A multifunctional porous interface bridging 3D architected electronics with skin

Abstract

Skin-integrated flexible electronics are rapidly advancing from short-term use to continuous, long-term wear to meet the demands of health monitoring, wound healing, and skin disease treatment. While three-dimensionally (3D) architected devices offer sensing capabilities and outstanding mechanical performances, the development of a conformal interface between these 3D electronic devices and skin for long-term comfortable wear remains challenging due to their geometric complexity and mechanical fragility. Here, we introduce a multifunctional porous interface that bridges 3D flexible electronics with skin through engineered microporous networks. Such a porous design enhances elastic stretchability and electrical reliability while providing moisture permeability, thermal buffering, impact mitigation, and drug delivery capability. Stiffness reduction and localized pore-wall-shell buckling synergistically underpin the outstanding mechanical resilience under large deformations. The proposed multifunctional porous interface allows development of a 3D closed-loop wound administration patch capable of multimodal sensing and on-demand therapy, markedly accelerating scald wound healing in vivo.

Microporous interface bridges 3D flexible electronics with skin for breathable, highly stretchable wear and drug delivery.

INTRODUCTION

Recent advances in skin-integrated flexible electronics have enabled their widespread uses in health monitoring (1–8), therapeutic feedback (9–13), energy harvesting (14–16), and human-machine interactions (17–22), driving a transition from short-term operation to continuous, long-term wear (23–25). This progression demands that flexible electronics should maintain high mechanical stretchability and robust physical and chemical stability to ensure functional reliability (26–28) while simultaneously providing skin-level comfort through breathability (29, 30), moisture permeability (31–33), and minimal mechanical constraint. The incorporation of advanced functionalities such as thermal management (34), impact buffering, and on-site therapy [e.g., drug delivery (35–37) and electrical stimulation (11, 38)] could further empower flexible electronic systems with versatile adaptability across complex scenarios. The emerging class of flexible electronics with three-dimensional (3D) architectures (39, 40) offers a compelling approach to accommodate the increasing demands of multifunctional integration and multimodal operation on the skin. Beyond enabling desired spatially distribution of functional components, well-designed 3D electronic architectures could introduce new capabilities, such as biomimetic mechanical responses (41–50); ultrawide field of view in visual sensing (51, 52); and decoupled tactile sensing of pressure, strain, and temperature (39, 53–57). Nevertheless, the inherent geometric complexity of 3D flexible systems presents substantial challenges in achieving mechanically/electrically stable (58), geometrically adaptive (59, 60), and long-term comfortable integration on the skin (33, 61).

Existing strategies for skin integration were mainly developed for planar or layered flexible electronic systems (62, 63). A common approach involves solid substrate-mediated interfaces, where flexible planar circuits are mounted on or embedded in ultrathin, soft substrates (e.g., polydimethylsiloxane, styrene-ethylene-butylene-styrene, and hydrogels) to achieve conformal contact with the skin (64–66). However, these solid interfaces often suffer from limited breathability and moisture permeability, leading to discomfort, irritation, or even serious dermatological issues during prolonged wear. Recent efforts have explored porous substrate designs (e.g., nanofibrous mats and fabrics) (29, 30, 32, 67, 68) to enhance skin compatibility and permeability. Semi-invasive designs represent another widely used strategy for skin integration, exemplified by microneedle-based systems that penetrate the superficial layers of the skin to establish direct and stable mechanical or bioelectrical coupling (69). While effective for interfacial coupling in planar systems, these approaches are highly challenging to implement in 3D flexible electronics with complex geometries. Key challenges arise from two factors: (i) 3D architectures usually do not allow conformal attachment to the relatively flat skin interface, thereby requiring an interfacial mediation layer to couple the flexible device with the skin surface. Conventional solid encapsulation (28) affords essential mechanical protection and water/oxygen shielding but compromises comfort during prolonged wear, while the nanofibrous mats and fabrics (30, 70) are incompatible with 3D electronics due to their inability to conformally nest within complex 3D microstructures. (ii) Compared with planar flexible structures, sophisticated 3D architectures are more susceptible to plastic deformations and fatigue failure. For 3D flexible electronics encapsulated in a solid matrix (27, 71), force transfer paths are highly complex under external loadings, leading to stress concentrations in suspended ribbons and branching regions, which could significantly reduce elastic stretchability [i.e., maximum applied strain to maintain structural deformations within the elastic regime (58, 66)] and compromise device reliability.

In this work, we propose a multifunctional porous interface to bridge 3D flexible electronic devices with skin, where rationally designed soft microporous networks are incorporated to ensure both electrical reliability and wearable comfort. In comparison to the conventional integration scheme using a solid encapsulation, the porously integrated 3D flexible electronics offers multidimensional advantages in enhanced lightness, softness, elastic stretchability, moisture permeability, thermal, and impact buffering, as evidenced by combined numerical simulations and experimental measurements. Constrained deformation behaviors are systematically analyzed for representative 3D polyimide/gold/polyimide (PI/Au/PI) mesostructures embedded in the porous matrix, revealing that the decreased effective modulus and local buckling of micropore walls/interconnects synergistically contribute to enhanced elastic stretchability. The design of a 3D flexible sensing system with biomimetic spatial arrangement of different sensors and engineered porosity distribution enables decoupled sensing of compressive force, tensile strain, and temperature. Leveraging the drug-loading and thermal buffering capacities of micropores within the encapsulation matrix, we develop the first 3D closed-loop wound administration electronic patch (3D-CURE Patch) for the scald wound, which accounts for ~50% of thermal injuries globally (72). The 3D-CURE Patch enables real-time monitoring of physiological symptoms (e.g., infectious swollen and inflammatory edema) of scald wounds and provides on-demand drug release according to the monitored status. The device can help mitigate key clinical challenges, including high infection risk, excessive exudate, and rapidly progressing wound areas that necessitate frequent and timely dressing changes and drug administration. Scald wound healing studies in a Sprague-Dawley rat (SD-rat) model demonstrate that the 3D-CURE Patch significantly accelerates skin regeneration, highlighting its potential for personalized health monitoring and wound care.

RESULTS

Multifunctional porous integration strategy

Figure 1 (A and B) illustrates the design concepts of multifunctional porous interface bridging 3D architected electronics and skin, exemplified by a 3D flexible electronic device consisting of helical and arc-shaped interconnects, suspended interdigital electrodes, and a flexible printed circuit board (FPCB) with integrated functional chips. Following the standard process of mechanically guided 3D assembly (73–75), the 2D precursor structure (fig. S1) of the flexible electronic device was selectively bonded with a biaxially prestretched elastomer substrate and then transformed into a bilayer 3D architecture (Fig. 1A, left) through controlled buckling deformations upon the release of prestrains. A 3D sacrificial template is created by stacking soluble microparticles (e.g., sugar microparticles) to fill the gaps within the assembled 3D flexible device, thereby offering excellent self-adaptability to intricate 3D geometries (figs. S2 and S3). After the negative pressure-assisted infusion and oven curing of the fluidic encapsulation material, the sacrificial template and bottom substrate are removed, yielding a porously encapsulated 3D flexible device (Fig. 1A, right) evidenced by scanning electron microscopy morphological characterization (figs. S4 and S5 and text S1). The porous network structure introduces densely distributed polymer-air interfaces and thin pore walls within the encapsulation matrix, notably enhancing the lightness and softness while also altering the thermal conductivity and moisture permeability. For example, in a wearable scenario for personalized medicine (Fig. 1B), the porous interface allows the 3D flexible device to be conformably mounted on curved skin surfaces of the human shoulder while maintaining high elastic stretchability to accommodate dynamic limb movements. The open porous network offers superior moisture permeability and thermal buffering compared to the case of conventional solid encapsulation, which enhances the adaptability to bodily fluids and resistance to external thermal fluctuations. Aside from the above advantages of porous design that are essential to the long-term integration with human skin for health monitoring, the micropores filled with thermoreversible phase-change gels could concurrently facilitate drug loading and on-demand release for potential wound treatment.

