A full-range sweat rate sensor with superhydrophobic heterophase shunt for psychology, skin health and body dehydration evaluating

Abstract

Sweat secretion provides insights into physiological and psychological states. However, the current single detection method is not compatible with the simultaneous monitoring of gaseous and liquid sweat, which restricts its application for exploring dynamic physiological processes. Herein, a fully integrated sweat sensor with a heterophase shunt for full-range sweat rate monitoring is reported. Through non-interfering integration of epidermal microfluidic conductivity sensing and humidity detection, continuous measurement across 0.182 ∼ 10 μL/(min·cm²) is achieved. The LiCl-doped UiO-66-NH₂ enables a highly sensitive and fast-responding humidity sensor. The biomimetic heterophase shunt, composed of a polydimethylsiloxane and polymethyl methacrylate co-electrospun membrane, exhibits superhydrophobicity (151.9°), gas-permeable [58.07 g/(H·m²)], and water pressure tolerance (1.1 kPa). Heterophase shunt effectively integrates sweat detection modes while providing protection for humidity sensor against sweat corrosion and skin contact-induced damage. Integrated with a specialized weak-signal processing module, the sensor allows applications in psychology, skin health and body dehydration evaluating.

INTRODUCTION

Sweat is considered as a frontier non-invasive bio-sample in personalized medicine, owing to its wealth of biological constituents. As a pivotal physiological mechanism for hydro-thermal regulation and autonomic nervous stress response, sweat rate directly reflects hydration, psychological stress, autonomic nervous function, and skin health. The advent of novel wearable electronic technologies has enabled real-time and dynamic monitoring of sweat rate, establishing it as a key research direction and facilitating continuous access to precise health metrics (1–7).

Sweat can be physiologically classified as sensible sweat and insensible sweat, and physically as liquid or gaseous (8–10). Human sweat rate and sweat volume dynamically vary across multiple orders of magnitude from low to high levels. This represents a fundamental physiological characteristic observed both inter-individually (across race, gender, age, health status, etc.) and intra-individually (among body regions, lesioned areas, and physiological states such as rest vs. exercise or changing ambient temperature). Detection of liquid sweat is the mainstream paradigm in wearable sweat sensors. Current methods for measuring sensible sweat rate—drawing inspiration from liquid level and flow measurement technologies—include electrical, thermal, and optical techniques (11–16). Electrical methods, particularly those based on conductivity or capacitance, are favored for their stability, accuracy, and ease of integration. However, such methods possess a notable limitation, in the initial stage of sweating, the low secretion rate leads to rapid evaporation, causing the detected volume to underestimate actual production (17–19). This measurement bias can delay warnings for dehydration or skin health conditions, thus increasing the risk of physical impairment.

Extending the lower detection range of sweat rate remains a key objective in this field. The detection limit can be significantly reduced by introducing hygroscopic materials, droplet-driven structures, or minimizing the characteristic dimensions of microfluidic channels. Although some studies have reported microfluidic sensors capable of low sweat rate and wide-range detection, they often suffer from incompatibility between the detectable sweat rate range and measurable sweat volume (20–26). Skin electrodermal activity (EDA) measurement offers an effective approach for monitoring low-rate sweat before it forms collectible droplets, providing sensitive response to dynamic changes, yet its detection capability is limited for gaseous sweat or high-volume liquid sweat and fails to establish a direct mapping relationship with sweat rate (27, 28). An alternative method assesses sweat rate by monitoring skin surface humidity (30% ∼ 100% RH) (29–33). Nevertheless, humidity-based detection is prone to signal saturation under high sweat rate, failing to effectively track the dynamic secretion process of sensible sweat. Consequently, the dependence on distinct detection methods due to the diverse forms of sweat is the fundamental reason why full-range sweat rate monitoring remains unrealized. As a conclusion, owing to the multiple physical forms of sweat and its multi-order dynamic variation characteristics, single-mode sensors can no longer meet multi-scenario application requirements. The development of universal full-range sweat rate sensing methods that enable non-interfering integration of different detection modalities is an inevitable choice for exploring the dynamic variation of important human physiological information and developing high-performance sensors and related electronic products.

Therefore, this work presents a full-range sweat rate sensor (FRSR sensor) integrates a humidity sensor with a microfluidic conductivity-based sensor at a single site based on the concept of heterophase sweat shunt (Fig. 1A). This heterophase sweat shunt enables independent detection of both insensible and sensible sweat without interference. For sensible sweat detection, a microfluidic conductivity measurement method is employed to enhance measurement stability (Fig. 1B). Humidity sensing is utilized for insensible sweat detection, employing a LiCl-doped UIO-66-NH₂ (LiCl@UIO-66-NH₂) metal-organic framework (MOF) as the hygroscopic element. Compared to traditional materials, LiCl@UIO-66-NH₂ exhibits significant advantages in response speed, recovery time, and stability due to its high specific surface area and exceptional water adsorption capacity, thereby overcoming the limitations of conventional hygroscopic materials in wearable applications (34–37). To protect the hygroscopic element from erosion by sensible sweat and interference from body movement, a superhydrophobic, breathable membrane fabricated via polydimethylsiloxane (PDMS) and polymethyl methacrylate (PMMA) superhydrophobic fibrous membrane (P@P SFM) was designed and prepared (Fig. 1C) as a heterophase sweat shunt. Through systematic optimization of the preparation processes for the hygroscopic material and the superhydrophobic fibrous membrane, and the development of a conductivity detection module and a personal interactive application, the effective operational periods for sensible and insensible sweat detection from traditional methods are ultimately combined (Fig. 1D). The resulting sensor achieves a sweat rate detection range of 0.182 ∼ 10 μL/(min·cm²), covering the full spectrum of human sweating. Validated across multiple scenarios, FRSR sensor demonstrated a stronger correlation with actual water loss in dehydration state alerts, thereby establishing a foundation for the practical application of wearable sweat rate monitoring technology. In summary, the proposed dual-mode architecture and heterophasic sweat shunting mechanism successfully realize the interference-free integration of humidity and microfluidic conductivity sensors, ensuring the independence and accuracy of the two detection modes at the same site under the same scenario. The developed sensor exhibits obvious advantages in key indicators including measurement range, application scenarios, and applicability (table S1), and is expected to promote the practical development of wearable sweat detection technologies.

