Printed Wearable Sweat Rate Sensor for Continuous in-situ Perspiration Measurement

Abstract

Sweat rate measures key physiological states such as hydration levels and heat tolerance. Incorporating wearable technology with sweat rate sensors allows individuals to conveniently monitor their health, optimize workouts, and enhance occupational safety. However, challenges persist in such integration techniques, including intricate manufacturing, non-linear responses to changes in sweat rates, and errors from the intermediate measurement of the distance sweat travels in the sensor. To address these issues, we present a comprehensive wearable platform that includes a fully printed, flexible sensor patch, readout electronics, and a mobile app for continuous, real-time monitoring of sweat rate. Utilizing direct 3D printing and scalable microfluidic fabrication, we produced a sensor patch measuring 700 mm² and weighing 380 mg. The microfluidic channels are 850 μm wide, and 164 μm thick with serpentine electrodes measure sweat rate using capacitance. Our custom readout electronics capture these changes in capacitance to accurately measure sweat rate, achieving a sensitivity of 0.01 μLmin⁻¹. We validated the sensor’s performance against analytical models, simulations, and on-body trials with commercial sensors. This cost-effective, flexible, and fully integrated sweat-sensing solution has significant potential in precision health.

Graphical Abstract

A wireless wearable sweat rate sensor system is presented, featuring digital 3D direct-write printing on a flexible substrate with microfluidic layers for continuous, real-time monitoring. Printed encapsulated metal electrodes are used for capacitance measurements, achieving high sensitivity (0.01 μL min⁻¹) while maintaining a compact and lightweight design. The sweat rate sensor is validated through analytical modeling, simulations, and experiments.

1. Introduction

Recent advancements in wearable technology have revolutionized personal healthcare and fitness monitoring from the real-time analysis of numerous physiological parameters in biofluids.^[1–5] Sweat is one of the most popular biofluids for such wearable technology-based devices because of its abundance on the human skin and its non-invasive acquisition. The concentration of sweat composition and sweat rate (SR) are closely interconnected. For example, sweat Na⁺, K⁺ concentrations and pH increase with increasing SR.^{[4, 6–8]} On the contrary, lactic acid, urea, and creatinine concentrations decrease with increasing SR.^[6] Incorporating SR with the sweat metabolomics biomarkers, improves the ion concentration measurement reliability.^[9] Hence, SR sensors are indispensable for reliable wearable technology.

Sweating is a natural and essential physiological process for regulating core body temperature during physical activity or exposure to a high-temperature environment. However, anomalies in sweating, such as hyperhidrosis (excessive sweating), hypohidrosis (reduced sweating), and anhidrosis (absence of sweating) can disrupt body temperature regulation and lead to medical emergencies including hyperthermia, heat exhaustion, and heat stroke.^[10–12] Therefore, it is essential to investigate SR and intervene promptly by proper fluid ingestion during exercise,^[13] athletic activities,^[14] and working in scorching environments.^[15] A wearable, cost-effective SR sensor patch can be used for this purpose, significantly improving personalized health assessment through continuous and real-time monitoring of sweating.

Traditionally, SR is measured by evaluating the change in body mass^{[16, 17]} using gauze or highly absorbent pads.^{[18, 19]} These methods require a bulky setup and are impractical for continuous SR monitoring. An example of a recent effort to develop wearable continuous SR sensors is hygrometer-based sensors.^[20–22] This type of sensor is vulnerable to environmental influence and post-processing of collected sweat, which hinders continuous SR measurement. Also, such sensors often involve chambers that create a humidity gradient, which are bulky and inconvenient to wear. Other recent developments described in previous literature include SR measurement by using image-based optical sensing^[23] or by detecting changes in electrical properties such as electrical admittance,^{[7, 24–27]} conductance,^[28] or capacitance.^{[22, 29]} However, these sensors use expensive and complex photolithography or off-the-shelf components for the manufacturing process, which restricts the large-scale production of such sensor patches. Especially, admittance-based SR sensors often use interdigitated electrodes that produce a spike in admittance or current once the sweat comes in contact with the electrode. Furthermore, such a design cannot capture changes in SR while the sweat travels between the electrodes.

In this work, we present a printed wearable continuous sweat monitoring patch (Figure 1 A). This development also includes custom-made readout electronics and a specialized mobile application that supports SR data processing, Bluetooth wireless communication, and real-time display on mobile devices (Figure 1 B, Supplementary Table 1 provides a comparative analysis of this work to the existing literature). Our sensor patch features a robust design and cleanroom-free fabrication process, making it reliable, scalable, cost-effective, and disposable (Figure 1 C, D). In the sensor design, we use silver (Ag) electrodes in a parallel-plate capacitor configuration. Microfluidic techniques are employed to create the sweat inlet and channel. The differential surface energies of the flexible substrate and the cover of the microfluidic channel induce a capillary force gradient, which guides sweat into and through the channel between the electrodes. As the channel fills with sweat, which has a higher dielectric constant than air, the capacitance changes, producing capacitance versus time plots with varying slopes based on the sweat production rate. These plots form a linear calibration curve (Figure 1 E). The performance of the SR sensor is confirmed through COMSOL simulations (Figure 1 F) and analytical modeling. We model the sensor behavior using a parallel-plate capacitor approach, which supports the observed linear relationship between the rate of capacitance change and the SR. The calibration curves from simulations, analytical modeling, and experimental data are consistent (Figure 1 G). By measuring changes in capacitance, an unknown SR can be determined from the calibration curve, enabling real-time continuous SR monitoring. Additionally, we demonstrate the durability of our capacitive SR measurements by testing our sensors under various environmental conditions. Most importantly, our solution tackles major challenges, including complex manufacturing processes, non-linear responses to changes in SR, and errors associated with the intermediate measurement of the distance sweat travels within the sensor. Our work presents a capillary force-based approach for dimension selection in SR sensors. We have designed the microfluidic layers to facilitate efficient movement of sweat within the channel. Additionally, we conducted COMSOL simulations and analytical modeling of the sensing system. Our fabrication process is inexpensive and does not require a cleanroom. Furthermore, we developed a custom wireless data collection and transmission system along with a corresponding smartphone application, which makes our sensor patch a complete solution for continuous on-body SR monitoring.

