Feed‐Draw Printing Enables Monolithically Integrated Flexible Sensors With High Interfacial Toughness and Wide Linear Range

ABSTRACT

Flexible sensors are essential for wearable electronics and robotic perception, yet their practical deployment is fundamentally constrained by the compromise between sensing sensitivity and mechanical durability, as well as the deficiency in achieving robust integration with other functional components. Here, we report a monolithic microcone capacitive sensor (MMCS) fabricated via a feed‐draw printing (FDP) strategy, which enables the direct fabrication of programmable microcone morphologies and in situ co‐curing across dielectric and electrode layers. This strategy generates continuous, covalently interlinked interfaces with interfacial toughness of up to 1 547 J m⁻², representing 3.97 times that of existing microcone counterparts. As a result, the MMCS exhibits outstanding durability, maintaining stable performance over 200 000 loading cycles. Meanwhile, the programmable microcone morphology offers on‐demand sensitivity tuning (reaching 0.29 kPa⁻¹), achieving a low limit of detection of 1 Pa and a wide measurement range from 0 to 450 kPa. Furthermore, the MMCS is monolithically printed into wearable bands for sports analytics, and within magnetic soft grippers for grasping perception. This work establishes a generalizable route for covalently bonded, sensitivity‐enhanced, and functionally integrated soft sensors, paving the way for high‐performance practical wearable and robotic interfaces.

A feed‐draw printing strategy enables monolithic microcone capacitive sensors with in situ co‐cured interfaces reaching 1 547 J m⁻² toughness and stable operation over 200 000 cycles. Programmable microcone morphologies offer on‐demand tunable sensitivity and a wide linear range, demonstrating seamless integration into wearable bands and magnetic soft grippers for robotic perception.

1. Introduction

Flexible sensors have become essential in emerging applications such as wearable health monitoring, sports analytics, and robotic manipulation. For real‐world applications, these devices should operate reliably under highly dynamic and mechanically demanding conditions [1, 2, 3]. For example, during high‐intensity athletic motion, wearable sensors are required to conform to rapidly deforming skin while withstanding high‐frequency impacts and shear forces. Even minor interfacial degradation causes signal drift or complete detachment under these conditions. Similar challenges also arise in robotic systems, where reliable grasping and manipulation require seamless mechanical coupling between actuators and sensing units to ensure stable force perception across unstructured environments. These demanding scenarios call for high sensitivity, mechanical robustness, and monolithic multi‐material integration [4]. However, addressing these requirements presents a significant challenge, for sensitivity enhancements often undermine robustness [5], and current fabrication strategies still struggle to ensure seamless integration of highly sensitive sensors with functional components to achieve system‐level stability.

To enhance sensitivity, one common strategy is introducing microstructures. For instance, foam‐like porous architectures can effectively reduce the effective Young's modulus between parallel electrodes, thereby enhancing sensitivity [6, 7, 8]. However, the single porosity feature often fails to achieve precise tunability. In contrast, well‐defined geometrical microstructures such as cubes [9, 10], cylinders [11, 12], domes [13, 14, 15, 16], pyramids [17, 18, 19], or cones [20, 21, 22] can provide predictable deformation patterns under compression, enabling controllable adjustment of sensitivity. However, most of these architectures rely on template‐based molding, which involves mold fabrication and demolding steps that increase manufacturing complexity and cost, while also hindering the formation of robust chemical bonds with the electrodes. Another strategy is incorporating conductive fillers into the dielectric matrix, avoiding complex molding while increasing permittivity and improving sensitivity [23]. Nevertheless, this method yields limited interfacial toughness enhancement, as the significant modulus mismatch between layers makes the interface prone to delamination under complex deformation [24]. In general, these two strategies usually prepare the dielectric and electrode separately and assemble them together afterward, where weak interfacial adhesion leads to unstable signals and premature failure during cyclic loading [5]. In real‐world scenarios, the weak interfacial toughness may emerge as a key bottleneck that limits both sensitivity optimization and device reliability.

Enhancing interfacial toughness is therefore pivotal for the reliability of soft devices. Conventional strategies, such as adhesives [25, 26, 27] and structural reinforcement [28, 29], primarily rely on physical bonding or mechanical interlocking. While initially effective, these methods often introduce materials with distinct mechanical properties, creating modulus mismatches that concentrate stress and accelerate fatigue failure under dynamic deformation [26]. To mitigate these mechanical discrepancies, chemical strategies, including surface activation [5, 30, 31] and covalent crosslinking [32, 33, 34, 35], have been developed to achieve molecular‐level continuity, significantly elevating interfacial toughness to values as high as 390 J m⁻². Although in situ cured or bonded microstructured interfaces have been explored to extend these benefits to structured surfaces, such methods remain limited to template‐defined geometries [33, 36] or low‐profile layered architectures [10]. Combining template‐free fabrication of high‐aspect‐ratio microcones with interfacial co‐curing remains a challenge.

Furthermore, real‐world applications impose stringent demands at the system level, necessitating compact and deployment‐ready structures where sensors need to mechanically couple with external modules such as actuators or wearable components [28, 34, 37]. Recent progress in decoupling strategies for multimodal soft sensors and integrated standalone sensing systems has further underscored the need for seamless mechanical and electrical integration at the system level [38, 39]. These integration requirements become even more stringent as soft sensors advance toward multimodal capabilities and deployment‐ready standalone systems, where reliable mechanical and electrical stability across functional layers is foundational [40, 41]. However, a manufacturing strategy that simultaneously achieves programmable microstructure geometries, high interfacial toughness, and monolithic multi‐material integration remains elusive.

