An ultrathin ionic thermoelectric cell design utilizing near body heat for self-powered wearable electronics

Abstract

Harvesting low-grade heat is a sustainable way to power wearable electronics, and quasi-solid-state ionic thermoelectric cells offer a flexible, low-cost option. Their use, however, has been limited by a key trade-off: miniaturization reduces the internal thermal gradient and compromises performance. Here, we address this challenge with an ultrathin asymmetric architecture that separates thermal energy harvesting from the conventional reliance on a sustained through-plane temperature gradient. The design couples thermally driven ionic modulation at one interface with engineered pseudocapacitive charge storage at the other. Our 1-mm-thick device delivers an open-circuit voltage of 0.1 V, a power density of 1.6 W m⁻², and an energy density of 1500 J m⁻² using near-body heat. An array of 20 cells generates 1.9 V and a peak power of 23 W m⁻², enabling continuous smartwatch operation. This strategy provides a practical route to ultrathin ionic thermoelectric cells for self-powered wearable systems.

Ionic thermoelectric cells suffer from compromised performance upon miniaturization due to reduced thermal gradients. Here, the authors propose an ultrathin asymmetric architecture that couples ionic modulation with pseudocapacitive storage, enabling high power density for self-powered wearables.

Introduction

Low-grade thermal energy (< 100 °C), which accounts for more than 65% of global energy dissipation, is abundant in sources such as industrial waste heat, environmental fluctuations, and human body heat^1,2. Efficiently converting this low-grade heat into electricity is of growing interest, especially for powering flexible and wearable electronics^3–6. Conventional solid-state thermoelectric materials, typically based on rigid inorganic semiconductors, have been explored for this purpose, especially in ultrathin or flexible formats^7,8. However, their practical deployment is constrained by inherently low Seebeck coefficients near body temperature (usually < 100 μV K⁻¹), mechanical rigidity that limits true conformability, and high manufacturing costs^9,10. As a compelling alternative, ionic thermoelectric cells (ITCs) have emerged, utilizing the thermodiffusion and thermogalvanic effects of ions in an electrolyte to generate giant thermopower on the order of mV K⁻¹ near room temperature, an order of magnitude higher than their solid-state counterparts^11–16. In particular, quasi-solid-state ionic thermoelectric cells (QS-ITCs) based on gel electrolytes offer enhanced safety, mechanical softness, and integration compatibility for wearable systems^17–21.

Despite these advantages, conventional ionic thermoelectric cells face their own critical and unresolved challenge, that is, a fundamental trade-off between device form factor and performance. For true wearability, devices must be ultrathin (< 2 mm), but this miniaturization inevitably collapses the through-plane thermal gradient, especially under mild heat sources such as human skin (ΔT < 5 K). This leads to a sharp decline in voltage and energy output, creating a miniaturization–performance trade-off that remains the principal barrier to practical application. Significant efforts have been dedicated to enhancing their performance by engineering electrolyte composition^22–25, tuning solvation structures^26,27, exploiting thermally induced phase transitions^28–30, and optimizing device architectures^31–33. These strategies have successfully pushed the energy output, with some systems reporting energy densities approaching 400 J m^−231,34,35. However, these advances have not resolved the core trade-off, as top-performing devices are typically several millimeters thick to maintain a sufficient ΔT, fundamentally limiting their integration into truly conformable, low-profile systems. Furthermore, the challenge is compounded in ultrathin QS-ITCs by sluggish ion transport and interfacial charge-transfer kinetics, which further suppress both power density and overall energy conversion efficiency. As such, achieving a high-performance QS-ITC that is both ultrathin and effective under sub-5 K gradients remains a critical challenge.

Here, we resolve this long-standing trade-off by introducing an ultrathin, asymmetric QS-ITC architecture that decouples thermal energy harvesting from the conventional reliance on a sustained through-plane temperature gradient. This architecture is enabled by synergistic interfacial engineering, coupling thermally driven ionic modulation at one electrode with spatially gradient redox activity at an asymmetric counter-electrode, which allows for directional charge accumulation and release. This temporally and spatially decoupled operation permits our 1 mm-thick device to efficiently harvest energy near body temperature (ΔT ≈ 3 K), storing it as a high-potential charge and subsequently delivering stable electrical output. As a result, the device achieves a record energy density of 1500 J m⁻² over 2 h and a steady-state output of 0.1 V under low ΔT conditions. Moreover, a 20-cell integrated array delivers 1.9 V and a peak power density of 23 W m⁻², successfully powering a commercial smartwatch using only body heat. This work provides a foundational strategy for overcoming the miniaturization bottleneck in ionic thermoelectric cells, unlocking their potential for self-powered, autonomous wearable systems.

Results

Design of ultrathin QS-ITC

Integrating multiple ITCs into arrays via series and parallel connections represents an effective strategy for enhancing output voltage and power^30,36,37. Conventional designs, reliant on vertical heat flow between the human body and environment, necessitate a thick, block-sandwich-like configuration to establish a sufficient thermal gradient^7,38,39. This arrangement imposes a fundamental trade-off between device miniaturization and effective thermal gradient (Fig. 1a and Supplementary Table 1).

