Personal health assistant HES-CHAT e-skins: integrated mechanosensitivity, electromagnetic shielding, and electrochemical energy storage

Abstract

Flexible electronic skins (e-skins) have emerged as a promising technology for various applications, including health monitoring. In this work, we present a novel study on a multifunctional, conductive hydrogel with integrated ChatGPT as an electronic skin to become a personal health assistant. The polyvinyl alcohol/polyethylene oxide–polyaniline hydrogel e-skin with ChatGPT (HES-CHAT e-skin) hydrogel demonstrated excellent mechanical flexibility and deformation response under certain strain conditions (500%), enabling it to function as a flexible sensor for monitoring human gestures. At the same time, it had good utility in the field of electromagnetic shielding, showing excellent electromagnetic shielding performance (∼59.7 dB) at a thickness of 1 mm. Furthermore, it could maintain a good level of shielding after a long period of time and high tensile deformation (47.0 dB after 56 days of storage, and 42.7 dB after 500% strain stretching). The incorporation of conductive materials with high electromagnetic interference shielding properties improved user safety and device functionality, making the HES-CHAT e-skin suitable for environments with high levels of electromagnetic interference. Additionally, the composite hydrogel demonstrated remarkable electrochemical energy storage properties. Symmetric supercapacitor devices with high volumetric capacitance (7848 mF cm⁻³ at a current density of 5 mA cm⁻³) and long cycle stability (81.3% capacitance retention after 10 000 cycles of testing), along with a high power density of 40 W cm⁻³ and a high energy density of 1090 mW h cm⁻³, were obtained. This multifunctional conductive polymer hydrogel provided a novel strategy for the development of flexible electronic devices in the field of smart hydrogels.

---

1 Introduction

The demand for wearable sensors in social life is increasing, driven by advancements in materials and multifunctional uses. Researchers working on electronic skins (e-skins), a new type of flexible wearable sensor, are engaging with the growing artificial intelligence industry, bringing new opportunities.^1–3 However, these opportunities are accompanied by challenges. The rapid development of artificial intelligence devices across multiple application fields has limited the application of electronic skins with a single function.^4–6 Therefore, there is an urgent need to develop a multifunctional and high-performance electronic skin.

In recent years, hydrogels have garnered significant attention from researchers as potential materials for electronic skin applications. Hydrogels are gels with a three-dimensional network structure, composed of hydrophilic polymer materials. Hydrogels with polymers as the framework material exhibit excellent mechanical properties.^7–9 Conductive hydrogels (CHs) are a type of emerging material that combines excellent electronic performance with the flexibility of soft tissues, making them ideal candidates for flexible epidermal sensors. Many CHs demonstrate exceptional biocompatibility, mechanical flexibility, and sensitivity, as well as the capability to effectively transduce physiological motion signals into electrical signals.^10–12 The versatility of CHs aligns well with the evolving concept of multifunctional electronic skin, and there is a growing trend towards their development as flexible energy storage devices.^13,14

Most CHs have good application prospects in the field of flexible energy storage. Xu et al.¹⁵ constructed a fully flexible and healable all-hydrogel prepared from a dynamically crosslinked polyvinyl alcohol@polyaniline (PVA@PANI) hydrogel, which not only could well detect and recognize body movements, but also showed a high specific capacity of 936.8 F g⁻¹ and could provide an energy density of 40.98 W h kg⁻¹. Liu et al.¹⁶ prepared H-Gel/AS-MWCNTs-PPy-AS hydrogels by a template degradation method and assembled them into flexible sensors with high sensitivity for mechanical detection, and assembled them into flexible solid-state supercapacitors with a capacitance of 75 F g⁻¹, and a capacitance retention of 98.1% after 5000 cycles in a flexed state. Additionally, with the increasing popularity of electronic devices and wireless communication systems, electromagnetic radiation exposure has become a concern regarding user safety and device functionality.^17–19 Therefore, incorporating electromagnetic shielding properties into electronic skins is essential to minimize potential risks and ensure reliable operation. Conductive hydrogel e-skins with electromagnetic shielding properties have a very promising future.^20–22 Hydrogels consist of a three-dimensional network skeleton and a continuous aqueous phase in which conductive substances form a network, showing potential in providing absorption-based electromagnetic interference (EMI) shielding properties.^23–29 Several studies have explored the incorporation of different conductive materials into hydrogels to enhance their EMI shielding properties. For instance, Zhao et al.³⁰ developed a conductive, stretchable, adaptable, self-healing, and biocompatible liquid–metal GaInSn/Ni-based composite hydrogel by incorporating magnetic liquid metal into a polyvinyl alcohol hydrogel backbone. This composite hydrogel exhibited significantly improved EMI shielding performance compared to pure hydrogels. Zhu et al.³¹ synthesized hydrogel-type shielding materials by combining MXene and polyacrylic acid, achieving high EMI shielding effectiveness and excellent reflection loss in a thin hydrogel. Zhang et al.³² synthesized three-dimensional porous charcoal composites for efficient electromagnetic wave absorption.

In this work, we introduced an innovative design for a polyvinyl alcohol/polyethylene oxide–polyaniline (PVA/PEO-PANI, PP-PANI) hydrogel electronic skin (e-skin) integrated with ChatGPT (HES-CHAT) functioning as a personal health assistant. This integration enabled the e-skin to not only detect and monitor physical stimuli but also provide intelligent and personalized health-related information and recommendations. The incorporation of electromagnetic shielding properties into the e-skin was essential to mitigate potential risks and ensure reliable functionality. Furthermore, we developed an e-skin with both electrochemical energy storage and electromagnetic shielding properties, which was equivalent to an e-skin capable of being assembled into an energy storage electronic device possessing electromagnetic shielding properties. While enhancing the electrical versatility of this hydrogel for use as a personal health assistant, we had also reduced the risk of high radiation exposure in its application and ensured operational feasibility. By integrating conductive materials with high EMI shielding effectiveness into the HES-CHAT e-skin, we enhanced its ability to block and absorb electromagnetic waves, protecting the user from potential hazards and enhancing the reliability of the integrated electronic components.³³ This additional functionality broadened the potential applications of the HES-CHAT e-skin, rendering it suitable for deployment in environments characterized by elevated levels of electromagnetic interference, such as industrial settings or areas with dense wireless communication networks.