Here, to address the aforementioned challenges, we propose a monolithic microcone capacitive sensor (MMCS) fabricated via a multi‐material feed‐draw printing (FDP) strategy. Inspired by the protein crosslinking that endows human skin's hierarchical structure with exceptional toughness (Figure 1a), the proposed sensor adopts quasi‐homogeneous materials—where CNT, NdFeB, and SiO₂ are uniformly dispersed within a PDMS matrix—to facilitate simultaneous co‐curing across multiple functional layers, thereby forming continuous covalent linkages. Constraining the interfacial contact interval to the early curing stage preserves active crosslinking sites, enabling subsequent co‐curing between the printed layers (Note S1). As a result, the sensors achieve an exceptional interfacial toughness of 1 547 J m⁻² (3.97 times that of state‐of‐the‐art microcone structure sensors [33]) and maintain high signal stability over 200 000 cyclic loadings. The feed‐draw printing process allows direct printing of microcones with programmable geometries, offering tunable sensitivity (up to 0.29 kPa⁻¹) and a wide pressure range (450 kPa). Furthermore, this strategy enables the monolithic integration of MMCS into wearable bands for real‐time pressure monitoring and magnetic soft grippers for grasping perception. Beyond the specific device demonstrated here, the proposed multi‐material feed‐draw printing strategy offers a versatile framework for achieving high interfacial toughness and functional integration in stable wearable and robotic interfaces (Table S1).

FIGURE 1.

Feed‐draw printing of monolithic microcone capacitive sensors with high toughness and seamless integration. (a) Biological inspiration from the architecture of human skin, where crosslinked protein networks provide exceptional interfacial toughness under high shear and repeated deformation. The illustration was generated with the assistance of Doubao (accessed December 2025) and subsequently modified by the authors. (b) (i) Cross‐sectional SEM images showing microcones bridging the top and bottom electrodes. (ii) Enlarged view reveals seamless, covalently bonded interfaces across layers (scale bar, 200 µm). (iii) Dielectric and electrode layers exhibit closely matched Young's moduli, ensuring mechanical compatibility. (c) Schematic illustration of the feed‐draw printing process for fabricating the microcone sensor. (d) Monolithic co‐cured fabrication enabling strong interfacial bonding. (e) Quasi‐homogeneous materials, PDMS continuous matrix embedded with CNT, SiO_2, and NdFeB nanoparticles. (f) (i) Relationship of sensor sensitivity and toughness with microstructure geometry. (ii) Comparison of interfacial toughness between state‐of‐the‐art microcone sensors and the MMCSs developed in this work. (g) Programmable microcone configuration enabling on‐demand tuning of sensor sensitivity and interfacial toughness. (h) Multifunctional applications of the sensor‐integrated structures, including a wearable pressure band for palm and foot pressure monitoring and a sensor‐integrated magnetic gripper for grasping perception.

2. Results and Discussion

2.1. Feed‐draw Printing of Monolithic Microcone Capacitive Sensors With High Toughness and Seamless Integration

Inspired by the hierarchical structure of human skin where interlayer protein‐chain crosslinking provides exceptional interfacial toughness under large force and deformation, we introduce quasi‐homogeneous materials and covalent interfacial bonding into multilayered sensors to achieve superior mechanical stability (Figure 1a). As shown in the SEM image (Figure 1b), the microcones form uninterrupted bridges between the upper and lower layers, while the dielectric and electrode interfaces are seamlessly fused. Furthermore, the similarity in Young's modulus between the two layers significantly improves their mechanical compatibility.

The microcones are fabricated via a feed‐draw printing strategy, as schematically illustrated in Figure 1c. Dielectric composite inks, consisting of 1 wt.% CNT and SiO₂‐modified PDMS, are loaded into syringes and extruded under controlled pneumatic pressure. During printing, the syringe feeds the ink while gradually lifting away from the substrate. Once the feeding stops, the nozzle continues to draw the ink until it breaks. These two processes are incorporated to shape the ink into a high‐aspect‐ratio microcone (Video S1). While feed‐only printing [42, 43, 44, 45] achieves increased height through continuous material deposition, it provides limited control over cone angle; draw‐only printing [46] creates tips through capillary bridge breakup, with height and angle coupled within a single drawing process. In contrast, the feed‐draw printing strategy integrates both processes in a synergistic manner. The feed segment primarily defines cone height via programmed displacement, and the draw segment primarily governs cone angle via regulated draw velocity, enabling active control of both parameters (Figure S4 and Table S2).

Existing methods require fully curing microstructures prior to adhesive or physical assembly. The feed‐draw process, by contrast, retains microcones in an uncured state for direct co‐curing with electrode layers, thereby eliminating discrete interfaces in the resulting monolithic sandwich‐type capacitive sensor (Figure 1d; Figure S5). Figure 1e presents the proposed quasi‐homogeneous materials, where PDMS serves as the elastic matrix, and 1% CNT and SiO₂‐modified PDMS act as the dielectric, while a composite material comprising 6% CNT is utilized as the electrode. Moreover, the feed‐draw printing strategy enables monolithic co‐cured fabrication and seamless integration of sensing elements with functional components, specifically, NdFeB‐filled PDMS for magnetic actuation and SiO₂‐modified PDMS for wearable structure. This approach allows sensing, actuation, and structural functions to be realized within a single printed architecture.

As shown in Figure 1f–i, the geometric evolution from cube‐like to conical/pyramidal morphologies enhances sensitivity via high aspect ratios but inevitably compromises mechanical robustness [10, 20]. In contrast, the feed‐draw printing strategy eliminates the need for molds and enables direct co‐curing of the printed microcones with the electrode layers, thereby reinforcing interfacial toughness by up to 3.97‐fold in comparison to state‐of‐the‐art microcone sensors [33] and maintaining structural integrity (Figure 1f–ii).

Furthermore, the feed‐draw printing enables on‐demand configuration of microstructures by independently modulating the microcone height gradient, areal density, and tilt orientation. (Figure 1g). Increasing the microcone density enhances connectivity and suppresses interfacial stress concentrations, thereby improving toughness, while introducing tilt or gradient geometries reduces the effective modulus of the structure, thereby enhancing sensitivity. These design parameters can be synergistically optimized to achieve customized sensitivity and robust interfacial toughness. As shown in Figure 1h, the printed sensors demonstrate stable operation under prolonged high dynamic loading and high shear stress, enabling their use in wearable pressure monitoring bands for sports analytics during intense exercise. Furthermore, incorporating NdFeB particles within the printable composite permits direct integration of sensors into magnetic soft grippers. This provides instant grasping perception and slippage detection capabilities, broadening its functional versatility.