Fig. 1. Ultrathin QS-ITCs architecture designed to decouple the trade-off between device miniaturization and effective thermal gradient formation.

a Conventional, bulk QS-ITC structure illustrating the vertical heat flow and the resulting thermal gradient. b Temperature distribution maps of quasi-solid-state gel electrolytes of varying thicknesses (H) under fixed boundary conditions: T_H = 36 °C, T_C = 10 °C, and an airflow velocity of 0.3 m s⁻¹. c Experimental measurement of the effective temperature difference (∆T_eff) generated by quasi-solid-state gel electrolytes of different thicknesses under a fixed heat source (T_H = 36 °C) and an ambient environment (T_amb = 12 °C). d Thermal field distribution and the resulting thermoelectric properties of bulk QS-ITCs. e Thermal field distribution and the resulting thermoelectric properties of thin QS-ITCs. f Thermal field distribution and the resulting thermoelectric properties of proposed ultrathin QS-ITCs. T_skin represents the near-body temperature. Source data are provided as a Source Data file.

To quantify this trade-off, the temperature-difference utilization (φ_tdu), defined as the ratio of the effective temperature difference (∆T_eff) to the available temperature difference (∆T_ava) is simulated using finite element analysis (Supplementary Note 1). The simulations are conducted for quasi-solid-state gel electrolytes of various thickness (1, 3, 5, 10, 15 and 20 mm) under conditions representative of wearable applications, encompassing environmental temperatures from 10 °C to 30 °C and airflow velocities from 0.1 m s⁻¹ to 1.0 m s⁻¹ (Fig. 1b and Supplementary Figs. 1, 2). The analysis reveals a positive correlation between ∆T_eff, gel thickness (H), ∆T_ava and airflow velocity. For gels thinner than 5 mm, maintaining a significant thermal gradient becomes challenging. For instance, even under a 10 °C ambient condition (∆T_ava = 26 °C) with an airflow velocity of 1.0 m s⁻¹, the maximum achievable ∆T_eff is limited to 3.6 °C, corresponding to a φ_tdu of only 12%. A further reduction in thickness to 1 mm results in a near-isothermal condition, which suppresses any substantive thermal gradient and thus precludes effective thermoelectric conversion. This trend is corroborated by experimental measurements (Fig. 1c). When subjected to a 36 °C heat source in a 12 °C ambient environment, a 10 mm-thick gel electrolyte attains a ∆T_eff of merely 7 °C. This value diminishes to 3 °C and 1 °C as the thickness is reduced to 5 mm and 1 mm, respectively. Such minimal thermal gradients are insufficient to generate substantial power, even with state-of-the-art ITC materials possessing high thermopower (e.g., 6.5 mV K⁻¹)²⁸.

From a thermodynamic perspective, establishing a sufficiently large ∆T_eff necessitates an adequate thermal diffusion length, which scales with the gel thickness (Fig. 1d). However, increasing the thickness of the QS-ITC introduces greater thermal resistance, resulting in a lower average temperature (T_ave) and extended ion migration paths. Both factors suppress ionic conductivity due to its Arrhenius temperature dependence and the elevated ionic resistance. In contrast, device miniaturization through thickness reduction enhances ion transport by increasing T_ave and shortening diffusion paths, yet simultaneously reduces ∆T_eff because of the limited spatial temperature gradient, thereby lowering the output voltage (Fig. 1e and Supplementary Note 2). As a result, the output power is constrained by the competing effects of reduced voltage and moderately enhanced conductivity. This intrinsic trade-off, rooted in the coupled limitations of heat conduction and ionic diffusion, represents a critical barrier to improving the thermoelectric performance of thin QS-ITCs for wearable energy harvesting.

To address this limitation, an ultrathin QS-ITC that achieves directional-dependent decoupling between the thermal field and potential field is proposed (Fig. 1f). This device is constructed using asymmetric electrodes with distinct thermo-electrochemical characteristics in the ionic electrolyte during the heating process (Fig. 2a). A dual-network copolymer hydrogel (PAM/DAC), which is endowed with elevated ionic conductivity, is synthesized via UV-initiated polymerization of acrylamide (AM) and acryloyloxyethyl-trimethyl ammonium chloride (DAC) monomers^40,41. A capacitive cathode (carbon cloth, CC) and a battery-type anode (polyaniline, PANI) are positioned in close proximity to each other on the same surface of the ultrathin gel. This configuration ensures intimate engagement with the ionic electrolyte while eliminating any inadvertent thermal gradient between electrodes (Supplementary Fig. 3). The resultant assembly of asymmetric electrodes and a laminar gel electrolyte constitutes the ultrathin QS-ITC.

Fig. 2. Principle of the ultrathin QS-ITC.

a Reaction mechanism at the interface between asymmetrical electrodes and gel electrolytes in the ultrathin QS-ITC during heating. b Schematic diagram illustrating the operational cycle of the ultrathin QS-ITC at different stages. Stage I: Initial stage (T_C); stage II: Thermal charging stage (T_H); stage III: Discharging stage (T_H); stage IV: Chemical recovery stage (T_C).

When the ultrathin QS-ITC resides at an initial T_C, the disparity in Fermi levels between the asymmetric electrodes establishes an intrinsic electrochemical potential difference, which manifests as the electrochemical voltage (V_EC) (Fig. 2b and Supplementary Note 3)⁴². Upon heating from T_C to T_H, a thermo-voltage (V_i-TE) arises from two synergistic processes. First, the capacitive CC cathode undergoes rapid and highly reversible pseudocapacitive reactions, wherein the elevated temperature promotes the chemisorption and desorption of electrolyte ions at its oxygen-containing functional groups⁴³. This Faradaic process subsequently lowers the cathode potential, thereby increasing the open-circuit voltage (V_OC). Concurrently, the PANI anode exhibits a modest potential decrease owing to its intrinsic positive temperature coefficient⁴⁴. The thermodynamically favorable reduction of Fe³⁺ to Fe²⁺ (Fe³⁺+ e⁻= Fe²⁺), coupled with the oxidation of reduced PANI chains, supplies electrons for the redox process and further modulates the electrode potential.