2 Results

2.1 Design of the PP-PANI hydrogel e-skin with ChatGPT (HES-CHAT)

Flexible e-skins have gained significant attention due to their potential in monitoring human health conditions. However, existing e-skins often lack the ability to provide real-time feedback and intelligent analysis for personal health management.³⁴ In this work, we proposed the integration of mechanosensitive sensors with ChatGPT, a state-of-the-art language model, to develop a comprehensive personal health assistant. The integrated e-skin system demonstrated excellent performance in both sensing capabilities and interactive functionality. The mechano-sensitive sensors exhibited high sensitivity, with a wide detection range, enabling accurate monitoring of physiological parameters and physical activities. Moreover, the integration of ChatGPT allowed for personalized health management, answering queries, providing health advice, and generating informative reports based on the collected data. Furthermore, e-skins had electrochemical energy storage and electromagnetic shielding properties, which enabled them to be assembled into flexible electronic devices and used as electromagnetic shielding devices in electronic skins used as personal health assistants (Fig. 1).

Fig. 1 Schematic of HES-CHAT e-skin. A multifunctional hydrogel electronic skin with mechano-sensitivity, electromagnetic shielding, and electrochemical energy storage, integrated with the ChatGPT personal health assistant.

In order to design excellent mechanical flexibility, electrical conductivity, and electromagnetic shielding properties all integrated into CPs, this work demonstrated an ingenious design idea and material composition. This work was mainly based on non-covalent chemistry (Fig. 1). Firstly, a heating and stirring method was employed to blend PVA with PEO, and a freeze–thaw process was utilized to prepare a three-dimensional porous hydrogel. The gel formed during low-temperature freezing was solidified into ice, causing a change in the physical structure of the hydrogel due to the arrangement of water molecules in an ice crystal structure. The volume space created by ice crystal formation served as the pore structure in the hydrogel for the subsequent introduction of conductive polymers. Upon thawing, the hydrogel underwent pore generation as a result of ice crystal melting, thereby forming an excellent three-dimensional porous framework. The PEO molecular chains, characterized by high crystallinity and molecular weight, could undergo physical entanglement cross-linking with PVA molecular chains, restricting their movement and enhancing the mechanical properties of the PVA hydrogel.³⁵ While the –OH groups in PVA did not form hydrogen bonds with the oxygen atoms in PEO,³⁶ both could interact with water molecules, providing a theoretical basis for the construction and application of the three-dimensional porous hydrogel framework. Secondly, conductive PANI monomers (aniline) were incorporated into the three-dimensional porous framework of the hydrogel material. Ammonium persulfate was used as an initiator to make it polymerize in situ to form conductive PANI, which grew freely in the three-dimensional skeleton of the hydrogel; water molecules formed intermolecular hydrogen bonds with the amino groups in the molecular structure, enhancing the mechanical stability of the hydrogel. Based on this the introduction of PANI not only enhanced electrical conductivity,³⁷ but also greatly strengthened the EMI shielding effectiveness of the hydrogel.

Furthermore, by changing the component ratios of PP-PANI composite hydrogels to simultaneously obtain excellent mechanical flexibility, electrochemical properties, and electromagnetic shielding properties, PVA molecules and PEO molecules physically crosslinked to form a hydrogel three-dimensional porous skeleton containing conductive polyaniline; these molecules could be formed with water molecules' hydrogen bonding to provide mechanical stability for the hydrogel's three-dimensional porous skeleton, which could be used as a flexible strain sensor to respond to human motion. The conductive polyaniline provided a path for electrons to move, making the hydrogel electrically conductive, and the three-dimensional porous skeleton of the hydrogel provided a pathway for electrolyte ions to move through, and electrochemical performance tests showed the potential for use as a supercapacitor electrode material with good electrochemical performance. The free electrons in the conductive polyaniline could respond to electromagnetic waves as they propagate through the hydrogel.³⁸ The pores in the three-dimensional porous skeleton structure formed by the freeze–thawing method provided more surface area for electromagnetic wave propagation and absorption, and enhanced the interaction with electromagnetic waves, and the composite hydrogel showed excellent electromagnetic shielding performance.

2.2 Morphological and structural characterization of the HES-CHAT e-skin

The composite hydrogel samples were processed by freeze-drying to form an aerogel, and the microscopic morphology of the samples was analyzed using a scanning electron microscope. The PP(PVA/PEO)-based hydrogel was a three-dimensional porous skeleton structure (Fig. 2a and S1†), which was well able to provide channels for electrolyte ion transport and storage, and it was an excellent substrate material for electrodes. Polyaniline was synthesized in PP hydrogel, and with the increase of polymerization time, the morphology of polyaniline transitioned from nanoparticle-like to fiber-rod-like (Fig. S1a–d†), and finally showed the spontaneous interweaving of polyaniline to form the polyaniline nanofiber three-dimensional network structure (Fig. 2b and S1e, f†). The polyaniline was close to the skeleton in the porous skeleton network structure of PP hydrogel. When the content of polyaniline increased, the storage and transportation of electrolyte ions would be affected to a certain extent due to the spatial limitation of the three-dimensional porous structure inside the PP, and the shape and the quality of the three-dimensional networks of the polyaniline nanofibers will be limited, so the contact between the polyaniline and the electrolyte ions was restricted, which resulted in limited contact between the PP-PANI composite. The electrochemical performance of the hydrogel would also be limited.

Fig. 2 Characterization of HES-CHAT hydrogel. (a) Cross-sectional SEM image of PP hydrogel. (b) Cross-sectional SEM image of PP-PANI60 hydrogel. (c) Infrared spectra of PVA, PP, PANI, and PP-PANI hydrogel. (d) XPS characteristic peak spectra of PP-PANI hydrogel. Electrochemical profile of flexible hydrogel supercapacitors: comparison of (e) CV curves and (f) GCD curves. Symmetrical supercapacitor device: (g) CV curves, (h) GCD curves, (i) capacitance retention after 10 000 cycles, (j) Ragone diagram.

