Continuous monitoring of vital/physiological signals over extended temporal windows plays a substantial role in a comprehensive perspective on the overall health status of an individual, enabling early disease prediction, self-directed diagnostics, personalized therapeutics and improved management of chronic health conditions9,10,11. Recent advances in wearable electronics, particularly those incorporated within skin-integrated or textile-based devices, have facilitated continuous on-body biosignal monitoring during daily activities12,13,14,15. However, despite notable progress in optimizing the mechanical and electrical properties of these wearable devices, several barriers to their widespread adoption and clinical applications remain16,17,18. One substantial challenge is the poor permeability of electronic materials/devices, leading to sweat accumulation at the skin–device interface after a certain duration of operation2,19. This hurdle not only causes thermophysiological discomfort to users but also deteriorates collected signal quality and adhesion property, thereby hindering precise signal monitoring over prolonged periods20.

Breathable electronics based on ultrathin7,21, porous5,22, nanofibre6,23, nanomesh4,24 or textile-based architectures8,25 enable the passive or active discharge of gases, vapours and sweat through their micro/nano openings. Thereby, they offer a user-friendly interface between the skin and the device, ensuring the possibility for more stable signal acquisition during long-term healthcare monitoring even under sweating conditions26,27,28. However, from the standpoint of practical applications, the development of breathable electronics is still in its infancy, as the functions of these devices are mainly associated with simple electrodes29, sensors30, antennas22, energy harvesters/storage31,32 and displays8 (Supplementary Fig. 1 and Supplementary Table 1). The state-of-the-art wearable electronics include much more complicated functional sensors, circuits and modules for signal acquisition, processing and transmission11. Therefore, it is still a notable challenge to realize high-level integrated and multifunctional wearable electronics, yet with a breathable format33,34,35,36.

Here we report a fundamental methodology from materials processing, device architecture and system integration for integrated permeable wearable electronics based on the concept of 3D LD (Fig. 1a), which can serve as a universal platform/substrate to integrate with well-established skin-integrated and wearable electronics. Distinguished from the reported works, the 3D LD does not rely on unique materials but adopts an in-plane liquid transport layer called the horizontal liquid diode (HLD) on top of the out-of-plane perspiration channels called the vertical liquid diode (VLD). Combining in-plane and out-of-plane sweat-transport features, the 3D LD not only offers great air/sweat permeability but also allows high-performance wearable electronics to be directly integrated on top of it, with no effect on perspiration (Fig. 1b and Supplementary Note 1). The 3D LD with nature-inspired microstructures directionally and spontaneously transport fluids to enable the continuous transport of sweat in a pre-programmed path37,38. It can rapidly self-pump sweat from the skin–device interface to the outlet, yet prevent backflow, which allows the wearable device to provide reliable biosignal monitoring, have robust adhesion strength and give great comfort free from irritation during long-term wear (Fig. 1c–e). Furthermore, the detachable design based on the magnetic coupling of 3D LD substrate and flexible/wearable electronics greatly enhances the practical usability and reduces costs (Fig. 1f).

Fig. 1: Integrated system-level sweat-permeable wearable electronics based on the concept of 3D LD.
figure 1

a, Comparison between conventional flexible electronics (left) and permeable electronics (right). Permeable design allows for enhanced breathability and sweat wicking, improving signal stability, adhesion strength and wearable comfort under sweating. b, Schematic of the integrated system-level sweat-permeable electronics, consisting of permeable electrodes, 3D LD and flexible circuit board. Blue arrow indicates the pathway of the sweat from the skin to the outlet. Exploded view illustrates the unidirectional sweat transport through the electrode, VLD and HLD. c, ECG signals recorded by commercial and permeable electrodes before and after exercise. d, Adhesion-strength evaluation of electrode–skin interface before and after a 20-min basketball workout. Bar height, mean; error bars, s.d.; n = 6 independent samples. e, Skin-irritation assessment of different electrodes after 3 days of wear. Bar height, mean; error bars, s.d.; n = 6 independent samples. f, Sweat-permeable skin-integrated and textile-based electronics for enhanced comfort and long-term monitoring. a.u., arbitrary units.

Source Data


Design of the permeable integrated system

The idea of integrated sweat-permeable electronics is illustrated in Figs. 1b and 2a, in which orderly integration of serpentine metallic interconnects serves as permeable electrodes and the 3D LD both serves as a substrate for directional sweat expulsion and supports multifunctional circuits/electronics with technical mature processing methods. Parameters of human sweat ducts are approximately associated with a diameter of 100 μm and a minimum interval of 400 μm (refs. 26,39) (Supplementary Fig. 2). We used serpentine meshed interconnects of polyimide (PI)/gold (Au) ribbons with a coverage ratio of 30% as the electrode (Fig. 2b). This design not only offers conformal contact with the skin for maintaining low skin–electrode impedance and stable electrical conductivity under deformation but also exhibits superior permeability of the electrode (Fig. 2c,d and Supplementary Figs. 35). The VLD, featuring a hydrophilicity gradient channel, can rapidly pump sweat generated at the skin–device interface to the back side (Fig. 2e, Supplementary Fig. 6a and Supplementary Note 2). The spatially distributed design of VLD channels enables sweat to form as droplets rather than disperse onto the surface, unlike the conventional Janus fabric, which is beneficial for subsequent detachment (Fig. 2f). To guarantee unobstructed sweat discharge towards the outlet, we also introduced a HLD with in-plane liquid-transport ability (Fig. 3a, Supplementary Fig. 6b and Supplementary Note 2). The HLD allows sweat to spread inside the channel with a default pathway to enable it to ultimately reach the sweat collector, at which it either evaporates into the surrounding environment or flows through the outlet and subsequently drips. Because all of the pores for dripping sweat are designed at the edge of the 3D LD substrate, the wearable electronics can be directly integrated into the middle of the substrate without impeding the perspiration pathway (Fig. 1b).