Upon connection to an external load, PANI is further oxidized, and electrons traverse the external circuit to the CC cathode, sustaining power delivery at T_H. The discharge capacity originates jointly from the valence-state transition of the PANI electrode and the Fe³⁺/Fe²⁺ redox couple. Although the accumulation of electrons at the CC causes its potential to rise and the discharge voltage (V_Dis) to diminish, the continuous consumption of Fe³⁺ adsorbed at the CC mitigates the rate of this voltage decline, thereby reinforcing the current output. Following the removal of the load and subsequent cooling to T_C, the accumulated Fe²⁺ is oxidized back to Fe³⁺ (Fe²⁺= Fe³⁺+ e⁻), while the released electrons reduce the PANI anode, regenerating its active state under acidic conditions. Simultaneously, ions desorb from the CC surface as the temperature decreases. These collective processes restore both electrode potentials and V_OC to their initial states, enabling cyclical and reversible operation of the ultrathin QS-ITC. The thermoelectric conversion in the QS-ITC is governed by an ionic thermoelectric effect, a mechanism fundamentally distinct from the classical thermogalvanic effect. It operates not via a spatial ∆T but is instead activated by uniform heating. The thermal input induces a synergy of asymmetric, temperature-sensitive interfacial processes (the pseudocapacitive response of the CC cathode, the intrinsic positive temperature coefficient of PANI, and the Fe³⁺/Fe²⁺ redox equilibrium) that collectively modulate the electrode potentials to generate electricity.

The fabrication of asymmetric electrodes of ultrathin QS-ITC

Commercial carbon cloth, derived from the high-temperature carbonization of polyacrylonitrile fibers, is a standard current collector valued for its exceptional electrical conductivity, chemical resistance, and mechanical flexibility. A primary constraint of the pristine material is its inherently low specific surface area and limited porosity, which restricts its electrochemical performance⁴⁵. Furthermore, the smooth, non-polar nature of its surface leads to negligible intrinsic electrochemical activity. To overcome these limitations, a mixed-acid thermal activation strategy is employed to modify the carbon cloth surface (Supplementary Fig. 4), a procedure that substantially enhances its electrochemical properties⁴⁶. During activation, oxygen-containing functional groups (e.g., C = O and O–C = O) are introduced onto the fiber surface. These groups facilitate the reversible chemisorption and desorption of protons and small-radius cations from the electrolyte, thereby enabling temperature-induced pseudocapacitive responses⁴⁷. Concurrently, the incorporation of polar functional groups (–COOH and –OH), along with the formation of a roughened, porous surface morphology, endows the activated carbon cloth with superhydrophilic properties. This modification promotes intimate contact at the electrode–electrolyte interface, accelerating interfacial mass and charge transport. In addition, the activated surface, with its high density of nucleation sites, serves as a conducive platform for the in situ growth of electroactive nanostructures.

The PANI/CC noncovalent composite electrode is fabricated via electrochemical deposition of polyaniline from an acidic aniline solution⁴⁸. This process utilizes the activated carbon cloth as the working electrode within a three-electrode system that includes a platinum counter electrode and a mercurous sulphate reference electrode (Supplementary Fig. 5). Compared with physical methods such as dip-coating, the in situ electrochemical growth of PANI on carbon fibers ensures intimate interfacial contact and strong adhesion. This strategy is highly controllable, providing a standardized and reproducible assembly principle: the loading mass can be precisely determined by the current density and deposition time, and the resulting layer shows exceptional distribution stability, thereby overcoming common interfacial limitations (Supplementary Figs. 6–8). During electrodeposition, polyaniline assembles into coral-like nanofibers that interweave densely across the carbon surface, forming a continuous, intricately entangled network (Fig. 3a, b). Energy dispersive spectroscopy (EDS) mapping confirms the abundant and uniform deposition of PANI on the carbon cloth (Fig. 3c). This porous, interconnected network structure enhances electrical conductivity and facilitates ion diffusion from the electrolyte into the interior, effectively shortening ion transport pathways and promoting rapid Faradaic reactions. The uniform and stable deposition of PANI is enabled by strong π–π stacking interactions between the polyaniline chains and the carbon surface, as well as by the presence of the oxygen-containing functional groups that serve as anchoring sites. This noncovalent integration yields a robust and electrochemically active interface, well-suited for efficient energy conversion.

Fig. 3. Preparation and structural characterization of asymmetric electrodes.

a SEM images of the activated CC. Scale bars: (i) 200 µm, (ii) 4 µm. b SEM images of the PANI/CC electrode. Scale bars: (i) 200 µm, (ii) 4 µm, (iii) 400 nm. c Elemental mappings of the PANI/CC electrode. Scale bars: 10 µm. d FTIR spectra of the CC electrode, PANI, and PANI/CC electrode. e Raman spectra of the CC electrode, PANI, and PANI/CC electrode. f XRD patterns of the CC electrode, PANI, and PANI/CC electrode. g XPS survey spectra of the CC electrode, PANI, and PANI/CC electrode. h XPS spectra for the fitted C1s peaks of the PANI/CC electrode. i XPS spectra for the fitted N1s peaks of the PANI/CC electrode. Source data are provided as a Source Data file.