Figure 1. Overview of the wearable printed capacitive SR sensor.

(A) The SR sensor is worn on the forehead of the user while exercising. (B) The sensor system has three major components: SR sensor patch, readout electronics, and Android phone app. (C) Photograph of the SR sensor patch. (D) The SR sensor patch consists of three layers: The first layer is the encapsulated Ag electrodes on a flexible polyimide substrate, the middle layer is the patterned microfluidics channel to direct the sweat, and the top layer is the microfluidics cover. The substrate and microfluidic cover layers are hydrophilic, exploiting capillary forces to direct the sweat into the channel. (E) Working principle of the capacitive SR sensor. Depending on the SR $\left(r_1<r_2<r_3\right)$, different channel areas will be filled at a certain time. $\Delta C$ is the change of capacitance of the channel at time $t$ adjusted for the capacitance of an empty channel and the effects of the sensor’s ambient surroundings. At time $t_1$, $\Delta C$ for rate $r_3$ is maximum and minimum for rate $r_1$. The linear calibration curve is generated by measuring the slope at different rates. This experimental calibration curve also agrees with the analytical modeling of the capacitive rate sensor system. This calibration curve can determine the unknown rate $r$ from a measured slope. (F) COMSOL simulation setup for the capacitive SR sensory system. (G) The calibration curve, slope vs rate, from simulation, analytical modeling, and experimental results are in agreement.

2. Results

2.1. Wearable Capacitive Sweat Rate Sensor

In the existing literature, admittance-based SR is measured from a polynomial behavior among Na⁺ ion concentration, the distance traveled in the channel at a particular time, and admittance.^[7] In this approach, Na⁺ concentration is measured in a reservoir, which is then mixed with the previously filled channel of the SR sensor. This mixing changes the overall concentration in the SR channel, which would be different from the reservoir Na⁺ concentration. Thus, Na⁺ concentration-dependent SR measurement is complex. Moreover, existing admittance and capacitance-based SR sensors require complex lithography-based microfabrication or require aligning multiple layers during fabrication.^{[7, 29]} Flow length measurement is also required during the SR calculation. In our development, we detailed the fabrication process of a three-layer capacitive SR sensor patch by utilizing large-area compatible, inexpensive 3D writing. Our sensor eliminates the need for measuring the sweat travel path and utilizes the whole physiological range of sweat concentration to generate a one-time calibration curve for continuous SR measurement.

2.2. Sweat Rate Sensor Fabrication

We utilize a flexible polyimide substrate to construct the parallel-plate-based capacitive SR sensor (Figure 2 A). Two parallel silver (Ag) electrodes are printed onto the substrate, and a dielectric layer is subsequently printed over the metal electrodes to encapsulate them and prevent direct contact with the sweat analyte. The metal electrodes and their encapsulation are fabricated using Direct 3D writing. The sweat inlet and channel are formed by patterning a double-sided microfluidic tape. This microfluidic layer is aligned and attached to the encapsulated metal electrodes. To complete the assembly, a microfluidic cover is added to guide sweat into the channel. A desktop cutting machine patterns both the microfluidic tape and the cover.

Figure 2. Fabrication and analytical modeling of the capacitive SR sensor.

(A) The fabrication process showing (i) direct 3D writing of two parallel Ag electrodes; (ii) direct 3D writing of dielectric encapsulation; (iii) assembly of a double-sided microfluidic tape, patterned by using a desktop cutter machine, with the already encapsulated Ag electrodes; (iv) assembly of a 3M^™ 9984 microfluidic cover, patterned by using the same desktop cutter machine, concluding the sensor fabrication. (B) Photographs of (i) direct 3D writing of Ag electrodes, (ii) fabricated SR sensor patch showing the sweat inlet, outlet area, and the connectors for readout electronics. Micrographs of (iii) dielectric encapsulation of the metal electrodes (the bar represents 1 mm) (iv) of the sensor outlet. The sensor outlet has a thickness of 164 μm and a width of 850 μm. (C) Contact angle, ${\theta }_c$, and surface properties. If ${\theta }_c<{90}^0$, then the surface is hydrophilic, ${90}^0\le {\theta }_c\le {150}^0$ for the hydrophobic surface, and ${\theta }_c>{150}^0$ for the super-hydrophobic surface. (D) SR sensor uses the hydrophilic polyimide $\left({\theta }_c={65.6}^0\right)$ as the substrate, hydrophilic side $\left({\theta }_c={18.0}^0\right)$ of 3M^™ 9984 microfluidic cover as the channel cover and hydrophobic side $\left({\theta }_c={91.0}^0\right)$ of the microfluidic cover is in the top side. (E) Capacitive system for the analytical modeling of the SR sensor. $d_d$ is the distance covered by the dielectric encapsulation, $d_f$ is the channel width, and $d=d_d+d_f+d_d$ is the total distance between two parallel metal electrodes. $A_f\left(t\right)$ and $A_a\left(t\right)$ are the area filled with sweat and air, respectively, in the channel. $C_d$ on either side is the capacitance formed between the metal electrode and the channel by the dielectric encapsulation. $C_f$ and $C_a$ are the capacitance formed by the sweat and air, respectively, inside the channel. The inset shows the change of slope with SR from analytical modeling.