Fig. 2: Characterizations of the permeable electrode and VLD.
figure 2

a, Cross-sectional representation of the 3D LD, demonstrating unidirectional sweat transport from the skin–device interface to the outlet. b, Design of serpentine interconnect to facilitate open channels above sweat pores. Scale bar, 0.5 mm. c, Permeable electrode maintains stable conformal contact on the fingertip under perspiration conditions. Scale bar, 2 mm. d, Magnified view illustrating sweat wicking within the open channels, with dashed yellow circles highlighting sweat droplets. Scale bar, 0.5 mm. e, Mechanism of the unidirectional sweat transport in the VLD. f, Photograph of the VLD in the sweat-wicking state. Scale bar, 5 mm. g, Anti-gravity sweat transport in a single channel of the VLD under a flow rate of 0.9 ml min−1. Scale bars, 1 mm. h,i, Analysis of sweat-transport rate (h) and spot size (i) of the single channel under various processing conditions. j, Sweat-rate mapping of the human body during physical activity and the VLD. k, WVTR comparison between pristine fabric (PF), superhydrophobic fabric (SF) and the VLD. Bar height, mean; error bars, s.d.; n = 5 independent samples. l, Sweat-transport rate of the VLD under stretching, bending and twisting. Bar height, mean; error bars, s.d.; n = 5 independent samples. m, Long-term stability assessment of the single channel in the VLD. Points, mean; error bars, s.d.; n = 5 independent samples.

Source Data

Fig. 3: Characterizations of the HLD and 3D LD.
figure 3

a, Unidirectional sweat-transport mechanism in the HLD. b, Scanning electron microscopy images of the supporting structure, pillars and surface coating in the HLD. Scale bars, 300 μm (support), 30 μm (pillar) and 2 μm (coating). Experiments were repeated three times with similar results. c, Computational modelling predictions of normal-force distributions in the HLD under stress conditions. d, Contact-angle measurements of various PDMS surfaces: pristine PDMS, plasma-treated PDMS, PVA-coated PDMS and PVA/SiO2 nanoparticles-coated PDMS. e, Time-dependent contact-angle measurements of PDMS surfaces. Points, mean; error bars, s.d.; n = 3 independent samples. f, Sweat-transport rate in the HLD with varying surface states. Bar height, mean; error bars, s.d.; n = 5 independent samples. g, Unidirectional sweat-transport capabilities of the HLD. Scale bars, 1.5 mm (cross-section views), 5 mm (top views). h, Comparison of PDMS membrane and 3D LD in terms of gas and sweat permeability. Scale bars, 0.5 cm. i, Continuous sweat transport in the 3D LD under bending and stretching conditions. Scale bars, 1 cm. j, Fluid simulation of sweat transport in the deformed 3D LD.


Characterizations of the VLD

The development of the VLD starts from the superhydrophobic treatment of a polyester fabric (150 μm thick; Supplementary Figs. 7 and 8). Single surface dip-coating of a water-based pressure-sensitive adhesive enables fabric exhibiting strong adhesion while maintaining its porous structure (Supplementary Figs. 912). Selective plasma treatment is used to generate a wettability gradient throughout the exposed channel vertical direction (Extended Data Fig. 1a). The exposed area (plasma-treated region) and plasma-treatment duration are closely associated with the performance of the VLD. We first measured the contact angles of a single exposed channel under varying treatment conditions. The results show that fabrics with larger exposed areas presented directional sweat-transport abilities at shorter treatment duration, whereas longer treatment duration breaks through the liquid diode (Extended Data Fig. 1b). A needle featuring controllable flow was used to evaluate the sweat-transport capabilities of the VLD under different treatment conditions (Fig. 2g, Supplementary Fig. 13 and Supplementary Video 1). A trade-off between the exposed area and treatment duration affects the sweat-transport rate, for which the maximum flow rate of a single channel can reach 1.4 ml min−1 (mask size of 1.5 mm, treatment duration of 0.5 min; Fig. 2h).

As well as the sweat-transport rate, the final size of the hydrophilic area (physically treated region, owing to the plasma diffraction) is another crucial factor influencing the performance of the VLD, the channel arrangement, density and surface tension. Figure 2i illustrates the dependence of hydrophilic area on treatment conditions, indicating that the practical area typically exceeds the exposed area. This phenomenon arises from plasma diffraction/diffusion within the channel along the porous fabric, as confirmed by surface-tension measurements (Supplementary Fig. 14). Despite the peak transport rate of a single channel at a mask size of 1.5 mm (1.4 ml min−1), the extensive hydrophilic area (1.76 mm diameter) complicates the close arrangement of the channels. Conversely, a mask size of 0.2 mm results in the smallest hydrophilic area (0.87 mm diameter), but the low sweat-transport rate (0.3 ml min−1) limits its efficacy in sweat discharge. Consequently, the optimal design parameters were found to be a mask size of 0.4 mm and a duration of 3 min, which simultaneously presented a high sweat-transport rate (0.9 ml min−1) and a small hydrophilic area (0.92 mm diameter). Further parameters, such as fabric thickness, dip-coating repetitions and superhydrophobic treatment duration, were also studied for their influence on the performance of the VLD (Supplementary Figs. 1517 and Supplementary Note 3). The channel-density arrangement was corroborated by the sweat-cohesion test of adjacent channels (Supplementary Fig. 18), which showed that sweat droplets more readily connect at neighbouring channels with small intervals. The results show that the VLD, with a 3-mm interval, concurrently presents a high channel-array density and a low droplet-cohesion trend, thus proving to be the optimized parameter for the VLD. With this optimal design, the VLD shows a high sweat-transport rate of 11.6 ml cm−2 min−1, which is three orders of magnitude superior to the human sweat rate during mild exercise (0.38–2.85 × 10−3 ml cm−2 min−1)40 (Fig. 2j and Supplementary Fig. 19).