Fig. 1. Design concepts of multifunctional porous interfaces bridging 3D flexible architected electronics and skin.

(A) Schematic illustrations of porous encapsulation seamlessly adapting to a 3D flexible electronic device that consists of complex bilayer 3D interconnects, curved electrodes and a central FPCB with integrated functional chips. (B) Demonstrations of outstanding properties and functionalities enabled by the porous design for 3D flexible devices in wearable scenarios, including high elastic stretchability, moisture permeability, thermal buffering, reliable sensing and on-demand therapy. (C) Mini-LED array interconnected by 3D buckled microribbons in the porous interface (top left) and the enlarged (top right) and cross-sectional (bottom) views of the microribbons and micropores. Scale bars, 2 mm (left panels) and 200 μm (right panels). (D) Comparisons of skin conditions after 1 hour of outdoor running and 7 days of daily wear. Scale bars, 10 mm. (E) Averaged illuminance of five regions under the pristine state and five practical scenarios. (F to I) Measured elastic modulus (E), impact resistance ratio (η) (F), thermal conductivity (λ) (G), and moisture vapor transmission rate (MVRT) (H) of the soft matrix, along with the elastic stretchability (I) of 3D helical mesostructures, as functions of the porosity (φ). (J) Comparison of key performance metrics between porous and conventional solid matrixes. Photo credit: Z.S. and X.C., Tsinghua University.

As a basic device-level demonstration of the porous integration strategy, Fig. 1 (C to E) presents a stretchable array of mini light-emitting diodes (mini-LEDs) embedded in the foam matrix, which can be conformally mounted on the forearm skin. An array of 5 by 5 red light mini-LEDs (SMD 0402) was electrically connected by a network of buckled copper/PI (Cu/PI) microribbons and encapsulated using a porous matrix (Ecoflex 00-20) with a porosity of 50% (i.e., φ = 50%) (fig. S6). Figure 1C shows the pore distributions on the bottom surface of the porous matrix (top right) and the interfaces between the porous matrix and buckled microribbons in a cross section (bottom). Micropores (size ranging from ~200 to 400 μm) are randomly distributed within the foam matrix and conform well to the curved surfaces of buckled microribbons without any noticeable delamination (fig. S7). Note that an ultrathin (~20 μm in thickness) layer of soft silicone conformally follows the 3D geometry of the electronic architecture, forming a protective shell that resists moisture and oxygen-induced degradation (figs. S8 and S9). The outstanding moisture permeability and high stretchability of porous interfaces ensure skin comfort during intense exercise and prolonged daily wear. After 1 hour of outdoor running exercise (ambient temperature of ~30°C) and 7 days of daily wear, the skin-contact regions exhibit negligible irritation reaction compared with the surrounding skin, while the solid silicone counterparts suffer from sweat accumulation and abnormal redness. Besides, intense pruritus and discomfort were reported throughout the daily wear period (Fig. 1D). These results demonstrate that the porous morphology remains well preserved, without evidence of pore blockage or structural collapse (fig. S10). This can be attributed to the breathable porous architecture, which facilitates moisture vapor transport and reduces the accumulation of exudates, together with the intrinsic chemical and mechanical stability of the silicone-based interface. The reliability of the porous encapsulation was validated by measuring the illuminance of five regions above five LEDs across five challenging scenarios (fig. S11), including cyclic stretching (10⁴ cycles at ~30% strain amplitude), hot water soaking (7 days at ~50°C), machine washing (1 hour), outdoor running (1 hour at ~30°C), and prolonged daily wear (7 days) on the forearm skin (Fig. 1E). Negligible degradation was observed in the optoelectronic device, as averaged illuminance variations remained below 10%, demonstrating its excellent mechanical durability and electrical stability under realistic operating conditions. Besides, the porously integrated device can withstand harsh mechanical loadings (including folding, stretching, and poking), as evidenced by the stable electrical performances under these loading conditions (fig. S12 and movie S1).

Porosity engineering serves as a powerful approach to tailoring the multidimensional physical properties of the porous interface. Based on the Gibson-Ashby theory of open-cell foams (76), the elastic modulus (E_φ) of porous encapsulation follows a power-law relationship with the relative density (1–φ), suggesting a notable reduction in modulus for a higher porosity (φ). On the basis of the uniaxial tensile test results (fig. S13), as the porosity increases from 0 to 50% and then to 70%, the elastic modulus of the encapsulation decreases from ~80 to ~25.9 kPa and further to ~12.9 kPa (Fig. 1F). The porous silicone network demonstrates excellent mechanical stability over repeated cycling, with an average hysteresis ratio of 7.18% and an SD of 0.12%, while the residual strain reaches ~3.04% after the 100th cycle (fig. S13, E and F). The compressive deformation behavior of porous encapsulation exhibits characteristic J-shaped stress-strain responses due to the densification effect, where micropores collapse and form a nearly solid matrix under large compressive strains (fig. S14). Leveraging the substantial energy absorption from pore-wall collapse, the porous matrix demonstrates superior impact resistance compared to solid matrix (Fig. 1F). In the drop weight impact test, the damping efficiency (η) was increased by 84.8 and 156.5% for porous encapsulations with 50 and 70% porosities, respectively, thereby offering enhanced protection for impact-sensitive fragile components (e.g., silicon chips and micro-electromechanical systems) (fig. S15). Maintaining a stable temperature and effectively buffering against external thermal shocks are critical requirements of thermal management for the normal functioning of 3D flexible electronics. The numerous air-polymer interfaces within the porous matrix extend the thermal transport path and confined porous microstructures interrupt continuous solid-conduction pathways while simultaneously restricting gas motion inside confined pores, which effectively reduce thermal conductivity (λ) from 0.162 to 0.0439 W/m·K as the porosity increases from 0 to 70% (Fig. 1G and figs. S16 to S19). In wearable scenarios, high moisture permeability allows porous interfaces to resist bodily fluids, improving the user comfort while mitigating potential delamination that could degrade the sensing performance of 3D flexible electronics. For a porous matrix with φ = 50% and 2-mm thickness, the diffusion channel networks formed by the open micropores facilitate a moisture permeability of ~715 g/m² per day (Fig. 1H and fig. S20), which aligns with the standard for medical gauze bandages (~500 g/m² per day at 23° ± 2°C and 50% relative humidity). The enhancement in elastic stretchability was demonstrated by an encapsulated helical mesostructure, which presents a 3.25-fold increase at 70% porosity compared to the case of solid encapsulation (Fig. 1I).

