset) and the compressive stress-strain curves of FTGA under different strain rates. d Cyclic stress-strain curves of the FTGA at tensile strains from 0% to 30%. e The variation of the thermopower (Se) after repeated stretchable cycles. f Photographs of a car compression test carried out by compressing a 35 (L)*25 (W)*2 (T) mm (where L is length, W is width, and T is thickness) FTGA specimen with a 1600 kg four-wheel car. g Thermopower of the FTGA before and after car compression, ΔT is the temperature difference.

Full size image

The superior mechanical performance of the FTGA can be further reflected by studying its mechanical behavior in the perpendicular direction relative to the aligned channels (Supplementary Fig. 14). The FTGA still exhibits an impressive elongation of ~300%, a high tensile strength of 2.41 MPa, Young’s modulus of 1.42 MPa, and toughness of 5.57 MJ m−3, surpassing previously reported tough quasi-state thermocells17,19. In addition to the tensile test, we also examined the compressive behavior of the different series of thermocells. As shown in Fig. 3b, our FTGA could be compressed to 80% without fracturing, and its compressive stress is higher than that of the other two series of thermocells, which is consistent with the results of the tensile test. The superior mechanical properties of our FTGA could be attributed to the synergistic effects of molecular engineering and structural design. First, the directional freezing-enabled aligned channel structure can facilitate closer packing, stronger entanglement and interactions of PVA chains. Second, during the subsequent salting-out process, the presence of SO42−, which is known as a strong salting-out anion in the Hofmeister series, effectively mitigates the negative effects of guanidine ion on the mechanical properties of the FTGA. This salting-out effect allows for efficient aggregation and crystallization of the PVA chains during this process. Therefore, the mechanical strength of the FTGA could be improved by coupling structural design and molecular engineering.

To manifest the merit of SO42−-induced salting-out effect in strengthening mechanical strength, we inspected the mechanical properties of the FTGA with different soaking salts ((Gdm)2SO4, GdmCl). As shown in Supplementary Fig. 15, in the absence of soaking treatment, the FTGA exhibits normal elongation, tensile strength, Young’s modulus, and toughness of 584%, 6.15 MPa, 3.56 MPa, and 26.8 MJ m−3, respectively. However, after soaking in a GdmCl aqueous solution, the corresponding properties of the FTGA decrease to 403%, 3.31 MPa, 1.94 MPa, and 8.45 MJ m−3, respectively. In contrast, when the FTGA is soaked in (Gdm)2SO4 aqueous solution with the same concentration, the corresponding properties increase to 861%, 11 MPa, 5.28 MPa, and 70.65 MJ m−3, respectively. This enhanced trend enabled by (Gdm)2SO4 is also observed in the compression test (Supplementary Fig. 16). These contrasting results are attributed to the Hofmeister effect of different salts, characterized by the ability to precipitate polymer chains25. As the most typical chaotropic cation in the Hofmeister series, Gdm+ directly bounds to PVA chains due to the salting-in effect, resulting in poor mechanical properties in the FTGA. In contrast, SO42− is a strong kosmotropic anion, which can polarize hydrated water molecules and interfere with the hydrophobic hydration of macromolecule chains due to the salting-out effect, leading to the aggregation and crystallization of polymer chains and eventually enhance the mechanical properties of the FTGA.

In addition to the impressive mechanical performance under static environments, our FTGA also manifests profound application feasibility under complex, dynamic working conditions. We assessed the dynamic impact resistance of the FTGA through the split Hopkinson pressure bar (SHPB) system, which is a powerful tool for studying the behavior of materials under high-strain rate deformation40. As shown in Fig. 3c (Supplementary Fig. 17), the SHPB system consists of an air gun system, a bullet, an incident bar, a transmission bar and a strain gauge, in which the strain gauge on the incident bar and transmission bar could convert the mechanical wave into an electronic signal and export the stress-strain curve during the impact process. When applying different high-strain rates from 11,000 to 12,800 and 13600 s−1 using the SHPB system, the FTGA still demonstrates superior resistance to these dynamic high-strain impacts. Remarkably, the highest compression stress of the FTGA reaches up to ~90 MPa at the strain rate of 13,600 s−1, which is a rough task for conventional hydrogels. Upon such high compression speeds, the energy absorption of the FTGA is quantified to be ~ 15.44, ~30.15, and ~46.01 MJ m−3 at high-strain rates of 11,000, 12,800, and 13,600 s−1, respectively. The combined flexible, tough, strong and high-energy adsorption properties empower the FTGA to be a potential armored material for smart wearable electronics where involving complex, dynamic working conditions.

Durability analysis

More valuable results are obtained by further coupling the mechanical and thermoelectric properties of the FTGA. In Fig. 3d, e, we performed a cyclic stress-strain and thermoelectric test on FTGA upon tensile strains ranging from 0% to 30% in the direction perpendicular to the aligned channels, which is sufficient to accommodate the natural movements of the human body for wearable electronics41,42. The results reveal that the FTGA exhibits superior mechanical stability without significant hysteresis behaviour during 100 cyclic tensile strains. Furthermore, even after undergoing 100 cyclic tensile cycles, the thermopower of the FTGA remains at a high level of 5.37 mV K−1 (Fig. 3e), corresponding to a low attenuation of only 3.42% during repeated deformation.