Furthermore, we assessed two key performances of the VLD, namely, breathability (water vapour transmission rate, WVTR) and stability (both mechanically and long-term operationally). In comparison with the pristine fabric, both superhydrophobic fabric and the VLD exhibit negligible differences in the WVTR (Fig. 2k). The VLD also showcases consistent sweat-transport capabilities under varying deformations and different batches (Fig. 2l and Supplementary Figs. 20 and 21). In the stability test over 30 days, artificial sweat at 0.6 ml cm−2 day−1, mimicking the daily sweat loss of football players41, passing through the VLD, the device consistently showed a high transport rate of 8.1 ml cm−2 min−1 even after one month (Fig. 2m). Note that this value is still far greater than the typical sweating rate of humans during physical exercise40.


Characterizations of the HLD

The unidirectional sweat-transport capability of the HLD is attributed to the meticulously engineered gradient hydrophilic micropillars with a careful mechanics design (Fig. 3a). A coating of superhydrophilic material (polyvinyl alcohol (PVA) and silica nanoparticles composite) on the polydimethylsiloxane (PDMS) micropillars enables long-term durable hydrophilicity to the surface (Fig. 3b and Supplementary Fig. 22). These spatially distributed micropillars exhibit uniform dimensions and height, whereas cross-shaped supports safeguard the micropillars against damage under vertical pressure (Fig. 3b,c). Finite element analysis (FEA) modelling corroborates the safeguarding role of the cross-shaped supports, for which it can be found that stress predominantly concentrates at the supporting structures when force is applied (Fig. 3c and Supplementary Fig. 23). The support parameters are established on the basis of an assessment of permeability and mechanical properties, showing that breathability of the HLD increases with the height of the supporting structure, achieving its maximum at a 4-mm interval (Supplementary Fig. 24). FEA findings reveal a substantial increase in film deflection under gravitational influence when the interval exceeds 4 mm, suggesting insufficient support for the film (Supplementary Fig. 25). Despite the 2-mm-interval support offering superior mechanical properties, the diminished permeable space substantially impedes the breathability of the HLD. The experimental and theoretical findings corroborated that the optimal interval is 4 mm. The micropillar arrangement notably affects the sweat dynamics of the HLD and we found the distance between adjacent rows of pillars to be the critical factor (Supplementary Fig. 26). Experimental evidence demonstrates that the transport rate of the HLD diminishes as the interval between rows increases, whereas the pinning breakdown position exhibits an inverse trend (Supplementary Figs. 2731). Taking this trade-off into account, an optimized parameter of 100 µm for the interval of the rows is established, resulting in both a high transport rate and an extensive breakdown distance.

PDMS is recognized as an advanced material for wearables and microfluidics, but its inherent hydrophobicity prevents its adoption in self-pumping liquid-transport systems (Fig. 3d). We used a composite coating (PVA/SiO2) for extended surface superhydrophilic treatment of the PDMS (Supplementary Fig. 32). The PVA/SiO2-coated PDMS surface showed superhydrophilicity over a 30-day testing period (Fig. 3e). We found that the contact angle of the surface would substantially affect the transport rate of the HLD, for which smaller contact angles lead to increased transport rates (Supplementary Figs. 33 and 34). The transport rate of the HLD with a superhydrophilic coating can greatly increase following surface moisturization by sweat (Fig. 3f,g, Supplementary Fig. 35 and Supplementary Video 2). Another determinant of the transport rate is the height of the micropillars, for which the transport rate increases as the micropillar height increases (Supplementary Fig. 36). However, a trade-off exists between the transport rate and structural stability. Height of the micropillars greater than 100 μm would cause broken pillars and bent morphology owing to the high aspect ratio (Supplementary Fig. 37). Moreover, pillars with a high aspect ratio also exhibit structural instability under sweat-flow flushing, and the experimental results are consistent with the FEA (Supplementary Figs. 3842). Ultimately, the diameter of the micropillars also emerges as a crucial factor to be optimized for amplifying the sweat-discharge-process efficiency, which modulates the dynamic flow within the HLD, especially in terms of fluid velocity and discharge flow rates (Supplementary Fig. 43). The flow velocity amplifies with the enhancement of the pillar diameter, whereas the flow rate peaks at a diameter of 50 µm. With optimized parameters (50 µm diameter, 100 µm height), the HLD demonstrates stable transport capability under deformation and prolonged sweat flushing (Supplementary Figs. 44 and 45).

Characterizations of the 3D LD

The 3D LD integrated with the VLD and the HLD can be realized by inserting an annular hydrophilic polyethylene terephthalate (PET) fabric, which serves as both the adhesive layer and the perspiration collector (Supplementary Fig. 46). The 3D LD shows superior capability in the transport of sweat from the skin to the designated outlets (Supplementary Fig. 47). FEA modelling of velocity magnitude, flow streamlines, pressure contours and spatiotemporal volume fraction also proves the high efficiency of the 3D LD (Extended Data Fig. 2 and Supplementary Video 3). Figure 3h shows the comparison of the perspiration performance between the 3D LD and the most commonly used PDMS substrate, in which a simplified artificial sweating model simulates the perspiration process (Supplementary Video 4). This model incorporates an internal chamber with a perforated surface and is connected to a syringe pump with a flow rate of 0.1 ml cm−2 min−1 (nearly 40 times the human sweat rate during physical exercise)40. We positioned the PDMS thin-film substrate and the 3D LD on top of the perspiring pores to examine fluid dynamics under continuous sweat flow. The continuous liquid injection heightened interfacial pressure, leading to the formation of a dome shape in the PDMS film because of its low gas and water permeability. In this scenario, further liquid injection caused the fluid to breach the interface between the PDMS and the artificial skin, resulting in a sweat leak and, thus, a delamination. Conversely, the 3D LD presents superior permeability, permitting unhindered transport of gas and sweat within the internal channels to the outlets, thus maintaining conformal contact with the artificial skin. Both experimental and FEA outcomes reveal that the 3D LD sustains stable sweat-transport capabilities under various deformations (Fig. 3i,j). Furthermore, cytotoxicity assessments indicate that the 3D LD exhibits great biocompatibility, with no discernible differences between the samples and the control group (Extended Data Fig. 3).