The composition of the materials was further determined from the IR (Fig. 2c) and XRD spectra (Fig. 2d) of PVA, PP, and PP-PANI hydrogels, and XPS spectra (Fig. S3†) of PP-PANI composite hydrogels. The characteristic peaks near 840 cm⁻¹ and 2940 cm⁻¹ corresponded to the stretching vibration of –CH₂ in PVA, and the characteristic peaks near 1430 cm⁻¹ and 3350 cm⁻¹ were caused by the stretching vibration of –OH in PVA.³⁹ When PEO was added to PVA, the peaks at both 1430 cm⁻¹ and 3350 cm⁻¹ became slightly flat and the stretching vibration of –CH₂ in PEO appears at 2887 cm⁻¹.⁴⁰ The main characteristic peaks of polyaniline were observed near 1600 cm⁻¹, 3450 cm⁻¹, and 1301 cm⁻¹ corresponding to the characteristic vibrations of the quinone structure, the N–H stretching vibration, and the C=N stretching vibration, respectively.⁴¹ In addition, in the X-ray diffraction pattern of the PP-PANI composite hydrogel (Fig. S3†), the characteristic peaks of the hydrogel were consistent with those reported previously. Pure PVA exhibited distinct diffraction peaks around 20°, and with the addition of PEO, the diffraction peak at 20° for the PP hydrogel became less pronounced, while a characteristic diffraction peak of PEO appears around 23°.⁴² As a result, the addition of PEO had led to a certain degree of alteration in the crystallinity of PVA. The diffraction peaks around 16°, 23°, and 26° corresponded to the characteristic peaks of the (011), (020), and (200) crystal planes of PANI.⁴³

Furthermore, in the XPS spectrum of the PP-PANI composite hydrogel, the signals of C1s, N1s, and O1s were consistent with the expected results. The N1s spike observed at 399 eV was attributed to PANI. The asymmetric peaks could be deconvolved into three peaks, quinone amine (–N=) at 398.1 eV, aniline (–NH₄⁻) at 399.6 eV, and the nitrogen cation radical (–N⁺) at 401.4 eV. The PP hydrogel appeared white and transparent, and it was compounded with polyaniline to form a conductive PP-PANI composite hydrogel, which changed from a white transparent state to a dark green state under natural light (Fig. S2a†). In the Raman spectrum presented in Fig. S4,† the D and G bands of the PP hydrogel were observed at approximately 1660 cm⁻¹ and 1450 cm⁻¹, respectively, which can be attributed to the C–C stretching vibrations of polyvinyl alcohol (PVA). In contrast, the PP-PANI hydrogel exhibits more pronounced peaks around 1570 cm⁻¹ and 1400 cm⁻¹. These peaks were ascribed to the C–C bond stretching vibrations associated with the benzene and quinone rings, as well as a notable peak corresponding to the C–N bond stretching vibrations. In a complete circuit with an illuminated LED, the PP-PANI composite hydrogel was used to replace the wire to connect to the battery device, and the light bulb was still observed to be illuminated (Fig. S2b and S5†). At this point, the hydrogel replacing the wires was manually stretched to change the morphology, and it was observed that the brightness of the bulb slowly decreased with the elongation of the hydrogel, which demonstrated that PP-PANI was a compliant hydrogel with the potential to be used as a supercapacitor electrode material. Therefore, the conductivity of hydrogels with different polyaniline content was tested, and it was detected that the conductivity of the PP hydrogel without polyaniline content was approximately 0 (Fig. S2c†), while the conductivity of the PP-PANI composite hydrogel increased with the increase of polyaniline content, but the conductivity of the composite hydrogel with too much content decreased, and the conductive abilities remained the same. The conductivity of the PP-PANI60 composite hydrogel was 28.16 mS cm⁻¹ and that of the PP-PANI80 composite hydrogel was 21.7 mS cm⁻¹, and the results further confirm the morphology presented in the scanning electron microscope (SEM) images. The micro-morphology and structure initially indicate that PP-PANI60 had more potential to be an excellent electrode material, which could also be confirmed by the subsequent electrochemical test results.

2.3 Electrochemical properties of HES-CHAT hydrogels

In order to evaluate the prepared HES-CHAT hydrogel electrode materials, the electrochemical properties of hydrogel materials with different polyaniline contents were measured in this work using a three-electrode device. At a scan rate of 10 mV s⁻¹ (Fig. 2e), the cyclic voltammetry (CV) curves of the PP hydrogel approximated a straight line, exhibiting its very small capacitance case. As the content of polyaniline in the hydrogel increased, the area of the CV curves of the hydrogel increased significantly when a small amount of polyaniline nanoparticles (PP-PANI40) was polymerized inside the three-dimensional porous skeleton structure of the hydrogel, thus demonstrating a substantial increase in the specific capacitance of the composite hydrogel with polyaniline nanoparticles (PP-PANI40), which was in line with the excellent specific capacitance of polyaniline reported to be capable of improving the specific capacitance of most electrode materials. In addition, as the content of polyaniline increased from 40% to 60%, there was a great leap in the area of the CV curves, and obvious redox peaks appeared in the rod-shaped polyaniline, which indicated a great improvement in the specific capacitance of PP-PANI60, which was mainly attributed to the large amount of pseudocapacitance due to the large amount of rod-shaped polyaniline. However, the area of the CV curves decreased when the content of polyaniline in the hydrogel was increased from 60% to 80%, which was mainly due to the increase in the polyaniline content and the increase in the polymerization density, which caused the three-dimensional porous skeleton-like structure of the polyaniline nanofibers to form, resulting in the accumulation of part of the polyaniline in the joints forming the three-dimensional structure.