Fig. 1. Design of the wearable FRSR sensor.

(A) Schematic diagram of the structure of the wearable FRSR sensor. (B) Schematic diagram of the sensible sweat sensing. (C) Schematic diagram of the insensible sweat sensing. (D) The advantage of the wearable FRSR lies in its combination of the effective detection duration of both humidity and flow sensors. (E) Schematic diagram of the insensible sweat detection module. (F) Schematic diagram of the sensible sweat detection module. (G) Wearable full-range sweat rate system and mobile APP.

RESULTS

Overview of the wearable FRSR sensor

Given that the physical origins of sensible and insensible sweat necessitate different detection modalities, a spatially separated but functionally integrated sensing strategy was employed. This scheme co-locates the humidity sensor for insensible sweat with the microfluidic inlet for sensible sweat in direct skin contact, ensuring measurement simultaneity and positional consistency. At low rates, sweat volatilizes instantly, diffusing as vapor through the pores and the PDMS-PMMA superhydrophobic membrane to be adsorbed and detected by the LiCl@UIO-66-NH₂-based humidity sensor (Fig. 1E and fig. S1A). Whereas at high rate, sweat forms a continuous liquid phase that enters the microfluidic channel through the action of the heterophase sweat shunt, and its flow velocity is determined electronically via the interdigitated electrodes (Fig. 1F and fig. S1B). Benefiting from the integrated design and flexible electronics fabrication strategy, both the insensible and sensible sweat detection modules exhibit excellent structural conformity and mechanical flexibility, as shown in fig. S2. The conductance signals originating from these two parallel detection pathways are subsequently acquired and processed by the dedicated circuit module (Fig. 1G), which relays the data wirelessly to a user terminal via Bluetooth, thereby completing an integral system for personalized sweat rate assessment.

Synthesis and Optimization of LiCl@UIO-66-NH₂

Humidity sensing is pivotal for full-range sweat rate monitoring. Conventional hygroscopic materials, including conductive polymers, transition metal compounds, and two-dimensional materials, often suffer from the trade-off between response/recovery time and long-term stability, limiting their wearable applications (38). Herein, LiCl@UIO-66-NH₂ is developed as the sensing material. The MOF’s amino groups (–NH₂) are strongly polar and hydrophilic, providing high water affinity, while its porous structure offers abundant active sites. The incorporated LiCl dissociates upon hydration, serving as an effective dopant to increase the ionic carrier concentration. At low humidity, limited water adsorption restricts LiCl dissociation, leading to insufficient proton conduction pathways and low conductivity. As humidity increases, enhanced dissociation raises the concentration of mobile ions (e.g., Li⁺, Cl⁻), which migrate freely through the porous framework at high humidity, significantly boosting ionic conductivity (Fig. 2A). Electrochemical impedance spectroscopy (EIS) under varying humidity levels confirms this dynamic responsiveness. Moreover, the MOF host effectively confines LiCl, preventing leakage and improving device durability and reliability (39).

Fig. 2. Synthesis and Optimization of LiCl@UIO-66-NH₂.

(A) The humidity sensing mechanism of LiCl@UIO-66-NH₂. (B) SEM image of UIO-66-NH₂. (C) SEM image of 12.5 wt% LiCl@UIO-66-NH₂. (D) XRD patterns of UIO-66-NH₂ and LiCl@UIO-66-NH₂. (E) XPS spectra for Li 1 s of UIO-66-NH₂ and LiCl@UIO-66-NH₂. (F) Nitrogen adsorption-desorption isotherms and pore size distribution of LiCl@UIO-66-NH₂. (G) Humidity response of 12.5 wt% LiCl@UIO-66-NH₂. (H) Comparison of humidity response sensitivity for LiCl@UIO-66-NH₂ with doping ratios of 5%, 10%, 12.5%, 15%, and 20%. (I) Response and recovery time of 12.5 wt% LiCl@UIO-66-NH₂.