The channel volume can be adjusted in three distinct ways. Firstly, the serpentine structure can be extended in length to modify the design (Supplementary Table 2). Secondly, by stacking additional layers of microfluidic double-sided tape, the height of the channel can be increased (Supplementary Table 3). Lastly, the distance between the electrodes and the channel width can be increased (Supplementary Table 4). The thickness of the double-sided microfluidic tape is 82 μm, which is the minimum thickness achievable for the microfluidic layer. In this project, we use an electrode separation of 1010 μm, two layers of microfluidic double-sided tape (164 μm), and a channel width of 850 μL (Figure 2 B). These parameters establish 16 μL channel capacity. The SR sensor patch features a circular inlet with a diameter of 4.5 mm. Details on sensor dimension selection, simulation, and fabrication are presented in Supplementary Figure S1 to Figure S5.

To guide the sweat inside the channel a capillary force gradient is created by using the surface energy of the substrate and the microfluidic cover. The contact angle serves as a measure of surface energy and the wettability of a surface (Figure 2 C, D). Two primary forces interact at the interface of a solid surface and a liquid: adhesive forces, which occur between the liquid and the solid surface, and cohesive forces, which act among the liquid molecules. Higher surface energy enhances the adhesive forces over the cohesive forces. This leads to capillary action that aids the movement of a liquid across a solid surface, characterized by a contact angle, ${\theta }_c$, less than 90°, indicating a hydrophilic surface. As surface energy decreases, cohesive forces become comparable to adhesive forces, rendering the surface hydrophobic, with a contact angle between 90° and 150°. In cases of very low surface energy, cohesive forces dominate over adhesive forces, making the surface super-hydrophobic with ${\theta }_c>150°$. In this work, our substrate is hydrophilic, with ${\theta }_c=65.6°$. The microfluidic cover used has one hydrophilic side with ${\theta }_c=18°$, which faces the sweat channel to promote capillary action, directing sweat through the channel. The opposite side of the cover is hydrophobic, with ${\theta }_c=91°$, preventing sweat from spilling over the channel. To visualize fluid flow within the channel, we use a colored solution and drop-cast it on the sensor inlet. This fills a certain volume of the channel. If there is no new input on the inlet, the sweat doesn’t move forward, so the capacitance remains fixed. Once there is new input at the inlet, the sweat moves forward, and capacitance changes (Supplementary Video 1).

2.3. Analytical Modeling

We model the capacitive SR sensor as a parallel plate capacitor system (Figure 2 E). A fixed area of encapsulation exists between two parallel metal plates. Here, $d$ is the distance between two metal electrodes, $d_d$ is the distance covered by the encapsulation, and $d_f$ is the channel width. So, $d=d_d+d_f+d_d$. The encapsulation width $d_d$ is assumed to be uniform along the entire channel. The total area of the channel is $A_{total}$. Initially, in its empty state, the channel is filled with air, resulting in an equivalent model of three capacitors in series formed due to layers of stacked materials of varying dielectric constants: two formed at the electrode-encapsulation interface $\left(C_d\right)$ on either side of the channel and the third capacitor formed by air $\left(C_a\right)$. So the total initial capacitance, $C_0$ is,

$$\begin{array}{l}\frac{1}{C_0}=\frac{1}{C_d}+\frac{1}{C_a}+\frac{1}{C_d}=\frac{2}{C_d}+\frac{1}{C_a}\\ =\frac{2d_d}{{ϵ}_d{A}_{total}}+\frac{d_f}{{ϵ}_a{A}_{total}}\\ C_0=\frac{{ϵ}_a{ϵ}_d{A}_{total}}{2{ϵ}_a{d}_d+{ϵ}_d{d}_f}\end{array}$$ (1)

Here, ${ϵ}_d$ and ${ϵ}_a={ϵ}_0$ are the dielectric constant of the encapsulation and the air respectively. With time, the sweat-filled area will increase when the sweat propagates along the channel, so the empty area with air will decrease accordingly. Now, $A_{total}=A_f\left(t\right)+A_a\left(t\right)$ is the total channel area where $A_f\left(t\right)$ and $A_a\left(t\right)$ is the area filled with sweat and empty area respectively at time $t$. For the total capacitance at time $t$, $C_s\left(t\right)$, we have two equivalent capacitors in parallel, one representing the filled portion and another the empty portion:

$$C_s\left(t\right)=C_{a,eq}\left(t\right)‖C_{f,eq}\left(t\right)$$ (2)

where,

$$\begin{array}{l}\frac{1}{C_{a,eq}\left(t\right)}=\frac{1}{C_d}+\frac{1}{C_a\left(t\right)}+\frac{1}{C_d}\\ =\frac{2d_d}{{ϵ}_d\left(A_{total}-A_f\left(t\right)\right)}+\frac{d_f}{{ϵ}_a\left(A_{total}-A_f\left(t\right)\right)}\\ C_{a,eq}\left(t\right)=\frac{{ϵ}_a{ϵ}_d\left(A_{total}-A_f\left(t\right)\right)}{2{ϵ}_a{d}_d+{ϵ}_d{d}_f}\end{array}$$ (3)