In addition, the excessive polyaniline buildup would also block the lattice channels in the three-dimensional porous skeleton structure of PP hydrogel. These factors would greatly reduce the effective contact area between the electrolyte ions and the active electrode materials, thus decreasing the electrochemical performance of the composites. In addition, from the comparison of the galvanostatic charge–discharge (GCD) curves, it could be observed that at the same constant-current charging and discharging currents (Fig. 2f), the discharge time of the PP-PANI composite hydrogel increased significantly compared with that of the PP hydrogel, which indicated that the specific capacitance of the hydrogel obtained after polyaniline doping increased significantly. When the polyaniline content was increased from 40% to 60% and then to 80%, the discharge time showed a tendency to increase and then decrease, and the GCD curves using hydrogels containing polyaniline showed a slight deviation from the symmetric triangular shape, which indicated that the pseudocapacitance was due to the presence of polyaniline as a way to further support the analytical results of the CV curves mentioned above. As shown in Fig. S6c,† the electrochemical impedance spectroscopy (EIS) plot was divided into a composition of semicircular arcs in the high-frequency region (charge-transfer resistance R_ct at the electrode/electrolyte interface) and sloped lines in the low-frequency region (Warburg impedance of the electrolyte ions diffusing from the electrolyte solution to the electrode interface). The PP hydrogel did not show a clear high-frequency and low-frequency region, and its extremely poor capacitance characteristics could be clearly observed. The PP-PANI composite hydrogel in the low-frequency region showed a high slope state and approximated a semicircular shape in the high-frequency region. Among them, the PP-PANI60 composite hydrogel had more complete and nearly semicircular arcs and nearly vertical slopes in the low-frequency region compared to the PP-PANI60 composite hydrogel. Due to the gap between the appropriate amount of polyaniline and the three-dimensional porous skeleton structure of the substrate hydrogel, the electrolyte ions had a smaller charge transfer resistance in the electrode material, and the porous skeleton structure provided a more convenient channel for the transportation of electrolyte ions. According to the linear relationship between i_p and v^1/2 shown in Fig. S7† and the above calculation formula, the diffusion coefficients of electrolyte ions in PP-PANI hydrogels with different contents are calculated to be 2.3087 × 10⁻¹⁴ cm⁻² s⁻¹, 1.318 × 10⁻¹³ cm² s⁻¹, and 2.0931 × 10⁻¹⁴ cm² s⁻¹, respectively. Therefore, based on the comparison of diffusion coefficients, it was determined that PP-PANI60 had the highest ion transport efficiency. Therefore, based on the results of the above comparative analysis of electrochemical properties, the electrochemical properties of the PP-PANI60 composite hydrogel were more prominent, and its composite material was selected for further investigation. In Fig. S8,† we subjected the PP-PANI60 hydrogel to varying durations of stretching, with the sample designated as PP-PANI60-100 being elongated 100 times under conditions of 100% strain. Subsequently, we evaluated the electrochemical performance of the modified hydrogel. The electrochemical performance data presented in Fig. S6† indicates a notable alteration in the hydrogel's electrochemical characteristics following the stretching treatment. This modification is likely to influence the physical structure of the hydrogel, resulting in deformation. Furthermore, the concentration of polyaniline within a unit volume of the three-dimensional framework diminishes, which in turn correlates with a reduction in the hydrogel's electrochemical performance. Concurrently, we subjected the PP-PANI60 hydrogel to varying durations of stretching, with the sample designated as PP-PANI60-100 being elongated 100 times under conditions of 100% strain. Subsequently, we evaluated the electrochemical performance of the modified hydrogel. The electrochemical performance data presented in Fig. S6† indicates a notable alteration in the hydrogel's electrochemical characteristics following the stretching treatment. This modification is likely to influence the physical structure of the hydrogel, resulting in deformation. Furthermore, the concentration of polyaniline within a unit volume of the three-dimensional framework diminishes, which in turn correlates with a reduction in the hydrogel's electrochemical performance. The CV curves of the PP-PANI composite hydrogel at different scanning rates of 5–100 mV s⁻¹ are shown in Fig. 2g, showing a PANI redox peak type. The voltage plateau in the GCD curve is consistent with the peak redox voltage in the CV curve, and it has a high volume-specific capacitance of 9036.67 mF cm⁻³ at a current density of 3 mA cm⁻³.

Then, the symmetric supercapacitor device was assembled for an electrochemical performance test. As shown in the figure, the device had excellent electrochemical energy storage and a volume-specific capacity of 7848 mF cm⁻³ at a current density of 5 mA cm⁻³ (Fig. 2h). At the same time, it also had a high power-density of 40 W cm⁻³ and a high energy density of 1090 mW h cm⁻³ (Fig. 2j). Moreover, at a current density of 10 mA cm⁻³, GCD curves still had a capacitance retention of 81.3% after 10 000 cycles of testing (Fig. 2i). It could be seen from the above that the hydrogel had good capacitance performance and excellent cycle stability. In addition, the high power-density and energy density could make it have certain practical application feasibility in the field of electrochemical energy storage devices.

2.4 Electromagnetic shielding properties of HES-CHAT e-skin hydrogels

The HES-CHAT electronic skin hydrogel could also be used as an electromagnetic shielding material. In the case where the electrical conductivity of PP hydrogel was nearly zero, its EMI shielding performance was similarly low. The EMI shielding performance of the PP-PANI composite hydrogel could be calculated according to the formula (in ESI†). Additionally, when conductive polyaniline was introduced into the PP three-dimensional network structure, the EMI shielding performance of the hydrogel could be effectively enhanced. The porous structure of the hydrogel increased the propagation path of electromagnetic waves, prolonged the interaction time between electromagnetic waves and materials, and enhanced the absorption loss. The content of polyaniline in the PP hydrogel influenced the electromagnetic shielding performance of the PP-PANI hydrogel, and the EMI shielding performance of the hydrogel in the case of the same thickness (Fig. 3a and g–i). When the PANI content changed, the absorption loss also changed obviously, but the reflection loss did not change obviously, and the total energy efficiency also changed correspondingly. The electromagnetic shielding performance of the hydrogel increased as the polyaniline content increased. However, this situation was consistent with the trend of electrochemical properties; with the increase in the content of polyaniline (40–60%), the EMI shielding performance of the composite hydrogel increased, but when it was increased in excess, the polyaniline would be in the three-dimensional porous skeleton structure of the PP hydrogel, the buildup of the situation occurred, reducing the electrically conductive pathway, hindering the flow of electrons, resulting in the decline of the performance of electromagnetic shielding (Fig. 3f). Therefore, the PP-PANI60 composite hydrogel containing an appropriate amount of polyaniline had a more favorable electromagnetic shielding performance (the thickness of the hydrogel was 1 mm, and it had an electromagnetic shielding effectiveness of 59.7 dB).

Fig. 3 Electromagnetic shielding properties of HES-CHAT hydrogel. (a) Composite hydrogels of the same thickness with different polyaniline contents. (b) Average SE_R, SE_A, and SE_T values of the PP-PANI60 composite hydrogels after being stretched for different times at different strains. (c) Average SE_R, SE_A, and SE_T values of PP-PANI60 composite hydrogels with different thicknesses. (d) Average SE_R, SE_A, and SE_T values of PP-PANI60 composite hydrogels after being preserved for different times. (e) Average SE_R, SE_A, and SE_T values of PP-PANI60 composite hydrogels in different states. (f) The shielding mechanism of HES-CHAT hydrogels. EMI SE curves of composite hydrogels with different polyaniline contents in the X-band, averaged over (g) SE_T values, (h) SE_A values, and (i) SE_R values.