The structural characteristics of the activated CC, the PANI active material, and the PANI/CC composite electrode are investigated using X-ray diffraction (XRD), Fourier-transform infrared (FTIR) spectroscopy, and Raman spectroscopy (Fig. 3d–f). The XRD pattern of activated CC exhibits broad diffraction peaks near 26° and 43°, corresponding to the (002) and (100) planes of graphite, respectively. The peak broadening suggests a low degree of crystallinity and a disordered graphitic structure. Both PANI and PANI/CC display consistent peaks at approximately 15°, 21°, and 26°, assigned to the (011), (020), and (200) planes, respectively, confirming the formation of the emeraldine salt form of PANI and its effective deposition onto the CC substrate⁴⁹. The graphite-related peak of CC at 26° likely overlaps with that of PANI, while the characteristic peak at 43° remains visible in the PANI/CC composite, indicating that some regions of the CC surface remain uncovered by the active material.

In the FTIR spectra, pristine CC exhibits negligible absorption due to its sp²-hybridized carbon framework and lack of polar functional groups, resulting in the PANI/CC spectrum closely mirroring that of pure PANI. The dominant vibrational bands of PANI appear near 1580, 1500, 1300, 1140 and 830 cm⁻¹, corresponding to C = C stretching in the quinoid and benzenoid rings, C–N stretching, C–H in-plane bending, and out-of-plane C–H bending of para-disubstituted benzene, respectively⁵⁰.

Raman spectral analysis of the activated CC reveals two characteristic peaks at 1345 and 1590 cm⁻¹, corresponding to the D and G bands of carbon materials, respectively. PANI presents multiple distinct bands at 812, 1160, 1210, 1480, and 1590 cm⁻¹, assigned to benzenoid ring deformation, in-plane bending of C-H in semi-quinonoid structures, C-N stretching in benzenoid units, C = N and C = C stretching in quinonoid units. The peak at 1320 cm⁻¹, attributed to C-N⁺ stretching vibrations of polarons, confirms the presence of doped PANI in the emeraldine salt (ES) state within the PANI/CC composite⁵¹.

X-ray photoelectron spectroscopy (XPS) analysis provides further substantiation of the surface composition and chemical bonding (Fig. 3g–i). The survey spectra of PANI and PANI/CC, relative to pristine CC, display additional signals for nitrogen (N1s, 399.5 eV) and chlorine (C12p, 532.5 eV), affirming the successful incorporation of HCl-doped PANI. The high-resolution C1s spectrum of PANI/CC is deconvoluted into four components, namely peaks at 284.8 eV (sp²-hybridized carbon, C–C/C = C), 286.3 eV (C–O/C = N/C-N), 287.4 eV (C = O/C-N⁺), and 288.6 eV (O–C = O)⁵¹. The N1s spectrum identifies four distinct nitrogen species: non-protonated imine (–N = , 398.3 eV), non-protonated amine (–NH–, 399.3 eV), protonated amine (–NH⁺–, 400.1 eV), and protonated imine (–NH⁺ = , 401.7 eV). The concomitant presence of these oxidized and reduced nitrogen species corroborates the formation of a conductive, doped PANI framework and suggests effective π-π conjugation with the carbon substrate.

Working mechanism of ultrathin QS-ITC

The asymmetric electrode configuration is integrated into an ultrathin QS-ITC, coupled with a charge-neutral polyacrylamide (PAM) gel containing FeCl₃ electrolyte (Supplementary Fig. 9). The device with Fe³⁺ electrolyte exhibits a noteworthy V_OC of approximately 0.1 V at 36 °C and a sustainable discharge behavior, with an initial current of 300 μA that stabilizes at ~ 100 μA after 600 s (Supplementary Fig. 10). In contrast, substituting FeCl₃ with KCl leads to a two-order-of-magnitude suppression in the discharge current, which rapidly decays below 1 μA within 30 s, a behavior indicative of a purely capacitive process. Furthermore, the utilization of Fe²⁺ as the primary cation results in a reversed V_EC that counteracts the desired V_OC, thereby constraining overall performance.

The comparative analysis of thermoelectric performance across various electrolyte systems yields several key insights: (a) the V_EC is governed by the specific identity of the specific ionic species present; (b) the V_i-TE is the result of coupled thermo-electrochemical reactions between the asymmetric electrodes and the electrolytes during heating; (c) the sustained discharge capability originates predominantly from the thermogalvanic effect of the Fe³⁺/Fe²⁺ redox couple. To validate these insights, temperature-dependent variations of individual electrode potentials are recorded using an Ag/AgCl reference electrode in systems incorporating FeCl₃, FeCl₂, and KCl as electrolyte.

The selective interactions between electrode materials and ionic species result in different Fermi level alignments at the interfaces, thereby generating differing initial V_EC⁴². Upon heating, the oxygen-containing functional groups (C = O/O-C = O) on the activated CC promote pseudocapacitive behavior by enhancing temperature-dependent ion adsorption at the electrode–electrolyte interface. The coordination of small-radius Fe³⁺ ions with these oxygen sites, along with proton interactions, yields the highest temperature coefficient in the FeCl₃ electrolyte (~ 1.9 mV K⁻¹) (Supplementary Fig. 11). The larger ionic radius and lower surface affinity of K⁺ and Fe²⁺ result in lower temperature coefficients of ~ 1.32 and ~ 1.0 mV K⁻¹, respectively (Supplementary Figs. 12, 13). Distinct from the Fe²⁺ and Fe³⁺ systems, K⁺ does not participate in redox reactions with the electrodes, and the resultant temperature coefficient of ~1.0 mV K⁻¹ is attributed to the combined effects of the CC-side pseudocapacitive behavior and the intrinsic positive temperature coefficient of the PANI electrode (~ 0.3 mV K⁻¹). Therefore, the substantial voltage and persistent high current in the FeCl₃-based system are a consequence of the synergistic effects of temperature-induced pseudo-capacitance at the CC side and redox interactions between the PANI electrode and the electrolyte.