UIO-66-NH₂ was synthesized via a one-step hydrothermal method, followed by the introduction of varying LiCl ratios through solution impregnation (fig. S3) to investigate the effect of doping concentration. SEM imaging revealed that UIO-66-NH₂ exhibits a typical octahedral morphology (Fig. 2B) (40, 41). Subsequent SEM and EDS analyses of the LiCl-doped samples confirmed the uniform dispersion of LiCl within the MOF framework (Fig. 2C and figs. S4 and S5). XRD patterns further verified that the characteristic diffraction peaks of the composites remained consistent with those of UIO-66-NH₂, with no detectable LiCl peaks at 30.095° or 34.882° (Fig. 2D), indicating preserved crystallinity and highly dispersed LiCl within the pores without aggregation. XPS analysis detected a Li 1 s peak at 56 eV in the composite material (Fig. 2E and fig. S6), confirming the presence of Li⁺. Furthermore, surface area and porosity analyses showed that the specific surface area of UIO-66-NH₂ decreased from 763.42 m²/g to 30.09 m²/g for the LiCl-doped sample (12.5 wt%), accompanied by a reduction in pore volume, which is attributed to the increased mass post-doping and the effective loading of LiCl within the pores (Fig. 2F and fig. S7).

The humidity response performance of the LiCl@UIO-66-NH₂ was evaluated under an excitation voltage of 50 Hz and 100 mV. The optimal sensitivity was achieved at a LiCl doping ratio of 12.5%, with the humidity response curve described by log(G) = 0.052·RH-9.223 (Fig. 2, G and H and fig. S8), where G represents the conductance value, and RH stands for relative humidity. Regarding response and recovery times, the 12.5% doping ratio exhibited a faster recovery time of 21.589 s, while its response time was comparable to ratios at 10.339 s (Fig. 2I and figs. S9 and S10). This performance trend can be attributed to the balance between effective ion concentration and material porosity accessibility. Specifically, insufficient doping leads to limited dissociable LiCl, constraining ionic conductivity, whereas excessive doping clogs the pores of UIO-66-NH₂, significantly reducing its specific surface area and available water adsorption sites, ultimately limiting further enhancement of humidity response performance. Consequently, a doping ratio of 12.5% LiCl was selected for the subsequent fabrication of the sensors.

Optimization for electro-spun P@P SFM

The inadequate adhesion at the interface between MOF materials and flexible substrates is recognized as a principal constraint on their implementation in wearable sensors (42–44). Additionally, the degradation of MOF film under operational conditions—specifically due to sweat flow and mechanical contact with the skin during sweat rate detection—must be considered (fig. S11). It has been established that a superhydrophobic and permeable heterophase shunt is indispensable for isolating the MOF material from the skin and sensible sweat, thereby enabling stable, long-term assessment of insensible sweat. Guided by the superhydrophobic mechanism observed on the lotus leaf (Fig. 3, A to C) (45, 46), a hybrid electrospinning methodology integrating PDMS and PMMA was developed (Fig. 3D). Within this configuration, PMMA was utilized as the core fibrous matrix owing to its superior humidity stability, while PDMS was applied as a hydrophobic surface layer. Subsequent microstructural and elemental distribution analyses of the PMMA and the P@P SFM composite (Fig. 3E) demonstrated that PDMS adhered to the PMMA fibers and generated microspheres, effectively increasing surface roughness and resulting in the formation of a superhydrophobic membrane (Fig. 3F). A comparison of the water contact angles and surface roughness for PMMA and P@P SFM fibers fabricated at different electrospinning voltages is presented in Fig. 3G. The results confirm that the hybrid PDMS and PMMA electrospinning achieves superhydrophobicity.

Fig. 3. Enhanced performance of the P@P SFM.

(A to C) Optical and SEM images of lotus leaf. (D) Fabrication of a superhydrophobic breathable membrane via PDMS and PMMA hybrid electrospinning. (E) SEM image, EDS image and XPS spectra of PMMA fibrous membrane and P@P SFM. (F) Optical characterization of the superhydrophobic properties of the thin P@P SFM. (G) Comparison of the contact angle and surface roughness between PMMA fibrous membranes and P@P SFM. (H) Effects of the PDMS/PMMA mass ratio, electrospinning voltage, and spinning distance on the contact angle. (I) Long-term stability assessment of the superhydrophobicity of P@P SFM. (J) Effects of the PDMS/PMMA mass ratio, electrospinning voltage, and spinning distance on the breathability. (K) Thermal imaging of skin-adherent PDMS, PET, and P@P SFM on fingertips. (L) Breathability comparison with or without P@P SFM.

A systematic optimization of the electrospinning process parameters—namely, spinning voltage, collection distance, and PDMS-to-PMMA mass ratio—was conducted to explore the optimal hydrophobicity and breathability (table S2). The SEM images of the fabricated P@P SFM under various conditions (figs. S12 to S14) indicated that the fiber diameter was negatively correlated with the collection distance. Conversely, the diameter was found to initially increase and then decrease with rising spinning voltage and PDMS ratio (figs. S15 to S18). A similar trend was observed for the water contact angle of the fibrous membranes, which reached an optimum value of 151.9° at the parameter set of 15 kV, a 2:2 mass ratio, and a 7 cm collection distance (Fig. 3H). The superhydrophobicity was maintained for one month, which is attributed to the inherent stability of PDMS (Fig. 3I). Breathability was identified as another key factor. Using the method depicted in fig. S19, it was discovered that the spinning distance and voltage had negligible effects, whereas a significant improvement was achieved at the 2:2 mass ratio (Fig. 3J). The breathability of the membrane fabricated under these conditions was found to be markedly superior to that of common flexible substrate materials such as PDMS and PET (Fig. 3K) and was comparable to the open-state condition at 58.07 g/H·m² (Fig. 3L). Furthermore, the P@P SFM exhibits a notable hydrostatic pressure bearing capacity of 1.1 kPa, which is significantly higher than the sweat breakthrough pressure, thereby ensuring structural integrity during wear (fig. S20). Therefore, the parameter combination of 15 kV, 2:2 mass ratio, and 7 cm distance was ultimately selected for the preparation of the superhydrophobic and breathable membrane.