Based on the PP-PANI composite hydrogel having good tensile properties, in this work, the PP-PANI60 composite hydrogel was subjected to tensile movement under different strains during the electromagnetic shielding test (Fig. 3c), and the EMI shielding performances of the 1 mm composite hydrogel after 100 and 500 cycles of tensile stretching at 100% and 500% strains were 55.6 dB and 42.7 dB, respectively, which indicated that this composite hydrogel still had good EMI shielding performance after some regular strain stretching. As shown in Fig. 3b, the EMI shielding performance of the PP-PANI60 composite hydrogel with different thicknesses was enhanced with the increase of thickness, and when the PP-PANI60 composite hydrogel was 4 mm, it had an electromagnetic shielding performance of 70.1 dB. In addition, the EMI shielding performance of this composite hydrogel was measured after it was stored in a constant humidity device for a period of time, and it could be observed that it still had a shielding performance of 47.0 dB after 56 days of storage (Fig. 3d). From the schematic diagram of the shielding mechanism (as shown in Fig. 3f), the synergistic effect of the conductive PANI network structure and the internal water-rich environment was the key to effectively shielding the incident electromagnetic wave and minimizing the reflection. It was an electromagnetic interference shielding mechanism based on absorption. First of all, due to the three-dimensional skeleton structure and good electrical conductivity of the hydrogel, which ensures proper impedance matching, electromagnetic waves penetrate directly into the hydrogel without obvious reflection.⁴⁴ Secondly, the incident wave propagated inside the three-dimensional skeleton of the hydrogel and was repeatedly scattered and reflected by the PANI conductive network structure, which could extend the path length of the wave and enhance its interaction with the interface before propagation.^45,46 Moreover, the combination of the termination of the hydrogel surface and the significant conductivity caused by the PANI conductive network structure could effectively increase the polarization loss and greatly attenuate the incident wave. The penetrating radiation energy would be concentrated in waters that could provide strong polarization loss and dielectric loss, and could induce a vortex current loop to dissipate electromagnetic energy further.

2.5 Mechanical and mechanosensitive properties of HES-CHAT hydrogels

Since mechanical properties are one of the important factors determining the ability of hydrogels to be used in multifunctional applications, the application of flexible strain sensors requires significant mechanical–mechanical strength in order to respond accordingly under different external deformations. Tensile tests were performed on all hydrogel samples in this work, and as shown in Fig. 4a–d, S9a, d, and S0e,† the strain of the pure PVA hydrogel was 310 ± 5.00% at 90 kPa ± 5.00% modulus. With an increase in the PEO content, the tensile strength of the hydrogel also increased; however, the fracture elongation exhibited an opposite trend. It showed that the addition of PEO would increase the elastic recovery of the PP hydrogel, but the corresponding tensile property enhancement showed an increase in the rigidity of the hydrogel, the material's softness gradually decreases, and the plasticity and tensile properties were poor. Therefore, the selection of PP (10 : 1) hydrogel as the substrate material for flexible strain sensors was a very wise choice. In addition, in the tensile test of the PP-PANI hydrogel, the tensile modulus of the hydrogel first showed an increase with the content of polyaniline, and later a decrease with the content of polyaniline. The tensile modulus of the PP-PANI60 hydrogel was 300 kPa, and its deformation became 600%, which was weakened compared with the strain of the PP hydrogel (900%) (tensile modulus 280 ± 10.00 kPa), which was due to the fact that in the hydrogel without polyaniline (PP 10 : 1 hydrogel), due to the molecular cross-linking and the existence of intermolecular hydrogen bonding, the three-dimensional network structure was not as good as that of the hydrogel with polyaniline. The three-dimensional network structure of the hydrogel was well stabilized due to molecular cross-linking and intermolecular hydrogen bonding. When polyaniline was added to the three-dimensional network structure of the hydrogel, the intermolecular force in the material system was enhanced, the strain of the hydrogel was reduced, and the change of tensile modulus increased and then decreased, which was due to the accumulation of polyaniline in the three-dimensional network structure of the hydrogel after too much polyaniline was added into the hydrogel, and the formation of nanofibers cross-linked with each other, which resulted in the material being more rigid and resistant to stretching. The result was a more rigid and tensile-resistant material. Therefore, the PP-PANI60 hydrogel was selected as the flexible strain sensor for subsequent performance investigation of its potential application.

Fig. 4 HES-CHAT for real-time monitoring of gestures. (a) Stress–strain curves of hydrogels. (b) Relative resistance changes (ΔR/R₀) of PP-PANI60 hydrogels under different strain conditions. (c and d) Relative resistance changes (ΔR/R₀) of PP-PANI hydrogels under different strains. (e) Corresponding deformation of hydrogels with increasing weights. (f) PP-PANI60 hydrogel's relative resistance changes under 20 min tensile cycling at 300% and 500% strain, respectively. (g) Schematic of HES-CHAT for real-time monitoring of gestures. (h) Signals of various gestures.

The swelling rate and water retention value of hydrogels were fundamental properties, which were generally related to the crosslink density and entanglement network.^47–49 Therefore, hydrogel swelling was tested for different ratios of PVA to PEO (Fig. S9b†). It could be directly observed that the swelling ratios of hydrogels with different ratios increased significantly with the increase in water absorption time. When the ratio of PVA/PEO was 10 : 1, the equilibrium swelling ratio of the PP hydrogel increased from about 15% to more than 60% in the pure PVA hydrogel, and then with the continuous increase of PEO, the swelling ratio of the hydrogel decreased instead, which indicated that the increase of the PEO content resulted in the three-dimensional porous skeleton structure of the PP hydrogel becoming gradually loose, and the interstices of the porous network were enlarged, which was able to provide space for the doping of conductive polyaniline, and provide space for the electrolyte solution to be doped. The excellent structure provided space for the doping conductive polyaniline, and a wider ion transport channel for electrolyte ions. In addition, the hydrogels were placed in a moisturizing device, and with the extension of time, the hydrogels gradually lost water, and after a period of time, the PP-PANI hydrogels all reached a water retention rate of more than 60% (Fig. S9c†), which had good water retention properties. As a result, it was shown that the PP (10 : 1) hydrogels had a high crosslink density and good water absorption/water retention properties, and the PP hydrogels at this ratio were used in this work to carry out subsequent investigations.