To elucidate the underlying thermodynamic mechanism, a series of ex situ characterizations are performed. The generation of Fe²⁺ during heating is first confirmed using the colorimetric assay o-phenanthroline (o-Phen), which forms an orange-red [Fe(o-Phen)₃]²⁺ complex (Fig. 4a and Supplementary Note 4). Asymmetric electrodes heated to 40 °C in the electrolyte produce a significant color change under both open-circuit and short-circuit conditions, confirming the redox reaction Fe³⁺+ e⁻= Fe²⁺ occurs within the device (Fig. 4b). Quantitative analysis via UV-Vis spectroscopy indicates that approximately 15–20% of the total iron is converted to the Fe²⁺ state, with a higher fraction observed under short-circuit conditions (Fig. 4c).

Fig. 4. Analysis of the working mechanism between the electrode and the electrolyte in ultrathin QS-ITC.

a Images of o-Phen indicator detecting Fe²⁺ at different concentrations in Fe³⁺ solution. b Verification and concentration comparison of Fe²⁺ generated during the heating process in open-circuit (①) and short-circuit (②) states under the solution system. c UV-vis spectra of FeCl₃ solutions with different concentrations of FeCl₂ and asymmetric electrodes in electrolyte solutions after heating in both short-circuit and open-circuit states. d Variation in Fe³⁺ concentration with temperature around the CC. e Variation in Fe³⁺ concentration with temperature around the PANI/CC. The inset shows a magnified view of the dashed region. f Variation in Fe²⁺ concentration with temperature near the CC. g Variation in Fe²⁺ concentration with temperature near the PANI/CC. h Comparison of Fe³⁺ concentration around the CC and PANI/CC at the same temperature. i Comparison of Fe²⁺ concentration near the CC and PANI/CC at the same temperature. j Ex-situ XPS survey spectra of the PANI/CC at different working stages. k XPS spectra fitting for the N1s peak of the PANI/CC at different working stages. Source data are provided as a Source Data file.

To determine the spatial distribution of this reaction, an H-type cell separated by a cation-exchange membrane (CEM) is employed. This experiment confirms that the reduction of Fe³⁺ to Fe²⁺occurs preferentially at the PANI/CC anode (Supplementary Figs. 14, 15 and Supplementary Note 5). This spatial bias is further investigated in the quasi-solid-state gel. Ex situ UV-Vis spectroscopy reveals that as the temperature increases, the concentration of the hydrated Fe³⁺ complex ([Fe(H₂O)₆]³⁺) diminishes near both electrodes, indicating its consumption (Fig. 4d, e, h). Concurrently, the concentration of Fe²⁺ shows a substantial increase near the PANI/CC anode but only a marginal increase near the CC cathode (Fig. 4f, g, i). This spatial disparity is maintained by a dynamic equilibrium: the continuous, thermodynamically favorable reduction of Fe³⁺ at the PANI anode creates a local concentration gradient. This gradient drives the diffusion of Fe³⁺ from the CC side to replenish the reactants at the anode, while the accumulated Fe²⁺ product diffuses from the anode toward the cathode. In contrast to the thermal charging stage, Fe²⁺ ions appear, and their concentration continuously increases near the CC electrode during external discharge (Supplementary Fig. 16). This observation confirms that a persistent reduction reaction occurs at the CC side. Upon circuit closure, the potential difference between PANI and CC drives electron transfer toward the CC electrode. At this stage, the further oxidation of PANI, coupled with the reduction of Fe³⁺ near the CC electrode, jointly sustains the power output.

Moreover, ex situ XPS is employed to monitor the chemical state transitions of PANI across a full operational cycle (Fig. 4j, k and Supplementary Fig. 17). The N1s spectra are deconvoluted into oxidized (-NH⁺=, -NH⁺- and -N = ) and reduced (-NH-) nitrogen species. During thermal charging (Stage I → Stage II), the fraction of reduced nitrogen (-NH-) decreases from 42.8% to 30.6%, while oxidized species increase, signifying the oxidation of PANI coupled with the reduction of Fe³⁺. Upon external discharge (Stage II → Stage III), the PANI is further oxidized, with the -NH- content declining to 26.9%. Finally, during self-recovery upon cooling (Stage III → Stage IV), the PANI electrode undergoes chemical self-reduction. The accumulated Fe²⁺ in the acidic environment donates electrons back to the PANI, restoring its reduced nitrogen components and returning the electrode to its initial state (Supplementary Notes 6, 7).

Overall, the thermo-voltage generation is a synergistic effect of the thermally responsive pseudocapacitive behavior of the CC electrode, the redox reaction between PANI and Fe³⁺, and the intrinsically positive temperature coefficient of PANI. During thermal charging, the reduction of Fe³⁺ occurs exclusively at the PANI/CC anode. Upon discharging, the high-capacity PANI electrode delivers electrons to the CC electrode through the external circuit. Simultaneously, Fe³⁺ near the CC electrode is reduced to Fe²⁺ by these incoming electrons. This reaction at the cathode not only mitigates the consumption of oxygen-containing functional groups on the CC surface, thereby enhancing reversibility, but also delays the voltage decay during discharging. The sustained oxidation of the PANI electrode, together with the continuous redox interactions between the asymmetric electrodes and the electrolyte, endows the ultrathin QS-ITC with a prolonged discharge duration.