2.2. Feed‐draw Printing of Tailored Microcones

Figure 2a illustrates the feed‐draw printing process used to fabricate microcones. The electrode and dielectric layers are first printed on a planar substrate, upon which microcones are subsequently feed‐draw printed. The printing nozzle begins at the dielectric surface (Figure 2a‐i) and moves upward gradually during feeding (Figure 2a‐ii). Subsequently, feeding is halted (Figure 2a‐iii) while the nozzle continues its upward motion. Finally, the ink filament breaks, forming a microcone (Figure 2a‐iv).

FIGURE 2.

Programmable feed‐draw printing of microcones with tunable morphologies. (a) Side‐view optical images of the feed‐draw printing process to create an individual microcone. (b) Viscosity–shear rate relationship of dielectric and electrode inks, demonstrating shear thinning. (c) Storage modulus (G′) and loss modulus (G′′) of the inks as a function of shear stress. (d) Schematic of segmented‐velocity feed‐draw printing: (i) dual‐segments and (ii) triple‐segments. (e) Effect of V ₁ and V ₂ on microcone morphology under dual‐segment procedure. (h = 0.3 mm) (f) Effect of V₁′ and V₁″ on microcone morphology under triple‐segment procedure. (h′ = 0.55 mm, h″ = 0.05 mm, V ₂ = 180 mm/min) (g) Side‐view optical images of microcones printed under (i) varying drawing velocities V* (0.1–1.0) at constant feeding height (0.3 mm) and (ii) varying feeding height (0.1–1.0 mm) at constant drawing velocity V*. (h) Microcone height and cone angle as a function of drawing velocity under constant feeding height (h = 0.3 mm). (i) Microcone height and cone angle as a function of extrusion height under constant drawing velocity (V* = 0.6). (j) Feed–draw printing of tilted microcones. (k) Side‐view optical images of microcones with programmable tilt angles. (l) Side‐view optical images of microcones printed under (i) varying h′ (0.40–0.80 mm) under constant h″ = 0.05 mm and (ii) varying h″ (0.05–0.45 mm) under constant h′ = 0.45 mm. (iii) combinations of vertical and tilted microcones. Scale bar, 300 µm.

To achieve stable feed‐draw printing, we formulated a dielectric ink by incorporating 10 wt.% fumed silica into a PDMS matrix loaded with 1 wt.% carbon nanotubes; a PDMS composite containing 6 wt.% carbon nanotubes served as the electrode ink. Both inks exhibit shear‐thinning behavior (Figure 2b). Oscillatory rheology reveals that under high shear stress, the storage modulus (G′) of both inks falls below the loss modulus (G″), indicating viscoelastic liquid behavior that facilitates smooth extrusion and high‐speed printing (Figure 2c). Under low shear stress, G′ exceeds G″, and the inks exhibit elastic solid characteristics that ensure shape retention after printing [47]. In contrast, the dielectric ink without SiO₂ shows negligible yield stress, as G′ remains below G″ throughout the low‐shear regime (Figure 2c), causing it to slump and lose geometric fidelity upon extrusion (Figure S6). With 10 wt.% SiO₂, the ink develops sufficient yield stress to maintain its shape, enabling self‐supporting 3D microcones.

To achieve precise control over the sensor architecture, we systematically evaluated the printing parameters governing the morphology of both the electrode‐dielectric base layers and the vertical dielectric microcones. A baseline feeding velocity (V₀), is adopted as 300 mm/min for ink feeding, unless otherwise specified (Figure S7). Regarding the printing of the base layers, the printed width of both electrode and dielectric inks decreases as the printing velocity increases (Figure S8). To ensure stable filament formation and higher resolution, the horizontal printing velocity was set to 4 times the feeding velocity. Subsequently, for the vertical microcones, the feed‐draw printing process is divided into a two‐segment procedure. In the first segment procedure, feeding and drawing occur simultaneously, with the upward drawing velocity and feeding height denoted as V₁ and h, respectively. In the second segment procedure, drawing proceeds without feeding, during which the velocity is defined as V₂ (Figure 2d‐i). As V₁ increases, the ink deposition decreases, resulting in slender microcones. Similarly, increasing V₂ results in microcones that are both slender and taller, as demonstrated in Figure 2e. By subdividing the V₁ segment into two sub‐segments, a wider range of microcone shapes can be obtained, as shown in Figure 2d‐ii. When the drawing velocity in the first sub‐segment () exceeds that in the second sub‐segment (), a shape with a constricted base and an expanded midsection can be achieved, as illustrated in Figure 2f.

As shown in Figure 2e, we first set V₁ = V₂ to simplify the analysis. The relative drawing velocity V*, defined as the ratio of the nozzle's drawing velocity to the ink feeding velocity, was found to be a key determinant of feed‐draw printing. At a fixed feeding height of 0.3 mm, increasing V* led to taller cones with smaller cone angles (Figure 2h,g‐i), indicating precise tunability of geometry through V*. At a constant relative drawing velocity (0.6), cone height increased with feeding height while maintaining a nearly constant cone angle (Figure 2i,g‐ii). By varying V* and the feeding height, microcones with diverse heights and cone angles can be reliably fabricated using feed‐draw printing (Figure S9).

Furthermore, guiding the nozzle along a tilting trajectory provides an additional degree of morphological control, as shown in Figure S10. By modulating the inclination of the printing path during drawing, microcones with programmable tilt angles can be fabricated (Figure 2j–k). This trajectory‐controlled printing strategy enables the programmable fabrication of a wide range of microcone architectures. When the feeding height of the second sub‐segment (h″) is held constant, increasing that of the first sub‐segment (h′) produces slender cones with a more pronounced midsection (Figure 2l‐i). Conversely, fixing h′ while increasing h″ yields shorter, broader cones with an enlarged midsection (Figure 2l‐ii). Furthermore, this approach allows vertical and tilted microcones to be integrated within a single construct (Figure 2l‐iii). Additional combinations of microcone geometries are presented in Figure S11.

