Recently, wearable sensors with the advantages of noninvasiveness, miniaturization, and low cost have made significant progress^[1–4]. They can be used for noninvasive monitoring of biological fluids (e.g.,  sweat, tears, and saliva) in order to enable the assessment of the wearer’s physical fitness and injury status^[5–9]. Among them, sweating behavior is an important physiological function for the body to cool down under heat stress. With the increasing demand for health monitoring, sweat has emerged as a potential noninvasive bio-specimen for real-time monitoring of human health status^[10]. Sweat contains a wealth of physiological information, and monitoring biomarkers in sweat is necessary as it can reflect the physiological states^[11]. The electrolyte is one of the important components for maintaining homeostasis within the body^[12,13]. Changes in electrolyte concentrations reflect the state of water–salt balance in the body, which has an important impact on neuromuscular function and cardiovascular health. Otherwise, excessive sweat without rehydration can lead to the loss of body fluids and electrolytes (e.g.,  sodium), resulting in disturbance of the body’s internal homeostasis and potentially contributing to heat-related illnesses^[8,14]. Therefore, using wearable sensors to accurately detect the electrolyte concentrations in sweat is important for monitoring individual health, exercise status, and disease diagnosis.

Self-driven sensors have advantages such as low energy consumption, high-speed response, low noise, and small size^[15,16]. (In,Ga)N materials are favored due to their long lifetime, small size, and tunable direct bandgap properties. However, the epitaxial substrates traditionally used to grow (In,Ga)N materials, such as Si and sapphire, are rigid and inflexible^[17,18]. That means conventional (In,Ga)N materials cannot be used to fabricate wearable sensors directly. In our previous work, a method with low cost was invented to detach GaN materials^[19]. However, up to now, wearable self-driven (In,Ga)N sensors have not yet been applied in sweat monitoring.

In this work, we realize a wearable self-driven sensor based on (In,Ga)N film, which is achieved by an economical and fast electrochemical (EC) etching method. Under both flat and bending conditions, the behavior of the generated current has been studied systematically. The stability, the underlying mechanism of the sensor, has also been investigated. Apart from the advantage of ultra-low energy consumption, wearable sensors can monitor the amount of sweat electrolyte concentration, which is crucial for maintaining fluid balance for human health during exercise.

In Fig. 1(a), the GaN-based epitaxial structure was grown on a silicon (Si) substrate. After that, an indium tin oxide (ITO) current spreading layer of 200 nm was deposited on the 5.08 cm surface. The specimens were sectioned into 10mm×10mm pieces and subsequently cleaned for 10 min using acetone followed by isopropanol. A 150 nm layer of SiO2 was deposited on the surface of the (In,Ga)N film. Then, a part of the SiO2 film was selectively removed to expose a certain area of the ITO layer and n-GaN layer. Next, 240 nm of Ti/Al metal film was selectively deposited on the ITO layer by electron beam evaporation. An annealing treatment was performed to make the Ti/Al layer highly conductive afterward.

On the surface of (In,Ga)N films, silver bars were immobilized using silver paste on the Ti/Al alloy layer. The specimens were maintained at a temperature of 130°C for a duration of 30 min to ensure the solidification of the silver paste. To prevent the silver strip from undergoing corrosion during EC etching, the silver strip was then protected by a high-temperature tape that was completely laminated to ensure that it would not be exposed. After melting the Sn/Pb alloys, the Si backside was connected to a conducting wire through soldering. To safeguard the Ti/Al alloy layer from corrosion during the EC process, it was coated with epoxy resin. During the EC etching, the Pt sheet served as the cathode while the sample acted as the anode, both immersed in a 1 mol/L solution of nitric acid (HNO3)^[20,21]. After the film was lifted off from the Si substrate, the prepared flexible devices are shown in Fig. 1(b). When the sensor is worn on the hand to monitor sweat data, a certain degree of bending occurs [Fig. 1(c)]. Under the actual wearing condition, the sensor needs to be secured to the wrist with a strap. Artificial sweat was used to mimic the human perspiration during the test, which is a synthetic testing reagent developed based on the composition of human metabolic sweat. It contains components such as NaCl, amino acids, lactic acid, and phosphates. More details about the epitaxial layer and measurements are shown in Sec. S1 in the Supplementary Material.

The clear (In,Ga)N/GaN multiple quantum wells (MQWs), (Al,Ga)N/GaN, and p-GaN interfaces can be clearly seen in the aberration-corrected scanning transmission electron microscope (AC-STEM) image [Fig. 1(d)]. No obvious material defects, such as growth dislocations and penetration dislocations, are observed. In Figs. 1(e)–1(g), clear lattice stripes of GaN and (In,Ga)N crystals are evident to signify good crystallinity within the active region of the MQWs. The thicknesses of the (In,Ga)N barrier layers and GaN potential wells are 3 and 10 nm, respectively, demonstrating good uniformity and aligning well with the intended design.

As the incident optical power intensity (Pin) increases, the photogenerated current (Iph) increases simultaneously [Figs. 2(a) and 2(b)], exhibiting the optical power-dependent characteristic [Fig. 2(a)]. The device demonstrates the capability for self-driven detection, as the measurements for detection are conducted under a bias voltage of 0 V^[22–24]. The responsivity (R) is up to 88.7μAW−1 when the incident optical power intensity is about 13.75mWcm−2 [Fig. 2(b)]. The factors that significantly impact the response current density include the restriction on the electron–hole pair counting, carrier scattering due to self-heating, and elevated charge complexation rate. As shown in Fig. 2(c), the rise time (Trise) is 0.04 s, and the decay time (Tfall) is 0.04 s. In addition, the sensor responds well to visible light (410 nm wavelength, Fig. S1 in the Supplementary Material), which indicates that the sensor can be used under sunshine.