Component optimization and thermoelectric performance

The influence of electrolyte composition and PANI electrode formulation on thermoelectric performance is systematically investigated using a custom-designed experimental setup (Supplementary Fig. 18). The distinct thermo-electrochemical behaviors of the asymmetric electrodes and the electrolyte establish a thermo-voltage that is positively correlated with temperature. This effect is attributed to the synergistic contribution of a pronounced positive temperature coefficient from the pseudocapacitive CC and the redox reaction between PANI and Fe³⁺ (Fig. 5a and Supplementary Fig. 19). For instance, in a 0.1 M FeCl₃ electrolyte, the V_OC of the ultrathin QS-ITC increases from 76 mV to 89 mV as the temperature rises from 278 K to 309 K.

Fig. 5. Composition optimization and thermoelectric performance of ultrathin QS-ITC.

a Temperature-dependent open-circuit voltage. b Effect of FeCl₃ concentration on the open-circuit voltage and maximum power density. c Influence of DAC content on the open-circuit voltage and maximum power density. d Effect of PANI loading mass on the open-circuit voltage and maximum power density. e Voltage curve across four operational stages, with 10 kΩ external resistance during power output stage. f Voltage variation over 20 charge-discharge cycles with an external 5 kΩ resistor in quasi-continuous mode. Inset shows the voltage curves of cycles 5–10. g Discharge process after thermal charging at 36 °C with a 500 Ω external resistance over 2 h. h Power curves during the 2 h discharge at varying external resistances, with the testing circuit schematic in the inset. i Energy density during the 2 h discharge for different resistances, calculated by integrating the power over time. j Comparison of the output performances and thermoelectric conversion efficiency with reported ITCs^54–57. k Comparison of the energy output density and corresponding required hot temperature with those of reported QS-ITCs^{12,17,32,34,35,52,53,57}. Source data are provided as a Source Data file.

The post-heating V_OC consists of the electrochemical voltage (V_EC) and the thermo-voltage (V_i-TE). The V_EC represents the intrinsic electrochemical potential difference, which is determined by the selective interaction between electrode materials and ionic species and by the reconfiguration of interfacial thermodynamic equilibria. These factors are influenced by ion species, concentration, solvation structure, and interfacial chemistry. As the concentration of FeCl₃ increases from 0.1 M to 1.5 M, the V_OC continuously declines from 89 mV to 48 mV, a trend attributed to the inverse correlation between ionic concentration and electrochemical potential (Fig. 5b and Supplementary Fig. 20). In contrast, the maximum power density (P_max) exhibits a non-monotonic dependence, initially increasing and then decreasing, with a peak at 0.5 M. Despite the reduction in V_OC at higher concentrations, the decreased internal resistance due to enhanced carrier density improves the ionic conductivity of the gel electrolyte, yielding the highest P_max of 0.34 W m⁻² at the optimal 0.5 M FeCl₃ concentration. To further enhance conductivity, DAC, which features quaternary ammonium groups, is introduced as a comonomer with PAM. This modification has a negligible effect on V_OC, but boosts P_max to 0.42 W m⁻² at an optimal DAC:AM mass ratio of 0.3, (Fig. 5c and Supplementary Fig. 21). Beyond this ratio, elevated crosslinking density constrains ion mobility and diminishes power output (Supplementary Fig. 22).

The PANI loading on the anode is also a pivotal parameter. Increasing the PANI loading enhances V_OC at 309 K due to a lower electrode potential of the PANI layer relative to the CC cathode, while the augmented electroactive surface area facilitates charge transfer (Fig. 5d and Supplementary Fig. 23). Consequently, P_max reaches 1.0 W m⁻² at a PANI loading of 2.8 mg cm⁻². Loading beyond this optimum, conversely, diminishes performance, as the excessively thick PANI layer extends ion diffusion paths and hinders electron transport, leading to increased interfacial and migration resistances. A decline in P_max is observed at a loading mass of 3.4 mg cm⁻². In addition, excessive loading could compromise electrode integrity, with the loosely bound PANI material being susceptible to detachment during operation (Supplementary Fig. 24). Moreover, reducing the inter-electrode distance from 10 mm to 5 mm significantly shortens the ion transport path, enhancing P_max from 1.0 W m⁻² to 1.6 W m⁻² (Supplementary Fig. 25).

The complete operational cycle of the ultrathin QS-ITC comprises four stages: thermal charging, power output, equilibration, and chemical recovery (Fig. 5e). Upon heating from 278 K to 309 K, a prompt thermo-voltage emerges from the asymmetric Faradaic reactions and saturates within three minutes. When connected to an external load, electrons flow from the battery-type anode to the capacitor-type cathode, diminishing the internal electrostatic field and causing a gradual decrease in discharge voltage. After disconnection, partial voltage recovery occurs, as Fe³⁺ reduction on the CC side preserves surface functionalities, though incomplete reversal of the PANI redox state limits full restoration. Finally, during cooling under open-circuit conditions, oxidized PANI reacts with Fe²⁺ to regenerate its reduced form, completing the self-regeneration cycle of both PANI and Fe³⁺.