Performance of the insensible sweat rate sensor

Building upon hygroscopic materials, the fabrication process of superhydrophobic fibrous membranes, and our previous research on sensible sweat detection (25, 47, 48), this work designed a sensor configuration for monitoring insensible sweat rate. As illustrated in Fig. 4A and fig. S21, the insensible sweat rate sensor was fabricated through a stepwise process. The sensing unit is situated within the sensible sweat collection chamber, with a defined spatial gap maintained between the LiCl@UIO-66-NH₂ composite and the inlet. The P@P SFM encapsulates the hygroscopic material and remains in direct contact with the skin. Insensible sweat evaporates in its gaseous state, permeates the P@P SFM layer, and is subsequently captured by the LiCl@UIO-66-NH₂ composite, thereby inducing a measurable change in electrical conductivity. Figure 4, B to E show the macro- and micro-scale optical images of the insensible sweat detection electrode respectively. The insensible sweat rate sensor is fabricated in a circular region with a diameter of 5 mm, featuring interdigitated electrodes with a finger width of 160 μm and an interfinger spacing of 175 μm. The diameter of the inlet of the microchannel is also 500 μm. The response of the humidity sensor encapsulated with P@P SFM is shown in Fig. 4F, where the sensitivity slightly decreased compared to the unencapsulated version, following the relationship log(G) = 0.0505 · RH − 8.4713. The response and recovery times increased to 31.87 s and 72.45 s, respectively (Fig. 4G). Physiological studies indicate that sweating onset typically involves a latent period of tens of seconds to minutes, followed by a peak sustained over minutes to tens of minutes, while post-excitation drying is similarly gradual. Steady-state sweating under environmental stimulation is reached within 30–60 min, and skin humidity recovery from high to low levels requires several minutes. Even on highly responsive palm skin, sweat rate changes occur on a minute scale (49–51). Therefore, the response and recovery times of the insensible sweat rate detection of FRSR sensor are adequate for monitoring daily activities. The sensor demonstrated excellent selectivity, showing no significant response to common volatile gases (Fig. 4H). Stability tests indicated consistent humidity responses on days 1, 3, and 5, as well as under different bending conditions (Fig. 4, I and J). Furthermore, stable performance was maintained throughout 50-cycle test (Fig. 4K). To evaluate the output stability under prolonged sweat exposure, the FRSR sensor was immersed in a 50 mmol/L NaCl solution for 30 min. Benefiting from the protection of the P@P SFM layer, its electrical output remained nearly unchanged before and after immersion (fig. S22). Three sensors from the same batch also exhibited uniform performance (fig. S23). The sensor output remained unaffected by temperature variations within the typical skin temperature range (figs. S24 and S25). The elongated microchannel designed for sensible sweat detection provided a semi-enclosed environment for insensible sweat monitoring, effectively suppressing interference from ambient humidity. Finite element analysis (FEA) and an experiment of the impact of ambient humidity on detection further validated the advantage of this semi-enclosed structure in mitigating ambient humidity effects (figs. S26 and S27).

Fig. 4. Performance of the insensible sweat rate sensor.

(A) Schematic illustration of the insensible sweat detection. (B and C) Optical image of the insensible sweat detection electrode. (D) Optical image of the insensible sweat detection electrode with LiCl@UIO-66-NH₂. (E) Optical image of the humidity sensor with P@P SFM. (F) Humidity response of 12.5 wt% LiCl@UIO-66-NH₂ with P@P SFM. (G) Response and recovery time of 12.5 wt% LiCl@UIO-66-NH₂ with P@P SFM. (H) Specificity of the insensible sweat rate sensor. (I) Multi-day reusability performance of the insensible sweat rate sensor. (J) Detection stability of the insensible sweat rate sensor under different bending conditions. (K) 50-Cycle stability test of the insensible sweat rate sensor. (L) Schematic illustration of the experimental setup for simulating insensible sweat secretion. (M) Conductance response corresponding to simulated sweat evaporation rate. (N) Fitting relationship between insensible sweat rate and sensor conductance.

To establish a quantitative relationship between relative humidity and sweat rate, a custom setup that simulates insensible sweat secretion is developed (Fig. 4L), the detailed design is provided in the Materials and Methods section (52). By controlling the liquid evaporation rate within a confined range and using electrical conductance as an intermediate parameter, we calibrated the conductance response to sweat rate (Fig. 4, M and N), obtaining the fitting equation: G = 0.541·v – 0.098. where G represents the conductance value and v is the insensible sweat rate.