To demonstrate the flexibility and stability of the self-driven sensor, we have conducted bending tests on the device [Figs. 3(a)–3(c)]. Sensors can be bent at angles of 18°, 25°, and 33°. After bending the flexible sensors in different degrees, a consistent and discernible optical switching performance was maintained. Under bending conditions with an angle up to 33°, the photocurrent exhibits an augmentation in response to an increase in the incident light power density [Fig. 3(d)].

In addition, the photocurrent produced during the 33° bent configuration resembles that observed in the unbent (flat) state [Fig. 3(e)]. The photocurrent value even rises slightly in bending conditions, which could be attributed to the lower distance between the sensor and the light source. As shown in Fig. 3(f), the sensor can still maintain a relatively clear response after multiple bending cycles. Hence, the sensor exhibits a good flexibility and stability under bending conditions.

To gain deeper insights into the operational mechanism of the device, Fig. 4 presents the schematic and energy band diagrams of the (In,Ga)N sensor under three conditions of the flat state [Fig. 4(a)], the bent states without the presence of liquid [Fig. 4(b)], and with the presence of liquid [Fig. 4(c)]. When external optical radiation is applied, photogenerated electrons and holes start to flow [Fig. 4(d)], resulting in the photogenerated current [Fig. 2(a)]. When the (In,Ga)N film undergoes bending and illumination [Fig. 4(b)], the positive piezoelectric charges are generated on the top surface of the film at the ITO/(In,Ga)N interface, while negative piezoelectric charges are generated on the bottom surface of the film^[25], forming a built-in electric field that causes the energy band to bend upward [red curve in Fig. 4(e)]^[26,27]. This bending of the energy band slows down carrier transport, leading to a reduction in optical responsivity. Additionally, after removing the rigid Si substrate through EC etching, the internal stress in the (In,Ga)N film is reduced, suppressing strain-induced bandgap shifts. The release of strain decreases the spontaneous and piezoelectric polarization within the material, thereby stabilizing the energy band structure^[28]. When the surface of the film is in contact with a liquid [Fig. 4(c)], the photogenerated holes are transported to the water, whereas the electrons are directed toward the n-GaN layer [Fig. 4(f)]. Under 365 nm light illumination, the generation of photocurrents is facilitated through a series of reactions:4H++4e−=2H2,(1)4h++2H2O=O2+4H+.(2)

When the GaN layer interfaces with the electrolyte of water, an EC equilibrium is established by the migration of excess carriers across both the top and lateral film/water boundaries. Subsequently, the photogenerated holes can readily migrate toward the film/water interface. In human sweat, carrier transport can be facilitated by the presence of electrolytes such as NaCl. As (In,Ga)N material is quite stable, the bending states have limited effects on the photogenerated carriers, leading to high stability [Figs. 3(e) and 3(f)].

Since the sensor maintains close contact with the skin in the wearing condition and the response time is rapid [0.04 s, Fig. 2(c)], the effect of evaporation can be suppressed. NaCl is the primary contributor to the electrolyte^[29]. Since the concentration of NaCl in human sweat is in the range of 40–80 mmol^[30] and is capable of reaching 100 mmol when dehydrated, the sensor has been measured in the concentrations from 10–160 mmol [Fig. 5(a)]. It is clear that the output current increases with increasing NaCl concentration in the range of 10–160 mmol [Fig. S2(a) in the Supplementary Material], demonstrating that the sensor is sensitive to electrolytes in this range. Figure 5(b) shows the variation of the output current obtained by the sensor in the same NaCl concentration (∼80mmol) with different volumes of artificial sweat. From Fig. S2(b) in the Supplementary Material, the output current shows a linear increase with the increasing volume of artificial sweat at the same electrolyte concentration. As a result, electrolyte loss in sweat can be detected. As previously reported^[31], the sodium concentration exhibits a positive correlation with the sweat rate during increased exercise intensity. When evaluating the feasibility of the device under passive sweating (low sweat volume) conditions, it is necessary to consider the challenges posed by insufficient sweat secretion and stability^[32,33]. In this study, the photocurrent of the self-driven flexible sensor exhibits a positive correlation with the sweat electrolyte concentration/volume (Fig. 5). Therefore, this wearable sensor has the potential to aid in formulating effective hydration plans for individuals during physical activity or exposure to high temperatures in warm climates. In future work, various biomarkers (e.g.,  glucose and lactate) in sweat would be worth studying. According to the long-time reliability measurement (Fig. S3 in the Supplementary Material), it is demonstrated that this sensor maintains stability after working 360 cycles.

In this study, we have developed a wearable self-driven sensor utilizing a lift-off (In,Ga)N film as its foundation successfully. Due to its exceptional flexibility, the self-driven sensor can perform in both flat and curved configurations. The sensor exhibits highly stable optical switching performance across various bending scenarios. Furthermore, the self-driven sensor can continuously detect sweat volume and NaCl concentration, which is crucial for human health during physical activity or exposure to high temperatures. Proper alignment of energy levels can enhance the separation and transfer of carriers at the interface between the film/electrolyte interface. Therefore, with the various advantages of small size, bendability, and stable performance, this sensor holds extensive potential application prospect in the fields of health monitoring, sports medicine, biosensors, etc.