A radar chart of performance metrics (Fig. 1J) can be constructed based on the measured results in Fig. 1 (F to I), highlighting the superior properties of porous interface (φ = 50%) compared to solid interface for 3D flexible electronics. In particular, the porous interface endows 3D electronic devices with light weight, improved softness, high elastic stretchability and moisture permeability, superior thermal, and impact buffering, which are beyond the capability of commonly used solid interfaces. Note that the elastic stretchability varies across different 3D structures, with 1.8-, 2.6-, and 18.7-fold enhancements achieved in tent-shaped, the helical, and cage-shaped mesostructures, respectively. Besides, the incorporation of advanced fabrication techniques enabled the creation of different types of micropores within the interface, such as open porous networks formed by sacrificial templating and freeze drying, as well as closed micropores using a foamed silicone with chemically blowing agent (text S2). These features greatly enhance the versatility of the porous design and its adaptability across diverse 3D flexible electronics and application scenarios.

Deformation mechanisms of 3D mesostructures in porous matrixes

A mechanics model was established on the basis of 3D finite element analysis (FEA) to capture complex interactions between 3D mesostructures and porous matrixes, allowing precise predictions of their elastic stretchabilities under in-plane tensile loadings. Figure 2A shows the computational results for 3D PI/Au/PI helical ribbons in solid (φ = 0%) and porous (φ = 50%) matrixes, where the plastic yielding can be observed in half of a certain cross section of Au layer when the applied stretching strains (ε_apl) reaches 20 and 52%, respectively. Here, the encapsulation matrix material is selected as Ecoflex 00-20 with an elastic modulus of ~80 kPa, and the thicknesses of the top PI layer, middle Au layer, and bottom PI layer are ~5.0, ~0.3, and ~5.0 μm, respectively. Note that randomly distributed polyhedral micropores are generated using the Voronoi algorithm with controlled porosity (φ = 50%), average pore size (D = 350 μm), and wall thickness (T_wall = 91 μm), in reasonable alignment with experimental observations (fig. S21).

Fig. 2. Deformation mechanisms of 3D PI/Au/PI microribbon embedded in porous matrixes.

(A) Computed 3D configurations of helical ribbons in solid (top) and porous (bottom) matrixes undergoing 20 and 52% uniaxial stretching, respectively. The color of the helical structure denotes the Mises stress of the metal layer normalized by the yield strength (σ_y) of the gold. (B) Cross-sectional (left and middle) and top (right) views of deformed configurations and strain distributions in the localized region of the helical ribbon and porous matrix (with a high porosity φ = 80%). (C) Computed 3D configurations of helical ribbons overlaid with normalized Mises stress contours under different stretching strains (ε_apl) and Micro-CT reconstructed results. (D) Ratio of the width in the dangerous cross section of the helical ribbon where the Mises stress of the metal layer exceeds the yield stress. (E) Measured resistance changing rate (∆R/R) of the Au layer versus the cyclic stretching number (N) of helical ribbons in solid and porous matrixes. (F) Distributions of cross-sectional average Mises stress (σ_avg) on helical ribbons in solid and porous matrixes and free state. (G) Elastic stretchability versus the averaged cell size of the porous matrix (φ = 50%) used for the encapsulation with a base material modulus E_φ = 80 kPa. Here, two solid matrixes (φ = 0%) with moduli of E₀ = 80 kPa and E_φ^eff = 25.9 kPa were included for comparison. Note that E_φ^eff is the effective modulus of the porous matrix with φ = 50% and E_φ = 80 kPa by assuming it as a homogeneous material. (H) Contour plot of the elastic stretchability of helical ribbons as a function of average pore diameter (D) and porosity (φ) for the assembly strain ε_asm = 33% and ribbon width w = 350 μm. Photo credit: Z.S. and X.C., Tsinghua University.

In the condition of conventional encapsulation using a solid matrix, the notable difference in tensile stiffness between the soft matrix material and embedded helical ribbons leads to a rapid increase in stress levels and subsequent plastic deformations (fig. S22). In the case of a porous matrix, the micropore networks substantially reduce the tensile stiffness of the soft matrix and only impose discrete deformation constraints on helical ribbons. Under the uniaxial tensile loadings, the thin pore walls interact with helical ribbons, inducing bending deformations of pore walls and localized shell wrinkling of helical ribbons, thereby lowering the overall stress level and notably enhancing the elastic stretchability (Fig. 2B, fig. S23, and movie S2). Micro–computed tomography (Micro-CT) was adopted to capture the deformed configurations and localized wrinkling of helical ribbons in the porous matrix (text S3), which agree well with FEA predictions (Fig. 2C and fig. S24). A wrinkle feature emerged at around 30% tensile strain and fully developed to form a region of stress concentration at 52% tensile strain. The localized wrinkling deformations in the helical ribbon are primarily governed by the porosity (φ) and average cell size (D) of the porous matrix, and distinct deformed configurations with different wrinkle patterns were characterized through FEA (fig. S25, left). The deformation constraints on helical ribbons decrease with increasing the porosity and the average cell size, resulting in a transformation of wrinkle patterns from the localized shell wrinkles with small wavelengths to global ones with much larger wavelengths (fig. S25, right).

Figure 2D presents the width ratio (ω) exceeding the yield stress in the dangerous cross section (fig. S26) of helical ribbons under different stretching strains, with the error band arising from the adoption of 10 randomly generated encapsulations in FEA, each having the same porosity (φ = 50%), averaged cell size (D = 350 μm), and wall thickness (T_wall = 91 μm) but varying micropore networks. For a width ratio (ω) of 50%, the SE remains below 9%, suggesting that the specific distribution of the micropore networks has a minor effect on the FEA results under given values of parameters (φ, D, and T_wall). Cyclic tests were carried out to measure the elastic stretchability of helical ribbons in solid and porous matrixes (Fig. 2E and fig. S27), with a fixed loading strain rate of 5% per second (text S4). On the basis of the measured relative resistance change more than 10,000 loading cycles, the helical ribbons in porous matrix endure the full cycles at a strain amplitude of 50% and become an open circuit after 6000 cycles at a higher strain amplitude (ε_apl = 55%). In contrast, the helical ribbons in solid matrix only endure a much lower strain amplitude (ε_apl = 18%) and the electrical connection fails at ε_apl = 20%. These experimental results are in good accordance with the FEA results at ω = 50%, which supports the prediction of elastic stretchability using the half-section yielding criterion proposed by the groups of Huang and Rogers (77–79). The deformation behaviors of arc-shaped ribbons were also analyzed through a combination of experimental and FEA analyses (figs. S28 and S29), which also adhere to the half-section yielding criterion, providing further evidence of elastic stretchability enhancement using the porous encapsulation. Figure 2F compares the distributions of cross-sectional average Mises stress (i.e., mean value of the Mises stress over a cross section, σ_avg) along the helical ribbons in three conditions (i.e., solid encapsulation, porous encapsulation, and free state). Embedding in the solid matrix leads to the maximum stress level surpassing the yield stress (σ_y), accompanied by multiple peaks that indicate pronounced stress concentrations. By contrast, embedding in the porous matrix produces a substantially lower and smoother stress profile, approaching that of the free ribbon in the arc length range of 0.4 to 0.7.