User study of the 3D LD

In the user study, we conducted comprehensive assessments to investigate the impact of sweating/moisture management of different wearable devices/substrates on both wearing comfort and device performance. The assessments include adhesion tests, evaluations of skin thermal comfort and comparison of device collected signal quality. To carry out adhesion evaluation, we affixed four different types of patch, including commercial electrocardiogram (ECG) patches, PDMS film, VLD and 3D LD, to the test subjects. The subjects then engaged in various physical activities while we captured images of the patch status (Supplementary Fig. 48a). Over the 30-min test period, we observed that the commercial ECG and PDMS patches detached from the skin during the test, whereas the VLD and 3D LD patches remained firmly attached to the body (Fig. 4a). We found that the adhesion strength of the patches under sweating conditions exhibits a positive correlation with the permeability and softness of the device (Fig. 4b and Supplementary Fig. 48b).

Fig. 4: Evaluation of moisture management on wear comfort and device performance.
figure 4

a, Adhesion test was performed on commercial ECG electrodes (1), PDMS (2), VLD (3) and 3D LD (4) when attached to the skin during physical activity. Scale bars, 5 cm. b, Adhesion strength between patches and skin at different time intervals during exercise. Points, mean; error bars, s.d.; n = 3 independent tests. c, Skin-irritation results for various patch types on the forearms of three volunteers. Insets show a closer view of the skin condition, magnified three times. Scale bars, 5 cm. d, Participants provided feedback on the wearing duration of different patches during a 12-h test period. Bar height, mean; error bars, s.d.; n = 10 volunteers. Statistical significance was assessed using two-sided unpaired t-tests. e, WVTR of the commercial ECG electrode, PDMS, VLD and 3D LD. Bar height, mean; error bars, s.d.; n = 5 independent samples. f, Infrared images of various patches on skin before and after exercise. Scale bars, 5 cm. g, Temperature variations of the skin were compared while wearing patches against that of bare skin after exercise. Bar height, mean; error bars, s.d.; n = 9 independent spots. h, Images show commercial electrodes (1), PDMS/PI/Au mesh electrodes (2) and 3D LD/PI/Au mesh electrodes (3) on skin before and after exercise. Scale bars, 5 cm. i, ECG signals recorded from different electrodes before and after exercise. a.u., arbitrary units; TEWL, transepidermal water loss.

Source Data

Next, we conducted a long-term wearing test, in which we affixed different patches to the forearms of three individual subjects for 3 days of continuous use. We observed no irritation or inflammation for the skin wearing the VLD and 3D LD patches, whereas the commercial ECG and PDMS patches caused obvious skin erythema (Fig. 4c). Furthermore, we applied ten 3D LD patches to various parts of a volunteer’s body and no irritation was observed during the 12-h wear period (Supplementary Fig. 49). We carried out another user test to assess the impact of permeability on user wear comfort. We affixed the four types of patch to the forearms of ten subjects (five male and five female) and instructed them to wear the patches for 12 h. The subjects were allowed to remove the patches if they experienced any discomfort and we recorded the duration that they wore each patch. The results indicated that the VLD and 3D LD last for the longest wear durations compared with the others (Fig. 4d). The excellent vapour and sweat permeability of the VLD and 3D LD contributed to their superior wear-comfort capability (Fig. 4e). Owing to the rapid transport and evaporation of sweat, the VLD and 3D LD demonstrated better thermal comfort during physical activity compared with the commercial ECG and PDMS patches (Fig. 4f and Extended Data Fig. 4). Infrared images revealed that, after running for 15 min, the average skin temperature increased by nearly 2.2 °C under the commercial ECG electrode, whereas the VLD and 3D LD caused an average temperature increase of only 0.4 °C (Fig. 4g).

Finally, we evaluated the effect of sweat accumulation on biosignal acquisition using three types of electrode: commercial ECG electrodes, PDMS/PI/Au mesh-based dry electrodes and 3D LD/PI/Au mesh electrodes. Although the integration of the Au mesh to the 3D LD marginally increases the skin–electrode impedance, the 3D LD allows superior adhesion and insulates against ambient interference (Supplementary Fig. 50). We affixed electrodes to a subject in a single-lead ECG configuration and tested the ECG signals of the subject before and after 10 and 20 min of exercise (Fig. 4h,i). All electrodes show high-quality ECG data before exercise/sweating. However, after 10 min of running, the signals of the commercial ECG and PDMS-based electrodes became unstable with clearly decreased signal-to-noise ratio. The signal-to-noise ratios of the commercial ECG and PDMS electrodes continued to deteriorate with longer duration of exercise/sweating, becoming progressively more difficult to read as more perspiration accumulated at the electrode–skin interface. By contrast, the 3D LD-based electrodes featured great moisture permeability, which provided stable skin–electrode impedance and ECG signals both before and after sweating (Supplementary Fig. 51).


