2.1 WeaVE design and working principle
Among all types of heat transfer pathways, radiation stands out as the ideal mechanism for wearable thermoregulation devices for several reasons. First, the tunable range is significant. Reports have shown radiative heat transfer accounts for 45–50% of human body heat loss in an indoor scenario28. Calculation based on ASHRAE Standard 55 also indicates that approximately 33% of heat loss is through thermal radiation. Second, the active material for tuning can be ultrathin and lightweight, which are particularly critical for wearable applications. This can be explained by the simplified Stefan-Boltzmann Law: \({\dot{q}}_{rad}= {e}_{surf}{\sigma }_{SB}({T}_{surf}^{4}-{T}_{amb}^{4})\), where \({\dot{q}}_{rad}\), is the radiative heat flux, esurf is the surface emissivity ranging between zero and unity, \({\sigma }_{SB}\) is Stefan Boltzmann constant, Tsurf is the surface temperature, and Tamb is the ambient temperature. Note that we assume the ambient area is much larger and completely enclosing the object so that the view factor is close to unity and the influence of ambient emissivity is neglected. According to Kirchhoff’s Law, emissivity (e) is equal to absorptivity (A) at thermodynamic equilibrium. The emissivity (absorptivity) is determined by the attenuation length, which can be as thin as microns or submicrons for mid-infrared (mid-IR) thermal radiation. For opaque objects, the transmittance (T) is zero, so \(e=A=1-R-T=1-R\), meaning that tuning reflectance (R) is the equivalent of tuning the emissivity and the radiative heat transfer.
As shown in Fig. 1B, our WeaVE device is a layered semi-solid electrochemical cell focusing on mid-IR electrochromism. The device consists of a working/counter electrode pair, and both are electrodeposited polyaniline (PANI) on an Au-sputtered nanoporous nylon membrane. In recent years, PANI and other conjugated polymers, such as PEDOT:PSS, have been viewed as promising materials for high-range non-volatile tunable optics29–35. Prior research in electrochromic applications of PANI-based material focused on spacecraft thermal control rather than personal thermal management36–47. In the working electrode, the underlying Au layer acts as a high-reflectivity (low-emissivity) layer. The PANI above will be electrochemically switched between a transmissive dielectric and a lossy metallic state, thereby realizing the mid-IR electrochromism that can vary emissivity to stabilize the radiative heat loss at a varying ambient temperature.
The optoelectrical property of PANI can be varied by switching continuously and reversibly between oxidation states. In acidic media, the three oxidation states of doped PANI are leucoemeraldine, emeraldine salt, and pernigraniline. Leucoemeraldine and emeraldine salt are the two favorable states in this device. The high transmittance of leucoemeraldine in mid-IR guarantees the low emissivity provided by the Au back reflector. On the other hand, when PANI is oxidized to emeraldine salt, its conductivity and sub-skin-depth thickness make it lossy and absorptive, therefore enhancing the emissivity of the device. The thickness of PANI was controlled by setting a specific polymerization charge during electrochemical deposition. The amount of PANI on the counter electrode is three times more than the working electrode, acting as a charge reservoir to ensure complete charge transfer and stable switching voltage14. Ionic conductivity is provided by the PVA/PEG-based gel electrolyte with sulfuric acid, which permeates in the Au-coated nanoporous nylon membrane. Prior to electropolymerization, kirigami patterns are defined by laser cutting, which is fast, precise, and scalable. In our earlier attempts, the laser-cut electrochemical cells were vulnerable to catastrophic crack propagation during stretching. Strong mechanical properties were accomplished by choosing nylon among other kinds of nanoporous membranes and implementing hot roll-pressing.
