Advanced Hub Main Navigation Menu
Transmembrane Inspired Mechano-Responsive Elastomers with Synergized Traction-Assisted Healing and Dual-Channel Sensing
Abstract
In the burgeoning field of bioinspired materials, the principles governing biological perception and self-healing drive advancements in biomimetic mechano-responsive materials, seamlessly integrating ionic signal sensing with self-healing. While current research often emphasizes individual functionalities, the concurrent enhancement of both self-healing and sensitivity in iontronic skins is often overlooked. Drawing inspiration from transmembrane proteins like TSP-15, Piezo 1 and Piezo 2, renowned for recruiting repair factors, multifunctional molecular-ionic regulatory sites are constructed within a polyurethane/ionic liquid composite system, leading to the development of a mechano-responsive elastomer (i-DAPU) that exhibited both rapid self-healing (72 µm min−1) and impressive sensitivity (7012.05 kPa−1). Leveraging the dual functionalities of i-DAPU in tandem with deep learning algorithms, a sophisticated system is devised for intelligently analyzing neural conditions in comatose patients based on muscle strength, achieving a remarkable 99.2% accuracy rate, holding significant promise for healthcare applications.
1 Introduction
In the dynamic landscape of molecular biology and genomics, rapid advancements are continuously revealing the microscale principles that govern biological perception and signal transduction.[1] This expanding knowledge base is providing critical theoretical foundations and pioneering design approaches for creating a new generation of biomimetic sensors, distinguished by their novel principles and functionalities.[2] Among these innovations, iontronic sensors, which used mechano-responsive elastomer as their dielectric material, constitute a forefront category in biomimetic tactile sensors.[3] These sensors not only replicated the soft touch and self-healing properties akin to human skin but also imitated the function of receptor cells. They modified the intracellular and extracellular potential difference in response to mechanical stress via ion transport, enabling effective pressure sensing. By harnessing these unique capabilities, biomimetic flexible sensors are leading a global technological shift in intelligent tactile perception and represent a significant advancement in the field.
To meet specific challenges in technological advancements, researchers are exploring the diverse functional mechanisms of biological systems, with the goal of more effectively integrating these insights into materials for enhanced biomimicry.[4] For instance, inspired by the hydrogen bond arrays in spider silk, Sun et al. have incorporated numerous hydrogen bonds into polyurethane segments, creating a flexible sensor that rapidly self-heals at room temperature and in aqueous solutions.[5] In a similar vein, Fu et al. devised a novel molecular strategy inspired by the mesh-like structure of human skin.[6] Their approach transformed the weak island structures in transparent, self-healing polyurea into robust dual continuous nano-phase separated structures, resulting in an optically transparent strain sensor with rapid self-healing capability and suitable for motion signal detection. However, further research efforts were focused on enhancing sensor sensitivity by mimicking the structural characteristics and sensing principles of human skin.[7] Inspired by the papillary ridges of the human epidermis, Ren et al. have fabricated a pressure sensor with biomimetic notochord microstructures, exhibiting high sensitivity and a broad linear range.[8] In the domain of cellular mimicry, our group and coworkers have emulated the sensing principle of Piezo2 proteins on Merkel cell membranes, leading to the fabrication of various highly sensitive iontronic skins.[3, 9] The latest study has leveraged synergistic effects from different biomimetic mechanisms, known as the coupled enhancement effect, to substantially improve device performance.[10] However, these efforts have predominantly concentrated on boosting sensing capabilities, with less emphasis on simultaneously improvement of self-healing and sensitivity in iontronic skin. Addressing this gap will offer an exciting avenue for novel breakthroughs by investigating new mechanisms of biological functionality.
Recently, research teams led by Xu et al. and Yu et al. have uncovered a pivotal role for the transmembrane protein tetraspanin in cellular self-repair.[11] Tspan 4, a member of this protein family, congregated at sites of cellular membrane disruption, forming micrometer-scale structures. These assemblies acted to limit the further damage expansion, thereby aiding rapid membrane repair.[11] Another member of the tetraspanin family, the transmembrane protein TSP-15, played a crucial role in promoting cellular membrane self-repair.[11, 12] In cases of significant cellular damage, TSP-15 recruited repair factors to temporarily fill the damage and formed structural domains to prevent wound expansion. This process continued until the self-assembly growth of the phospholipid bilayer was complete, after which TSP-15 was internalized and shed in vesicular form, restoring the cell membrane to its initial state. Furthermore, Zeng et al. have identified a critical role of the projecting neurons in the petrous and cervical ganglia in maintaining blood pressure stability.[13] This crucial role was attributed to the simultaneous expression of the channel proteins Piezo 1 and Piezo 2 on their terminals, enabling sensitive detection of minute blood pressure fluctuations. This discovery significantly enriched our understanding of sensory mechanisms.