Here, $A_a\left(t\right)=A_{total}-A_f\left(t\right)$. Similarly,

$$C_{f,eq}\left(t\right)=\frac{{ϵ}_a{ϵ}_f{A}_f\left(t\right)}{2{ϵ}_f{d}_d+{ϵ}_d{d}_f}$$ (4)

Now, from equation (2), using equations (3) and (4),

$$\begin{array}{l}{C}_s\left(t\right)=\frac{{ϵ}_d{ϵ}_a\left(A_{total}-A_f\left(t\right)\right)}{2{ϵ}_a{d}_d+{ϵ}_d{d}_f}+\frac{{ϵ}_d{ϵ}_f{A}_f\left(t\right)}{2{ϵ}_f{d}_d+{ϵ}_d{d}_f}\\ =C_0-\frac{{ϵ}_d{ϵ}_a{A}_f\left(t\right)}{2{ϵ}_a{d}_d+{ϵ}_d{d}_f}+\frac{{ϵ}_d{ϵ}_f{A}_f\left(t\right)}{2{ϵ}_f{d}_d+{ϵ}_d{d}_f}\\ =C_0+\left(\frac{{ϵ}_d{ϵ}_f}{2{ϵ}_f{d}_d+{ϵ}_d{d}_f}-\frac{{ϵ}_d{ϵ}_a}{2{ϵ}_a{d}_d+{ϵ}_d{d}_f}\right)A_f\left(t\right)\end{array}$$ (5)

Now, the sweat-filled area, $A_f\left(t\right)$, depends on the SR, $SR\left(t\right)$. We use μL min⁻¹ as the unit of $SR\left(t\right)$. So,

$$A_f\left(t\right)=\frac{1}{d_f}{\int }_0^{t}SR\left(t\right)dt$$ (6)

Using equation (6) in equation (5),

$$\begin{array}{l}{C}_s\left(t\right)=C_0+\\ \left(\frac{{ϵ}_d{ϵ}_f}{2{ϵ}_f{d}_d+{ϵ}_d{d}_f}-\frac{{ϵ}_d{ϵ}_a}{2{ϵ}_a{d}_d+{ϵ}_d{d}_f}\right)\left(\frac{1}{d_f}{\int }_0^{t}SR\left(t\right)dt\right)\end{array}$$ (7)

Differentiating both sides of equation (7) with respect to time,

$$\frac{d{C}_s\left(t\right)}{dt}=\left(\frac{{ϵ}_d{ϵ}_f}{2{ϵ}_f{d}_d{d}_f+{ϵ}_d{d}_f^2}-\frac{{ϵ}_d{ϵ}_a}{2{ϵ}_a{d}_d{d}_f+{ϵ}_d{d}_f^2}\right)SR\left(t\right)$$ (8)

Equation (8) correlates change of capacitance with sweat rate, $SR\left(t\right)$. Figure 2 E inset shows the variation of the slope, $\frac{d{C}_s\left(t\right)}{dt}$, with the changing SR. For the shaded region, ${ϵ}_f$ has been varied from $30{ϵ}_0$ to $1000{ϵ}_0$ and the solid line represents ${ϵ}_f=80{ϵ}_0$.

2.4. Sensor Characterization.

To characterize the SR sensor, we designed an experimental setup (Figure 3 A) with an electroosmotic pump to precisely control the rate of artificial sweat solution and deliver it to the inlet of the sensor. To study the effects of varying ion concentrations in sweat, we test on both the lower and upper bounds of the physiological ranges of sweat ionic concentrations. We use an LCR meter to measure the capacitance. We also measure the resistance of the system to determine admittance. Admittance is measured to compare the sensor performance with the performance of existing rate sensors in the literature. We choose 400 kHz as the measurement frequency which is the maximum frequency achievable in the readout electronics system for the on-body wearable application. In the capacitive SR sensor, we want to minimize the effects of resistance and less variation in capacitance due to measurement frequency, both of which are attainable by choosing a measurement frequency of 400 kHz (Supplementary Figure S6).

Figure 3. Characterization of the SR sensor.