Systematic computational analyses indicate that the elastic stretchability of the assembled helical ribbon is primarily governed by the averaged cell size (D) and porosity (φ) of the matrixes (Fig. 2, G and H). Under a given porosity (φ = 50%) and base material (Ecoflex 00-20, Young’s modulus E_φ = 80 kPa), the elastic stretchability increases from ~31 to ~52% as the averaged cell size grows from 50 to 350 μm, owing to reduced mechanical constraints from the soft matrix on the ribbon. However, as the cell size further increases (e.g., D > 400 μm), the resulting thick cell walls induce stress concentrations in the embedded ribbon segments, which compromises the overall mechanical performance (Fig. 2G). Two additional cases are considered in Fig. 2G: (i) helical ribbons embedded in solid matrix with Young’s modulus E₀ = 80 kPa and (ii) helical ribbons embedded in solid matrix with Young’s modulus E_φ^eff = 25.9 kPa, which corresponds to the effective modulus (E_φ^eff) of the porous material when approximated as a homogeneous medium. Note that the stretchability of helical ribbons in porous matrix is much higher than the counterparts in case i and also surpasses that in case ii when D > 50 μm. This observation indicates that the enhancement of the elastic stretchability arises not only from the stiffness reduction of soft matrix but also from the microstructural buckling in the ribbons and free pore space that alleviates stress concentrations. Figure 2H and fig. S29 present contour plots of the elastic stretchability in terms of the averaged cell size (D) and porosity (φ), as well as the assembly strain (ε_asm) and dimensionless width (w/L), respectively. With fixed D, a higher porosity and a larger assembly strain typically correspond to a higher elastic stretchability.

Incorporating the porous design into 3D mesostructures with more complex geometries could offer even higher enhancements of elastic stretchability when compared to solid design. Taking the 3D cage-shaped mesostructure as an example, the biaxial elastic stretchability reaches ~28% in the porous matrix (φ = 50%) based on the results of FEA and cyclic tests (figs. S30 and S31), representing an 18.7-fold of magnitude enhancement over that of the solid encapsulation. Such a significant enhancement can be attributed mainly to the instability of thin pore walls (fig. S30, D and E). As evidence, FEA results reveal that merely reducing material modulus in a solid matrix leads to only a ~1.1-fold and a ~1.2-fold increase in biaxial and uniaxial elastic stretchability, respectively (fig. S30F). In the solid matrix, the chiral geometry of the cage-shaped mesostructure induces twisting deformations in the six curved ribbons under tensile loadings, as observed in the distributions of rotational displacements in FEA results (fig. S30E, middle and right). However, because of the strong constraint imposed by the solid matrix, substantial shear stress accumulates, causing stress concentrations in the twisting regions and lastly leading to failure of the metallic layer. In contrast, the micropores in the porous matrix mitigate the twisting deformation by promoting the bending and buckling of thin pore walls, which helps alleviate stress concentrations in the metallic layer (fig. S30E, left). Figure S30G analyzes the influence of the arc angle of the 2D precursor structure on the elastic stretchability, highlighting the critical role of twisting deformation in governing the constrained deformation behaviors. As the arc angle increases, the chiral characteristics of the assembled 3D cage-shaped mesostructure become more pronounced, resulting in a more pronounced twisting deformation in solid matrix, thereby reducing the elastic stretchability. FEA results show that the porous design notably enhances the elastic stretchability of cage-shaped mesostructures across a broad range of arc angles (90° ≤ θ ≤ 180°).

A 3D multimodal sensing system with engineered porous encapsulation

Engineering the spatial porosity distribution in the matrix allows the development of a wearable 3D multimodal sensing system capable of decoupled measurement of compressive force, tensile strain and localized temperature. Figure 3A and fig. S32 illustrate the 2D precursor design of the multimodal sensing system, featuring a circumferentially arrayed eight-arc pattern (region I), a straight ribbon with one middle bonding site (region II), and three straight ribbons each with a central circular plate (region III). Micrometer-width (~8 μm) gold (Au) traces were strategically arranged in regions I and II and sandwiched by a thin (~2 μm) top-layer PI and a relatively thick (~5 μm) bottom-layer PI, allowing a high sensitivity to structural deformations after the 3D assembly. A pair of semicircular gold electrodes were patterned on the central plate of Region III, with two square openings etched for the integration of negative temperature coefficient (NTC) thermistor. Leveraging the controlled compressive buckling deformations, the 2D precursor structure was assembled into a 3D architecture (Fig. 3B and Materials and Methods), encompassing a 3D cage-shaped force sensor, a bi-arc strain sensor, and three table-shaped temperature sensors. Quantitative mechanics simulations show that the metallic layer undergoes elastic deformations under typical loadings of stretching (e.g., 25%) and twisting (e.g., ~60°), indicating a high mechanical reliability of the 3D multimodal sensing system (fig. S33).

Fig. 3. Design and assembly of a bioinspired 3D multimodal sensing device with engineered porous interface.

(A) Illustrations of the PI/Au/PI-trilayered 2D precursor structure of the multimodal device (top), and optical images (bottom) of the Au grid patterns in the force (region I) and strain (region II) sensors and semi-circular Au electrodes for the integration of NTC thermistors (region III). Scale bars, 200 μm. (B) Computed 3D configurations of the assembled multimodal sensing device (top) and distributions of the maximum principal strain (bottom) in the Au layers of three flexible sensors. (C) Illustrations of the mechanoreceptors in human skin (left) and the skin-inspired multilayer construction of three types of 3D flexible sensors with varying heights (right) for decoupled measurements of normal force, tensile strain, and temperature. (D) Encapsulation illustrations of the multimodal sensing device with an engineered porosity distribution, featuring localized solid (φ = 0%), cavity (φ = 100%), and porous (φ = 50%) matrixes for the force sensor, strain sensor, and thermal sensors, along with bottom-view optical images of the force and strain sensors acquired through the substrate, respectively. (E to G) Illustration of the engineered encapsulations (top) of three types of 3D flexible sensors and relative resistance changes (bottom) in response to stepwise out-of-plane compressive loadings (E), in-plane uniaxial stretching (F), and cyclic thermal loadings (G). (H) Decoupled measurements of the tensile strain and normal force in complex loading conditions. Photo credit: Z.S. and X.C., Tsinghua University.

Among the various mechanoreceptors in human skin, the Merkel cells in the epidermis detect static pressure and fine touch; the Ruffini endings in the dermis sense skin stretch and sustained force; and the free nerve endings respond to temperature and pain (Fig. 3C). The depth-dependent distribution forms a comprehensive sensory network for mechanical and thermal perception, inspiring the development of a multilayer construction of three types of 3D flexible sensors with varying heights (Fig. 3D). Note that the temperature sensor was not incorporated in our previous work (39). In this work, the decoupled measurement of compressive force, in-plane tensile strain, and localized temperature was enhanced by engineering the porosity distribution (Fig. 3D). By adding a drop of molten polyethylene glycol around the bi-arc strain sensor, a microcavity (φ = 100%) can be formed after removing the soluble templates, effectively isolating the strain sensor from out-of-plane compression. The near-closed configuration of the 3D cage-shaped force sensor blocks soluble microparticles, enabling a localized solid matrix (φ = 0%) beneath the force sensor and minimizing the influence of in-plane strain on force sensing. Under out-of-plane loadings, the relatively high (~950 μm) 3D cage-shaped sensor first detects the compressive force, and the solid silicone material (Ecoflex 00-20) provides stable mechanical support across a broad force range. Other regions, including the stretchable interconnects and temperature sensors, follow a standard porous design (φ = 50%) with sugar sacrificial microparticles with prescreened diameters.