Integrated system-level sweat-permeable electronics

We demonstrate the universality and scalability of the 3D LD technology in two exemplary wearable electronic devices: a thin, soft ECG monitoring system and a textile-integrated weather station (Fig. 5a,b). The ECG monitor uses a detachable design through a magnetic-coupling method. Component 1 is the vapour/sweat-releasing substrate, with a thickness of 0.65 mm, weight of 1.05 g, bending stiffness of 55 kPa and an elastic modulus of 106 kPa. Component 2 represents the flexible, waterproof circuitry layer, with a thickness of 2.1 mm and weight of 1.55 g including the battery (Supplementary Figs. 5254). The soft, stretchable and wireless circuit was developed by well-established skin-electronics technologies11,42, in which a vertical interconnect access connects the skin electrodes to the electronic connectors. (Fig. 5c and Supplementary Figs. 55 and 56). The sophisticated circuit design ensures stable functionality under various deformations (Extended Data Fig. 5 and Supplementary Video 5). We affixed the device on the chest of the subject and the ECG data were wirelessly transmitted to a smartphone (Fig. 5d and Supplementary Video 6). We compared the motion artefacts between our device and the conventional cable-connected type in daily activities and exercises, including walking, jumping and cycling (Extended Data Figs. 6 and 7 and Supplementary Video 7). Our device exhibits superior ECG readouts under both conditions, owing to its low weight, conformable design and strong adhesion during the activities. The infrared images also exhibit lower thermal accumulation in our device (Extended Data Fig. 8). To conduct long-term and user-friendly ECG monitoring, we used our ECG monitor for this purpose, as it exhibits exceptional moisture permeability and conformability. Owing to battery-storage limitations, we designed a magnetic charging cable to charge the device when needed (Supplementary Fig. 57). Throughout the testing period, parameters such as heart rate (HR), QT interval (QTO) and heart-rate variability (HRV) were successfully monitored (Fig. 5e). On a day (Wednesday) during the week when the subject exercised, our device demonstrated superior capabilities in maintaining dry skin conditions and producing stable signals during perspiration (Extended Data Fig. 9). Furthermore, an increase in HR and a decrease in QTO and HRV were observed as a result of the physical activity (Fig. 5f). By increasing the number of layers of the circuit, the footprint of the circuit can be minimized and the device can be further integrated into a more compact format and, therefore, used as a multifunctional sensor, that is, for continuous electromyography (EMG) monitoring (Extended Data Fig. 10). Another demonstration involves textile-integrated moisture-permeable electronics, as conventional textile electronics typically present a trade-off between permeability and functionality or integration level. To address this issue, we developed a battery-free, flexible, permeable, multifunctional and wireless textile-integrated electronic device, functioning as a T-shirt-based weather station capable of monitoring temperature, humidity, ultraviolet (UV) index, atmospheric pressure and altitude (Fig. 5g and Supplementary Fig. 58). Using near-field communication technology, weather data can be transmitted to a mobile device, as demonstrated during a hiking expedition (Fig. 5h,i). These integrated system-level sweat-permeable wearable electronic implementations indicate the adaptability and scalability of our 3D LD technology.

Fig. 5: Breathable and sweat-permeable skin-integrated and textile-based electronics.
figure 5

a,b, Exploded-view illustrations of the skin-integrated (a) and textile-based (b) devices featuring a detachable design, highlighting key layers. Component 1 represents the vapour/sweat-discharging substrate and component 2 refers to the flexible circuitry layer. A magnified view shows the magnetic-coupling mechanism between the two parts. c, Optical images of the permeable ECG monitor in a detached configuration (i) and a twisting state (ii). Scale bars, 1 cm. d, Photograph of the permeable ECG monitor on the upper chest, with signal transmission occurring wirelessly to a mobile phone. Scale bar, 5 cm. e, Daily ECG outcomes collected using the proposed device over a 1-week period. Standard deviation of R-R (SDRR) intervals reflects the HRV. Squares, mean; centre lines, median; box limits, upper and lower quartiles; whiskers, 1.5× interquartile range; points, outliers; n = 72 independent spots. f, Changes in HR, QTO and SDRR before and after physical activity. Points, mean; error bars, min–max; n = 6 independent spots. g, Photographs of the permeable textile-integrated weather station. Scale bars, 5 cm (i), 2 cm (ii). h, Optical image of a subject wearing the T-shirt weather station while hiking. Scale bar, 5 cm. i, Real-time meteorological data captured during the hiking excursion.

Source Data



The 3D LD offers integrated system-level moisture permeability, surpassing human perspiration rates by three orders of magnitude, which enables seamless, continuous and thermally comfortable healthcare monitoring. Comprehensive user studies show its superior performance in maintaining a stable and comfortable skin–device interface during sweating conditions. The ECG monitor based on the 3D LD concept demonstrates reduced motion artefacts compared with conventional counterparts, yielding reliable signals over a week-long duration. This positions 3D LD technology as a promising candidate for user-friendly, long-term healthcare monitoring wearables. Optimizing functional materials, device architecture and manufacturing techniques is essential for enhancing its capabilities and achieving mass production. Incorporating a washable and recoverable adhesive backing is expected to improve the reusability and cost-effectiveness of the device.



Fabrication of the 3D LD

The fabrication process is divided into the fabrication of the VLD43 and the HLD and assembly processes. Fabrication of the VLD began with laser structuring of the hydrophilic polyester fabrics (LPKF ProtoLaser U4), followed by cleaning with acetone, isopropyl alcohol and deionized water and then drying at 50 °C for 2 h. 4 ml of 1H, 1H, 2H, 2H-perfluorooctyltriethoxysilane, 10 g of P25 TiO2 nanoparticles and 1 ml of deionized water were dispersed into 200 ml of ethanol by means of stirring for 30 min to form a suspension. The tailored fabrics (150 μm thick) were immersed into the suspension, stirred for 2 h and dried at 50 °C for 2 h to yield the superhydrophobic fabric. Single surface dip-coating in the water-based acrylic pressure-sensitive adhesives formed the permeable adhesive network (Supplementary Fig. 10) and the optimal coating repetition is five times. Silicone PET release film served as the protective layer of the back adhesive and laser-structured PI tapes (50 μm) served as the masks of the fabrics for selective oxygen plasma treatment (200 W, 200 mTorr, O2 flow rate of 50 sccm, PlasmaTherm 790).