2.3. Mechanical Properties of the Co‐cured Quasi‐homogeneous Material Interfaces

To evaluate the mechanical compatibility of the printed sensor, we first characterized the tensile behavior of the cured dielectric and electrode composites. As shown in Figure S12, the Young's modulus of the dielectric composite increases systematically with SiO₂ content. At the formulated loading of 10 wt.%, the cured dielectric reaches 3.1 MPa, closely matching the electrode modulus of 2.1 MPa (Figure 3a,b). This mechanical compatibility is essential for reducing interfacial stress mismatch, thereby preserving structural integrity during cyclic deformation. Furthermore, both composites demonstrate high fracture toughness (939 J m⁻² for the electrode and 1 337 J m⁻² for the dielectric; see Figure 3c). These high toughness values indicate strong resistance to crack initiation and propagation, thereby enabling the materials to preserve structural integrity and avoid catastrophic failure under large deformations during device operation.

FIGURE 3.

Mechanical properties of the co‐cured quasi‐homogeneous material interface of sensors. (a) Tensile stress‐strain curve of the electrode or dielectric samples. (b) Young's moduli for the material of the electrode and dielectric samples. (c) Force‐displacement curves of notched and unnotched samples. (d) Schematic of three peel tests of the sensor's interfaces: (i) separately cured with PDMS bonding, (ii) sequentially cured, and (iii) co‐cured. (e) Peel force as a function of displacement for electrode and dielectric samples based on different fabrication methods. (f) Interfacial toughness of each sample. (g) Schematic of three peel tests of sensors: (i) separately cured with PDMS bonding, (ii) sequentially cured, and (iii) co‐cured. (h) Peel force as a function of displacement for sensors based on different fabrication methods. (i) Interfacial toughness of each sensor. (j–l) Finite element simulations of interfacial stress distributions under 50 kPa pressure. Comparisons are shown for: (j) the co‐cured interface; (k) an interface bonded with a stiffer adhesive (with a modulus 10 times that of the dielectric); and (l) an interface bonded with cyanoacrylate glue.

To investigate the interfacial mechanics governing device reliability, we examined three representative interfacial configurations: (i) separately cured dielectric and electrode layers bonded with PDMS, (ii) sequentially cured layers, and (iii) co‐cured layers (this work), as illustrated in Figure 3d. The corresponding peel force–displacement curves (Figure 3e) and the calculated results (Figure 3f) reveal that co‐curing enhances the interfacial toughness by more than an order of magnitude, from ∼100 J m⁻² to over 1 600 J m⁻². Mechanistically, the co‐curing strategy constrains the degree of cure at the interface to remain low during contact (Note S1), allowing the newly deposited ink to interact with an uncured polymer. Under this condition, unreacted polymer chains can interdiffuse across the interface prior to gelation. These chains participate in crosslinking reactions across the boundary, resulting in a high interfacial toughness. As a result, the interfacial adhesion energy approaches the bulk fracture energy, leading to a transition in failure mode from interfacial delamination to cohesive fracture.

To further quantify the contribution of co‐curing to interfacial reinforcement of microstructures, we compared the bonding behavior of flat and microconed surfaces (microcone density: 1 mm⁻²) under the same three interfacial configurations mentioned above (Figure 3g). The sensor interface during the peeling test is shown in Figure S13. The corresponding peel curves (Figure 3h) and toughness values (Figure 3i) consistently demonstrate that co‐curing yields a substantial improvement in interfacial toughness of 1 347 J m⁻²—up to 9.4 times that of the separately cured counterparts. Figure S14 reveals that the microcones predominantly fracture at their mid‐sections rather than delaminating from the planar base, demonstrating robust interfacial bonding. Furthermore, as illustrated in Figure S15, the microcone structures effectively inhibit catastrophic crack propagation, thereby significantly enhancing the interfacial toughness of the sensor.

Finite element simulations were performed under an applied pressure of 50 kPa (Figure 3j–l) and over a pressure range of 0 to 400 kPa (Figure S16 and Video S2). The simulations show that the co‐cured interfaces display a markedly more uniform stress distribution than those bonded with stiffer adhesives or cyanoacrylate glue, effectively mitigating local stress concentration and suppressing interfacial delamination during prolonged operation.

2.4. Sensing Properties of Monolithic Microcone Capacitive Sensors

To establish the geometry–property relationship, we tune sensor performance by independently adjusting three microcone parameters: areal density, tilt orientation, and height gradient. These geometric degrees of freedom determine the effective modulus E and thus the pressure sensitivity. In the low‐pressure regime (P ≪ E), sensitivity scales as S ≈ 1/E (Note S2). Reducing areal density from 2 to 1 mm⁻² lowers E and raises initial sensitivity from 0.016 to 0.027 kPa⁻¹ (Figure 4a). This gain modestly reduces interfacial toughness, from 1 547 to 1 347 J m⁻² (Figure S17). Tilting microcones shifts the dominant deformation mode from axial compression to bending, which further reduces E and increases initial sensitivity to 0.033 kPa⁻¹ while maintaining a toughness of 1 466 J m⁻² (Figure S18). The height gradient design (Figure S19) achieves a high sensitivity and wide detection range. It exhibits a multi‐stage linear response: a high initial sensitivity of 0.29 kPa⁻¹ in the low‐pressure regime (0–10 kPa), a stable sensitivity of 0.07 kPa⁻¹ (R² = 0.999) in the medium range (10–300 kPa), and a maintained sensitivity of 0.028 kPa⁻¹ (R² = 0.999) in the high‐pressure range (300–450 kPa). This broad linear range arises from a progressive contact mechanism (Figure S20). As pressure increases, shorter microcones sequentially engage the sensing area. The continuous recruitment of new capacitive units prevents early saturation under high‐pressure loads [48]. The finite‐element model quantitatively reproduces the experimental capacitive response in the low‐pressure regime, validating its predictive capability in the initial linear regime (Figure S21).

FIGURE 4.

Sensing characteristics of monolithic microcone capacitive sensor. (a) Normalized capacitance change as a function of pressure for sensors with different microcone height ratios (Uniform: 1:1:1:1, Gradient: 1:2:3:4) and microcone density (1 and 2 microcones mm⁻²). (b) Normalized capacitance change under low pressure (1 Pa). (c) Response and relaxation times of the sensors. (d) Normalized capacitance change under dynamic pressure gradients ranging from 10 to 400 kPa. (e) Normalized capacitive response under three cycles per pressure at 10, 100, 200, and 400 kPa. (f) Normalized capacitive response at different loading rates of 50, 100, and 150 mm/min. (g) Long‐term stability tests of the sensor over 200 000 loading/unloading cycles at 20 kPa.