Drawing inspiration from the intricate mechanisms of cellular repair and sensing, we have engineered a multifunctional molecular-ionic regulatory site within a polyurethane/ionic liquid (PU/IL) composite system. This innovation combined traction-assisted self-healing and dual-channel synchronous sensing to develop a high-performance iontronic skin characterized by rapid self-healing and impressive sensitivity. Herein, we have integrated donor(D)–acceptor(A) self-assembly groups into the main chain of polyurethane and coblended them with the ionic liquid [BMIM]+[PF6]−, forming a novel mechano-responsive elastomer (i-DAPU). The molecular-ionic interactions networks between the polyurethane (DAPU) molecules and [BMIM]+[PF6]−, laid the groundwork for traction-assisted self-healing and dual-channel synchronous sensing. Upon damage, disrupted D–A groups re-linked due to mutual attraction, while the electrostatic interactions between cations and anions facilitated the self-assembly of D–A groups. Under pressure, the D and A groups released anions and cations, mimicking membrane protein actions. Utilizing i-DAPU as the dielectric material, the fabricated iontronic sensor (DA-skin) achieved rapid self-healing and impressive sensitivity. Leveraging this dual-enhanced performance of DA-skin, we have developed an intelligent analysis system for assessing the neural condition of comatose patients through muscle strength analysis. This work not only introduced new design concepts and research strategies for high-performance mechano-responsive elastomers and biomimetic tactile sensors but also showed significant potential in the medical diagnostics, enhancing precision in disease diagnosis.
2 Results and Discussion
2.1 Physiological Mimicry of Cell Membranes
In this study, we leveraged the molecular-ionic interaction network, formed through the interplay between donor–acceptor (D-A) self-assembly and ionic liquid (IL) interactions within the DAPU/IL composite system, to mimic cellular self-healing mechanisms and mechanosensing. This approach underpinned the development of a new mechano-responsive elastomer, i-DAPU, capable of traction-assisted self-healing and dual-channel synchronous sensing, as depicted in Figure 1a. The chemical structure of DAPU, along with its nuclear magnetic resonance (NMR) and Infrared (IR) spectra, were illustrated in Figure S1 (Supporting Information). Donor groups typically comprised conjugated macrocyclic structures adorned with electron-donating substituents such as methyl, ether, and amino groups, rendering the groups electron-rich. In this work, these were referred to as amino-substituted naphthalene groups. Conversely, acceptor groups often carried electron-withdrawing substituents, such as nitro, halogen groups and imide groups, which further decreased the electron density within the conjugated π-bonds, resulting in an electron-deficient state; herein, they were referred to as imide-substituted benzene groups. Due to the electron-rich nature of donor groups and the electron-deficient state of acceptor groups, their interaction led to mutual attraction, neutralizing the electron cloud density and facilitating self-assembly. In the DAPU, this self-assembly occurred both intermolecularly and intramolecularly within molecular chains. Donor and acceptor groups on the same DAPU molecular chain attracted each other during the entanglement process, forming intramolecular donor–acceptor (D–A) self-assemblies. Similarly, donor and acceptor groups on different DAPU molecular chains also attracted each other when in proximity, resulting in intermolecular D–A self-assemblies. Through these multiple D–A self-assembling interactions, DAPU exhibited outstanding mechanical properties and self-healing capabilities.
Inspired by the cellular membrane transmembrane proteins Tspan 4 and TSP-15, which facilitated post-injury restriction and rapid self-repair, our design emulated these biological mechanisms through a molecular-ionic interaction network in i-DAPU. This network enabled the self-healing capability, where the D–A groups in the polymer chain attracted [BMIM]+ and [PF6]− ions to facilitate healing. Upon damage, these groups were drawn together by the long-range electrostatic interactions between the ions, enhancing the self-healing process. Additionally, under mechanical pressure, i-DAPU mirrored the action of Piezo 1 and Piezo 2 channels in cellular membranes, with the D and A groups releasing [BMIM]+ and [PF6]− ions. This mechanism facilitated dual-channel synchronous sensing, significantly improving the sensitivity.
Previous studies have reported the self-healing characteristics of donor–acceptor self-assembling groups.[14] Here, we employed density functional theory (DFT) calculations to demonstrate the formation of the molecular-ionic interaction networks between D group, A group, [BMIM]+ and [PF6]−. As illustrated in Figure 1b, the binding energy of [BMIM]+ and [PF6]− decreased from the initial value of −83.94 to −80.18 kcal mol−1 and −80.34 kcal mol−1, respectively. Notably, introducing both D and A groups concurrently substantially weakened the binding energy, reducing it to −69.43 kcal mol−1. This reduction indicated that D and A groups could adsorb cations and anions, forming the molecular-ionic interaction networks and consequently weakening the binding capacity of ions. Further analysis of the adsorption force strength was conducted using Multiwfn 3.8 (dev). The results presented in Figure 1c,d were visualized with VMD 1.9.3, which colored the reduced density gradient (RDG) isosurfaces. Here, blue and green regions indicated attractive interactions (with blue signifying strong adsorption and green indicating weaker adsorption), while red regions denoted repulsive interactions. It was evident that whether D or A group predominantly exhibit weak van der Waals forces of attraction towards the cation and anion. Enhanced adsorption was observed when both groups were introduced, as marked by denser regions indicated with red arrows in the figures (sign(λ2) ρ ≈ 0). However, the system consistently exhibited a lack of strong adsorption. This balanced state of adsorptive force allowed the D–A groups to maintain a traction force for assisted self-healing, while also enabling the release of ions under pressure for dual-channel synchronous sensing.