2.2 Mid-IR electrochromic property characterization
As shown in Fig. 2A, PANI is electrochemically biased between − 0.1 to 0.7 V versus Ag/AgCl, and its reflectivity was measured by the Fourier transform infrared (FTIR) spectrometer equipped with a diffuse gold integrating sphere. As there is no transmission in the presence of the Au layer, emissivity is calculated as 1-reflectance for each wavelength. At -0.1V, the PANI is in leucoemeraldine form, which is more IR-transparent, revealing the low emissivity of underlying Au, with an emissivity of ~ 0.38. As the potential increases, the active layer is oxidized to the emeraldine salt form, enhancing the optical conductivity and emissivity simultaneously. At 0.4 V, the emissivity of PANI/Au reaches the maximum (~ 0.65), which indicates the transformation into emeraldine salt is finished. The emissivity starts to decrease when applying a voltage larger than 0.4V. The decrease of emissivity is mainly due to the presence of insulating pernigraniline, which is unfavorable because of the irreversibility and instability. The low conductivity of pernigraniline makes it more difficult for switching. Note that the thickness of the PANI-based conductive polymer lies in the sub-skin depth region, and the enhancement of emissivity is attributed to the increase in conductivity. Cyclic voltammetry is also conducted to further understand the oxidation/reduction behavior of PANI. (Fig. S1)
Because of the small thickness of PANI, emissivity is determined by both PANI and the underlying Au layer together. At mid-IR frequency, Au is very close to a perfect electric conductor (PEC), but the thickness of PANI needs to be optimized to achieve large emissivity contrast, Δe. As shown in Fig. 2B, samples with different thicknesses were prepared by varying the deposition areal density. Their emissivity spectra at -0.1 and + 0.5V (versus Ag/AgCl) were measured and weighted-averaged based on black body radiation at 34℃, the typical human skin temperature. The trade-off behavior between a large absorption coefficient and mid-IR transparency can be observed. Low electropolymerization charge density (0.144 C/cm2) provides higher reflection from the gold, but the thickness is not enough to provide high emissivity (absorptivity). On the other hand, when increasing thickness, both states are too dense and lossy, reducing the overall performance of the device. Consequently, Δe is maximized at an intermediate value of 0.192 C/cm2.
The in-situ emissivity tuning kinetics and cycle life measurement of WeaVE is conducted. In Fig. 2C, the representative cycle data shows its fast switching and stability as an electrochromic layer. We can see PANI can reach 90% of the dynamic range within 16 seconds. Figure 2D shows the in-situ cyclic spectra measurement. Each voltage is applied for 45 seconds (90 seconds per cycle). The experiment results showed a 5-percentage decrease in Δe after 100 cycles of switching. The decrease in performance may result from the drying of the electrolyte, or the drifting of equilibrium electrochemical redox potential of PANI. The drying issue can be solved by adding a passivation layer on the surface of the device, and the voltage drifting can be avoided by preconditioning the materials before assembly.
2.3 Skin temperature stabilization
To experimentally quantify the adaptability of WeaVE device, we first demonstrate the ability of active tuning of the WeaVE device by placing it under a controlled varying temperature environment (Fig. 3A). WeaVE or traditional textile (58% cotton, 37% polyester, and 5% spandex) is attached on the top surface of a heating unit, which supplies a constant heat flux upward. The heating unit consisted of two stacked polyimide heaters with the same temperature, separated by a PDMS spacer. To ensure the heat flow is one-dimensional and upward, a bottom auxiliary heater that follows the variation of the top heater with PID control is used. The temperature of skin, ambient, top heater, and bottom heater, denoted as Tskin, Tamb, Ttop, Tbot, are measured and monitored by thermocouples. When Ttop = Tbot, there is no heat flow within the PDMS, and all the power supplied to the top heater will be upward, which is 105.2 W/m2 throughout the experiment.
The thermoregulation of WeaVE is demonstrated in Fig. 3C by comparing with traditional textile in Fig. 3B. Both traditional textile and WeaVE device are exposed to the same ambient temperature condition. When covered with traditional textile, the temporal variation in Tskin temperature has the same trend as the controlled ambient temperature, except for the lag in time due to the thermal inertia, which can be observed by the ~ 7 minutes difference in the occurrence of the minimum temperature. With a 6.2°C transient variation, ranging from 23.3°C to 17.1°C, traditional textile shows a 5.0°C temperature difference. On the other hand, WeaVE can stabilize the artificial skin temperature by adjusting the emissivity states at different voltages: begin with + 0.7 V in the cooling state, and the voltage was decreased stepwise (0.1 V interval) to -0.7 V in heating state, and back to 0.7 V in the end. The time for applying different voltages is manually determined by observing the amount of Tskin decrease when conducting the experiment. By adjusting the emissivity and thus the radiative heat loss, WeaVE can reduce the artificial skin temperature fluctuation to 1.3°C while the ambient temperature fluctuates by 6.2°C. This means the WeaVE can expand the thermal adaptability by 6.2–1.3 = 4.9°C (8.8°F). To put this number in a daily life perspective, this 4.9°C (8.8°F) of expansion means the user will feel the same thermal comfort no matter the ambient temperature is 22.0°C (71.6°F) or 17.1°C (62.8°F).