The mechanical-thermoelectric properties of the FTGA can be further reflected under extreme compressive impact. A specimen with dimensions of 35 (L)*25 (W)*2 (T) mm was prepared for a car compression test (Fig. 3f and Supplementary Movie 1). Remarkably, even after being compressed by a car with a total weight of 1600 kg, the FTGA still keeps flexible with no fracture or irreversible deformation. Moreover, it maintains a high thermopower of 4.35 mV K−1 (Fig. 3g), which remains higher than many previously reported flexible quasi-state thermocells19,20,22. These results demonstrate the exceptional mechanical and thermoelectric stability of FTGA, further highlighting its potential for use in robust and sustainable applications such as wearable electronics.

Potential applications of FTGA device and comprehensive performance comparison

The high and stable mechanical and thermoelectric properties of the FTGA enable the exploitation of versatile, sustainable applications in wearable electronics (Fig. 4a). A key application is to serve as flexible armor that leverages the high mechanical toughness and strength of the FTGA to provide protection and enhance the durability of various devices and structures. Moreover, the FTGA is flexible and can be applied to sustainably harvest low-grade heat energy from the human body. By leveraging the temperature difference between the environment (cold source) and the human body (hot source), the FTGA can directly convert heat energy into electrical energy. This energy could then be used to power smartwatch or other wearable devices for monitoring various aspects of human health, including blood sugar levels, blood oxygen levels, heart rate, and blood pressure, etc. Such a self-powered system would enhance the convenience of human-machine interaction by eliminating the need for external power sources.

Fig. 4: Potential application demonstration and comprehensive comparison of the FTGA.
figure 4

a Schematic illustration of the FTGA served as an armor for safeguarding and a flexible thermoelectric generator to power a smartwatch through harvesting thermal energy from human skin. b Photograph of the FTGA module fixed on a human arm for harvesting heat energy from human body. c The output voltage-time curve from harvesting human body heat, the temperature difference (ΔT) of the hot and cold side of the electronics was about 5 °C. d The FTGA module could power various electronic devices with the assistance of a passive voltage booster, including an LED array (d1) and a smartwatch (d2). e Comprehensive comparison of the Carnot-relative efficiency (ηr) versus stretchability for various thermogalvanic cells.

Full size image

To demonstrate the scalability and practical value of the FTGA, a sizeable electronic module consisting of 36 units of the FTGA (size: 5 (L)*5 (W) *2 (T) mm) was designed (Supplementary Fig. 18) and prepared through a three-dimensional (3D) printing method (Supplementary Note 4). To harvest thermal energy from human body (ΔT ~ 5 K), the assembled module was tightly fixed onto a human arm through an adhesive airtight elastomer (VHB 4910, 3 M Company) (Fig. 4b), an output voltage of 0.58 V was generated, demonstrating the ability of the module to sustainably convert thermal energy into electrical energy (Fig. 4c). Furthermore, the assembled module equipped with a voltage amplifier could drive various electronic devices, including an LED array and a smartwatch (Fig. 4d, Supplementary Movie 2). These results highlight the scalability and practicality of the FTGA, demonstrating its potential for large-scale production and utilization in real-world applications, particularly in the field of energy harvesting and powering electronic devices.

Besides being used as a flexible armor for body heat harvesting and safeguarding purposes, the FTGA can also be used to combine with photovoltaic (PV) panel to achieve enhanced output performance (Supplementary Fig. 19). The PV panel can convert solar energy into electricity, however, its efficiency is largely hindered by the huge amount of heat generation during the working process. Therefore, it is  important to find efficient ways to cool the device to sustain its efficiency. Interestingly, the FTGA could serve as an ideal heat sink for PV panel since it contains a lot of water, and achieve fast heat dissipation by water evaporation. Meanwhile, FTGA, itself, could generate electricity by harnessing the waste heat from the PV panel. As a proof of concept, a FTGA was attached on the back surface of the PV panel, and the surface of the PV panel was reduced ~4.3 °C during the working process (Supplementary Fig. 19a). Consequently, the output voltage of the PV panel increased by 2.14% with the assistance of FTGA (Supplementary Fig. 19b). In addition to improving the performance of PV panel, the FTGA can also utilize the temperature difference between the PV panel and the environment, further generating an output voltage of ~9.7 mV (Supplementary Fig. 19c). These results ulteriorly confirm the generality and functionality of the FTGA.