2.2 Mechanical Properties
Figure 1 in our study theoretically demonstrated the adsorption of the IL by D-A groups, a crucial aspect for our sensor research. By comparing the mechanical properties of DAPU with varying contents of D and A groups (stress-strain curves in Figure S2 (Supporting Information), toughness in Figure S3 (Supporting Information), and a summary of mechanical properties in Table S1 (Supporting Information), we selected DAPU-2, which exhibited outstanding toughness, as the base material. This base material was co-blended with the ionic liquid [BMIM]+[PF6]−, referred to as i-DAPU-xx (xx denoting the IL content), and its stress-strain curve was shown in Figure 2a. The results indicated that as the IL content increased, the strength and modulus of i-DAPU gradually decreased, while the elongation increased due to the plasticizing effect of the IL in the polyurethane chains. The increased spacing between chains allowed easier movement of polymer molecules, resulting in the observed reduction in strength and increase in elongation. However, at IL contents of 40% or higher, the elongation rate of i-DAPU transitioned from increasing to decreasing. To investigate this phenomenon, we examined i-DAPU films with varying IL contents under a microscope, as shown in Figure S4 (Supporting Information). We observed that the films remained uniform and free of apparent defects when the IL content was increased from 0% to 30%. Yet, at 40% IL content, noticeable pores began to appear on the films, becoming more pronounced with further increases in IL content. This indicated that while DAPU and IL are highly compatible at IL contents below 40%, an excess of IL led to self-aggregation beyond this threshold. Consequently, we identified 30% as the optimal IL concentration for our purposes, and henceforth, i-DAPU-30 would be simply referred to as i-DAPU.
Then we tested its mechanical recovery capability, as shown in Figure 2b. Initially, the i-DAPU material was stretched to 500%, and it returned to the original point, then stretched to 1000%, recovered again, and finally extended to 1500%, reaching the breaking point. These tests clearly demonstrated the exceptional fatigue resistance of i-DAPU. Notably, multiple stretching tests did not significantly alter the mechanical performance of i-DAPU, indicating the stability of the molecular chains. Furthermore, the mechanical performance of i-DAPU remained largely unchanged even after self-healing, likely due to the combined effects of D–A self-assembly and the molecular-ionic interaction network between D--A self-assembly and IL. Under the influence of these factors, i-DAPU also exhibited excellent tear resistance ability (Figure 2c). The notches on the stretched samples were blunt, with no swelling observed during elongation, indicating superior tear resistance ability. This prevented further enlargement of small damages, while the undamaged areas retained their properties. This characteristic was crucial for flexible sensors using this material as the dielectric, as it ensured the sensor maintains stable operation despite minor damages.
2.3 Self-Healing Capability
The tear resistance of our i-DAPU material effectively prevented the further expansion of notches, while its excellent self-healing capability enabled the spontaneous healing of these notches without external force, thus not affecting subsequent use. Owing to the presence of the molecular-ionic interaction networks, i-DAPU exhibited exceptional self-healing properties. Figure 2d displayed the progressive disappearance of a notch in i-DAPU at 25 °C, as captured in 2D optical microscopic images and 3D surface mapping microscopic images, with the notch depth illustrated in Figure S5 (Supporting Information). Initially, a clear notch was present on the i-DAPU film. Over time, driven by the dual forces of the D–A self-assembly interactions and the electrostatic interactions between the ionic liquids, the notch gradually diminished until it completely disappeared. The 3D surface mapping microscopic images and the scratch depth chart detailed this process, where the notch depth decreased from an initial 240 to nearly 0 µm. Furthermore, the mechanical performance of i-DAPU remained almost unaffected after self-healing, as shown in Figure S6 (Supporting Information).
Figure S7 (Supporting Information) compared the self-healing speeds of mechano-responsive elastomers, including DAPU, with varying levels of DA contents. The self-healing speed was calculated using the formula: healing speed = notch depth/healing time. It was evident that the self-healing performance of DAPU increased with an increase in the soft segment content, as a higher soft segment results in a reduced number of rigid and non-planar rings, enhancing molecular chain mobility despite a reduced number of D–A self-assembly sites, thereby increasing the healing speed. As the temperature rise, particularly above 50 °C, the self-healing speed increased significantly due to the enhanced movement of molecular segments. The dynamic mechanical analysis (DMA) of DAPU was shown in Figure S8 (Supporting Information). From the spectra, a faint glass transition peak (Tg) was observed at approximately 50 °C, indicative of the hard phase, while a pronounced Tg peak was evident at around −20 °C, representing the soft phase. Consequently, as the temperature rose to 60 °C or higher, the hard phase of DAPU was disrupted, no longer constraining the movement of molecular chain segments.
Figure 2e and Figure S9 (Supporting Information) compared the self-healing speeds of i-DAPU and DAPU without ionic liquid at different temperatures. The green font represented a fold increase in self-healing speed after the IL addition. It was evident that the IL inclusion significantly accelerated the overall self-healing speed of the material. Particularly at room temperature, DAPU, which previously could not self-heal, gained breakthrough self-healing capability upon the IL addition. This was attributed to the traction-assisted self-healing action in i-DAPU. Similar to the role of Tspan 4 and TSP-15 proteins in promoting cellular membrane self-repair, the D–A self-assembly in i-DAPU provided the fundamental self-healing action. Compared to DAPU without IL and lacking a traction-assisted self-healing mechanism, the electrostatic interactions between [BMIM]+ and [PF6]− ions adsorbed by the D–A self-assembly in i-DAPU could drive the D and A groups closer after the disruption of the D–A self-assembly, thereby accelerating the reconnection of the polyurethane molecular chains and even surpassing the original self-healing temperature limit. Moreover, this traction-assisted self-healing action strengthened as the IL content increased. Therefore, with the IL addition, i-DAPU not only significantly improved its self-healing speed compared to DAPU but also achieved room temperature self-healing. Thus, the self-healing speed of i-DAPU increased with higher temperatures, an increased IL content and a higher D–A functional density, and at a rate far exceeding that of regular DAPU (Figure S10, Supporting Information). For most polymers, mechanical performance and self-healing efficiency often exhibited a trade-off (Figure 2f and Table S2, Supporting Information). However, our synthesized i-DAPU combined exceptional mechanical performance with efficient self-healing capability.