To evaluate the limit of detection (LOD) of the sensor, the sensor was placed inside a custom‐built glass isolation chamber to eliminate signal fluctuations induced by ambient airflow. We then placed a lightweight piece of paper weighing 10 mg on the sensor, simulating the application of an extremely small pressure. Figure 4b demonstrates that the sensor achieves an LOD of 1 Pa. Such high sensitivity also enables the detection of minute pressure stimuli, such as the gentle pulse at a fingertip (Figure S22a). In pulse monitoring, the device captured both the amplitude and frequency of arterial pulsations, each cardiac cycle producing a distinct capacitive spike at a rate of ∼73 beats per minute—consistent with normal resting values for healthy adults.

Rapid loading–unloading tests further confirmed the device's fast temporal response (Figure 4c). The capacitance increased immediately upon pressurization and stabilized within ∼80 ms, while release induced a return to baseline within ∼70 ms. These rapid responses underscore the suitability of the sensor for real‐time, dynamic pressure monitoring.

Figure 4d demonstrates that the sensor has good repeatability at different pressures (10 to 400 kPa). Meanwhile, as shown in Figure 4e, the stability of the sensor under various loads was investigated by applying low pressure (10 kPa) to high pressure (400 kPa). The sensors exhibited a rapid and stable response under different load conditions, and the normalized capacitance quickly returned to the baseline when the pressure was withdrawn. To investigate the response of the capacitive signal at different application rates, tests were carried out at different loading rates (i.e., 50, 100, 150 mm/min). The pressure sensor exhibits a stable capacitive signal response (Figure 4f) that is independent of loading rate.

High‐performance capacitive sensors are pivotal components for wearable healthcare. Here, physiological monitoring capabilities were assessed across diverse body sites. Mounted on the cervical region, the sensor distinguished syllabic stress during phonation of “sen‐sor” (Figure S22b). When adhered to the abdomen, it captured continuous respiration rates of ∼15 breaths per minute (Figure S22c). Furthermore, it enabled motion perception, from monotonic wrist flexion (Figure S22d) to complex gesture recognition (e.g., “6”, “OK”, “thumbs‐up”, and “yeah”) via finger joint movements (Figure S23).

The sensor also exhibits exceptional stability during cyclic testing. Long‐term mechanical durability was evaluated through cyclic fatigue testing, as shown in Figure S24. The sensor was subjected to 200 000 compression cycles at a constant pressure of ∼20 kPa, simulating prolonged operational conditions. Throughout the test, capacitive output remained stable, with no significant drift or degradation (Figure 4g). Comparative analysis of response signals from the first, middle, and final five cycles reveals virtually identical profiles, attesting to the sensor's exceptional repeatability and mechanical robustness. The sensor further maintained stable output over 30 000 loading/unloading cycles within a pressure range of 0–100 kPa, conducted in two sessions separated by 40 days (Figure S25), demonstrating stable performance under elevated pressure and long‐term storage conditions. Figure S26 shows that when the sensor is sheared for 10 000 cycles under a shear stress of ∼20 kPa, no obvious change in the signal waveform or amplitude is observed (Figure S27). Such enduring performance is paramount for applications demanding sustained functionality in challenging practical environments, ensuring long‐term reliability and precision. Beyond long‐term cyclic stability, sample‐to‐sample and batch‐to‐batch evaluations confirm consistent initial capacitance across independent fabrications (Figure S28). The MMCS exhibits negligible capacitance drift across temperature (25°C–85°C) and relative humidity (60%–90% RH) variations as shown in Figure S29. Furthermore, Lap‐shear tests measured a peak shear strength of 365.3 ± 14.6 kPa (Figure S30). The high shear strength enables reliable detection of heavy loads (2.1 and 4.2 kg) during dynamic transport on a conveyor belt (Figure S31).

2.5. Monolithic Integration of Wearable Bands for High‐Dynamics Monitoring

Conventional wearable sensors are commonly fabricated by separately printing the structural substrate and functional sensing layers, followed by post‐assembly [49, 50]. Even when the internal interfaces within the sensor are reinforced, the interface between the sensor and the wearable structure often remains a critical mechanical weak point, resulting in signal degradation or even delamination under long‐term loading. These limitations severely restrict the reliability required for high‐dynamics human–machine interactions. We develop a wearable pressure monitoring band by directly co‐curing an MMCS with a 14 wt.% SiO₂‐modified PDMS structural layer (Figure 5a; Figure S32). The SiO₂ content was primarily selected to ensure printable rheology, imparting the yield stress and solid‐like behavior required for self‐supporting 3D architectures (Figures S33, S34). This formulation concurrently elevates the Young's modulus of the cured structural layer (Figure S35), ensuring mechanical robustness under dynamic loading. As revealed by cross‐sectional SEM imaging (Figure 5b), a continuous interfacial connection is achieved between the sensor and the structural matrix. This monolithic architecture endows the interface with an interfacial toughness of 1 884 J m⁻²—11 times higher than that of the separately cured counterpart (Figure S36) —effectively eliminating delamination risk while ensuring stable signal under repeated deformation.

FIGURE 5.

Applications of the 3D‐printed wearable pressure monitoring band. (a) Schematic of the wearable pressure‐sensing band structure. (b) Cross‐sectional SEM images of the electrode–band interface. Scale bar, 25µm (c, d) Real‐time capacitance change during boxing at different impact forces. (e, f) Real‐time sensor response capturing dynamic palm pressure while climbing horizontal bars. (g, h) Real‐time monitoring of plantar pressure and gait dynamics across varying running speeds (3, 6, and 9 km/h). (i) Stable sensor response for plantar pressure during continuous running for 80 min. The illustrations in c, e, g were generated with the assistance of Doubao (accessed December 2025) and subsequently modified by the authors.