As a promising self-powered electronic device, our FTGA features high flexibility and energy conversion efficiency (Fig. 4e)11,12,17,19,20,21,22,38,43,44,45,46,47. This result indicates that our FTGA not only addresses the leakage risk and rigid properties associated with liquid thermocells, but also substantially surpasses the ηr of the flexible thermocells reported in the literature and exceeds the commercialization threshold (~5%) (Supplementary Table 5), which shows a huge potential for commercial use of our device in future studies. Furthermore, our well-designed FTGA exhibits impressive comprehensive performance compared with other reported quasi-solid thermocells (Supplementary Fig. 20)17,19,20,21,22, mainly including its Carnot-relative efficiency (ηr), electrical conductivity, thermopower (Se), stretchability and force impact resistance properties (Supplementary Table 6). These results convincingly prove the superiority of our strategy for designing high-performance and sustainable thermogalvanic devices.

Discussion

We developed an ultrastrong, flexible thermogalvanic armor that features remarkable Carnot-relative efficiency (8.53%), specific output power density (11.90 mW m−2 K−2), stretchability (~900%), strength (11 MPa) and toughness (70.65 MJ m−3) through the integration of molecular engineering and corresponding structural design. This ingenious device overcomes the trade-off between thermoelectric and mechanical properties that exist in conventional quasi-solid thermocells, thereby offering the potential to efficiently harvest body heat for sustainably powering wearable electronics even after various mechanical deformations, including cyclic tension, car compression, and force impact. From a broader perspective, we propose that our approach may represent a paradigm shift toward enhancing the energy conversion efficiency and mechanical behavior of thermal energy harvesting systems, including quasi-solid thermocells, thermal charging cells48, and thermal regenerative electrochemical cycles49, etc. Based on the synergistic strategies demonstrated in this study, we anticipate significant advancements in energy harvesting and conversion from abundant and omnipresent low-grade heat.

Methods

Materials

The following materials were used: Poly(vinyl alcohol) (PVA, Mw: 146,000–168,000, 99% hydrolyzed, Sigma), Potassium ferricyanide (K3Fe(CN)6, Aladdin), Potassium ferrocyanide trihydrate (K4Fe(CN)6*3H2O, Aladdin), Guanidine sulfate ((Gdm)2SO4, Aladdin), Guanidine hydrochloride (GdmCl, Aladdin), Deionized water (18.2 MΩ cm−1) was obtained from a Millipore-Q system. All materials were used directly without any purification process.

Fabrication of FTGA

FTGA was fabricated through three-step procedures: (1) preparation of 10 wt.% PVA viscous solution, (2) repeated directional freezing and thawing process to form the crosslinked hydrogel network, (3) solution absorption method for redox couples and salts diffusing into hydrogel networks to form FTGA. For step (1), the PVA powders were firstly dissolved in DI water at 95 °C under vigorous stirring to form a homogeneous solution (10 wt.%). Afterward, the uniform PVA aqueous precursor was poured into a homemade mold with polytetrafluoroethylene thermal insulation and a Cu billet bottom for good thermal conduction. For step (2), the mold containing PVA solution was lowered into liquid nitrogen, where half of the Cu billet was submerged into it, and ice could nucleate and grow vertically and generate ice pillars in parallel. The directionally frozen PVA solution was then melted at room temperature. The directional freezing and thawing process was repeated for three times to form the PVA hydrogel. Finally, for step (3), the PVA hydrogel was soaked into the solution of 0.4 M K3Fe(CN)6/K4Fe(CN)6*3H2O and different molar amounts of salts ((Gdm)2SO4) for 24 h to prepare FTGA. In this process, the ions in redox couples and salts would diffuse into gel network, and the thermopower, electrical conductivity and mechanical strength of the hydrogel could be strengthened simultaneously.

Characterization

The structures and morphology of the FTGA were characterized by environmental scanning electron microscope (FEI/Philips XL30 Esem-FEG), the samples were first soaked into deionized water for 24 h, then frozen by liquid nitrogen and freeze-dried by a freeze-dryer (LGJ-18). Voltage-time and current-voltage curves were measured by a Keithley 2400 instrument and an electrochemical station (CHI 660E), respectively. The commercial Platinum wires were used as electrodes. The thermopower measurement was illustrated in Supplementary Note 5. During the voltage and current measurements, the temperatures were controlled by a homemade setup using the TEC temperature control module (21E06), and commercial thermocouples were used to measure the temperature changes in real time too. The current-voltage curves were measured from 0 V to the open-circuit voltage and the corresponding power-current curves were calculated by the product of the corresponding current and voltage values. The effective electrical conductivity (σeff) was calculated from the slope of the current-voltage curve. The effective thermal conductivity (κeff) was measured through a thermal conductivity analyzer (TRA-4A). The tensile curves were recorded on a universal mechanical test machine (Instron 68SC-2) at room temperature. The compressive tests were conducted on a UTM system (Instron series 3382). The cyclic tensile tests were conducted by an ElectroPulsTM (Instron E1000). The Split Hopkinson pressure bar (SN-SHPB-P16) was used to test the high-velocity impact behavior of the FTGA (Supplementary Note 6).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.