In addition, the dynamic storage modulus G′ and loss modulus G′′ of hydrogels with different polyaniline contents in the frequency range of 0.1–100 rad s⁻¹ are shown in Fig. S10a.† The present form of polyaniline in PP hydrogels was related to the polymerization time, and there was no stable cross-linking behavior with the hydrogel's three-dimensional network. The values of G′ and G′′ decreased gradually, with the increase of polyaniline content in the frequency range. In the frequency range of all hydrogels, G′ was much higher than G′′, which was consistent with the solid-phase elastic nature of hydrogels. In the strain scanning measurements (Fig. S10b†), a significant strain-dependent viscoelastic response was found for all hydrogels. The G′ values of PP-PANI composite hydrogels were always higher than G′′ in the 1–100% strain linear viscoelastic range. Meanwhile, at 100% –10 000%, the hydrogels exhibited irregular linear viscoelastic behavior. The G′ was interlaced with G′, which was due to the collapse of the internal network of the hydrogels. This fully indicated that the addition of polyaniline would cause large-scale deformation of the network structure and de-entanglement of the cross-linking points of the hydrogel induced by shear strain, and its mechanical properties were limited to a certain extent. This further verified the results analyzed in the tensile test.

PP-PANI hydrogels were acted on using weights of different masses. The hydrogel underwent a certain degree of deformation with the increasing mass of the weights, and when the weights were increased to 100 g, the deformation of the hydrogel increased significantly (Fig. 4e and f), which had a certain degree of sensitivity. From this, it could be conjectured that this hydrogel had the potential to be a strain sensor. Sensitivity is a key factor in evaluating the performance of a sensor. The sensitivity (GF) of CPs could be determined after calculating the slope of the relative resistance change ΔR/R₀ (ΔR = R − R₀) against strain curve. The ΔR/R₀ value increased with the increase of the strain (Fig. 4b), and the GF of the hydrogel varied linearly, which was mainly divided into two response zones, namely, 0–200% and 200–500%, and their corresponding GF values are 0.40 and 0.95, respectively, demonstrating the high strain sensitivity and tensile properties of the hydrogel. As shown in Fig. 4c and d, no significant signal fluctuation of ΔR/R₀ occurred under different strain conditions (20–400%), indicating that the signal stability of this hydrogel was good. In addition, the PP-PANI60 composite hydrogel as a sensor showed good continuity and reliability by maintaining stable amplitude motion and waveform transformations when continuously running tensile cycles at 300% strain and 500% strain for 20 min, respectively (Fig. 4f).

The HES-CHAT hydrogel possessed excellent mechanical properties, high electrical conductivity, and excellent sensing properties that enabled it to be used as a strain sensor for detecting a variety of body movements (Fig. 4g). The composite hydrogel was remarkably soft and flexible and could be easily attached to different joints of the body. When this composite hydrogel sensor was attached to the human throat (Fig. 4h), the subtle movements based on the compression effect generated by the vibration of the vocal cords, and the precision of the swallowing action during drinking could be accurately indicated, and this sensor could be applied to a pharyngeal motion recognition device. This hydrogel sensor could accurately express the vibration behavior of the larynx, and swallowing and vocalization can be accurately recognized. The hydrogel was attached to the index finger and wrist, and the signals changed significantly when the fingers were bent at different angles and the wrists were twisted in different directions, and their respective relative resistance (ΔR/R₀, ΔR: change in resistance under strain stimulation; R₀: original resistance)-time curves showed remarkable stability and reproducibility. This hydrogel strain sensor not only accurately recognized and monitored the subtle movements of humans, but also responded to the movements of the major joints of the body (shoulder, elbow, and knee joints) (as shown in Fig. 4h). This hydrogel was attached to the various joints of the body in order to detect human body movements, and it could be clearly observed that the relative resistive signals were even and the changes were obvious, which demonstrated that the reliability of the PP-PANI composite hydrogel strain sensors for detecting human body movements was good, and it had great potential as a flexible and wearable electronic device for the monitoring of the real-time human body health.

2.6 HES-CHAT e-skin for gesture signals interpretation

The HES-CHAT e-skin interface pipeline enabled seamless integration of electronic devices with human skin for various applications such as health monitoring, wearable electronics, and human-machine interfaces. The process could be further elaborated in the following steps. For starters, the e-skin device, made of a highly-mechanosensitive hydrogel electrode, was attached to the human skin. This device was designed to be highly sensitive to even the minutest changes in the skin's deformation. The recorded signals had high dimensionality, making it challenging to process and analyze them effectively. To address this, the pipeline employed the Principal Component Analysis (PCA) algorithm. PCA transformed the high-dimensional data into a lower-dimensional representation, preserving the essential features and patterns in the data while reducing noise and redundancy. With the dimensionality of the signals reduced, the pipeline used the state-of-the-art Generative Pre-trained Transformer (GPT) model for further analysis. This deep learning model was capable of understanding complex patterns in the data and making accurate predictions. By fine-tuning the GPT model using the two-dimensional signal matrix obtained from the PCA, the pipeline could generate valuable insights and predictions about the user's health or other relevant factors. The resulting two-dimensional signal matrix was then utilized for fine-tuning and prediction using the GPT model (Fig. 5a). The dimension reduction analysis of the raw signals acquired by HES-CHAT revealed three distinct functional domains (Fig. 5b). Function domain 1 predominantly captured the signals associated with swallowing and speaking. Function domain 2 exhibits a concentration of Finger-90 (degree), Finger-45 (degree), and Shoulder signals. On the other hand, function domain 3 is enriched with Elbow-LR and Elbow-UD signals. Notably, the similarity heatmap reflects a similar trend, where the speaking and swallowing signals exhibited high inter-group similarity (Fig. 5c). These findings highlighted the remarkable capabilities of HES-CHAT in terms of high mechano-sensitivity, enabling the detection of subtle differences in various gestures.

Fig. 5 Chat window-based HES-CHAT and GPT interface for gesture signal recognition. (a) Pipeline of HES-CHAT for signal acquisition (A/D conversion), dimension reduction, and GPT fine-tuning. (b) t-SNE (t-distributed Stochastic Neighbor Embedding) plot of the first two PCA components of various gesture signals. (c) Similarity heatmap. (d) Prompt based fine-tuned interface. (e) Confusion matrix of the GPT recognition accuracy.