The HLD system fabrication was based on photolithography and replica moulding of PDMS (10:1 of base to curing agent, Sylgard 184, Dow Corning). The layout was designed with computer graphics software (AutoCAD, Autodesk) and transferred by chrome masks. Fabrication started from spin casting a 100-μm-thick film of negative photoresist (SU8 2050) on an O2-plasma-treated 4″ silicon wafer followed by soft baking at 65 °C for 3 min and 95 °C for 20 min on a hot plate. Exposing the wafer to UV light through a mask patterned the photoresist. The wafer was post-exposure baked at 65 °C for 5 min and 95 °C for 10 min, followed by immersing in the developer solution for 10 min. Laminating laser-patterned PET films to the wafer served as the mould. Moulds were treated with octadecyltrichlorosilane (OTS) by vapour silanization to facilitate the release of the PDMS micropillars. PDMS was poured into the mould, followed by degassing under vacuum for 1 h and then baking at 80 °C for 30 min. After release from the mould, a mechanical punch aligned by the laser-structured PET mask defined the outlet holes. 0.1 g of hydrophilic fumed silica and 0.05 g of sodium dodecyl sulfate were dispersed into 10 ml of 1 wt% PVA solution to form the hydrophilic coating after sonication for 10 min. Spray coating the solution onto the plasma-treated PDMS microfluidic film performed on a hot plate at 110 °C yielded the long-term superhydrophilic layer.

Laser-structured polyester fabric (150 μm thick) served as the sweat collector after cleaning with acetone, isopropyl alcohol and deionized water and then drying at 50 °C for 2 h. The VLD, sweat collector and HLD were bonded using the Sil-Poxy silicone adhesive.

Fabrication of the PI/Au open-mesh electrode

Spin casting PDMS (500 rpm, 30 s) on a temporary glass sheet was followed by baking at 100 °C for 10 min. 20-μm-thick PI film was laminated on the PDMS and patterned into open-mesh architectures (trace width of 50 μm) by laser cutting (LPKF ProtoLaser U4). Water-soluble tape (WST) was used to pick up the completed patterns from the glass substrate. Ti/Au (5/50 nm) thin film was deposited on the patterned PI films by electron-beam evaporation. WST was removed by immersing it into deionized water, forming the self-supported open-mesh electrode.

Fabrication of the wireless ECG monitor

The ECG monitor includes two parts, the vapour/sweat-discharging substrate and flexible-circuit components. Two-step casting PDMS into the mould and assembling laser-patterned iron discs (0.25 mm diameter and 0.15 mm thickness; LPKF ProtoLaser U4) formed the HLD. The thickness, length, width and weight of the VLD are 150 μm, 65 mm, 30 mm and 0.3 g, respectively, whereas the HLD measures 500 μm in thickness, 65 mm in length, 30 mm in width and weighs 0.6 g. After assembling as the 3D LD, three through holes (0.5 mm in diameter) in PDMS were defined by a mechanical punch. PI/Au open-mesh electrodes passed through the fabric and through holes and were bonded with iron discs by the silver paste followed by baking at 80 °C for 30 min. The punched-out PDMS pillars were restituted to the through holes and bonded by the Sil-Poxy silicone adhesive. PET release film was attached to the bottom of the substrate to protect the adhesive.

The flexible circuit began with laminating the laser-structured PET sheet to the silicon wafer as the mould, followed by placing the magnets (1.5 mm diameter and 0.5 mm thickness) and pouring PDMS into the mould and curing at 80 °C for 10 min. Moulds were treated with OTS by vapour silanization to facilitate release of the PDMS film. Spin casting PDMS (500 rpm, 30 s) on a temporary glass sheet was followed by curing at 100 °C for 10 min. Cu/PI sheet (Cu 18 μm, PI 12.5 μm) was laminated on the prepared PDMS/glass substrate and then structured into the designed pattern by laser cutting (LPKF ProtoLaser U4). The pattern was picked up by WST and Ti/SiO2 (5/50 nm) thin film was deposited by electron-beam evaporation. The first layer of patterned PI/Cu interconnection was directly laminated on the moulded PDMS substrate after activating by UV ozone for 5 min and heated at 80 °C for 10 min. The WST was removed by deionized water. Laser-patterned PET stamps (0.6 mm diameter and 1 mm thickness) were placed using thermally sensitive adhesive on the Cu pads. Repeated pouring of PDMS into the mould and curing at 45 °C for 12 h was followed by removing the PET stamps at 60 °C, forming the vertical interconnect accesses through the PDMS. The second layer of the PI/Cu pattern was transferred onto the PDMS in the same way. Soldering the electrical components (microcontroller, resistors, capacitors, antenna, light-emitting diode and so on), magnets, battery and Cu interconnected pads was carried out using low-melting-point solder paste (Supplementary Fig. 59). The circuit was sealed using PDMS (about 100 μm thick) and cured at 80 °C for 10 min, followed by removal from the mould and laminating the PET release film. A PDMS cap was moulded to protect the magnets from ambient sweat. For operation, magnets in the flexible circuit and iron discs in the substrate were automatically aligned and mechanically coupled, building the electrical conduction from the electrode to the circuit.

Fabrication of the magnetic charging cable

The top and bottom covers were fabricated by photocuring 3D printing. The USB cable was assembled with the top cover and soldering the Cu wires onto the magnets (1.5 mm diameter, 1 mm thickness) was completed with low-melting-point solder paste, followed by installing the bottom cover.