Aside from thermoregulation, the heat transfer and electrochemical measurement also demonstrate the energy efficiency of WeaVE. In contrast with traditional HVAC systems that actively provide work input, WeaVE modulates the skin temperature by changing the emissivity, the intrinsic material property of the electrochromic layer. It is important to point out that WeaVE requires no energy to maintain its emissivity. We can calculate the effective power provided by WeaVE by comparing the heat transfer coefficient. The heat transfer coefficients in the low-e and high-e states are 8.98 and 6.79 W/m2K. That is, when the ambient and skin temperature difference is 15.9K, the effective power is equivalent to \(\left(8.98-6.79\right) \text{W}/{\text{m}}^{2}\text{K}\times 15.9 K=34.8 W/{m}^{2}.\) To put this energy consumption in the real-life perspective, if using an electric heater to supply the same heating power for the area of 1 m2, this electric heater will drain a flagship smartphone battery (4500 mAh, 3.7 V) in less than 30 minutes. In contrast, WeaVE can indefinitely provide the warming effect with zero energy input because it is the heat transfer coefficient that is being tuned. We further include the energy for switching by considering the time series of current density and voltage of the redox reaction. As shown in Figure S2, the switching energy per cycle is approximately 5.58 mJ/cm2, which means an 1 m2 large WeaVE can switch more than 1,000 times by the aforementioned smartphone battery. Even if we conservatively assume the WeaVE is switched 5 times a day, it can still last for 200 days or several months of usage. This calculation clearly shows that WeaVE is orders of magnitude more efficient than traditional active thermoregulation due to its non-volatile and reversible tuning mechanism.
To realize the functionality of autonomous tuning, we incorporate the sensing and controlling electronic components with WeaVE, including Arduino Uno controller (Fig. S6), a thermistor, a humidity and temperature sensor (DHT11), a Bluetooth module, and smartphone application (Fig. S7). The control loop diagram is shown in Fig. 3D. The mobile app contains the user interface items, including the input of control parameters and the information of current system status (ambient and skin temperature, WeaVE emissivity, estimated heat loss, applied voltage, relative humidity). Desired upper and lower limits of temperature tolerances are entered into the smartphone app and transmitted to the Arduino controller via Bluetooth. Skin temperature is detected by the thermistor and transmitted into Arduino Uno from the thermoresponsive unit (Fig. S8), and then emissivity and applied voltage to WeaVE are calculated. Ambient temperature and humidity, which are used to calculate the heat loss, are measured by DHT11. The working principle illustration and the picture of the system are shown in Fig. 3E and 3F, respectively. The whole system is powered by three 9V alkaline batteries. The photographs of the WeaVE integrated with the personal thermoregulation system in both heating and cooling state are shown in the inset of Fig. 3F, where the upper and lower limits are at 23.0°C and 33.0°C, with ambient temperature at 22°C. The applied voltage is varied linearly within the temperature limits.
The heterogeneous deformation and the complex curved configuration on various parts of the human body result in the trade-off between the large contact area (device-human skin interface) and conformability, especially during deformation. Therefore, kirigami patterns are utilized to fit various body parts of humans. Here, different from existing wearable devices, we harness the correlation between the morphology (the principal curvature) of human skin and the maximum tensile strain in the attached device to guide the general design of the scalable and personalizable WeaVE device with zero-energy-cost emissivity adaptivity coupled with conformable capability.