(A) Experimental setup with LCR meter and electroosmotic pump to measure the capacitance at different rates using artificial sweat. (B-D) Measurement on the physiological lower bound of sweat ionic concentration (10 mM NaCl, 3 mM KCl, 0.2 mM CaCl₂, and 0.02 mM MgCl₂). (B) Change in capacitance, $\Delta C\left(pF\right)$, with time at rates varying from 0.25 μL min⁻¹ to 5.0 μL min⁻¹. (C) Calibration curve derived from measuring the slope (pF min⁻¹) from B. Error bars represent standard deviations for n=3 measurements. (D) Change in admittance is not linear with time for the lowest concentration artificial sweat. (E-H) Measurement on the physiological upper bound of sweat ionic concentration (80 mM NaCl, 15 mM KCl, 2 mM CaCl₂, and 0.3 mM MgCl₂). (E) The input rate of the highest concentration of artificial sweat is varied over time. The rate is determined using the calibration curve of G. (F) Change of capacitance, $\Delta C\left(pF\right)$, with time at rates varying from 0.25 μL min⁻¹ to 5.0 μL min⁻¹. (G) Calibration curve derived from measuring the slope (pF min⁻¹) from F. Error bars represent standard deviations for n=3 measurements. (H) Change in admittance is again not linear with time for the highest concentration of artificial sweat.

Human sweat contains several micronutrients.^[30] Among the salts, Sodium Chloride (NaCl), Potassium Chloride (KCl), Calcium Chloride (CalCl₂), and Magnesium Chloride (MgCl₂) have the highest percentages. To mimic the sweat properly, we prepare two artificial solutions - one with the lowest concentration (Figure 3 B–D) of salts (10 mM NaCl, 3 mM KCl, 0.2 mM CaCl₂, and 0.02 mM MgCl₂) found in sweat and another with the highest concentration (Figure 3 F–H) of salts (80 mM NaCl, 15 mMKCl, 2 mM CaCl₂, and 0.3 mMMgCl₂). From Figure 3 B (Figure 3 F), the change in capacitance of low-concentration (high-concentration) artificial sweat depends on SR and increases linearly over time. Human SR varies from 1 to 20 nL/min/gland and typically there are 150 glands/cm².^{[31, 32]} So, for an inlet with 4.5 mm diameter, we expect the SR to be as high as 1.8 μL min⁻¹ for our sensor during the on-body in-situ SR measurement. To capture the whole physiological range, we use 0.25 μL min⁻¹, 1.0 μL min⁻¹, 1.5 μL min⁻¹ and 5.0 μL min⁻¹ as the artificial SR. Figure 3C (Figure 3 G) shows the slope measured in pF min⁻¹ exhibits linear behavior with low concentration (high concentration) artificial SR. At 400 kHz, the change in admittance over time is not linear for low-concentration (high concentration) artificial sweat, as shown in Figure 3 D (Figure 3 H). To use admittance in determining the SR calibration curve, measurement of distance traveled by sweat is required. This additional measurement renders admittance-based SR calibrations prone to error, as in most cases, such measurements are made by visual inspection of photos of the sensors taken at a certain time interval.

In Figure 3 E, high-concentration artificial SR varies over time. Initially, we set the rate to 0.6 μL min⁻¹ for 60 s. Then in next 90 s the rate is varied from 0.6 μL min⁻¹ to 4.0 μL min⁻¹ and kept fixed for 60 s. The rate varies from 4.0 μL min⁻¹ to 2.8 μL min⁻¹ in next 50 s and stays at that rate for 50 s more. In next 60 s the rate is changed from 2.8 μL min⁻¹ to 3.8 μL min⁻¹ and kept there for 60 s. Rate is reduced from 3.8 μL min⁻¹ to 0.4 μL min⁻¹ over the next 180 s and kept at this lower rate for 120 s. We choose rates different from the one we use to determine the calibration curve, and the time is randomly varied. The calibration curve of Figure 3 G has been used to determine the rate from the sensor. The actual rate and the rate determined from the sensor agree, validating the sensor performance for real-time continuous sweat monitoring. We also validated the sensor performance for different artificial solutions, different radii of curvature surfaces, temperatures, and horizontal and vertical orientations (Supplementary Figure S7 to Figure S9). The sensor is re-usable by emptying the channel using absorbing material (Supplementary Video 1, Supplementary Figure S10).

2.5. System Design and Calibration for Wearable SR Sensor.

We design a custom-built front-end readout electronics system for the SR sensor and an Android phone app (Figure 4 A, E). This system will collect, process, wirelessly transmit (by using Bluetooth), and display the SR sensor data. In the system design, we configure the capacitance to digital converter, FDC1004, in single-ended measurement mode. Among the two electrodes of the sensor, one is connected to the CIN1 pin of the FDC1004, and the other end is connected to the ground. We use the FFC connector to connect the SR sensor with the readout electronics. This setup will measure the capacitance between the two Ag serpentine electrodes, which are modeled as a parallel-plate capacitor. The CAPDAC feature of FDC1004 has been utilized to increase the capacitance reading range up to 100 pF. The digitized capacitance value is sent to the Nordic nRF52832, which establishes a connection with the mobile phone via Bluetooth. We develop a customized Android phone app by using Cordova platform. A 70 mAh battery is used to power up the electronics readout system. The Power Measurement Unit (PMU) is built with a battery input port and Low Dropout Voltage Regulator (LDO) to deliver a constant 3 V. The Nordic chip has a built-in temperature sensor. The readout electronic system drains 7 mA current. The 70 mAh battery can power up the continuous measurement for 10 h with a single charge. From a partially filled SR sensor patch, we continuously measure the capacitance, temperature and battery potential for 165 minutes. The battery potential was reduced to 2.8 V over this time period (Supplementary Figure S11). Since the readout electronics are directly in contact with the skin, we use the temperature sensor to monitor body temperature.