Fabrication of the textile-integrated wireless weather station

The wireless weather station includes the textile-integrated vapour/sweat-discharging substrate and flexible circuit. Fabrication of the vapour/sweat-discharging substrate followed the same method as the ECG monitor but without creating the back adhesive and electrode. The diameter of the VLD (46 mm) was larger than that of the HLD (36 mm). Laser structuring clothing with a hole (36 mm diameter) was followed by bonding the 3D LD to the clothing using the Sil-Poxy silicone adhesive. A customized flexible circuit board was prepared and then electrical components and magnets (1.5 mm diameter and 0.5 mm thickness) were soldered on (Supplementary Fig. 60). The flexible circuit board was placed into the 3D-printed mould and PDMS with white pigments (Silc Pig, Smooth-On, Inc.) was poured, followed by degassing and curing at 80 °C for 10 min. PI shadow masks were used to insulate the sensors (temperature, humidity, UV and pressure) from the PDMS.

Contact-angle measurement

Contact angles were measured using an optical contact-angle meter. The PDMS films were mounted on the glass sheets and the fabric strips were fixed to the 3D-printed frame. Liquid drops with a volume of 10 μl were used to check the contact angle and transport direction. A linear motor was used to measure the contact angle of the fabrics at various strains.

Anti-gravity sweat-transport test of the VLD

VLDs were fixed on the 3D-printed frame and a needle (0.2 mm inner diameter) was placed beneath the selectively plasma-treated spot with a gap of 1 mm. Continuous artificial sweat flow with a controlled rate was generated from a syringe pump. A camera was placed in front of the fabric to record the sweat-transport behaviour.

Sweat-transport test of the HLD

Artificial sweat droplets (20 μl) with red dye were dripped onto the HLD with different surface properties. A camera was set above to capture the sweat-transport process and sweat-transport rates were calculated by dividing the transport distance by the time. For the cross-section view, the HLD was cut radially and the camera was set in front. The real-time morphology variation of the micropillars during sweat flow was captured using an inverted microscope (ECLIPSE Ts2R, Nikon).

Sweat-transport test of the 3D LD

Spin casting PDMS with flesh pigments (Silc Pig, Smooth-On, Inc.) on a temporary glass sheet was followed by baking at 80 °C for 10 min. Using laser structuring, holes (1 mm diameter and interval) through the PDMS was completed. Pouring the PDMS with flesh pigments into the 3D-printed mould and baking at 80 °C for 30 min formed the substrate with an open chamber. This was followed by the assembly of the perforated PDMS film, silicon tube (0.3 mm diameter) and PDMS substrate using Sil-Poxy silicone adhesive, to simulate the sweat glands. The imitative sweat glands were covered by the 3D LD and continuous artificial sweat flow at a controlled rate was produced using a syringe pump. Two cameras were placed in front and above the device to record the deformation of the 3D LD and the sweat-flow process. For the sweat-wicking test, a drop of artificial sweat droplets (20 μl) with red dye was dripped onto the forearm of the volunteer. The 3D LD was covered on the droplet for 4 s and then it was removed. A camera was set above to capture the sweat flow in the 3D LD.

WVTR measurement

The WVTR test was conducted at 17 °C with a humidity of 67%. 30 ml of deionized water was placed in a 50-ml bottle (15 mm diameter) and completely sealed by the samples. The water mass lost was tested after 12 h. The WVTR was calculated by dividing the mass lost by the dimensions of the opening area and the time duration.

Stability test

A stability test was conducted on the VLD and HLD over a period of 30 days. The skin surface area of a person is estimated at 1.8 m2 and the sweat-loss volume is approximately 10 l during long-term physical work41. Thus, the daily sweat use in the test is set to 0.6 ml cm−2. The VLD was securely fastened onto a 3D-printed frame and a needle (0.2 mm inner diameter) was placed beneath the selectively plasma-treated spot with a gap of 1 mm. Continuous artificial sweat flow with a controlled rate was generated from a syringe pump. The working area of the HLD in Fig. 3g is 2.2 cm2, hence the daily volume of sweat needed is estimated at about 1.3 ml. Continuous artificial sweat flow (0.1 ml min−1) was generated from a syringe pump.

Resistance stability of the open-mesh electrode

Spin casting PDMS (500 rpm, 30 s) on a temporary OTS-treated glass sheet was followed by baking at 100 °C for 10 min. The PI/Au open-mesh electrode was picked up by WST and Ti/SiO2 (5/50 nm) thin film was deposited by electron-beam evaporation. The PI/Au open-mesh electrode was directly laminated on the PDMS substrate after activating by UV ozone for 5 min and heated at 80 °C for 10 min. The PDMS substrate was fixed onto a biaxial stretching system and the resistance of the electrode at various strains from 0% to 30% was measured.

Cell cytotoxic test

The mouse embryo fibroblast cell line, NIH/3T3 (ATCC), served as the model for evaluating in vitro cytotoxicity of the VLD and the 3D LD. These cells (1 × 104 cells ml−1) were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Hyclone) supplemented with 10% foetal bovine serum and 1% penicillin/streptomycin (Procell) at 37 °C, in a humidified incubator with 5% CO2. Specimens were cut into 1-cm-diameter discs, soaked in culture medium overnight and seeded with NIH/3T3 cells in a 24-well plate. Cells were cultured for 24, 48 and 72 h. After removing the medium, a phosphate-buffered saline solution containing 2 μM calcein-AM and 8 μM EthD-1 was added and the samples were incubated for 30 min before examination with a fluorescent microscope. For cell-proliferation evaluation, a CCK-8 assay (Beyotime Biotechnology) was conducted as per the manufacturer’s instructions. Briefly, CCK-8 solution was added to each well at 10% of the culture medium’s volume, followed by a 4-h incubation. The absorbance at 450 nm, indicative of cell metabolic activity, was measured using a microplate reader.