Performance of the sensible sweat rate sensor

Sensible sweat detection was implemented using an epidermal microfluidic-based interdigitated electrode conductivity sensing technique, with a design similar to our previous report (47, 48). The FRSR sensor was fabricated using flexible electronics processing techniques such as laser cutting and multilayer alignment lamination (Fig. 5A). In the absence of sweat, the interdigitated electrodes behave as an open circuit, resulting in near-zero overall conductance. The detailed structure for sensible sweat rate detection is shown in Fig. 5B. The uneven surface of the glitter film (Fig. 5, C and D and fig. S28) makes the sweat-filled sections clearly visible, allowing the wearer to monitor the measurement in real time. The maximum detectable sweat volume of this sensor is 88 μL. The structural parameters are as follows: channel width of 500 μm, interdigitated electrode width of 100 μm, adjacent electrode spacing of 6 mm (Fig. 5, E and F). Detailed electrode layouts and the underlying design rationale for the FRSR sensor are provided in fig. S29. Additionally, the asymmetric channel arrangement on the patch integrates both mechanical and golden ratio design principles, as detailed in fig. S30. The prototype adopts a golden-ratio-inspired asymmetric channel layout. FEA under bending indicates reduced strain concentration versus symmetric design, though the idealized model neglects anatomical variability, muscle deformation, adhesion, and placement misalignment. Thus, the asymmetric layout is a pragmatic prototype-specific compromise, not a universal optimum. In practical wearing scenarios, placement is preferentially performed on relatively flat body sites, such as the chest, forehead, and back, to minimize the risk of detachment. To mitigate compensatory sweating caused by pore occlusion when the sensor is attached to the skin, a hollowed-out structural design was developed, as illustrated in the Fig. 5G. Detailed design methodologies are provided in note S1. As shown in Fig. 5H, the sensible sweat detection signal exhibits a characteristic step-like profile, where a higher sweat rate leads to a shorter plateau duration. Differentiation of the step signal yields a pulse sequence (Fig. 5I), with the pulse interval corresponding to the plateau duration. A quantitative relationship between the interval ΔT and sweat rate v was established as: ΔT = 85.67v^–1.16, with an R² value of 98.95%, indicating a good consistency as shown in the Fig. 5J. FEA method confirmed uniform flow distribution in the microchannel, with consistent velocity profiles across different channel sections (fig. S31). As shown in fig. S32, a support structure was designed to secure the sensor. The sensor was then incrementally bent inward by 1 mm and 2 mm, which correspond to bending angles of 20° and 25°, respectively. As illustrated in Fig. 5K, a change in the electrical conductivity response was observed when the bending angle reached 25°. This indicates a certain limitation in the application of the sensor designed in this work. Consequently, in all subsequent on-body experiments, the sensor was attached to relatively flat areas of the body that do not undergo significant bending during movement. Figure 5L and fig. S33 presents on-body testing results, showing the original step signal, its derivative pulse signal, and the converted sweat rate profile from top to bottom. It is worth noting that a decrease in conductance was occasionally observed during sensible sweat detection. The underlying causes were investigated through multiple comparative experiments. Given that the final readout is derived from the derivative of the conductance curve, such transient fluctuations do not affect the accuracy of the detection results, as detailed in note S2. During high-intensity basketball exercise, the conductance signal exhibited a monotonically increasing trend, thereby confirming the stability of the detection (fig. S34).

Fig. 5. Morphological and performance characterization of the sensible sweat rate sensor.

(A) Fabrication process of the sensible sweat sensor. (B) Optical image of the sensible sweat detection module and schematic diagram of the detection mechanism. (C and D) Optical image of the glitter film. (E and F) Optical images of microfluidic channel and interdigitated electrodes. (G) Optical image of the hollowed structure in the 3 M adhesive layer for reducing the effects of compensatory sweating. (H) Conductance response of sensible sweat rate sensor to different flow rates. (I) Differential signal of different flow rates. (J) Quantitative relationship between sweat rate and temporal interval. (K) Influence of sensor bending on conductance response for sensible sweat rate measurement. (L) Experimental validation of sensible sweat monitoring through on-body testing.

Implementation of a FRSR sensor across multiple scenarios

To enable synchronous acquisition of both sensible and insensible sweat rates, this work developed a universal dual-channel miniature conductance signal detection and transmission circuit. The circuit module structure achieves electrical connection with the sweat sensor through contact points (Fig. 6A and fig. S35). As shown in Fig. 6B and fig. S36, the system employs dual-channel inputs to separately process sensible and insensible sweat signals, integrating dedicated signal excitation and modulation circuitry. The sensor-integrated patch and the detection module are electrically connected via spring-loaded contacts. The sensor itself is designed for single use; after each instance of perceptible sweat detection, the patch is replaced with a new one. This separable configuration between the detection module and the disposable sensor significantly reduces the overall economic cost. This sensing system was successfully applied to multi-scenario sweat monitoring.

Fig. 6. FRSR sensor system and applications.

(A) Schematic diagram and optical images of the FRSR sensor system. (B) Schematic diagram of the miniature conductance signal detection circuit. (C and D) CCK-8 assay of NHDF cells for cell viability testing of the P@P SFM after culturing 1, 3, and 5 days. (E) Detection of sweat rate on the skin surface of a psoriasis patient. (F) Physiological response of human sweat rate at various ambient temperatures. (G) Physiological response of human sweat rate under various real-life conditions. (H) Simultaneous monitoring of EEG and sweat rate in volunteers watching a horror movie. (I) Occipital EEG responses during horror video viewing, with inset showing the corresponding heatmap of the occipital region. (J) Insensible sweat rate in response to horror video stimulation.