Figure 4. System design for readout electronics and calibration.

(A) Photograph of the SR sensor patch with the readout electronics and custom-built Android phone application (the bar represents 4 mm). (B) Measurement by using the readout electronics of E. The change in capacitance with time at rates varying from 0.25 μL min⁻¹ to 5.0 μL min⁻¹ for the lowest concentration (10 mM NaCl, 3 mM KCl, 0.2 mM CaCl₂, and 0.02 mM MgCl₂) artificial sweat. (C) A calibration curve (error bars represent the standard deviation for n = 3 measurements) correlating the slope, derived from B, and the rate (μL min⁻¹) for the lowest concentration artificial sweat. (D) Temperature measurement while the capacitance was measured for the lowest concentration artificial sweat at a rate of 0.25 μL min⁻¹. Temperature remains constant at 27.7°C. (E) Experimental setup and readout electronics, where the equivalent parallel-plate capacitance is measured and converted to a digital value by the FDC1004 IC. A Nordic nRF52832 transceiver establishes the Bluetooth connection and sends the measured capacitance values to a mobile phone. The temperature sensor of nRF52832 is used to measure the surface temperature. The Power Measurement Unit (PMU) contains a battery and Low Dropout Voltage Regulator (LDO) to power up the readout electronics. (F) Change in capacitance for the highest concentration artificial sweat (80 mM NaCl, 15 mM KCl, 2 mM CaCl₂, and 0.3 mM MgCl₂) while the rate is varied from 0.25 μL min⁻¹ to 5.0 μL min⁻¹. (G) A calibration curve (error bars represent the standard deviation n=3 measurements) correlating the slope, derived from F, and the rate (μL min⁻¹) for the highest concentration artificial sweat. (H) A calibration curve was determined from the setup of E. Dotted lines (red and blue) represent the upper and lower boundaries for the two different concentrations of artificial sweat. The black dotted line is derived from averaging the upper and lower boundary of the experimental results and is used for on-body measurements. The black solid line from analytical modeling (with dielectric constant, ${ϵ}_f=80{ϵ}_0$) coincides with the average of the calibration curve derived from experimental results.

The readout electronics require a new calibration curve for continuously measuring the SR in real-time. This is a one-time calibration and does not need to be repeated if the sensor dimensions are the same. Also, the user can easily input this one-time calibration parameter on the Android application. From Figure 4 B (Figure 4 F), we use low concentration (high concentration) artificial sweat in the electroosmotic pump along with the custom built readout electronics and phone app to measure the change of capacitance with time at rates of 0.25 μL min⁻¹, 1.0 μL min⁻¹, 1.5 μL min⁻¹ and 5.0 μL min⁻¹. The slope of the capacitance vs. time at each rate has been determined. The calibration curve is determined by plotting the slope against the corresponding rate. From Figure 4 C (Figure 4 G), slope in pF min⁻¹ is linear with rates for low concentration (high concentration) artificial sweat.

Figure 4 D shows the temperature while the low concentration artificial SR is kept fixed at 0.25 μL min⁻¹. Since the measurement is done at room temperature, the temperature shows an average of 27.7°C. Figure 4 C and 4 G represent the lower and upper boundary in the calibration curve. From Figure 4 H, these boundaries align with the analytical model. The shaded region represents the analytical modeling data while we change the dielectric constant of the artificial sweat from ${ϵ}_f=30{ϵ}_0$ (coincides with the lower bound of the experimental data represented by the blue dashed line) to ${ϵ}_f=1000{ϵ}_0$ (coincides with the upper bound of the experimental data represented by the red dashed line). However, the human sweat is represented by ${ϵ}_f=80{ϵ}_0$.^[29] We determine the average (represented by the black dashed line) of the upper $\left({ϵ}_f=1000{ϵ}_0\right)$ and lower $\left({ϵ}_f=30{ϵ}_0\right)$ bound found from the experimental data. The averaged calibration curve coincides with the analytical modeling (represented by the black solid line) done with ${ϵ}_f=80{ϵ}_0$. We use the averaged line from the experimental data as the calibration curve for the readout electronics to convert capacitance measurements to SR and to continuously monitor the on-body in-situ perspiration. The Android application facilitates monitoring the continuous SR with a single click and can also store the data for future processing if required (Supplementary Video 2).

2.6. Continuous On-body in-situ SR Measurement.

To wear the sensor on the body and to avoid sweat leaking, we use double-sided tape to secure the placement of the sensor on the skin. We also insulate the readout electronics to avoid direct contact with sweat.

The on-body performance of the SR sensor has been validated by inducing sweat through pilocarpine iontophoresis (Figure 5 A, B). We perform this chemical sweat induction on both forearms. Then, we use our SR sensor on the left forearm and the Macroduct sweat collection device on the right forearm. We ensure the inlet for both sensors is in the middle of the area where we apply the pilocarpine (inset of Figure 5 A). We use the same adhesive tape to wear both sensors. Our sensor uses the calibration curve of Figure 4 H (Black dashed line) to determine the SR. For the Macroduct sweat collection, we determine the sweat volume at a known time and then calculate the SR (Supplementary Figure S12). Our sensor and Macroduct sweat collection device give similar SR (Figure 5 B). We also validate our sensor performance with another commercial sensor, $G^X$ Sweat Patch (Gatorade) (Supplementary Figure S13).