Punching force is critical for both boxing training and competitive performance [51]. When attached to the dorsal aspect of the hand, the wearable pressure monitoring band effectively converts impact forces during boxing into distinct capacitive peaks, thereby enabling the real‐time quantification of punching strength and frequency (Figure 5c). As the impact force increases, the sensing response is correspondingly increased (Figure 5d), providing intuitive visual feedback for athletes and coaches. Notably, the sensing signal exhibits highly reversible and reliable behavior during boxing without noticeable degradation (Figure S37b and Video S3), confirming its dynamic robustness. In contrast, the separately cured sensor undergoes signal failure under punching conditions (Figure S37c), demonstrating that the co‐curing strategy is essential for maintaining interfacial integrity.

While the boxing scenario demonstrates robustness against high‐impact transients, applications involving intense shear forces and sustained heavy loads present different challenges. To evaluate performance under such conditions, we conducted a horizontal bar climbing experiment where local hand forces exerted by a 65 kg climber can reach ∼650 N [52] imposing stringent mechanical demands on the sensor–band interface. Benefiting from its monolithic architecture, the band affixed to the palm reliably withstands these high loads while accurately capturing temporally varying pressure (Figure 5e; Video S4). During movement across the horizontal bars, the climber's center of gravity undergoes continuous transition. This weight shifting redistributes the effective contact force applied to the localized sensing area, generating the signal reductions as shown in Figure 5f. By accurately detecting this dynamic weight transfer, the sensor demonstrates its capability to monitor unstable, complex contact states under extreme mechanical conditions.

Beyond withstanding extreme static and shear loads, the sensor should also demonstrate long‐term fatigue resistance under high dynamic loading. To verify this, we focused on plantar pressure monitoring, which is essential for evaluating lower‐limb load during locomotion [53]. When fixed in the shoe sole, the sensor records plantar pressure variations whose frequency increases with running speed, thus enabling gait and performance monitoring (Figure 5g,h; Video S5). It is noteworthy that the sensor maintains a stable and continuous signal output over 80 min of uninterrupted operation (Figure 5i), thereby substantiating its exceptional durability and reliability for wearable applications. In contrast, the separately cured sensor undergoes signal failure during running (Figure S38), underscoring the importance of robust interfacial bonding for long‐term wearable sensing applications.

Collectively, these results illustrate the versatility and resilience of the integrated sensing architecture, which captures mechanical stimuli spanning from high‐impact strikes to gait dynamics. Such performance positions it as a promising platform for continuous physiological monitoring, sports analytics, and intelligent health management.

2.6. Monolithic Integration of Magnetic Soft Grippers for Tactile Perception

Tactile perception and proprioceptive feedback are fundamental to the dexterity of the robotic hand, enabling adaptive manipulation in unstructured environments. However, conventional fabrication strategies frequently entail constructing actuators and sensors as discrete units that are subsequently bonded with adhesives. This inevitably undermines structural integrity. Here, we further fabricate a 3D‐printed magnetic soft gripper that monolithically integrates actuation and multi‐site sensing (Figure 6a; Figure S39). The magnetic actuation layer comprises NdFeB/PDMS (6:1 mass ratio, 86 wt.%), and its ink exhibited shear‐thinning behavior with a yield stress; oscillatory tests confirmed G′ > G″ at rest and a liquid‐like response under high shear, consistent with self‐supporting direct‐ink‐writing (Figure S40). The gripper comprises a magnetic actuator, four pressure sensors embedded along the inner fingers for tactile sensing, and a bending sensor positioned at the outer region to enable proprioceptive feedback. Cross‐sectional SEM imaging (Figure 6b; Figure S41) reveals a seamless interface between the sensing and actuation layers, validating the structural coherence afforded by the co‐printing process. This monolithic architecture endows the interface with an interfacial toughness of 1 907 J m⁻²—11.5 times higher than that of the separately cured counterpart (Figure S42). Room‐temperature hysteresis measurements showed high remanent magnetization and coercivity, confirming hard‐magnetic behavior for persistent actuation (Figure S43). Initial capacitance showed no difference across standalone sensors, co‐cured unmagnetized grippers, and magnetized grippers (n = 4, Figure S44), indicating that magnetic integration and magnetization do not introduce parasitic capacitance or electromagnetic interference. Switching of the external magnetic field induces negligible drift, indicating that magnetic actuation does not perturb the capacitive readout (Figure S45).

FIGURE 6.

Applications of the 3D‐printed magnetic soft gripper with integrated sensors. (a) Schematic of the magnetic soft gripper structure. (b) Cross‐sectional SEM image of the gripper–sensor interface. Scale bar, 25 µm. (c, d, e) Grasping cylindrical objects with different weights and orientations. (f) Grasping objects of various sizes. (g) Grasping a leaking balloon. (h) Detection of object release using the magnetic soft gripper with integrated sensors.

The integrated sensors enable real‐time feedback on object state. The gripper demonstrates object weight perception and object recognition across diverse grasping scenarios. When handling cylindrical objects of varying weights and orientations (Figure 6c–e), the pressure sensors exhibit distinct capacitive responses proportional to the applied load, thereby enabling the estimation of grasping force and contact position.

During manipulation of objects with different sizes (Figure 6f), the bending sensor records decreasing capacitance with increasing object diameter, consistent with the reduced bending curvature required for larger targets. When grasping a leaking balloon (Figure 6g; Video S6), the progressive decrease in capacitance reflects the shrinking state of the balloon and the corresponding bending state of the gripper, effectively translating external deformation into signal changes.

Furthermore, slippage is identified through the combined analysis of tactile and proprioceptive signals in object slippage detection tests (Figure 6h). Specifically, the pressure sensors capture a rapid decrease in contact pressure as the contact region migrates along the finger surface when the object begins to slip. At the same time, the bending sensor detects an increase in finger curvature, accompanied by a sharp reduction in the bending signal, resulting from the displacement of the object within the grasp. The simultaneous occurrence of decreasing pressure signals and rapidly diminishing bending signals, which together reflect contact migration and curvature variation, provides a reliable indicator of object slippage. During operation, upon detecting these correlated signal variations, indicative of insufficient grasping force, the gripping system adaptively increases the magnetic strength, thereby restoring grasp stability and effectively preventing further slippage (Video S7).