2.4 High Sensitivity
Consequently, we selected i-DAPU as the dielectric material for our subsequent iontronic skin. In Figure 1, through theoretical simulations, we confirmed the existence of a molecular-ionic interaction network between D–A groups and the IL. In Figure 3a, we explored how this molecular-ionic interaction network, resulting from the attraction between D–A groups and IL, could be disrupted under pressure, laying the foundation for dual-channel synchronous sensing. In i-DPU, only one peak was observed in the ultraviolet absorption spectrum, which corresponded to the interaction between Donor groups and [BMIM]+. This peak gradually diminished under pressure and then returned to its original state once the pressure was released. In contrast, the ultraviolet absorption spectra of i-(D+A) PU and i-DAPU showed two distinct peaks at 359 nm (IL peak) and 430 nm (DA peak), with the latter showing a notably weaker overall peak intensity in i-(D+A) PU than in i-DAPU. As pressure increased, the intensity of these two peaks gradually decreased, with the peak at 359 nm diminishing more rapidly than the one at 430 nm, as detailed in Figure S11 (Supporting Information). Upon pressure release, both peaks could recover to their original state. It was important to note that in DAPU, although the intensity of the peak also decreased under applied pressure and recovered upon pressure release, the phenomenon of two distinct peaks did not occur. But the force could disrupt the D–A self-assembly. According to the UV absorption spectra of D-PU, A-PU and DAPU (Figure S12, Supporting Information), the data demonstrated that D-PU containing only donor groups, and A-PU containing only acceptor groups, neither of these polyurethanes exhibited a discernible peak at 430 nm. However, this specific UV absorption peak at 430 nm was observed in DAPU, corresponding to the self-assembly of donor and acceptor groups within and between the polyurethane molecules.[14] When pressure was applied to DAPU and subsequently increased, the peak at 430 nm gradually diminished, indicating disruption of the donor–acceptor self-assembly. This suggested that the applied force led to the dissociation of the self-assembly structure.
Additionally, we conducted an analysis of the initial capacitance (C0) and final capacitance (C) under identical pressure conditions for i-DAPU, i-DPU, i-APU and i-TPU, as depicted in Figure S13a (Supporting Information). In iontronic skin, capacitance was contingent upon the concentration of free ions. It was observed from the figure that i-DPU and i-APU, which incorporated a single trapping group, exhibited substantially lower C0 compared to i-TPU, which lacked trapping groups. Notably, i-DAPU, which featured dual trapping groups, demonstrated the lowest C0. This phenomenon could be attributed to the capacity of donor and acceptor groups to trap cations and anions, respectively, thereby diminishing the free ion concentration at the initial state of the iontronic skin. Subsequently, when subjected to the same pressure, i-DAPU, i-DPU, i-APU and i-TPU all exhibited nearly identical final capacitances. This outcome confirmed that both donor and acceptor groups could release the trapped ions under pressure, thus enhancing the final capacitance of the iontronic skin. By comparing the capacitance before and after the application of pressure, the ion trap-release capabilities of the donor and acceptor groups were substantiated. Moreover, Figure S13b (Supporting Information) illustrated that repeated application of the same pressure did not alter their capacitances, affirming the stability and reversibility of the ion trap-release process across multiple cycles. Thus, this implied a unique interaction in i-DAPU, fundamental to its advanced sensing capabilities, where the molecular-ionic networks responded distinctly to mechanical stimuli, a feature absent in DAPU. This differential response underlay the enhanced functionality and sensitivity of i-DAPU as a material for iontronic skins.
This distinction was due to the fact that in D-PU, only one type of donor (Donor) group existed, which lacked self-assembled structures. When mixed with the IL, D-PU solely adsorbed the [BMIM]+ cation from the ionic liquid, effectively functioning as a single-channel protein, capable of transporting only one type of ion. This process was reversible, allowing for restoration to the original state once the pressure was removed. Unlike i-DPU, which contained only one type of group, i-(D+A) PU and i-DAPU feature both donor and acceptor groups, forming D–A self-assembly between them. Moreover, these donor and acceptor groups attracted the cations and anions from the ionic liquid, respectively, creating the molecular-ionic interaction networks analogous to cellular membranes housing two types of channel proteins for different ion transport. So, the D–A self-assembly was disrupted by the IL, cleaving into two peaks (Figure 3a and Figure S14, Supporting Information). The peak at 430 nm corresponded to the UV absorption peak of D–A self-assembly (DA peak), while the one at 359 nm corresponded to the blue-shifted UV absorption peak of D–A self-assembly (IL peak) after dual trapping with cations and anions, which intensified with increasing IL content (Figure S15, Supporting Information).[14]
Similar to the process in cellular membranes where mechanical changes triggered the simultaneous opening of two channel proteins for ion transport, under pressure, not only was the D–A self-assembly disrupted, but also the molecular-ionic interaction networks. This resulted in the release of more trapped ions at the same pressure, transitioning them into a free state and achieving dual-channel synchronous ion release, thus altering the ionic concentration within the material. Consequently, this led to a significant enhancement in sensitivity (Figure 3b). However, the UV absorption peak intensity and associated sensitivity in i-(D+A) PU were inferior compared to i-DAPU. This discrepancy arose due to the inherent structural limitations of (D+A) PU, fabricated through a physical blend of D-PU and A-PU polyurethanes. Such a blend restricted the formation to predominantly intermolecular D–A self-assemblies. In stark contrast, i-DAPU, incorporating both D and A groups within its principal molecular chain, facilitated a more complex self-assembly landscape. This included not only intermolecular but also intramolecular D–A self-assemblies. Consequently, the overall density of D–A self-assemblies in i-DAPU markedly surpassed that in (D+A) PU. This increased density of D–A self-assemblies significantly enhanced the capacity for dual-channel synchronous ion release in i-DAPU. Upon pressure stimulation, the change in ion concentration was more pronounced in i-DAPU, resulting in superior sensitivity. Therefore, in materials containing trap-groups, i-DPU, i-(D+A) PU, and i-DAPU represented single-channel release of a single ion, weak dual-channel synchronous release and strong dual-channel synchronous release, respectively. Consequently, i-DAPU exhibited the highest sensitivity, reaching up to 90.98 kPa−1.