We carried out characterizations of the sensing performance of the 3D flexible system, including both the single physical signal (Fig. 3, E to G, and figs. S34 and S35) and multisignal (Fig. 3H) measurements. Following the increase of the compressive force from 1 to 4 N in 0.5-N intervals, the relative resistance change of the 3D force sensor increased from ~0.16 to ~0.64% in ~0.08% intervals, with a sensitivity of 0.16%/N. Similar electrical responses can be observed in the unloading process, showing a moderate hysteresis (~0.070) (Fig. 3E and fig. S35A). In a stepwise testing on strain sensing performance, uniaxial tensile loading and unloading were applied to the porously encapsulated electronic system, with the strain ranging from 0 to 20% in 4% intervals. The relative resistance change of the bi-arc strain sensor spans from 0 to 0.38% with an interval of ~0.08% (Fig. 3F and fig. S35B). On the basis of the measured results, the strain sensor exhibits a good linear response with a gauge factor more than ~1.9% and an acceptable hysteresis (~0.079). Owing to the high thermal sensitivity (2%/°C) of the NTC thermistor, the 3D sensor in Fig. 3G measured the localized temperature changes near the top surface of the porous encapsulation over 20 cycles of heating and cooling. Note that a digitally controlled heating plate was used to regulate the device temperature. This setup allows accurate setting of the target temperature at different stages, but the heating and cooling rates cannot be precisely programmed. Figure 3H presents measured tensile strains and compressive forces during a sequential loading process: (i) applying compressive force, (ii) applying in-plane tensile strain, (iii) uniaxially stretching under a fixed compressive loading, and (iv) applying a compressive loading under a fixed tensile strain. In addition, fig. S36 demonstrates four different measurements with complex loading conditions, involving multilevel compressive loadings and multilevel tensile strains. These results provide validation of the decoupled sensing capability of compressive force and tensile strain.

A 3D closed-loop wound administration electronic patch (3D-CURE Patch)

Scald burns, caused predominantly by hot liquids, represent one of the most common forms of thermal injuries globally, accounting for ~60% of pediatric burn injuries (72). The scald wounds particularly affect vulnerable populations such as children and the elderly, which contribute evidently to patient morbidity and health care burden worldwide. These wounds present critical clinical challenges, including high infection risk with compromised skin barrier integrity and weakened antimicrobial defenses, abundant exudate promoting delayed healing, and rapidly progressing wound areas within the first 24 to 72 hours complicating prompt intervention. Clinical assessment typically relies on periodic observation of indicators such as local temperature changes and tissue condition (e.g., swelling or contraction). However, these evaluations are usually performed intermittently (e.g., during dressing changes), which may delay the detection of early-stage deterioration and could introduce secondary injury in dressing replacement.

Integration of the above 3D multimodal sensing system with a flexible heating circuit allows development of a wearable 3D electronic patch for the closed-loop scald wound administration, serving as a biomedical application of the porous integration strategy. The 3D closed-loop wound administration electronic patch (3D-CURE Patch) consists of a PI-sandwiched, serpentine-shaped heating circuit, a porously encapsulated 3D multimodal sensing system, and cefazolin sodium–loaded agarose gel within the micropores (Fig. 4A). The 3D-CURE Patch is connected to a FPCB circuit (Fig. 4, B and C), which includes the control, communication, power supply, and feedback modules (Fig. 4C and figs. S37 to S40), with the operation logic illustrated in Fig. 4D. In an SD-rat model with a deep second-degree scald, the 3D-CURE Patch was conformally attached to the circular scald wound at the dorsal skin, without interfering with normal physiological activities of the SD-rats (figs. S41 and S42 and movie S3). Note that no observable false-positive responses were detected during normal activities throughout the treatment period. In addition, the 3D-CURE Patch (~8.8 g) is substantially lighter and smaller than representative commercial wearables (table S1), and a 400-mAh Li battery can support the intended sensing, communication, and on-demand heating functions for over 3 days under the current operating protocol.

Fig. 4. Design, control, and characterization of a 3D closed-loop wound administration electronic patch (3D-CURE Patch).

(A) Illustrations of the functional layers in the 3D-CURE Patch for the scald wound of a SD-rat model, including the heating circuit, 3D multimodal sensing device, and porous matrix loaded with agarose gel-drug composites. (B) Illustrations of the feedback loop of the 3D-CURE Patch on an SD-rat model, consisting of the monitoring and therapeutic phases. The multimodal sensing device monitors infection-related physiological symptoms (e.g., inflammatory edema and infectious swollen) (monitoring phase) and activates the heating circuit to induce the phase transition of the agarose gel, enabling controllable release of therapeutic drugs to the wound area (therapeutic phase). (C) Block diagrams that illustrate the working principle of the closed-loop wound care system, enabling real-time monitoring of the scald wound and providing real-time therapeutic intervention. MCU, microcontroller unit; GPIO, general-purpose input/output; ADC, analog-to-digital converter. (D) Flowchart of the operational logic. (E) Optical (left) and infrared (right) images of the 3D-CURE Patch in an activated state performing the therapeutic intervention. Scale bars, 20 mm. (F) Measured temperature by the thermal sensor following the activation of the heating circuit in the 3D-CURE Patch with porous and solid interfaces. (G) Measured cumulative mass of cefazolin sodium that is released from the agarose gel over four controlled heating cycles. (H to J) Temporal recordings of sensor signals (H) under two typical wound symptoms: inflammatory edema (I) and infectious swollen (J). Inset fluorescent images in (H) confirm drug release on the scald wound using the 3D-CURE Patch, triggered by force/temperature changes induced by these two typical symptoms in (I) and (J). Scale bars, 10 mm. Photo credit: Z.S. and X.C., Tsinghua University.

During an operational cycle, the 3D multimodal sensing system continuously monitors the compressive force, tensile strain, and localized temperature of the scald wound at a predefined frequency (150 ms per cycle). This enables rapid acquisition of physiological signals, which is critical during the early stages of scald wound progression. The mini-LED indicator illuminates once the relative resistance changes exceed the preset thresholds for the three different types of sensors, providing a visible alert for potential inflammation (e.g., localized swelling and temperature rise). The green light and blue light activate when the relative resistance changes of the force/strain sensors exceed 0.15%, respectively, while the red light turns on when the temperature rises above 1.5°C. Note that the alert signals are collected and assessed over a predefined period (~20 s) before activating the heating circuit, so as to mitigate transient fluctuations caused by routine motion or environmental disturbances. After activating the heating circuit and maintaining for a preset period (~160 s), the embedded agarose gel softens into a semi-solid state at ~50°C due to a phase transition, enabling controlled drug release to the scald wound. Following each treatment session, the 3D-CURE Patch switches to a low-power mode and remains on standby for 2 hours to allow the drug to take effect. Subsequently, the multimodal sensing system could evaluate the wound condition by analyzing the swelling and temperature signals to determine the required cycles of drug administrations (Fig. 4, B and D). Note that the sensing cycle interval of the 3D-CURE Patch can be readily extended by adding the cycle time in the control module to meet different clinical requirements.