To train a personalized GPT robot capable of comprehending the inherent signal differences, a prompt-based fine-tuning process was employed using the 2-dimensional feature matrix (Fig. 5d). Following the training phase, the GPT model exhibited exceptional accuracy (Fig. 5e), achieving 100% accuracy in most gesture signals, including speaking, swallowing, Finger-45, Elbow-LR, Elbow-UD, knee, and shoulder. However, it was important to note that the HES-CHAT system encounters challenges when it came to wrist signal recognition. Approximately 80% of the wrist signals were incorrectly predicted as Elbow-LR signals. This misclassification could be attributed to the similarity in motion trajectory and features shared by these two gestures. This particular limitation highlighted the need for further optimization and refinement in distinguishing between wrist and Elbow-LR signals. By exploring innovative approaches such as incorporating additional features or refining the classification algorithm, it was possible to enhance the accuracy and overcome the current difficulties faced by HES-CHAT in recognizing wrist signals. In summary, the integration of the aforementioned components in the HES-CHAT e-skin circuit demonstrates its promising potential for advanced applications. The utilization of a hydrogel electrode with exceptional mechano-sensitivity allowed for efficient A/D conversion. The subsequent PCA-based dimension reduction technique enhances signal processing efficiency. Finally, fine-tuning and prediction using the GPT model further improved the system's overall performance. These results collectively emphasized the ability and significance of the HES-CHAT e-skin circuit in detecting and interpreting a wide range of gestures.

3 Discussion and conclusion

In this work, we developed the HES-CHAT e-skin as a multifunctional personal health assistant. When the conductive hydrogel of the e-skin was configured into a symmetric supercapacitor device, it exhibited a specific volume capacity of 7848 mF cm⁻³ at a current density of 5 mA cm⁻³, a power density of 40 W cm⁻³, and an energy density of 1090 mW h cm⁻³. Additionally, this electronic skin demonstrated commendable energy efficiency in the EMI shielding field, achieving a high shielding performance of 59.7 dB at a thickness of 1 mm, and a shielding performance of 42.7 dB was still achieved after 500 cycles of stretching at 500% strain. Notably, our integrated e-skin system exhibited exceptional performance in both sensing capabilities and interactive functionalities. The mechanosensitive sensors exhibited high sensitivity, with a wide detection range, enabling accurate monitoring of physiological parameters and physical activities. The integration of ChatGPT allowed for personalized health management, including answering queries, providing health advice, and generating informative reports based on the collected data. Moreover, the electromagnetic shielding feature ensured the reliability and safety of the e-skin system, making it suitable for everyday use. The HES-CHAT e-skin, which we developed for gesture signal interpretation, showcased remarkable capabilities in terms of high mechano-sensitivity. It successfully detected subtle differences in various gestures, achieving high accuracy in most gesture signals. However, challenges were encountered in distinguishing between wrist and Elbow-LR signals, where misclassifications occurred due to similarities in motion trajectory and features. Future optimization and refinement, such as the incorporation of additional features or enhancements to the classification algorithm, might enhance accuracy and address these challenges.

In summary, the HES-CHAT e-skin design, which integrated mechanosensitive sensors with ChatGPT, provided a sophisticated personal health assistant characterized by advanced functionalities. The incorporation of EMI shielding properties enhanced users' safety and expanded the potential applications of the e-skin. The demonstrated performance in gesture signal interpretation highlights the promising potential of the HES-CHAT e-skin. Collectively, this research advanced the development of flexible electronic skins that incorporated both intelligence and functionality, thereby facilitating progress in healthcare monitoring and related domains.

Ethical statement

The primary purpose of this manuscript involving human participants is to contribute to the development of flexible electronic skins with integrated intelligence and functionality, paving the way for advancements in healthcare monitoring and other related fields. These purposes can never take precedence over the rights and interests of individual research participants. Additionally, we guarantee that informed consent was obtained for the experiments involving human participants.

Data availability

The data are available from the corresponding author on reasonable request.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (22078184, 22378249); Key R&D Plan of Shaanxi Province (2024GX-YBXM-335); China Postdoctoral Science Foundation (2019M653853XB); and Natural Science Advance Research Foundation of Shaanxi University of Science and Technology (2018QNBJ-03).