Collectively, these results underscore the efficacy of co‐printing in achieving integration of magnetic actuation and sensing. The resulting system exhibits not only high mechanical robustness but also adaptive responsiveness, paving the way for intelligent, self‐regulating soft robotic platforms.

3. Conclusion

Taken together, this work establishes a generalizable paradigm for constructing monolithically integrated soft sensing systems via a feed‐draw printing strategy, enabling the co‐fabrication of sensors and functional components within a unified manufacturing process. By engineering quasi‐homogeneous materials, we achieved continuous covalent linkages across the interfaces, yielding exceptional interfacial toughness (up to 1 547 J m⁻² at the sensor‐level interface, and 1 907 J m⁻² at the system‐level interface) that secures stable response over 200 000 cycles. The combination of these quasi‐homogeneous materials with the feed‐draw printing process enables programmable fabrication of microcones with tunable morphologies, eliminating the need for multistep templating or post‐bonding processes that typically introduce interfacial discontinuities. The resulting MMCS achieves an enhanced sensitivity (0.29 kPa⁻¹) and an ultralow detection limit (1 Pa) with a wide linear range (up to 450 kPa) and rapid temporal response (∼80 ms). The monolithic and compact integration of these sensors into wearable bands and soft grippers has been demonstrated. This underscores their practical deployability in real‐world scenarios, where sensitivity, mechanical robustness, and ease of system‐level integration are critical. Future efforts may extend this unified manufacturing framework to incorporate additional sensing modalities and more complex soft structures, including decoupling strategies for multimodal sensing and integrated standalone sensing systems, while preserving mechanical compatibility and system‐level robustness.

4. Experimental Section

4.1. Preparation of Quasi‐Homogeneous Inks

The composite matrix was prepared using PDMS base and curing agent (Sylgard 184, Dow Corning Co., Ltd.) at a 10:1 weight ratio. The multi‐walled carbon nanotubes (MWCNTs, purity ≥ 95%, 10–20 nm diameter, 10–30 µm length; Nanjing Xianfeng Nano‐technology Co., Ltd.) were first dispersed in n‐hexane (Aladdin, ≥ 99%) by ultrasonication (the ultrasonic material disperser, LC‐JY92‐IIN, 240W) for 2 h. The concentration of CNTs in n‐hexane was 5 mg mL⁻¹. For the electrode, the ratio was set to 6 wt.% CNTs in the composite. For the dielectric layer, the CNT content was reduced to 1 and 10 wt.% fumed silica was additionally incorporated to tune the rheology (yield stress and shear thinning for direct ink writing) and to mechanically reinforce the cured composite. For the wearable structural layer, 14 wt.% SiO₂ (relative to PDMS) was incorporated into the PDMS matrix to simultaneously enable self‐supporting 3D printing and increase the structural modulus for robust system‐level integration. For the magnetic actuation layer, NdFeB microparticles and PDMS were combined at a 6:1 mass ratio. Both mixtures were homogenized using a planetary centrifugal mixer (THINKY ARE‐310) at 2 000 rpm for 5 min. All chemicals were used as received without further purification.

4.2. 3D printing of the Sensor and In situ Co‐Curing

The 3D printing of the sensor was performed on a custom‐built 3D printer. The syringe needle was selected to have a diameter of 300 µm. The G‐code, generated from the designed model, was post‐processed with a Python script to generate the microcone printing path and thereby control both extrusion and build‐platform movement (Figure S46). The sensor consists of two independently printed components: the first integrates the bottom electrode, dielectric layer, and microcone array; the second integrates the top electrode and its dielectric layer. These components were aligned so that the cone tips contacted the opposing dielectric layer, forming a sandwich architecture. In most cases, the post‐assembly microcone height was set to the feeding height plus 0.3 mm. The assembly was completed within 2 min, while the ink remained in the early curing stage. This enabled covalent bonding across the dielectric–electrode and microcone interfaces, yielding a monolithic sensor with high interfacial toughness. Co‐curing proceeded at 80°C for 4 h (Figure S5).

4.3. Rheological Properties

Rheological properties of inks were measured by a rotational rheometer (Physical MCR 301, Anton Paar, Austria) under ambient conditions with a fixed gap of 500 µm. Viscosity was measured with a shear rate ranging from 0.01 to 1 000 s⁻¹. The oscillation test was carried out at 1 Hz with stress from 0.3 to 5 000 Pa.

4.4. Morphology Properties

SEM and EDS mapping images were acquired with a scanning electron microscope (Hitachi, SU8220). The printed fiber diameter was measured by using a digital microscope system (Mshot, MS60).

4.5. Mechanical Properties

A universal mechanical testing machine (Shimadzu AG‐X plus machine) was used to assess the composites’ tensile properties (sample size: 50 mm × 4 mm × 2 mm). Lap‐shear tests were conducted on 5 mm × 5 mm samples backed with flexible but inextensible BOPP (Biaxially‐oriented Polypropylene) films; peak shear stress was recorded as the shear strength. The test speed was 10 mm min⁻¹. Each test was repeated at least three times.

4.6. Notch Tests for Fracture Toughness

The fracture toughness of the dielectric and electrode samples was quantified using a standard notch tensile test. Rectangular specimens (30 mm × 10 mm × 1 mm) were prepared with and without a pre‐introduced notch. A sharp notch (∼6 mm in length) was created using a razor blade. The specimens were subjected to uniaxial tensile loading at a constant speed of 50 mm min⁻¹. The force–displacement response of the unnotched specimen was recorded to determine the mechanical work input, while the notched specimen was used to identify the critical displacement corresponding to crack initiation and propagation [54, 55]. This method quantifies the intrinsic energy required to propagate a crack through a homogeneous bulk material.

4.7. Peeling Tests

We performed the 180° peel tests on the universal testing machine. Electrode‐Dielectric interface samples, of length 30 mm and width 10 mm, and sensors of 10 mm × 10 mm were prepared and subjected to a standard 180° peel test. The peeling speed was controlled at 50 mm min⁻¹. All the samples are attached to BOPP sheet as the flexible but inextensible backing layer. The interfacial toughness for the 180° peel test was the ratio 2F _ss/B (the peel force reaches a plateau F _ss, B is the width of the elastomers) [56]. This value represents the apparent system‐level energy release rate required to advance a crack steadily along the interface.