Hence, the iontronic sensor of DA-skin, fabricated using i-DAPU as the dielectric material, exhibited remarkable flexibility in signal transmission. The compositional structure of DA-skin has been incorporated into Figure S16a (Supporting Information). DA-skin adopted the conventional sandwich structure, with the outermost two electrode layers and the intermediate dielectric layer. The electrodes were composed of DAPU coated with AgNWs and liquid metal, and the dielectric layer consisted of i-DAPU filled with 30% IL. Within the entire sensor structure, the sensitivity of DA-skin was primarily determined by the diffusion quantity of ions in the i-DAPU. Due to the presence of a molecular-ionic interaction network in i-DAPU, the stronger ion trapping force of i-DAPU resulted in a very low initial concentration of free ions. The formation of the Stern layer in the EDL was influenced by electrostatic adsorption, a relatively stable interaction, while the Differ layer resulted from ion thermal motion. When the free ion concentration was extremely low, the mean free path of ion thermal motion was extended, causing free ions in the EDL to be more inclined towards the Differ layer, thereby reducing the initial capacitance of the sensor and increasing the magnitude of the change in the number of free ions before and after pressure application, thereby endowing DA-skin with ultrahigh sensitivity. As shown in Figure S16b (Supporting Information), when the deformation ranged from 0% to 100% stretching, the resistance increased slightly from 3.34 to 3.41 Ω, without significant resistance fluctuations in response to external forces, thereby maintaining signal transmission stability and not affecting the transmission of capacitance signals in the EDL.
The responding speed of DA-skin was 25 ms per time, as shown in Figure S17 (Supporting Information), which exhibited temporal stability. Under sustained pressure, multiple tests revealed that the responding speed of the DA-skin remained remarkably swift, consistently at 25 ms per time even after repeated presses, without any noticeable relaxation phenomenon. This indicated that prolonged pressure did not compromise its responding speed. The DA-skin continued to retain its rapid responsiveness even after multiple uses. Furthermore, DA-skin not only boasted a rapid response time, but also possessed a low detection limit, capable of detecting minute objects such as tea leaves (Figures S18 and S19, Supporting Information), showcasing DA-skin's high sensitivity. Moreover, DA-skin maintained stable signal output both before and after self-healing. When cyclically applying a pressure of 30 kPa to the DA-skin, the variation of its capacitance over time was depicted in Figure S20 (Supporting Information). It was evident that the capacitance output signal of DA-skin maintained a high level of stability during long-term cyclic usage, exhibiting only a minor decline of 2%, which was nearly negligible. Furthermore, it retained the same level of stability even after self-healing. Hence, this result further demonstrated the excellent capacitance stability of i-DAPU, suggesting broad prospects for its application.
Upon integrating a dome-like microstructure on the surface of the dielectric material, the results, as depicted in Figure 3c and the SEM image and the size of dome-like microstructure was shown in Figure S21 (Supporting Information), revealed that when the same pressure was applied, stress concentration occurred at the microstructures, resulting in a more pronounced double-layer effect. Consequently, its sensitivity could be elevated to 7012.05 kPa−1. Currently, most sensors maintained high sensitivity only within a low-pressure range (Figure 3d and Table S3, Supporting Information). Once the pressure exceeded 10 kPa, the sensitivity tended to decrease. However, our DA-skin not only exhibited superior sensitivity but also maintained this high level of sensitivity across a broad pressure range (0–50 kPa). This characteristic endowed DA-skin with tremendous potential for fabrication into flexible sensors suitable for multiple application scenarios.
Upon calculating the sensitivity enhancement ratio, it was surprisingly discovered that the sensitivity of DA-skin is 19.26 times higher than that of iontronic skin fabricated with conventional i-PU as the dielectric material. To account for this remarkable sensitivity of DA-skin, we formulated a hypothesis regarding its underlying mechanism. As illustrated on the left side of Figure 3e, the equivalent circuit of iontronic skin comprised a sandwich-structured plate capacitor (PC) and the electric double layer (EDL) at the electrode ends. The capacitive value Cpc, mainly influenced by the mechanical properties of the dielectric layer, was relatively consistent for materials within the same system. Hence, the sensitivity differences observed among various iontronic skins originated primarily from CEDL. As shown on the right side of Figure 3e, the magnitude of CEDL was closely related to the number of mobile ions within the material, affecting both the capacitance of the Stern layer (CS) and the Differ layer (CD). Consequently, based on in situ UV spectroscopy results, we illustrated the changes in mobile ion quantities in i-DPU, i-(D+A) PU, and i-DAPU before and after pressure application (Figure 3f). It was evident that the variations among these materials were significant, especially for i-DAPU, where the change in ion quantity was several times higher than the other materials, leading to a marked increase in the sensitivity of DA-skin.