References

1. K. Kim, J.-H. Hong, K. Bae, K. Lee, D. J. Lee, J. Park, H. Zhang, M. Sang, J. E. Ju, Y. U. Cho, K. Kang, W. Park, S. Jung, J. W. Lee, B. Xu, J. Kim and K. J. Yu, Sci. Adv., 2024, 10, eadr1099.
2. X. Guo, Z. Sun, Y. Zhu and C. Lee, J. Phys.:Condens. Matter, 2024, 36, 2406778.
3. Y. Xia, X. Zhi, M. Guo, Y. Zhang, S. Ma and X. Wang, Adv. Fun. Mater., 2024, 35, 2414936.
4. Y. Song, W. Ren, Y. Zhang, Q. Liu, Z. Peng, X. Wu and Z. Wang, Small, 2023, 19, 2303301.
5. J. Qi, T. Luo, J. Yu, C. Wang, F. Chu and J. Wang, Chem. Eng. J., 2024, 499, 156561.
6. J. N. Niu, X. Bai, J. W. Wang, Y. Y. Chen, B. Q. Zhao, W. Sha, Y. Long, Z. L. Wang and W. G. Hu, Adv. Fun. Mater., 2024, 33, 2209114.
7. D. S. Liu, Y. F. Cao, P. Jiang, Y. X. Wang, Y. Z. Lu, Z. Y. Ji, X. L. Wang and W. M. Liu, Small, 2023, 19, 2206819.
8. M. Paswan, S. Patel, V. Prajapati and B. Z. Dholakiya, Int. J. Biol. Macromol., 2023, 253, 126979.
9. S. P. Quah, A. J. Smith, A. N. Preston, S. T. Laughlin and S. R. Bhatia, Polymer, 2018, 135, 171–177.
10. X. X. Le, W. Lu, J. W. Zhang and T. Chen, Adv. Sci., 2019, 6, 1801584.
11. J. Y. Chong, C. H. Sung, K. S. Nam, T. W. Kang, H. Kim, H. S. Lee, H. Park, S. Park and J. H. Kang, Nat. Commun., 2023, 14, 2206.
12. F. F. Fu, J. L. Wang, H. B. Zeng and J. Yu, ACS Mater. Lett., 2020, 2, 1287–1301.
13. R. Elashnikov, O. Khrystonko, T. Jilková, S. Rimpelová, J. Prchal, I. Khalakhan, Z. Kolská, V. Švorčík and O. Lyutakov, Adv. Fun. Mater., 2024, 34, 2314420.
14. S. X. Zhou, Y. J. Zhao, K. X. Zhang, Y. R. Xun, X. Y. Tao, W. T. Yan, W. Zhai and J. Ding, Nat. Commun., 2024, 15, 337.
15. M. Xu, J. Zhu, J. Xie, Y. Mao and W. Hu, Small, 2024, 20, 2305448.
16. X. H. Yu, Y. Zheng, H. P. Zhang, X. S. Fan and T. X. Liu, Chem. Mater., 2021, 33(15), 6146–6157.
17. Y. Yang, N. Wu, B. Li, W. Liu, F. Pan, Z. Zeng and J. Liu, ACS Nano, 2022, 16, 15042–15052.
18. C. Y. Xiong, C. M. Zheng, Z. Zhang, Q. Xiong, Q. S. Zhou, D. P. Li, M. X. Shen and Y. H. Ni, J. Materiomics, 2025, 11(1), 100841.
19. L. Zhang, B.-W. Liu, Y.-Z. Wang, T. Fu and H.-B. Zhao, Composites, Part B, 2022, 238, 109944.
20. J. Liu, L. McKeon, J. Garcia, S. Pinilla, S. Barwich, M. Möbius, P. Stamenov, J. N. Coleman and V. Nicolosi, Adv. Mater., 2022, 34, 2106253.
21. M. Hao, Y. Wang, L. Li, Q. Lu, F. Sun, L. Li, X. Yang, Y. Li, M. Liu, S. Feng, S. Feng and T. Zhang, Soft Matter, 2021, 17, 9057–9065.
22. Y. Z. Zhang, K. H. Lee, D. H. Anjum, R. Sougrat, Q. Jiang, H. Kim and H. N. Alshareef, Sci. Adv., 2018, 4, eaat0098.
23. Q. F. Guan, Z.-M. Han, K.-P. Yang, H.-B. Yang, Z.-C. Ling, C.-H. Yin and S.-H. Yu, Nano Lett., 2021, 21, 2532–2537.
24. Y. Zou, Z. Liao, R. Zhang, S. Song, Y. Yang, D. Xie, X. Liu, L. Wei, Y. Liu and Y. Song, Carbohydr. Polym., 2025, 348, 122788.
25. Y. Yang, B. Li, N. Wu, W. Liu, S. Zhao, C. J. Zhang, J. Liu and Z. Zeng, ACS Mater. Lett., 2022, 4, 2352–2361.
26. S. Ganguly, P. Bhawal, R. Ravindren and N. C. Das, J. Nanosci. Nanotechnol., 2018, 18, 7641–7669.
27. P. Bhawal, S. Ganguly, T. K. Das, S. Mondal, S. Choudhury and N. C. Das, Composites, Part B, 2018, 134, 46–60.
28. A. A. Isari, A. Ghaffarkhah, S. A. Hashemi, S. Wuttke and M. Arjmand, Adv. Mater., 2024, 36, 2310683.
29. H. Zhang, J. B. Wan, R. Q. Wu, Y. Y. Chen, H. Z. Yu and S. W. Shi, Carbon, 2024, 218, 118716.
30. B. Zhao, Z. Y. Bai, H. L. Lv, Z. K. Yan and Y. Q. Du, Nano-Micro Lett., 2023, 15, 79.
31. Y. Y. Zhu, J. Liu, T. Guo, J. J. Wang, X. Z. Tang and V. Nicolosi, ACS Nano, 2021, 15, 1465–1474.
32. F. Zhang, W. Cui, B. B. Wang, B. H. Xu, X. H. Liu, X. H. Liu, Z. R. Jia and G. L. Wu, Composites, Part B, 2021, 204, 108491.
33. J. C. Dong, X. W. Tang, Y. D. Peng, C. H. Fan, L. Li, C. Zhang, F. L. Lai, G. J. He, P. M. Ma, Z. C. Wang, Q. F. Wei, X. P. Yan, H. L. Qian, Y. P. Huang and T. X. Liu, Nano Energy, 2023, 108, 108194.
34. Z. Lou, S. Chen, L. Wang, R. Shi, L. Li, K. Jiang, D. Chen and G. Shen, Nano Energy, 2017, 38, 28–35.
35. F. Yoshii, Y. Zhanshan, K. Isobe, K. Shinozaki and K. Makuuchi, Radiat. Phys. Chem., 1999, 55, 133–138.
36. J. Zhang, Y. Li, H. Wu, C. Wang, K. M. Salleh, H. Li and S. Zakaria, Polymers, 2022, 14, 4473.
37. Z. Guo, J. Wang, P. Yu, M. Li, L. Huang, Z. Hu, Y. Wang and Z. Song, Adv. Energy Mater., 2023, 13, 2301520.
38. G.-C. Xu, Y. Nie, H.-N. Li, W.-L. Li, W.-T. Lin, Y.-R. Xue, K. Li, Y. Fang, H.-Q. Liang, H.-C. Yang, H. Zhan, C. Zhang, C. Lü and Z.-K. Xu, Adv. Mater., 2024, 36, 2400075.
39. M. F. El-Kady, V. Strong, S. Dubin and R. B. Kaner, Science, 2012, 335, 1326–1330.
40. H. Liu and B. Q. Zuo, Appl. Sci., 2018, 8, 296.
41. Y. Sun, Z. Fang, C. Wang, A. Zhou and H. Duan, Nanotechnology, 2015, 26, 374002.
42. C. Y. Xiong, T. H. Li, Y. C. Zhu, T. K. Zhao, A. Dang, H. Li, X. L. Ji, Y. D. Shang and M. Khan, J. Alloys Compd., 2017, 695, 1248–1259.
43. H. Xiao, Z. S. Wu, L. Chen, F. Zhou, S. H. Zheng, W. C. Ren, H. M. Cheng and X. H. Bao, ACS Nano, 2017, 11, 7284–7292.
44. X. Yao, S. F. Zhang, L. W. Qian, N. Wei, V. Nica, S. Coseri and F. Han, Adv. Fun. Mater., 2022, 32, 2204565.
45. C. Y. Xiong, T. H. Li, A. L. Dang, T. K. Zhao, H. Li and H. Q. Lv, J. Power Sources, 2016, 306, 602–610.
46. P. P. Zhang, Y. Li, G. Wang, F. X. Wang, S. Yang, F. Zhu, X. D. Zhuang, O. G. Schmidt and X. L. Feng, Adv. Mater., 2019, 31, 1806005.
47. R. Zhu, D. Zhu, Z. Zheng and X. Wang, Nat. Commun., 2024, 15, 1344.
48. R. Huang, J. Hua, M. Ru, M. Yu, L. Wang, Y. Huang, S. Yan, Q. Zhang and W. Xu, ACS Nano, 2024, 18, 15312–15325.
49. P. Liu, Y. Zhang, Y. Guan and Y. Zhang, Adv. Mater., 2023, 35, 2210021.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5ta00500k

‡ These authors contributed equally to this work.

---