The ultrathin QS-ITC demonstrates excellent rechargeability and stability under both quasi-continuous and continuous operation. In quasi-continuous mode, the ultrathin QS-ITC charges to saturated voltage at 309 K and undergoes 20 charge-discharge cycles using a 5000 Ω load (10 s per discharge) (Fig. 5f). The voltage is capable of recovering within 3 minutes between cycles, retaining over 98% of its initial value and indicating high reversibility (Supplementary Figs. 26, 27). Moreover, the QS-ITC exhibits excellent thermo-response stability over 50 repeated heating and cooling cycles, maintaining its output voltage with only minor decay (Supplementary Fig. 28). This robust reversibility is primarily attributed to the highly reversible redox behavior of PANI in acidic conditions, which enables sustained interfacial reactions and stable ionic transport. In continuous operation, the ultrathin QS-ITC discharges over 2 hours under a 500 Ω load after saturation at 309 K (Fig. 5g). Despite the rapid attenuation of both output voltage and current attributable to ohmic losses and other parasitic effects, a state of quasi-steady-state output is eventually achieved. By integrating the voltage-current profiles over the 2 h duration, the quantified energy density across a series of external resistances reveals a parabolic dependence (Fig. 5h, i and Supplementary Fig. 29). The device achieves a peak energy density of 1500 J m⁻² for a 5 mm electrode distance and 500 Ω load, significantly surpassing those recorded at 50, 1000, and 1500 Ω (250, 900, and 750 J m⁻², respectively), and markedly exceeding the saturated energy density at a 10 mm electrode distance (600 J m⁻²). By integrating thermally induced interfacial ionic modulation with spatially engineered electrochemical activity, the ultrathin QS-ITC (1 mm) addresses the critical challenge of achieving high energy output in an ultrathin form factor under the small temperature gradients with thin/bulk QS-ITCs. This strategy demonstrates excellent performance of a V_OC of 100 mV, a P_max of 1.6 W m⁻², a relative Carnot efficiency of 4%, and an E_2h of 1500 J m⁻² at 309 K, representing a comprehensive enhancement in thermoelectric performance that markedly outperforms existing ITCs (Fig. 5j, k, Supplementary Figs. 30–32, Note 8 and Tables 2, 3)^{12,17,32,34,35,52–57}.

Proof-of-concept demonstration and stability in wearable applications

To highlight the advantages of the ultrathin asymmetric architecture, a comparative evaluation of diurnal output stability is conducted between the ultrathin QS-ITC and a high-performance conventional thermogalvanic cell [Fe(CN)₆^3-/4- with Gdm⁺] under identical environmental conditions (Fig. 6a). With the T_H fixed at 36 °C to simulate body temperature, the conventional thermogalvanic cell exhibits a substantial voltage fluctuation, ranging from 10 mV to 67 mV, as the ambient temperature varies between 11.7 °C and 28.1 °C over a 12 h period. This instability is exacerbated by higher thermopower, which renders the device more sensitive to temperature perturbations. In stark contrast, the ultrathin QS-ITC maintains a highly stable output voltage within the narrow range of 85–100 mV under the same conditions. This exceptional thermal resilience is a direct result of its operation mechanism, which generates voltage from asymmetric interfacial reactions on an isothermal plane, rather than being dictated by a directional thermal gradient. The design strategically decouples the temperature field from potential generation, reconciling the competing demands of an ultrathin architecture and effective thermal utilization, thereby achieving stable, high-performance output in dynamic ambient environments.

Fig. 6. Output stability and proof-of-concept wearable device.

a Comparison of open-circuit voltage stability between the ultrathin QS-ITC and a bulk QS-ITC under ambient temperature fluctuations. b Images of an integrated 5-unit wearable device. c Output voltage curves of wearable devices with varying unit numbers at 36 °C. d Scalability of the output voltage with the increased unit numbers. Data are presented as mean values ± S.D. For the measurements of output voltage, n = 3. e Current-voltage and power curves of the wearable devices. f Comparison of output performance of this work with previously reported wearable flexible thermoelectric devices for body heat harvesting^{17,24,25,31,34,53,55}. g Direct power supply to an electronic watch through body heat. h Powering a hygrometer using a 20-unit wearable device at 36 °C without requiring a voltage amplifier. Source data are provided as a Source Data file.

Flexible, wearable ultrathin QS-ITC arrays composed of 5, 10, and 20 units are assembled via straightforward serial integration to demonstrate scalable low-grade heat harvesting (Fig. 6b and Supplementary Fig. 33). These serially connected devices exhibit pronounced thermoelectric responses, with the V_OC increasing linearly with the number of units, reaching 0.48 V, 0.94 V, and 1.9 V at 36 °C for arrays comprising 5, 10, and 20 units, respectively (Fig. 6c, d). Although the short-circuit current density shows a gradual decline due to the cumulative interconnection resistance, the 20-unit array nonetheless achieves a peak power density of 23 W m⁻², underscoring the excellent scalability of the ultrathin QS-ITC architecture (Fig. 6e). When benchmarked against existing wearable devices for human body heat harvesting, this ultrathin QS-ITC configuration delivers not only stable output but also a more than a five-fold enhancement in power output at comparable voltage levels (Fig. 6f and Supplementary Table 4)^{17,24,25,31,34,53,54}.

Benefiting from its stable, high-performance output and the inherent mechanical flexibility of all components, the 20-unit array achieves full conformability with the human body and can effectively harvest body heat to directly power commercial electronics, such as a smartwatch (Fig. 6g and Supplementary Movie 1). Furthermore, the high V_OC of 1.9 V enables the device to directly operate a digital hygrometer/thermometer, a task that typically requires a voltage amplifier for other QS-ITCs (Fig. 6h and Supplementary Fig. 34). In addition, to evaluate the biocompatibility of the device for potential wearable applications, a skin irritation test is performed (Supplementary Fig. 35). It is demonstrated that no visible erythema, edema, or irritation is observed on the skin surface after direct contact with the device, confirming its excellent dermal compatibility. Moreover, the epithelial tissue remains intact with no inflammatory cell infiltration even after 72 h of continuous contact with the device, indicating negligible cytotoxicity and excellent biocompatibility. These demonstrations establish the ultrathin QS-ITC as a high-performance, mechanically compliant platform for wearable thermal energy conversion, offering significant promise for powering next-generation electronics.