Owing to the excellent biocompatibility of the sensor materials (Fig. 6, C and D), the system is suitable for managing the progression of typical dermatological conditions such as psoriasis. Tests on patient’s normal skin, lesional skin, and drug-treated lesional skin indicated that perspiration is predominantly insensible. The sweat rate was lower in normal skin, significantly higher in lesional skin, and decreased after drug treatment (Fig. 6E and fig. S37). This is primarily attributed to psoriasis compromising the epidermal barrier function, leading to increased transepidermal water loss, potentially exacerbated by bacterial infection further elevating local humidity. The protective membrane formed after drug treatment effectively ameliorates these conditions (53, 54). Consequently, the FRSR sensor can monitor the treatment response and infection status in inflammatory skin diseases like psoriasis, providing assistance for therapy (55). Furthermore, the system can detect human sweat rates under different ambient temperatures (Fig. 6F) and accurately distinguish sweating variations induced by drinking hot water, consuming spicy food, work load, and rest states (Fig. 6G). In emotion recognition (fig. S38), simultaneous recording of electroencephalography (EEG) and sweat signals from volunteers watching a horror movie revealed that the sweat sensor detected a significant increase in sweating approximately 20 seconds after the occipital cortex responded to the frightening scenes (Fig. 6, H and I and fig. S39). This indicates the capability of FRSR sensor to sensitively reflect mental stress states, holding potential for applications such as psychological state assessment and lie detection (56).

An additional significant advantage of full-range sweat rate detection lies in its ability to provide more precise threshold determination for dehydration alerts. To validate this, on-body tests were conducted on multiple body sites of volunteers, including the forehead, dorsal forearm, ventral thigh, chest, and back (Fig. 7, A to C). First, the regional sweat rate (RSR) was converted to the whole-body sweat rate (WBSR) using the method detailed in the note S3. As shown in the top panel of Fig. 7B, during the high-intensity phase of early exercise, sensible sweat had not yet initiated a response, whereas insensible sweat had already detected sweating signals. When the sensible sweat signal became activated, the readings from the insensible sweat detection channel stabilized, and the sweat rates measured by both methods were comparable. When exercise intensity was reduced during periods of heavy sweating, only the sensible sweat channel responded, with no significant change observed in insensible sweat. This phenomenon was consistently observed across all tested body sites (Fig. 7C). It is particularly noteworthy that the ventral thigh, due to its low sweat rate, cannot form flowing sensible sweat, it can only be effectively monitored by the insensible sweat sensor, which providing crucial technical support for dehydration warning in low-sweating populations. Compared with a single sensible sweat rate detection method, the FRSR sensor, owing to its advantage of multimodal detection, achieves detection capability in the transition regions of different sweat morphologies, with an 82.00% improvement in performance (fig. S40 and Fig. 7D). Furthermore, measuring body weight change before and after exercise to reflect actual fluid loss, comparisons show that the dehydration level estimated by the FRSR sensor is closer to the actual body weight change than that estimated by sensors relying only on sensible sweat detection or only on insensible sweat detection (fig. S41 and Fig. 7E), with prediction accuracy improved by 32.44% and 16.57%, respectively. It is worth noting that measurement errors in in-vivo testing are evaluated through cross-validation between independent sensible and insensible sweat sensors, rather than by comparison with standardized reference methods; this is primarily due to the difficulty of obtaining data on the body’s real-time physiological state.

Fig. 7. Performance verification of the FRSR sensor in precise dehydration alerting.

(A) Optical image of a volunteer during forehead sweat rate monitoring. (B) Simultaneous recording of exercise intensity and forehead sweat rate monitoring. (C) The whole body sweat rate calculated from dosrsal forearm, chest, ventral thigh and back. (D) Compared with sensible sweat detection methods, the detection performance of the FRSR sensor in different forms of sweat transition regions. (E) Comparative evaluation of sensible sweat rate sensors and FRSR sweat sensors for predicting sweat loss.

DISCUSSION

In summary, this work presents an innovative sensor configuration for full-range sweat rate monitoring. By non-interferingly integrating sensible and insensible sweat detection modules, the sensor achieves continuous monitoring of human sweat rate ranging from 0.182 to 10 μL/(min·cm²) at a single location, overcoming the detection range limitations of existing sensors. LiCl@UIO-66-NH₂ is employed as the hygroscopic material, demonstrating high sensitivity along with excellent response and recovery speeds. The specially designed P@P SFM serves as a heterophase sweat shunt, effectively ensuring the stability of the hygroscopic element in complex usage environments. Sensible sweat detection utilizes an established microfluidic conductivity-based approach, enabling long-term monitoring of up to 88 μL of sweat. Benefiting from its wide dynamic detection range and the developed dual-channel miniature conductance signal monitoring and transmission module, this FRSR sensor is suitable for various daily scenarios, such as skin health assessment, routine activity monitoring, and mental stress feedback. Furthermore, during exercise or high-intensity physical labor, it can provide more accurate sweat loss data and dehydration warnings. Looking ahead, the capability of full-range sweat rate detection will provide unique and in-depth insights and value for physiological research and clinical applications, including sweat gland function evaluation, early diagnosis and monitoring of autonomic neuropathy and neurodegenerative diseases (e.g., Parkinson’s disease, diabetic neuropathy), drug development and therapeutic efficacy evaluation, personalized health management, exercise training optimization, and quantitative assessment of stress responses (see note S4 for details). Furthermore, establishing standardized reference methods is important to future research aimed at enabling the detection of a wider range of physiological states.