The magnitude of sensitivity enhancement in DA-skin potentially stemmed from the stronger ion adsorption force of i-DAPU, resulting in a very low initial concentration of free ions. The formation of the Stern layer in the double layer was influenced by electrostatic adsorption, a relatively stable interaction, while the Differ layer resulted from ion thermal motion. When the free ion concentration was extremely low, the mean free path of ion thermal motion was extended, causing free ions in the double layer to be more inclined towards the Differ layer, thereby reducing the initial capacitance of the device. Initial capacitance tests under identical conditions for devices based on i-DPU, i-(D+A) PU, and i-DAPU showed that the first two had capacitances of 30.745 and 20.714 nF, respectively, while the capacitance of DA-skin was as low as 17.346 nF, significantly different from the others and aligning with our analysis. This not only indicated that a dual-channel synchronous sensing design was an effective way to enhance the sensitivity of iontronic skins but also helped in refining the electrical mechanism for high-sensitivity perception in these devices.
2.5 Clinical Application: Muscle Strength Recognition
In practical clinical settings, monitoring the muscle strength recovery in comatose patients poses a significant challenge for physicians. Muscle strength refers to the force exerted during active muscle movement. Normally, sensory neurons generate nerve impulses in response to stimuli and transmit these impulses to the spinal cord, triggering a spinal reflex. These impulses are then relayed from the spinal neurons to the muscles, releasing neurotransmitters that stimulate muscle cell contraction. This process is illustrated in Figure 4a. Medically, muscle strength is graded on a scale of 0 to 5, where grade 0 indicates no muscle response to pain stimuli, grade 1 indicates muscle contraction under pain stimuli but without lifting capability, and grade 2 indicates that muscles can momentarily lift the fingers off the bed under pain stimuli. The transition from grade 0 to 1 involves subtle changes in muscle strength, often difficult to discern visually for physicians, potentially impacting the assessment of a patient's coma degree and the determination of treatment plans. Our sensor, DA-skin, with its exceptionally high sensitivity, can translate these subtle pressure signals into discernible capacitive signals, making it applicable in this specialized scenario to assist doctors.
To validate the feasibility of using DA-skin in clinical medicine for detecting patient muscle strength recovery, we engaged a neurosurgeon as a volunteer, relying on their expertise to evaluate the sensor. As illustrated in Figure 4b, the volunteer doctor (the Affiliated Lihuili Hospital of Ningbo University) laid flat on a hospital bed with DA-skin attached to their fingertip. When the doctor slightly lifted their finger, the capacitance decreased and then increased as the finger returned to its original state. This process was repeated twice, with results presented in Figures S22 and S23 and Movie S1 (Supporting Information). With DA-skin affixed to the fingertip, a muscle strength of grade 0 naturally applied minimal pressure to the bed surface. As muscle strength recovered to grade 1, slight finger twitching caused a small decrease in capacitance. When muscle strength reached grade 2, allowing the finger to be slightly lifted, the pressure on the bed surface from the fingertip disappeared momentarily, and the capacitance almost returned to zero. Hence, by monitoring changes in capacitance, DA-skin attached to the fingertip could effectively track muscle strength recovery.
To further develop a system capable of sensing the distinct muscle strength recovery of different fingers, we attached DA-skin to the fingertips of various fingers and measured the capacitance values corresponding to muscle strengths of grade 0, 1, and 2, as shown in Figure 4c. At grade 0, each finger naturally rested on the table, exerting pressure on the surface. Due to variations in weights of each finger and contact areas with the table, the resulting capacitive signals differed in intensity. At grade 1, with slight finger tremors, the pressure on the table slightly decreased, and our highly sensitive sensor DA-skin immediately captured this signal, resulting in an immediate decrease in capacitance. However, at grade 2, when fingers were fully lifted, removing pressure on the table, DA-skin on each finger returned to its initial capacitance (C0). Consequently, the capacitive signals generated by each finger at grade 2 were nearly identical. To distinguish changes in muscle strength more straightforwardly in real-life applications, we employed circuit design to visualize these changes in electrical signals through LED lights, as illustrated in Figure 4d and Figures S24 and S25 (Supporting Information), and the progress was demonstrated in Movies S2–S6 (Supporting Information). When the right hand was naturally placed on the table, each finger exerted pressure on the surface. We designated the signals generated in this state as representative of grade 0 muscle strength, at which point LED-1 lit up. When muscle strength changed from grade 0 to 1, allowing for slight finger tremors, the capacitance slightly decreased, and LED-2 illuminated. As muscle strength progressed from grade 1 to 2, enabling fingers to lift and detach from the table, the pressure on the sensor disappeared, and LED-3 lit up. Therefore, by observing the illumination of different bulbs, one could intuitively determine the changes in muscle strength.