Discussion

In summary, we present a high-performance, ultrathin (1 mm) quasi-solid-state ionic thermoelectric cell that resolves the long-standing miniaturization–performance trade-off in flexible thermal energy harvesting. By decoupling thermal-to-ionic conversion from the conventional reliance on a sustained through-plane temperature gradient through a planar, asymmetric architecture, our system operates efficiently under minimal temperature gradients (about 3 K) without sacrificing energy output or flexibility. This is enabled by synergistic interfacial engineering—combining thermally modulated ion dynamics with spatially graded redox activity—which together achieve stable voltage generation (0.1 V), exceptional energy density (up to 1500 J m⁻²), and robust operation under fluctuating ambient conditions. A scalable 20-cell integrated array further delivers 1.9 V and a record-high power density of 23 W m⁻² at near-body temperatures (about 36 °C), directly powering commercial wearable electronics. This work establishes a design paradigm for soft thermoelectric systems, offering a universal strategy to unlock efficient, autonomous, and conformable thermal energy harvesting for next-generation wearable and distributed electronics.

Methods

Materials

The following raw materials were used: Ferric chloride (FeCl₃), Ferrous chloride (FeCl₂), Dibasic Sodium Phosphate (Na₂HPO₄), Acrylamide (AM), 2-hydroxy-2-methylpropiophenone (HMPP), Poly(ethylene glycol) diacrylate (PEGDA), Acryloyloxyethyl-trimethyl ammonium chlorid (DAC), Aniline were purchased from Aladdin industrial Corporation. HCl, H₂SO₄, HNO₃ were purchased from Sinopharm Chemical Reagent Co., Ltd. 1,10-Phenanthroline Monohydrate (o-Phen) was purchased from TCI (Shanghai) Development Co., Ltd. The purity of all chemicals regents has reached analytical grade, and they can be directly employed in experiments without further purification. Carbon cloth (W0S1011) was purchased from CeTech CO., Ltd., China.

Fabrication of the PANI/CC electrode

Carbon cloth (CC) was sequentially cleaned with acetone, sulfuric acid, deionized water, and ethanol, and then vacuum-dried at 40 °C for 3 h. Subsequently, the dried CC was immersed in a mixed acid solution of sulfuric acid and nitric acid (volume ratio of 3:1) and heated in an 80 °C oil-bath for 12 h to prepare the activated CC. PANI was electrodeposited onto activated CC in a three-electrode configuration using a mercury sulfate electrode as reference and Pt foil as counter electrode. The electrodeposition electrolyte contained 0.2 M aniline in 1 M HCl. Galvanostatic polymerization was conducted at a 1 mA cm⁻². The loading mass of PANI was controlled by adjusting the constant current and deposition time. Subsequently, the obtained PANI/CC electrode was rinsed with ultrapure water and ethanol under mild agitation, vacuum-dried at 40 °C for 12 h and immersed in a 1 M HCl solution for 48 h to ensure full protonation.

Fabrication of the ultrathin QS-ITC

The quasi-solid-state electrolyte is denoted as PAM/x DAC/m FeCl₃, where x is the mass ratio of DAC to AM and m is the molar concentration of FeCl₃. For PAM/0.3 DAC/0.3 FeCl₃, 2.986 g AM monomer was dissolved in 7 mL of deionized water, followed by the addition of PEGDA (10 μL) and HMPP (10 μL) as crosslinker and photoinitiator. After stirring in the dark for 10 min, 0.9 g DAC solution was added and stirred for another 10 min to obtain the precursor solution for the PAM/0.3 DAC gel. The precursor solution was then poured into a glass mold and exposed to UV light (365 nm) for 5 min, resulting in the formation of the PAM/0.3 DAC hydrogel substrate. An ionic electrolyte solution was prepared by adding 1 vol% of 3 M HCl to 0.3 M FeCl₃. Unless otherwise stated, the electrolyte solution is prepared with HCl at this concentration. Finally, the PAM/0.3 DAC gel was immersed in the electrolyte solution for 24 hours to facilitate ion exchange, thereby producing a quasi-solid-state electrolyte. For ultrathin QS-ITC assembly, activated CC (15 mm in length and 5 mm in width) and PANI/CC electrode of identical dimensions were fixed to the same surface of the quasi-solid-state electrolyte (20 mm in length, 5 mm in width, and 1 mm in thickness), with a 10 mm gap between the electrodes. The assembled ultrathin QS-ITC was then encapsulated with 0.5 mm thick VHB tape.

Characterization

The morphology of the electrodes and the quasi-solid-state electrolyte were examined by scanning electron microscopy (SEM, Zeiss Ultra PLUS). The crystal structure and phase composition of PANI were analyzed by X-ray diffraction (XRD, Rigaku D/max-2550VB). The chemical composition and structural features of the electrodes were characterized using Fourier transform infrared (FTIR, Thermofisher Nicolet IS50) spectroscopy and Raman spectroscopy (HORIBA XploRA PLUS). The elemental states and redox reaction mechanisms on the surface of the PANI/CC were investigated via X-ray photoelectron spectroscopy (XPS, Escalab 250Xi). Variations in ion concentration within the electrolyte were measured using ultraviolet-visible (UV-vis, UV-2600) spectroscopy.

Author contributions

W.G., Y.C., and H.M. conceived the idea and designed the project. W.G. and Y.C. supervised the project. H.M. performed the experiments and processed the data. W.G., Y.C., and H.M. analyzed the data, drafted, and revised the paper. All authors discussed the results and commented on the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data supporting the findings of this study are available within this article and the Supplementary Information or from the corresponding author upon request. The data generated in this study are provided in the Supplementary Information/Source Data file. Source data are provided in this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-71286-2.