MATERIALS AND METHODS

Materials

ZrCl₄ and LiCl were achieved from Aladdin (Shanghai, China). 2-Aminoterephthalic Acid (NH₂-H₂BDC) was obtained from Macklin (Shanghai, China). PMMA, tetrahydrofuran (THF) and N,N-Dimethylformamide (DMF) were achieved from Sinopharm Group Chemical Reagent Co. Ltd., China. PDMS was achieved from Dow Corning Co. Ltd., USA. Polyethylene terephthalate (PET) membranes and PET membranes with double-sided tape were provided by Suzhou Leanstar Electronic Technology Co., Ltd., China. All other chemicals were commercially available and of analytical reagent grade. All reagents were used as received. The aqueous solutions were prepared freshly with deionized water (1 μS/cm). Normal human dermal fibroblasts (NHDF) cells were achieved from BeNa Culture Collection (Henan, China). Dulbecco’s modified eagle medium was provided by Viva cell (Shanghai, China). Fetal Bovine Serum and Penicillin\Streptomycin(100×) were obtained from Keycell Biotechnology Co., Ltd. (Hubei, China). PBS biological buffer was purchased from Hyclone (State of Utah, USA). Cell Counting Kit-8 was achieved from Beyotime Biotechnology Co., Ltd. (Shanghai, China). Live-Dead Cell Viability/Cytotoxicity Assay Kit was achieved from KeyGEN Biotechnology Co., Ltd. (Jiangsu, China).

Characterization

The SEM images from HITACHI S-4800 was used to observe the surface morphology of the UiO-66-NH₂, LiCl@UiO-66-NH₂ and P@P SFM. The phases of UiO-66-NH₂ and the LiCl@UiO-66-NH₂ composite were characterized by X-ray diffraction (XRD, SmartLab 3KW) from 5° to 85° at a scan rate of 5°/min. X-ray photoelectron spectroscopy (XPS, Thermo Escalab 250Xi) and energy-dispersive spectroscopy (EDS, Regulus 8100) characterized the surface elemental composition and distribution of the LiCl@UiO-66-NH₂ composite. An accelerated surface area and porosimetry system (ASAP 2010, Micromeritics) was used to characterize the UiO-66-NH₂ and the LiCl@UiO-66-NH₂ under N₂ adsorption −195.8°C. The surface roughness of both PMMA fibrous membranes and P@P SFM was characterized by the confocal laser scanning microscope (LEXT OLS4000). The electrical conductance signal corresponding to the sweat rate was measured in the laboratory using an LCR meter (Keysight E4980AL).

Synthesis of LiCl@UiO-66-NH₂ composites

A mixture was prepared by sequentially adding 53.1 g of DMF, 470 mg of ZrCl₄, and 362 mg of NH₂-H₂BDC into a 50 mL reagent bottle. Finally, 2 mL of hydrochloric acid was introduced. The bottle was sealed and magnetically stirred at room temperature for 30 minutes. The resulting mixture was then transferred to a polytetrafluoroethylene-lined autoclave and reacted at 120°C for 24 hours. After cooling to room temperature, a yellow suspension was obtained. The solid product was collected by alternating centrifugation (10000 rpm, 10 min) with DMF and ethanol. This washing procedure was repeated three times to remove unreacted reactants and solvents. The product was subsequently dried under reduced pressure at 100°C for 12 hours in a vacuum oven, yielding yellow UiO-66-NH₂ powder.

A precursor solution was prepared by adding 78.9 g of anhydrous ethanol into a reagent bottle, followed by the addition of an appropriate amount of LiCl. The bottle was sealed and stirred at room temperature for 4 h. Subsequently, 20 mL of this precursor solution was transferred to another reagent bottle, and 100 mg of UiO-66-NH₂ was added. The mixture was stirred at room temperature for 48 h. The final LiCl@UiO-66-NH₂ composite was obtained by drying the product in an oven at 100°C for 12 h.

Preparation of P@P SFM

A mixed solvent was prepared by combining 10 mL of THF and 10 mL of DMF in a reagent bottle. Subsequently, 2 g of PMMA was added to the solvent mixture and stirred at room temperature for 24 h until completely dissolved, yielding a homogeneous solution. Then, 2 g of PDMS was incorporated into the mixture. The resulting blend was sonicated for 1 h followed by additional stirring for 3 h until full dissolution was achieved, producing the final electrospinning precursor solution.

The P@P SFM was fabricated via a hybrid electrospinning process (DP30, Yunfan Technology Co., Ltd., China). A portion of the precursor solution was drawn into a plastic syringe (10 mL) and allowed to stand for 3 min to eliminate air bubbles. A 22-gauge metallic needle (inner diameter: 0.4 mm) was then attached. The drum collector of the electrospinning apparatus was covered with aluminum foil. Key processing parameters—including applied voltage, working distance, and solution flow rate—were set according to experimental requirements, with the drum rotation speed fixed at 300 rpm. Electrospinning was conducted under ambient conditions of 40% relative humidity and 30°C for 6 h, ultimately yielding a P@P SFM.

Fabrication of the FRSR Sensor

The microfluidic structure was assembled through multiple precision alignment steps. The flash film, microchannel layer, electrode layer, and 3 M adhesive were first patterned into the desired geometries via laser cutting. The electrodes on the electrode layer were batch-fabricated at scale by Suzhou Leanstar Electronic Technology Co., Ltd. (Suzhou, China). Subsequently, 50 mg of the LiCl@UiO-66-NH₂ composite was placed into a vial, followed by the addition of 200 mg of deionized water. The mixture was stirred until a homogeneous paste was formed. Then, 6 μL of the paste was transferred onto the interdigitated electrodes designated for insensible sweat rate detection using a micropipette, with a controlled distance maintained from the microchannel inlet. Finally, the device was placed in an oven and thermally treated at 100°C for 30 min to ensure firm adhesion of the LiCl@UIO-66-NH₂ composite to the electrode surface. Finally, the P@P SFM film was trimmed to match the dimensions of the collection chamber and then placed over the LiCl@UIO-66-NH₂ layer.