To assist physicians with rapid diagnosis, we employed a deep learning algorithm to intelligently classify muscle strength levels based on the signals. The utilized algorithm was a multi-layer perceptron (MLP), as illustrated in Figure. 4e, comprising fully connected input, hidden, and output layers. The input layer consisted of a single layer with 24 points; the hidden layer had two layers, each with 15 points; and the output layer comprised one layer with three points, corresponding to grades 0, 1, and 2 (G0, G1, G2). This network algorithm was implemented using Python, with the specific program code provided in the Supporting Information. The training set for this network consisted of 300 samples from the data shown in Figure 4c, while the test set comprised 250 samples extracted from the data in Figures S26 and S27 (Supporting Information), evenly distributed across five channels. The final output results were shown in Figure 4f. During 90 training iterations, the accuracy of our MLP gradually increased from 67.5%, and the trained MLP demonstrated a high accuracy rate of 99.2% in actual testing. To understand what limited further improvements in accuracy, we displayed the confusion matrix from the last iteration's test results in Figure 4g. Most samples showed consistency between predicted and true labels, resulting in a diagonal distribution in the confusion matrix. However, two samples classified as G0 were misjudged as G1, preventing the final accuracy from reaching 100%. This was due to the very subtle motion caused by grade 1 muscle strength, leading to weak capacitive signals that could be mistaken for the noise generated by grade 0 muscle strength (as shown in Figure 4c). Overall, this system can rapidly and accurately identify changes in muscle strength in different fingers, which holds significant importance in clinical medicine for assisting physicians in assessing patients' muscle strength recovery.
3 Discussion
In summary, drawing inspiration from tactile perception channel proteins Piezo 1 and Piezo 2, as well as the repair proteins Tspan 4 and TSP-15 in the human body, we introduced both donor and acceptor groups into the main chain of polyurethane and co-blended them with the ionic liquid of [BMIM]+[PF6]−. This innovation led to the creation of a novel mechano-responsive elastomer, i-DAPU, integrating traction-assisted self-healing and dual-channel synchronous sensing. The self-healing property of i-DAPU, significantly enhanced compared to DAPU, was attributed to the synergistic effects of intrinsic self-assembly, achieving room-temperature self-healing capabilities. Moreover, the molecular-ionic interaction network formed between the D-A self-assembly and [BMIM]+[PF6]− was fundamental to dual-channel synchronous sensing ability of i-DAPU. This network could be disrupted under pressure to release a substantial number of ions, analogous to the process in cells where dual-channel proteins transmit ions upon stimulation. The observed reduction in ionic liquid binding energy confirmed the formation of the molecular-ionic interaction network. Under pressure, donor and acceptor groups released ions and the molecular-ionic interaction network was disrupted, reforming upon pressure relief. It was the presence of this dual-channel synchronous sensing that endows the DA-skin sensor, fabricated with i-DAPU as the base material, with an initial capacitance lower than that of conventional sensors, thus exhibiting superior sensitivity, as high as 90.98 kPa−1. We further applied this sensor in clinical medicine to monitor subtle changes in muscle strength, typically challenging for physicians to detect. Utilizing deep learning algorithms for signal processing, we achieved intelligent muscle strength level classification, with an accuracy rate as high as 99.2%. These findings demonstrated the immense potential of sensor in applications for analyzing muscle strength in paralyzed patients.
4 Experimental Section
Materials
Pyromellitic dianhydride, ethanolamine (AR, 99%), 1,5-naphthalene diisocyanate (GC, ≥ 98.0%), isophorone diisocyanate (IPDI, 98%), dibutyltin dilaurate (DBTDL, 95%), N,N-dimethylformamide (DMF, anhydrous, 99.8%), and 1-butyl-3-methylimidazolium hexafluorophosphate ([BMIM]+[PF6]−, 97%) were purchased from Aladdin (China). Toluene (99.5%) was purchased from Sinopharm Chemical Reagent Co., Ltd. Polycaprolactone diol (PCL diol, Mn = 2000 g mol−1) was purchased from Shanghai Macklin Biochemical Technology Co., Ltd. These materials were used without further purification.
Synthesis of Chain Extender A
First, 20 g of finely powdered pyromellitic dianhydride crystals were added to a 500 mL three-neck flask, followed by the addition of 200 mL of N,N-dimethylformamide and magnetic stirring at room temperature to obtain a clear solution. Subsequently, 18 g of ethanolamine (in excess) was added to a dropping funnel. It was then slowly added dropwise to the solution of pyromellitic dianhydride, and the mixture was continuously stirred magnetically at room temperature for 12 h to obtain a homogeneous emulsion. Next, a 50 mL separating funnel and a straight condenser were assembled, fixed with an iron stand and condenser clamp. Toluene was added to the separating funnel's water outlet, and 50 mL of toluene was added to the milky solution as a coboiling agent (to remove water generated during the reaction at high temperatures). The system temperature was raised to 170 °C, and stirring continued until the water layer in the separating funnel ceased to increase. The reaction was then stopped, and toluene was removed by rotary evaporation. The residue was precipitated in deionized water, and filtered to obtain a white product, which was dried under vacuum at 60 °C for 8 h.