Figure 4E presents the optical and infrared (IR) images of the 3D-CURE Patch with the heating circuit activated. This response is triggered when the compressive force and temperature signals exceed their normal ranges, indicating inflammation progression at the scald wound. Notably, the mechanical and thermal trigger thresholds were defined based on the sensor-response deviations from the healthy state, after different wound states had first been identified according to routine clinical criteria. The temperature on the top surface of the 3D-CURE Patch rises rapidly under a heating power of ~0.38 W and stabilizes within the range of 65° to 70°C. The porous encapsulation layer between the heating circuit and the scald wound surface serves as a thermal barrier, effectively slowing the heat transfer and ensuring controlled thermal distribution (Fig. 4F). Specifically, the porous interfacial structure helps improve the uniformity and controllability of drug release by acting as a thermal buffer that interrupts direct heat-conduction pathways and suppresses localized hotspots, as evidenced by the IR images and the temperature profiles along the midline cross section (fig. S43). During the heating process (~160 s), the temperature near the scald wound gradually rises and stabilizes between 47° and 52°C for ~160 s. This sustained heating is optimal for triggering the phase transition of the agarose hydrogel, enabling controlled drug release while preventing excessive heat exposure to the wound. In contrast, the 3D-CURE Patch with a solid encapsulation leads to rapid overheating with the temperature near the wound exceeding 55°C within ~40 s, posing a great risk of the secondary burn injury (Fig. 4F). The controlled drug release of the 3D-CURE Patch was quantitatively measured using a ultraviolet-visible spectrophotometer (Fig. 4G and text S5). Specifically, the heating circuit was initially activated for 3 min, resulting in a cumulative drug release of ~200 μg, after which no notable increase in drug mass was observed upon deactivation. Subsequently, three drug release cycles were performed, with a 60-s heating period, followed by a 120-s deactivation. The incremental drug mass during three cycles was 22, 17, and 16 μg, respectively. The measured cumulative drug release profile offers insights for optimizing the heating duration, allowing customization to accommodate different wound types and treatment requirements. Note that the interface temperature can be switched on demand by adjusting the pulse-width modulation duty cycle of the heating output, which provides an additional means to improve thermal safety and controllability. For future clinical translation, lower transition temperature gel systems could also be adopted to achieve the same function at a reduced triggering temperature, together with more comprehensive evaluation based on interface temperature-time profiles.

In an SD-rat model with deep second-degree scalds, two typical clinical symptoms were identified by visual inspection and gentle palpation according to routine clinical criteria (fig. S44): (i) edema after scald injury, manifested as wound elevation/swelling; and (ii) a local infectious inflammatory response in the wound area, characterized by localized erythema and an obvious increase in local skin temperature. The force and strain sensors effectively monitored the swelling deformation at the wound site and detected a remarkable increase in compressive force accompanied by subtle variations in strain for symptom I [Fig. 4, H (left) and I]. For symptom II, the temperature sensors precisely tracked the rise in surface temperature of the scald wound and recorded an increase over 1.5°C [Fig. 4, H (right) and J], indicating the onset of inflammation. In both cases, the heating circuit in the 3D-CURE Patch was activated to trigger the release of cefazolin sodium particles for antibacterial and anti-inflammatory treatment (Fig. 4J). The inset fluorescence images in Fig. 4H show the drug particles remaining on the wound surface after removing the 3D-CURE Patch, providing further validation of the on-demand drug release.

Animal experiments were divided into three groups: (i) blank group, which received no treatment after the scald injury; (ii) positive group, which received conventional antibiotic treatment by applying cefazolin sodium powder to the scald wound every other day; and (iii) device group, which underwent close-loop antibiotic treatment using the 3D-CURE Patch device. A standardized scald model was established on the dorsal skin of 6 SD-rats in each group, following the same procedure, with the day of model creation designated as day 0. Changes in wound area were monitored to evaluate the progression of scald wound healing throughout the skin regeneration process.

Localized swelling was observed around the scald wound in more than 60% of SD-rats between day 2 and day 3, and all of wounds in the control groups exhibited varying levels of local inflammation, characterized by visible redness around raised border of the wound in blank groups. Figure 5A compares the progression of scald wound healing across three groups, demonstrating that the 3D-CURE Patch effectively suppressed inflammation and accelerated the skin regeneration process. On day 4, the blank group exhibited unscabbed wounds with significantly elevated local skin temperatures, indicating the sustained inflammation [Fig. 5, A (top) and B (left)]. In the positive group, scab formation was observed along with a slight increase in wound temperature, suggesting that the inflammation remained despite the antibacterial and anti-inflammatory effects of cefazolin sodium [Fig. 5, A (middle) and B (middle)]. In contrast, the device group showed well-formed scabs along with a lower wound temperature, highlighting the effectiveness of the 3D-CURE Patch [Fig. 5, A (bottom) and B (right)]. The continuous monitoring of inflammation and prompt intervention enable the 3D-CURE Patch to suppress inflammation at an early stage, thereby accelerating the wound healing. The wound healing rate, defined by the wound area ratio to initial state (Materials and Methods), serves as a quantitative measurement of the skin regeneration (Fig. 5C). On day 4, the wound healing rate in the device group was 3.1- and 1.6-fold higher than the blank group and positive group, respectively; On day 14, it was 1.57- and 1.36-fold higher than the blank group and positive group, respectively. Comparison across those three groups reveals that the wound care with on-demand treatment using the 3D-CURE Patch effectively inhibits wound infection and accelerates the healing process, particularly in the early stages (e.g., before day 4).

Fig. 5. Evaluation of treatment effectiveness in an SD-rat model with an infected scald wound using the 3D-CURE Patch.

(A) Optical images of the skin regeneration process in an SD-rat model with deep partial-thickness scald wounds under three different treatment approaches: (i) blank group: natural healing without treatment; (ii) positive group: typical application of cefazolin sodium powder to the scald wound every other day; and (iii) device group: treatment with the 3D-CURE Patch. Scale bars, 10 mm. (B) Infrared images of scald wounds in the above three groups on day 4, showing an increase in skin temperature that confirms the onset of the inflammatory response. Scale bar, 10 mm. (C) Quantitative analysis of wound closure to evaluate the therapeutic efficacy of the 3D-CURE Patch. (***P < 0.001). (D to F) Hematoxylin and eosin–stained histological longitudinal sections of scald wounds at post-operative day 21, evaluating skin regeneration outcomes across three groups: blank group (D), positive group (E), and device group (F). (BV, blood vessels; HF, hair follicle; SG, sebaceous glands). Scale bars, 1 mm (top panels) and 200 μm (bottom panels). Photo credit: Z.S. and X.C., Tsinghua University.