Figure 5. On-body SR sensing.

(A) Photograph of the sensor being worn on the left forearm and a snapshot of the phone application. The inset shows the area for the iontophoresis. (B) The SR is measured by using the rate sensor, which is in agreement with the SR measured by using the Macroduct sweat collection system. (C) Photograph of the sensor being worn on the forehead while the subject exercises on a stationary bike and a snapshot of the phone application. (D) Power of the stationary bike while the subject exercises. On-body SR measurements of 2 subjects. The subject starts to sweat after a certain time. After reducing the bike power to zero the SR reduces. On-body temperature measurement. Temperature increases for both of the subjects at the onset of sweating. (E) Subject has been fasting 5 hours and SR is measured while the subject was biking. The same subject has a sufficient liquid intake for 5 hours. SR is measured again while the subject is biking. Proper hydration increases the SR.

We use the SR sensor while the subject exercises on a bike and collect the data using the readout electronics and the phone app (Figure 5 C, Supplementary Figure S14). The subject starts biking with no power; the power is increased to 100 W at 100 s and again reduced to 0 W at 900 s (Figure 5 D). We collect SR data from 2 subjects. It takes 5 to 10 minutes for the subject to start sweating, and SR varies from subject to subject with the same biking power. At the onset of sweating, the body temperature increases. To evaluate the effect of proper liquid intake on SR, one subject is fasting for 5 hours, and we measure the SR while the subject bikes (Figure 5 E). Since the subject is dehydrated, the SR is very low. The same subject then takes proper liquid intake for 5 hours, and we repeat the biking experiment. The SR increases considerably when the subject is properly hydrated.

3. Discussion.

We have developed a wearable capacitive SR sensor system, incorporating microfluidics for ongoing SR monitoring. The sensor’s performance aligns with both simulation and analytical models of the capacitive system. Our design bypasses the need for intermediate length measurements to determine SR, enhancing its practicality for wearable use. We confirm the accuracy of our sensor by comparing its output with data from an electroosmotic pump. The system includes a readout electronics setup and an Android application that continuously displays the SR. It requires a single calibration, after which it can continuously determine, display, and log data for future analysis if needed. We have also validated our system’s on-body performance against a commercial SR sensor. Our current readout electronics can measure capacitance up to 100 pF, which constrains the sensor length and volumetric capacity. Future improvement could include adopting an alternative capacitance measurement system to overcome this limitation in wearable electronics. Additionally, we have not explored measurement frequencies above 400 kHz due to hardware constraints, although higher frequencies could potentially improve system performance. A potential solution for considering the effects of sweat ionic concentration on capacitance change includes introducing a separate parallel plate capacitor with known dimensions at the inlet of the microfluidic channel. The reading from this capacitor will allow the calculation of the dielectric constant of the incoming sweat. This has the added benefit of indirectly conveying information regarding relative ion levels present in sweat (Supplementary Figure S15). Our fabrication process is cost-effective and straightforward, utilizing only three layers in the sensor patch, which simplifies integration with existing commercial health-monitoring devices without the need for a separate tracking module. This SR sensing system is designed to support user-friendly, continuous SR monitoring tailored for personalized healthcare applications.

4. Materials and Methods

Rate Sensor Electrodes Fabrication

We used a flexible polyimide as the substrate for the SR sensor. For direct 3D writing of metal electrodes and dielectric encapsulation, the Voltera printer was used. We used Ag ink (FE3124 from ACI Materials) for metal parallel plate electrodes. Nozzle diameter of 150 μm, print speed of 500 mm min⁻¹, trace space of 150 μm, print height of 100 μm, dispense pressure of 450, relief pressure of 50 and preheat temperate of 35°C had been used for Ag ink. Ag ink was cured at 130°C for 20 minutes. For the dielectric encapsulation, we used a clear heat-curable dielectric ink (SunTronic IME SB Dielectric Clear). Nozzle diameter of 100 μm, print speed of 400 mm min⁻¹, trace space of 120 μm, print height of 150 μm, dispense pressure of 650, relief pressure of 40 and preheat temperate of 35°C had been used for the dielectric ink. Dielectric encapsulation was cured at 120°C for 15 minutes.

Microfluidics Fabrication

We used a desktop cutter tool, Silhouette cutter (CAMEO 4), to pattern the microfluidic channel and cover. We used double-sided microfluidic tape (3M^™ Microfluidic Diagnostic Tape 9965) for the channel. We used blade depth 4, force 21, passes 2, and speed 3 to pattern the microfluidic double-sided tape. The patterned tape was assembled with the previously fabricated metal electrodes to form the channel. We used a hydrophilic microfluidic cover (3M^™ 9984 Diagnostic Microfluidic Surfactant Free Fluid Transport Film) on top of the channel. We used blade depth 3, force 20, passes 1, and speed 3 to pattern the cover. The patterned cover was then assembled with the double-sided microfluidic tape.

Sensor Characterization

The fabricated SR sensor was characterized by using an electroosmotic pump (EOP-Driven Micro Pumping Unit IBP Series, TAKASAGO ELECTRIC, INC.). Flow of this pump was controlled by using a power source (KEYSIGHT E36234A). The capacitance was measured by an LCR meter (KEYSIGHT E4980AL) at 400 kHz. LABVIEW software platform was used to program the KEYSIGHT instruments. For the on-body sensor characterization, custom-built readout electronics were used.