Human experiment

Volunteer experiments were conducted to assess the impact of sweat on both device performance and the epidermis, including adhesion test, biocompatibility test, skin-temperature evaluation and ECG signal collection. Four samples (commercial ECG electrode, PDMS, VLD and 3D LD) were attached to the chest, upper back and right forearm for the adhesion test. The volunteer played basketball outdoors (28 °C, 72% humidity) for 30 min and pictures of the samples were captured at 5, 15 and 30 min. To assess the adhesion strength between various patches and the skin on the back, we performed a 90° peel test using a customized test system (Supplementary Fig. 48b). This system consists of a clip connected to a highly sensitive force gauge (resolution 0.001 N) by means of a 3D-printed connector. The stretching speed was regulated at nearly 50 mm min−1 and the adhesion force was denoted by the peak value measured before the complete detachment of the patches from the skin. First, we evaluated the adhesion strength of each patch type against dry skin. Then, six sets of samples (commercial ECG electrodes, PDMS, VLD and 3D LD) were affixed to the back of a volunteer. Following every 5-min interval of a basketball exercise session, the patches were peeled off and the corresponding peeling forces were documented. The skin-irritation test was carried out on human skin. Four samples (commercial ECG electrode, PDMS, VLD and 3D LD) were attached to the forearms of the volunteers. Patches were removed after being worn for 3 days and skin was captured before and after the test. Skin-reddening percentage was further quantitatively determined using ImageJ software. Ten volunteers participated in the wear comfort test and the test duration is 12 h. Four samples were attached to the forearms and the volunteers removed the patches if they felt any discomfort, then the wearing duration of each sample was recorded. The skin-temperature evolution under the patches after exercise was performed with a thermal imaging camera (FLK-TIS60, Fluke). Four patches (commercial ECG electrode, PDMS, VLD and 3D LD) were attached to the back of the volunteer, removed after 15 min of running and thermal imaging was carried out. The ECG test of a volunteer after exercise was performed with a commercial data-acquisition system (PowerLab). Patches (commercial electrode, PDMS/PI/Au mesh and 3D LD/PI/Au mesh) were attached to the chest and abdomen. The ECG signals of the volunteer at rest and after running for 10 and 20 min were recorded. For long-term ECG monitoring, the sweat-permeable ECG monitor was attached to the left chest of the volunteer. ECG signals were recorded from 9 am to 9 pm, considering that the working period of the monitor is 5 h with a fully charged battery. A power bank with magnetic charging cable is used to charge the battery after the monitor had been working for 4 h. The ECG curves were analysed by Origin and MATLAB. For wearable weather station operation, the volunteer wore the T-shirt and hiked on the Beacon Hill trail, Hong Kong, recording the weather parameters every 2 min. All human experiments were conducted by protocols approved by the Human Subjects Ethics Sub-Committee of the Research Committee, City University of Hong Kong, Hong Kong, China, and guidelines were followed. Informed consent from all participants was obtained before inclusion in this study.

Mechanical simulation

To numerically analyse the mechanical response and optimize the design parameters of various parts of our electronic, FEA was conducted. The 3D models were first constructed by SOLIDWORKS 2022 and then meshed by MSC Apex 2022. All objects were meshed into linear hex elements. The mesh size was one-fifth of the width of the electrode, integrated circuit and silicone support models, ensuring convergency and accuracy. The mesh files were imported to Marc Mentat 2022 for finite element model construction. Friction was neglected in the simulation. The elastic modulus (E) and Poisson’s ratio (ν) applied in the simulation were ECu = 119 GPa and νCu = 0.34 for copper; EPI = 0.8 MPa and νPI = 0.5 for PI; EPDMS = 145 kPa and νPDMS = 0.5 for PDMS; and EAu = 77.2 GPa and νAu = 0.42 for gold.

Fluid simulation

We used FEA commercial software COMSOL to verify and optimize the design and performance of the 3D LD. To optimize the structural stability of the pillars under liquid flow, we created a 2 × 2 pillar array located in the bottom centre of a cube with a side length of 1 mm. The diameter of the pillar is 50 μm and the height (H) and interval between adjacent pillars (L) were varied. The density, Young’s modulus and Poisson’s ratio were 970 kg m−3, 970 kPa and 0.49, respectively. The density and viscosity of the fluid were 1,000 kg m−3 and 1.01 × 10−3 Pa s, respectively. The inlet velocity was set as Vinlet = 15sin(2πt) mm s−1 and the outlet pressure was given as 0 Pa. All simulations used an incompressible laminar flow model based on the Navier–Strokes equations. The displacement along the centreline of the pillar was analysed. 3D models were constructed to simulate the sweat-discharging process of the 3D LD. The arrangement of the sweat-discharging patch array was imported from its AutoCAD file and used to generate a layered structure consisting of a top layer of PDMS (0.5 mm thickness), a micropillar array (0.1 mm height) and a bottom layer textile (0.15 mm thickness) with an inlet (0.9 mm diameter). The diameter of the patch used for the simulation was 10 mm. The inlet flow rate and outlet pressure were set as 0.5 mm s−1 and 0 kPa, respectively. The simulation was set to consider the laminar two-phase flow of sweat (water) and its surrounding air (gas). To reveal the dynamic characteristics of the sweat-discharging process, we used the phase field module for time-lapsed computation. The temperature, reference pressure level and surface tension between the liquid–gas phase were 293.15 K, 1 atm and 72.8 mN m−1, respectively. Gravity of each phase was considered in the flow field simulation and the contact angle between liquid phase and the PDMS wall was 10°.