Figure 5 (D to F) and fig. S45 provide hematoxylin and eosin (H&E)–stained histological longitudinal sections of the scald wounds on day 21 postoperation. Experimental results show improved skin regeneration in both the positive and device groups, as evidenced by the presence of mature epithelial structure with several blood vessels in Fig. 5 (E and F). In contrast, these structures were scarcely observed in the longitudinal sections of the blank group (Fig. 5D). Compared to the positive group, the device group exhibited a more obvious improvement in skin regeneration, with a higher density of newly regenerated blood vessels, hair follicles, and sebaceous glands. Note that the scar appeared evidently in the wound section of the control groups, while the scar was obviously smaller and more akin to the adjacent uninvolved skin in the device group. These findings validate the superior skin regeneration capability of the 3D-CURE Patch, highlighting its promising potential for advanced wound care treatments.

DISCUSSION

The design concepts and fabrication techniques reported here allow the development of a multifunctional porous interface that bridges 3D flexible electronics with skin while offering comfort, mechanical robustness, and electrical reliability during continuous wear. In particular, embedding complex 3D electronic mesostructures within engineered microporous networks renders substantially enhanced elastic stretchability and multifunctional capabilities (such as moisture-penetrability, thermal buffering, impact mitigation, and drug delivery) inaccessible to conventional solid encapsulation and 2D epidermal electronics (text S6 and table S2). Combined computational and experimental analyses revealed that stiffness reduction and localized pore-wall-shell buckling synergistically underpin the improved elastic stretchability. By engineering the porous encapsulation at different spatial locations, a 3D multimodal sensing system capable of simultaneous measurements of compressive force, tensile strain, and temperature was demonstrated. This multimodal sensing system allowed development of a 3D closed-loop wound administration electronic (3D-CURE) patch that achieved multimodal, real-time scald wound monitoring and on-demand therapy, markedly accelerating healing in vivo. These findings establish porous interface as a versatile platform for next-generation wearable electronics in personalized healthcare and beyond.

MATERIALS AND METHODS

Fabrication of porous encapsulation using sacrificial templates

The porous encapsulation process for 3D assembled electronic device started with enclosing the device with a 3D printed mold (Young’s modulus: ~2.5 GPa; wall thickness: 1 mm; wall height: ~130% higher than the maximum height of 3D device). Water-soluble sugar microparticles with a relatively uniform size distribution, prepared by grinding and sieving, were loaded into the 3D printed mold. Applying slight in-plane vibration facilitated adaptive particle packing within complex 3D geometries, enabling uniform filling of mold cavities when the feature size was larger than the particle size. This process produced a densely packed template conformal to the complex 3D geometries of the flexible electronic device (fig. S2, step 1). Then, gently pouring a silicone elastomer precursor (Ecoflex 00-20, Smooth-On, USA) over the sugar-filled region, followed by a vacuum degassing, ensured thorough infiltration of the elastomer precursor into the interstitial voids between the microparticles and around the 3D electronic device (step 2). After oven curing at ~60°C for 4 hours, the 3D printed frame was carefully demolded. Immersing the encapsulation in deionized water for over 72 hours, with use of a mild vacuum (~10 Pa), allowed efficient dissolution of the sugar microparticles (step 3 and text S7), resulting in a porous elastomer structure conformally encapsulating the 3D flexible electronic device. Last, the porous encapsulation was dried in a convection oven at ~60°C for ~12 hours to remove residual moisture and stabilize the materials/structures for further use (e.g., the encapsulated cage- and table-shaped mesostructures in fig. S46). Note that the metal interconnects in 3D flexible devices exploited a sandwich design, where the metal layer is fully encapsulated between two PI layers (figs. S47 and S48). This configuration provides effective protection against water penetration, enabling high electrical stability even under extended immersion (fig. S49).

Fabrication of 2D precursor structures of 3D flexible structures and electronics

The fabrication of 2D precursor structures started with spin coating a sacrificial layer of polymethyl methacrylate (PMMA; ~200 nm in thickness) and then a layer of PI (~2 μm in thickness) on a silicon wafer (fig. S47, step 1), followed by thermal curing in a convection oven at 250°C for 2 hours (step 2). Photolithography using AZ 5214 photoresist defined the layout of the circuit in the flexible electronic devices (steps 3 and 4), followed by sequential deposition of Cr (~5 nm in thickness) and Au (~80 nm in thickness) using electron beam evaporation (step 5). Then, a lift-off process enabled patterning of the metal interconnects (step 6). Spin coating another layer of PI (~5 μm in thickness), followed by a similar thermal treatment, led to the formation of the electrically insulated construction (i.e., PI/Au/PI) (step 7). The 2D precursor geometry of the 3D flexible device was then defined through the photolithography (step 8), deposition of a layer of Cu (~150 nm in thickness) serving as hard mask (step 9), lift-off (step 10), and reactive-ion etching of top and bottom PI layers (step 11). The Cu hard mask was removed using wet etching (step 12), and the PMMA layer was dissolved in acetone solution (step 13), allowing a smooth transfer of 2D precursor structure from the silicon wafer to the PVA soluble tape (step 14).

Compressive assembly of the 3D multimodal sensing system

The assembly of the 3D multimodal sensing system began with bonding three NTC thermistors (SMD0201; 10 kΩ; TDK, Japan) onto the metal pads of thermal sensing units using the silver conductive adhesive, followed by silicone encapsulation of the thermistors for waterproofing and mechanical protection. A soft silicone elastomer substrate (Ecoflex 00-20; Smooth-On, USA) that was biaxially stretched to a prestrain (ε_pre = ΔL/L, where ΔL is the increase in length and L is the initial length; ε_pre = 33.3% in the case of 3D multimodal system) and then exposed to ozone to generate a uniform coverage of surface hydroxyl groups served as a platform that guided the mechanical assembly. Note that the prestrain is different from the assembly strain, which is defined as ε_asm = ΔL/(L + ΔL). Transfer printing of the 2D precursor structures, whose bonding sites are selectively deposited with SiO₂ assistant with a steel mask, allowed its integration with the prestrained substrate. The treated surface enabled strong and spatially selective adhesion through covalent linkages formed via condensation reactions at hydroxyl-terminated silicon regions. Comparatively weak van der Waals forces dominated interfacial interactions in the rest regions. Releasing the substrate to its original shape induced large compressive forces on the precursor structures through the bonding sites, which initiated a controlled buckling process that lifted the weakly contact regions of the structures out of the substrate surface, completing the process of 3D assembly (fig. S48).

Encapsulation of 3D multimodal sensor system with engineered porosity distribution

A sequential encapsulation process was used to achieve the spatially distributed porosity in the 3D multimodal sensing system. This encapsulation started with dispensing the molten polyethylene glycol (PEG; melting point of ~60°C) onto the arc-shaped strain sensor. Driven by the capillary force, the PEG solution rapidly infiltrated narrow slits between 3D micro-arcs and the elastomeric substrate and subsequently solidified upon cooling. A standard porous encapsulation process was then carried out for the 3D multimodal sensing system using a sacrificial sugar template with an average microparticle size of ~350 μm. Owing to the nearly closed 3D configuration (the maximum gap of ~300 μm between two adjacent ribbons) of the cage-shaped force sensor, sacrificial particles were blocked outside while allowing the infiltration of silicone precursor. Following the dissolution of sugar microparticles and PEG material, three distinct encapsulation architectures were formed: a cavity encapsulation (φ = 100%) around the arc-shaped strain sensor, a solid encapsulation (φ = 0%) beneath the cage-shaped force sensor, and a porous encapsulation (φ = 50%) in thermal sensors and other regions.