Fabrication, and testing of simulating insensible sweat secretion device

A mold was fabricated using 3D printing technology. Subsequently, 7 mL of deionized water was injected into the mold, ensuring a fixed distance between the liquid level and the mold outlet. A microporous array with a radius of 100 μm was created on a PET membrane via laser cutting, and this membrane was securely attached to the mold outlet. To simulate varying insensible sweat rates, an adjustable heating plate was positioned beneath the mold, allowing for temperature control within the range of 25 ∼ 40°C. The change in water mass inside the mold was monitored in real-time using a precision electronic balance (accuracy: 0.1 mg), from which the simulated insensible sweat rates under different temperature conditions were calculated. The conductance of the test sensor, which was fixed on the polyporous PET membrane with the integrated P@P SFM, was measured at different simulated sweat rates. This testing was carried out in a climate-controlled chamber (25 ± 0.5°C, 27 ± 2% RH) to ensure stable conditions and reproducible results.

On body test

The designated area on the human subject was cleaned with ethanol and deionized water, then dried using lint-free paper. The flexible patch, integrated with the sensor module, was securely applied to the site. Subjects performed exercises of varying intensity on a stationary bicycle (Keep K0102C) in an ambient environment (22 ∼ 27°C, 20 ∼ 30% RH). Real-time heart rate was monitored using a smart sports watch (HUAWEI WATCH GT4). Exercise intensity was controlled by adjusting both the bicycle resistance and the subject’s pedaling frequency. Data for sweat rate were transmitted in real-time to a mobile application via a Bluetooth wireless module. EEG signals were acquired using the waveguard EEG cap (EEMAGINE, 32channels, Tyco68) and processed with the open-source toolbox EEGLAB. Of note, sampling by the sensor module is stopped prior to removal of the patch after the test.

The testing procedure for participants with psoriasis was as follows: Prior to testing, participants cleaned the dorsal surfaces of both hands with water and gently dried them. Two sensors were then attached—one to the normal skin on the dorsum of the left hand and the other to the lesioned skin on the dorsum of the right hand—to obtain baseline sweat measurements before any drug intervention. Following this, an antimicrobial agent was applied to the affected skin area. After allowing approximately two hours for absorption, the patches were replaced and sensors were reapplied to the same locations on both hands to collect post-treatment sweat data.

CCK-8 cell viability and cytotoxicity assessment

NHDF cells were resuscitated from liquid nitrogen by rapid thawing in a 37°C water bath with gentle agitation. After complete dissolution, the cell suspension was transferred to a centrifuge tube containing 5 mL of culture medium and centrifuged at 1000 rpm for 5 min. The supernatant was discarded, and the cell pellet was resuspended in complete medium supplemented with 10% fetal bovine serum, followed by seeding into culture dishes and gentle mixing via pipetting. Cultures were maintained at 37°C under 5% CO₂ with saturated humidity. For experimental assays, cells in the logarithmic growth phase with optimal viability were plated at densities of 4 × 10³ cells/well in 96-well plates and 1 × 10⁵ cells/well in confocal dishes, followed by overnight incubation under the same culture conditions. P@P SFM was sterilized under UV irradiation, followed by co-culture with cells. A 96-well plate pre-seeded with cells was used to accommodate the sterilized samples for co-culture. The experimental design included three observation time points (1, 3, and 5 days), with experimental groups, control groups, and blank control groups established for analysis.

Following the designated incubation period, 10 μL of CCK-8 solution was added to each well of the 96-well plate under light-protected conditions. After 3 hours of incubation, the absorbance at 450 nm was measured using a microplate reader, and the cell viability was calculated accordingly. The Calcein-AM and PI stock solutions were allowed to equilibrate at room temperature for 30 minutes. A working solution was prepared by first adding 5 μL of 16 mM PI stock to 10 mL of PBS, followed by vortex mixing to obtain an 8 μM PI solution. Then, 5 μL of 4 mM Calcein-AM stock was introduced into the PI solution and thoroughly mixed by vortexing. The resulting working solution (2 μM Calcein-AM and 8 μM PI) was ready for direct cell staining applications. Prior to the experiment, cells were gently rinsed with PBS to remove residual esterase activity from the culture medium, followed by aspiration of the supernatant. An adequate volume of the aforementioned working solution was added to fully cover the cell monolayer. After incubation at room temperature for 30–45 min, the staining solution was aspirated to terminate the reaction. Cellular morphology was examined under a confocal microscope using XY-plane scanning mode. The relative cell viability (%) for each group was calculated using the formula: (OD_exp − OD_back)/(OD_cont − OD_back) × 100.

Ethics declaration

Healthy control human subject studies were approved by the Institutional Review Board of the Suzhou Institute of Nano-Tech and Nano-Bionics, CAS with number SINANO/EC/2025–036. All healthy subjects and one psoriasis patient understood the guidelines of the test and offered written informed consents.

Supplementary Materials

The PDF file includes:

Figs. S1 to S50

Tables S1 to S3

Supplementary Notes S1 to S4

Legend for data S1

References

Other Supplementary Material for this manuscript includes the following:

Data S1