Synthesis of DAPU
In a glove box filled with Ar (99.999%), PCL diol (Mn = 2000), 1,5-naphthalene diisocyanate, chain extender A, and anhydrous DMF were sequentially added to an open reactor equipped with a mechanical stirring device. After sealing, the mixture was heated and stirred at 60 °C until the solution became clear and transparent. Subsequently, a specific amount of IPDI and DBTDL was injected into the open reactor using a syringe. The temperature was then raised to 80 °C, and the reaction was stirred for 8 h. After completion, the product was poured into deionized water and washed several times, soaked in methanol under ultrasound for 30 min, followed by ultrasound soaking in deionized water for an additional 30 min, and finally vacuum-dried at 60 °C. The molar ratio between the hard and soft segments varied as 5/1.5 (DAPU-1), 5/2 (DAPU-2), 5/2.5 (DAPU-3), and 5/3 (DAPU-4). This progress is shown in Figure S28a (Supporting Information).
Preparation of i-DAPU Film
A certain amount of DAPU-2 was taken and added to DMF for stirring until completely dissolved. Once the solution became clear and transparent, a specific quantity of [BMIM]+[PF6] − was introduced, and the mixture was magnetically stirred at room temperature for 24 h. The resulting solution was then poured into a polytetrafluoroethylene mold and dried on a heating plate at 80 °C until complete evaporation of DMF. The IL content was calculated as the IL weight divided by the sum of the IL weight and the weight of DAPU-2. This progress is shown in Figure S28b (Supporting Information).
Preparation of DA-Skin
The i-DAPU was cut into a rectangular shape to serve as the dielectric layer. Subsequently, the flexible electrodes were cut into two pieces, slightly smaller rectangles than the i-DAPU. The i-DAPU was then clamped between the two electrodes, with a copper wire placed in the middle. Finally, the assembly was encapsulated using tape.
Characterization
The 1H NMR spectra were acquired using a 400 MHz Bruker AVANCE III instrument (Bruker, Switzerland), and deuterated dimethyl sulfoxide (DMSO-d6) was chosen as the solvent. FTIR spectra were obtained through attenuated total reflection Fourier-transform Infrared spectroscopy (ATR-FTIR) using an Agilent Cary 600 series FT-IR instrument. Molecular weight and molecular distribution of DAPU were determined by gel permeation chromatography (GPC, Waters-2690) with N,N-dimethylformamide as the mobile phase. Tensile and cyclic compression tests were conducted on a 1 KN universal material testing machine (UTM, Zwick Instruments, Model: Z1.0), with samples shaped according to the national standard Type IV. To ensure the reliability and representativeness of the mechanical tensile tests, a minimum of three specimens were selected for the experiments. Each specimen underwent detailed tensile testing to obtain mechanical performance data, including modulus, tensile strength, elongation at break, and toughness. The mean and standard deviation of the test results were calculated. 2D images and ionic liquid compatibility images of i-DAPU were obtained using an optical microscope equipped with a heating stage (Olympus / BX 51). 3D images of DAPU self-healing were acquired using laser confocal microscopy (VK-X200K, Japan). Impedance and ion conductivity data were obtained using an electrochemical workstation (PGSTAT302N, Metrohm Autolab), with a frequency scan range of 0.1 Hz to 100 kHz and a voltage of 10 mV. UV absorption spectra were obtained using a UV–visible–near-infrared spectrophotometer (LAMBDA) (Lambda 950, America). All capacitance data were obtained using an LCR meter (UC2878 from Youce, China).
The Measurement Method and Steps of Testing Sensitivity of DA-skin
The steps for testing its sensitivity are as follows: 1) without applying any pressure, the initial capacitance C0 of the sensor was measured using LCR testing; 2) the sensor was placed in a UTM while maintaining its connection to the LCR. Gradual pressure was applied by using the UTM, and the LCR recorded real-time capacitance signals at different pressures; 3) by calculating the contact area of the sensor, pressure was converted into pressure intensity. Utilizing the formula “Capacitance change = (C-C0)/C0 = ∆C/C0,” the capacitance change corresponding to different pressure intensities was calculated; 4) pressure intensity was plotted on the horizontal axis, and capacitance change was plotted on the vertical axis to generate a curve. The slope of the fitted curve represents the sensitivity of the sensor. The DA-skin sensor was only temporarily affixed to the epidermis of the participants' fingertips. Typically, non-invasive studies involving minimal risk, particularly when the device is used for brief, superficial contact, may be exempt from the requirement for ethical approval. Given the nature of its use and the brief exposure time, the doctors who are affiliated in the Affiliated Lihuili Hospital of Ningbo University involved in this work believe that such an experimental setup does not necessitate a declaration for ethical approval.
Muscle Strength Recognition Assisted by MLP
The input layer of the MLP used in experiments were pressure distribution images with the size of 24, the hidden layers contained 2 kernels with the size of 15, the output layer were 3 neurons that symbolized the types of muscle strength grade needs to be recognized. During recognition, the data set was collected by sensor array based on DA-skin with 5 channels. It had a total of 550 samples, of which 300 samples were used as training sets and 250 samples were used as test sets. The program code of algorithm is placed in the Supporting Information, and was run by Python software to compute.
Acknowledgements
This work was supported by the Brain Pool program (RS-2023-00222619) and the international cooperation program (2022K2A9A2A06046004) funded by the Ministry of Science and ICT through the National Research Foundation of Korea. This work was also supported by the National Natural Science Foundation of China (52211540393). The authors appreciate the volunteer doctor Pandi Chen who are affiliated in Department of Neurosurgery, the Affiliated Lihuili Hospital of Ningbo University for assisting us in the demonstration of DA-skin applied to test muscle strength and the related medical support.
Conflict of Interest
The authors declare no conflict of interest.
Open Research
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