Fabrication of the 3D-CURE patch

Fabrication of the 3D-CURE Patch started by infusing molten agarose hydrogel premixed with drug into the porous encapsulation layer of the engineered 3D multimodal sensing system via vacuum-assisted infiltration, ensuring uniform filling of the porous structure. Subsequent controlled cooling induced gelation of the hydrogel within the porous structure, enabling stable retention of the drug-loaded gel in the encapsulation layer. Last, a flexible heating circuit, consisting of a copper conductive layer sandwiched between PI films and interconnected by serpentine shaped ribbons, was gently laminated onto the bottom surface of the 3D multimodal sensing system, completing the assembly of the patch.

Finite element analysis

3D FEA based on the commercial software (ABAQUS) allowed predictions of constrained deformation modes, strain/stress distributions, and the elastic stretchability of the encapsulated 3D flexible electronic devices. The 3D mesostructures and porous encapsulation were modeled with four-node shell elements (S4R), and the elastomeric substrates and solid encapsulation adopted the eight-node 3D solid elements (C3D8RH). The geometric nonlinearity was taken into account in the FEA, and mesh refinement was applied to ensure computational accuracy. Given the presence of large nonlinear deformations and complex contact interactions between the encapsulation material and 3D mesostructures, an explicit solver was used in the FEA, and its convergence reliability was validated by an energy analysis (fig. S50). Linear elastic constitutive models were adopted for PI and gold layers, where the elastic modulus (E) and Poisson’s ratio (v) are E_PI = 2.5 GPa and v_PI = 0.34 for PI and E_Au = 79 GPa and v_Au = 0.44 for gold and a yield strength (σ_y) of 237 MPa [corresponding to ~0.3% yield strain (40, 58, 77)]. The Mooney-Rivlin constitutive model was exploited for the elastomer substrate and encapsulation material to account for their nearly incompressible behavior, where E_Ecoflex = 80 kPa and v_Ecoflex = 0.49. The first-order buckling mode obtained from the linear buckling analysis was added as initial geometric imperfection in the 2D precursor structures, which triggers the post-buckling deformations in the compressive assembly process. Note that the solid encapsulation beneath the cage-shaped force sensor was constructed based on the 3D configurations of the sensor, ensuring a node-to-node bonding interface between the 3D sensor and silicone encapsulation.

A transient FEA-based heat transfer model was used to analyze thermal conduction behaviors of porous and solid encapsulations. Four-node linear heat transfer elements (DC3D4) and four-node heat transfer quadrilateral shell (DS4) were applied to soft encapsulation and 3D electronic devices, respectively. Natural convection with an ambient environment was modeled by applying air-convection boundary conditions on all exposed surfaces of the devices. The mass density (ρ), thermal conductivity (λ) and heat capacity (C) were ρ_PI = 1200 kg·m⁻³, λ_PI = 1 W/ (m·K), and C_PI = 1090 J/(kg·K) for PI; ρ_Au = 19300 kg·m⁻³, λ_Au = 314 W/ (m·K), and C_Au = 129 J/(kg·K) for gold; ρ_Ecoflex = 1080 kg·m⁻³, λ_Ecoflex = 0.162 W/ (m·K), and C_Ecoflex = 1460 J/(kg·K) for Ecoflex 00-20; and ρ_air = 1.18 kg·m⁻³, λ_air = 0.026 W/ (m·K), and C_air = 4180 J/(kg·K) for surrounding air.

In vivo wound healing experiments

In vivo wound healing experiments were performed on the dorsal skin of male SD-rats. A total of 18 8-week-old SD-rats (average weight of ~300 g) were obtained from the Laboratory Animal Resources Center, Tsinghua University. All animals were housed individually under standard laboratory conditions: ambient temperature of 25° ± 1°C, a 12-hour light/12-hour dark cycle (lights on at 7:30 a.m.), and free access to food and water. All surgical procedures were conducted under general anesthesia induced by inhalation of isoflurane mixed with oxygen (1.5 to 3% isoflurane at a flow rate of 2 L/min), and the anesthetic depth was continuously adjusted based on the animals’ responses to ensure safety and adequate sedation.

After shaving and disinfecting, full thickness scald wounds with a diameter of 2 cm were created by using a custom-designed scalding apparatus. The scalding parameters were: probe temperature of 90°C, applied load of ~300 g; and probe-to-skin contact duration of 8 s. Immediately after injury, the wounds exhibited characteristic features of deep second-degree burns, including pale opalescent discoloration and a dry surface. Following injury induction, SD-rats were randomly assigned into three experimental groups (six rats per group), including (i) blank group: no treatment was administered post-injury; (ii) positive group: wounds were treated with cefamezin hydrogel applied every other day, starting on day 1 postinjury; and (iii) device group: rats received a closed-loop antibiotic treatment via the 3D-CURE Patch. To ensure stable device attachment throughout in vivo studies and prevent displacement caused by animal scratching, biting, or rubbing, the 3D-CURE Patch was further secured with a soft, breathable, transparent commercial medical dressing film after being placed over the scald wound. Moreover, the device was first zero-calibrated after being mounted on the rat, and throughout the entire testing period, it maintained stable sensing responses and overall functionality under the above fixation condition. Wound progression was evaluated and documented on the day of injury and at subsequent time points, where key indicators such as wound size, color, and scab formation were visually inspected and recorded. The percentage of wound closure was calculated by comparing the wound area (M_i) at each time point (day i) to the original wound area (M₀) on the day of modeling

$$M=(M_0-M_i)/M_0$$ (1)

All the animal experiments involved in this work were approved by the Animal Committee of Tsinghua University (approval no. 23-ZYH1.G24-1).

H&E staining and skin maturity quantification

On post-operation day 21, the full-thickness skin tissues encompassing the scald wound area were harvested from each rat, and these excised samples were immediately fixed in 4% paraformaldehyde for 72 hours at room temperature. Tissue dehydration was performed using sequential immersion in 15 and 30% sucrose solutions for a total duration exceeding 48 hours. Subsequently, the tissue samples were embedded in an optimal cutting temperature (OCT) compound and stored at −80°C until further processing. The embedded samples were cryo-sectioned into 10-μm-thick longitudinal slices using a freezing microtome (CM1860, Leica, USA). Representative sections spanning the wound center and surrounding healthy tissue were selected for histological analysis. H&E staining was performed following standard protocols to visualize tissue morphology. Stained sections were imaged using a digital slide scanning system (Axio Scan Z1, Zeiss, USA) for subsequent analysis of epidermal and dermal regeneration, inflammatory cell infiltration, and skin structural maturity.

Supplementary Materials

The PDF file includes:

Supplementary Text S1 to S7

Figs. S1 to S53

Tables S1 to S3

Legends for movies S1 to S3

References

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S3