Wearable SR Sensor Demonstration

On-body measurements were performed at the University of Southern California, LA, in compliance with the human research protection program (HS-24-00019) and approved by the USC Institutional Review Board. There were three participants in the on-body measurements and informed written consent of all participants was obtained. We used a double-sided tape (Daily Wear A Contour 3M^™ 1522 Clear Adhesive Tape) to wear the sensor patch. We encapsulated the electronics by using another double-sided tape (3M^™ Medical Silicone Tape 2477P). This encapsulation ensured that the sweat did not get in contact with the electronic components. During the exercise, a stationary bike (Exerpeutic Folding Exercise Bike) was used.

Artificial Sweat Preparation

All the chemicals were bought from Sigma-Aldrich and used without further purification.

NaCl Solution

10 mL, 1 M NaCl solution was prepared by adding 584.4 mg of NaCl with 10 mL of DI water (Millipak 0.22μm). This solution was diluted to 10 mL of 100 mM by adding 1 mL of 1 M NaCl solution with 9 mL of DI water. 10 mL 40 mM solution was prepared by adding 4 mL of 100 mM NaCl solution with 6 mL of DI water. 10 mL 20 mM Nacl solution was prepared by adding 5 mL of 40 mM NaCl solution with 5 mL DI water.

High Concentration Artificial Sweat

10 mL, 1 M NaCl solution was prepared by adding 584.4 mg of NaCl with 10 mL of DI water. This was diluted to 10 mL 320 mM NaCl solution by adding 3.2 mL of 1 M NaCl solution with 6.8 mL of DI water. 10 mL, 1 M KCl solution was prepared by adding 745.5 mg of KCl with 10 mL of DI water. This was diluted to 10 mL 100 mM KCl solution by adding 1 mL of 1 M KCl solution with 9 mL of DI water. This was further diluted to 10 mL 60 mM KCl solution by adding 6 mL of 100 mM KCl solution with 4 mL of DI water. 10 mL, 100 mM CaCl₂ solution was prepared by adding 110.98 mg of CaCl₂ with 10 mL of DI water. This was diluted to 10 mL 8 mM CaCl₂ solution by adding 0.8 mL of 100 mM CaCl₂ solution with 9.2 mL of DI water. 10 mL, 100 mM MgCl₂ solution was prepared by adding 95.2 mg of CaCl₂ with 10 mL of DI water. This was diluted to 10 mL 10 mM MgCl₂ solution by adding 1 mL of 100 mM MgCl₂ solution with 9 mL of DI water. This was further diluted to 10 mL 1.2 mM MgCl₂ solution by adding 1.2 mL of 10 mM MgCl₂ solution with 8.8 mL of DI water. The high concentration artificial sweat was prepared by adding 10 mL 320 mM NaCl, 10 mL 60 mM KCl, 10 mL 8 mM CaCl₂, and 10 mL 1.2 mM MgCl₂. This prepared a 40 mL solution containing 80 mM NaCl, 15 mM KCl, 2 mM CaCl₂ and 0.3 mM CaCl₂.

Low Concentration Artificial Sweat

10 mL, 1 M NaCl solution was prepared by adding 584.4 mg of NaCl with 10 mL of DI water. This was diluted to 10 mL 100 mM NaCl solution by adding 1 mL of 1 M NaCl solution with 9 mL of DI water. This was further diluted to 10 mL 40 mM NaCl solution by adding 4 mL of 100 mM NaCl solution with 6 mL of DI water. 10 mL 1 M KCl solution was prepared by adding 745.5 mg of KCl with 10 mL of DI water. This was diluted to 10 mL 100 mM KCl solution by adding 1 mL of 1 M KCl solution with 9 mL of DI water. This was further diluted to 10 mL 12 mM KCl solution by adding 1.2 mL of 100 mM KCl solution with 8.8 mL of DI water. 10 mL 100 mM CaCl₂ solution was prepared by adding 110.98 mg of CaCl₂ with 10 mL of DI water. This was diluted to 10 mL 10 mM CaCl₂ solution by adding 1 mL of 100 mM CaCl₂ solution with 9 mL of DI water. This was further diluted to 10 mL 0.8 mM CaCl₂ solution by adding 0.8 mL of 10 mM CaCl₂ solution with 9.2 mL of DI water. 10 mL 100 mM MgCl₂ solution was prepared by adding 95.2 mg of CaCl₂ with 10 mL of DI water. This was diluted to 10 mL 10 mM MgCl₂ solution by adding 1 mL of 100 mM MgCl₂ solution with 9 mL of DI water. This was further diluted to 10 mL 1 mM MgCl₂ solution by adding 1 mL of 10 mM MgCl₂ solution with 9 mL of DI water. This was further diluted to 10 mL 0.08 mM MgCl₂ solution by adding 0.8 mL of 1 mM MgCl₂ solution with 9.2 mL of DI water. The low-concentration artificial sweat was prepared by adding 10 mL 40 mM NaCl, 10 mL 12 mM KCl, 10 mL 0.8 mM CaCl₂, and 10 mL 0.08 mM MgCl₂. This prepared a 40 mL solution containing 10 mM NaCl, 3 mM KCl, 0.2 mM CaCl₂ and 0.02 mM CaCl₂.