To examine the mechanical stability and durability of the WeaVE device, we have conducted both the uniaxial tensile test (Fig. 4A) and the cyclic test of three different types of samples (Fig. 4B) with an introduced small notch (inset of Fig. 4A), including roller-pressed nylon (the materials used for WeaVE device) and pristine nylon and polyethersulfone (PES) for comparison. Figure 4A shows the stress-strain curves of the three samples. Compared with PES, both nylon samples show a much higher fracture strain and stiffness, thus, leading to a higher toughness. The toughness of the nylon and roll-pressed nylon are measured to be 85.69 kJ/m2 and 97.34 kJ/m2, respectively (Supporting Information), indicating that the toughness of the nylon sheet is enhanced by roll pressing (Fig. 4A). Moreover, the low cycle fatigue tests under strain control of the three notched sheets show that all three types of samples exhibit negligible hysteresis and permanent deformation, as shown in Fig. 4B. After 1,000 cycles of loading and unloading, the residual strain is about 0.1% (Fig. 4B, i-iii), and the maximum stress of each curve almost does not change (Fig. 4B, iv). The stability and notch-insensitivity of the device support the application of kirigami designs to improve conformability.
Here, we choose two different patterning cuts, triangular and parallel, to make the device conformable to different parts of the human body that undergo different levels of bending strains (Fig. 4C-4D). As validated by both the finite element method (FEM) simulation (Fig. 4D) and the experiments, the WeaVE device deforms with the human skin conformably. The parallel cuts facilitate bending for directions both along and perpendicular to the cuts and make each ribbon a geodesic curve of the human skin, where conformability and adaptivity arise naturally. Thus, the parallel-cut pattern is especially suitable for joints such as knees and elbows. As shown in Fig. 4D, i, the parallel-cut precursor is first bent to fit and wrap around the curved configuration of the human knee, where the normalized bending curvature \({\stackrel{-}{\kappa }}_{1}\) increases from 0 to 1.78 (i.e., the curvature of the knee in the direction 1). Then when the knee bends, it bends with the knee smoothly with the bending curvature \({\stackrel{-}{\kappa }}_{2}\) increasing from 0 to 1.09 (the curvature of the knee in the direction 2 when the bending angle is 90\(^\circ\)). \({\stackrel{-}{\kappa }}_{1}\) and \({\stackrel{-}{\kappa }}_{2}\) are the principal curvatures normalized by the width of the human knee. On the other hand, the triangular-cut pattern exhibits an approximate isotropic-morphology feature, and its morphology varies smoothly with the inhomogeneous deformation of the human skin (Fig. 4D, ii). When stretched by the human skin, the triangular network opens progressively to fit the skin with both \({\stackrel{-}{\kappa }}_{1}\) and \({\stackrel{-}{\kappa }}_{2}\) increasing from 0 to 1.78.
Further, to make our device scalable and personalizable, a general and scale-independent correlation between the normalized principal curvature \({\stackrel{-}{\kappa }}_{i}\) and the maximum principal tensile strain is derived both theoretically and numerically. As shown in Fig. 4C, the maximum principal tensile strain \({\epsilon }_{max}\) in the device increases linearly with increasing curvature \({\stackrel{-}{\kappa }}_{i}\) of the human skin. It is also noteworthy that the maximum tensile strain of both patterns is much smaller than the fracture strain (Fig. 4B), which shows the stability and robustness of the WeaVE device.
In the fabrication process of kirigami WeaVE, a SEBS passivation layer was coated on the surface to prevent electrolyte overflow and drying. In Fig. 4E-H, the thermal image of two volunteers wearing kirigami WeaVE was captured in an ambient temperature of 17℃ at 0.5V and − 0.5V, respectively. The image visualizes the apparent temperature change when WeaVE device is in both states. Compared to bare skin, the area covered with WeaVE has a tunability of apparent temperature of ~ 2.5℃, representing the adjustable thermal insulation. Moreover, because the mid-IR wavelength range greatly differs from visible light, we can apply a visibly colored, mid-IR-transparent layer to render the WeaVE device with different visual appearance for aesthetic preference without affecting the radiative thermoregulation performance (Fig. S5).