4.8. Sensing Stability Testing

The sensor was subjected to compression (Figure S24) and shear (Figure S26) cycle tests at 20 kPa for 200 000 and 10 000 cycles, respectively. Sensors (10 mm × 10 mm) were fixed to specifically designed holders using a layer of cyanoacrylate glue (Krazy Glue, 3M CA40H), ensuring they were stable during testing. The holders were controlled using a computer‐controlled stage with a force gauge (Shimadzu AG‐X plus machine). Environmental stability testing was performed by placing the sensor in a temperature‐humidity chamber. Environmental stability was further assessed by monitoring the normalized capacitance using an LCR meter at 1 MHz and 1 V bias under varying temperatures (25°C–85°C) at constant relative humidity (60%), and varying relative humidity (60%–90% RH) at constant temperature (25°C).

4.9. Electrical Properties of the Sensor

The capacitance was measured using an LCR meter (E4980AL, KEYSIGHT) at 100 kHz and 1 V. The external pressure was applied and measured using a computer‐controlled stage (Shimadzu AG‐X plus machine). The response time and relaxation time were measured via the LCR meter at a testing frequency of 1 MHz.

4.10. FEA Modeling

The MMCS models were developed in COMSOL Multiphysics to investigate the influence of adhesive modulus and microcone density on stress distribution through numerical simulation. In order to replicate the experimental conditions, the model was compressed between a fixed lower plate and a vertically loaded upper plate. Contact pairs were defined at all potential interfaces to prevent interpenetration, and the mesh size was set to “Extremely Fine” to ensure numerical accuracy. The MMCS modulus was determined from experimental data (E = 3.1 MPa), while a Poisson's ratio of 0.49 was assumed. A 3D coupled structural–electrostatic model was built in COMSOL. Microcones were assigned a Neo‐Hookean law, and the capacitance–pressure curve was extracted via Augmented Lagrange contact together with parametric displacement sweeps. The relative permittivity is set to 4.0 [57].

4.11. Portable Wireless Readout System

To facilitate portable monitoring, a custom wireless readout system was developed (Figure S47). The system incorporates a high‐precision capacitance‐to‐digital converter (CDC, PCAP01AD) to capture capacitance variations with a 17‐bit resolution. The acquired data were processed by a microcontroller unit (MCU, STM32F103) and subsequently transmitted wirelessly via Bluetooth to a personal computer for visualization and analysis.

4.12. Fabrication and Signal Acquisition of the Wearable Pressure Band

A multi‐material 3D printing technique was employed for the fabrication of the sensor‐integrated wearable pressure monitoring band. Specifically, the wearable structural layer was printed first, followed by the printing of the electrode layer on its surface. Microcone structures were subsequently fabricated via feed‐draw printing, and the assembly was completed by bonding with a separately printed electrode layer. After curing, the final wearable pressure monitoring band was obtained (Figure S32). For dynamic testing scenarios including boxing, climbing, and running, capacitive signals were acquired in real‐time using the portable wireless readout system.

4.13. Fabrication and Signal Acquisition of the Sensor‐Integrated Magnetic Gripper

Multi‐material 3D printing was also employed for the fabrication process of the magnetic soft gripper with integrated sensors (Figure S38). A magnetically actuated layer was first printed, onto which four electrode layers were printed. Microcone structures were then fabricated using feed‐draw printing, followed by assembly with an electrode layer and curing. Finally, an additional bending sensor was adhered to the rear side of the magnetic gripper to enable bending detection. The gripper was then magnetized along its longitudinal direction by applying an impulse magnetic field (2.8 T) generated by an impulse magnetizer (M20‐2020, HLT Co. LTD, China) to magnetically saturate the embedded NdFeB particles. The magnetic gripper contained 86 wt.% magnetic particles to ensure sufficient magnetic grasping force. Actuation of the magnetic soft grippers was achieved using two‐axis Helmholtz coils. The applied magnetic field strength was precisely tuned by adjusting the current magnitude, whereas the opening and closing of the gripper were controlled by reversing the current direction. The magnetic field strength for grasping objects ranged from 9 to 27 mT. During experimental characterization, slip detection tests were conducted using the portable wireless readout system, while all other electrical measurements were performed using the LCR meter (1 MHz). Prior to gripper testing, a standalone MMCS was placed within the two‐axis Helmholtz coils. The magnetic field was switched on and off at a strength of 54 mT to assess the magnetic interference on the capacitive readout.

4.14. Ethical Statement

All the volunteers have given informed consent to the experiments. The characterizations conducted were in vitro, non‐clinical, and non‐toxic, primarily intended for research on wearable devices. According to the institutional policies applicable to this type of low‐risk research, formal ethical approval was not required.

4.15. Statistical Analysis

Quantitative data are expressed as mean ± standard deviation based on independent replicates (n ≥ 3). Data analysis and visualization were performed in OriginPro 2023. Sample‐to‐sample variation was evaluated using 5 sensors from the same batch, while batch‐to‐batch reproducibility was assessed using three independent batches.

Author Contributions

Conceptualization: X.L., M.L., Y.S. Methodology: X.L, M.L., Z.H. Investigation: X.L., Z.H., B.S., J.W., Z.D., H.J., Y.L. Funding acquisition: M.L., S.Z, T.P. Visualization: X.L., D.L., M.W. Writing – original draft: X.L., H.Z., W.L. Writing – review & editing: M.L., S.Z. Supervision: M.L., T.P., S.Z.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: adma74018‐sup‐0001‐SuppMat.pdf.

Supporting File 2: adma74018‐sup‐0002‐VideoS1.mp4.

Supporting File 3: adma74018‐sup‐0003‐VideoS2.mp4.

Supporting File 4: adma74018‐sup‐0004‐VideoS3.mp4.

Supporting File 5: adma74018‐sup‐0005‐VideoS4.mp4.

Supporting File 6: adma74018‐sup‐0006‐VideoS5.mp4.

Supporting File 7: adma74018‐sup‐0007‐VideoS6.mp4.

Supporting File 8: adma74018‐sup‐0008‐VideoS7.mp4.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.