Electrical tuning of radiative cooling at ambient conditions

Banerjee et al. demonstrate electrical thermoregulation at ambient conditions via tuneable passive radiative cooling, which works as a thermostat for radiative cooling technologies. They show that the temperature of their devices can be reversibly modulated by electrochemically tuning the redox state and the corresponding thermal emissivity of a conducting polymer.


INTRODUCTION
Energy efficient means to regulate temperatures of buildings, vehicles, and devices is essential for the transformation to a sustainable society.][6][7][8][9][10][11][12][13][14] Of particular interest is the development of tuneable and adaptive systems, [9][10][11][12][13]15,16 with electrical tunability being desirable due to, for example, its potential for remote operation. Our goup 17 and others 11 demonstrated the use of electrical emissivity tuning to vary apparent device temperatures as seen by thermal cameras, but such measurements do not provide evidence of real temperature variations because they are sensitive to the emissivity change itself.Instead, absolute temperature control via electrical tuning of radiative cooling power was, to our knowledge, limited to nonambient conditions in vacuum.15 Hence, the challenge remains to demonstrate tuning of device temperatures at ambient conditions by electrically varying the radiative cooling power.In turn, adding such a tuning knob could transform the concept of radiative cooling from static cooling to adaptable thermoregulation of buildings, vehicles, and devices.Furthermore, dynamic control of thermal radiation from surfaces holds promise for improving indoor comfort at reduced or elevated temperatures by the regulation of thermal energy flows between individuals, walls, and objects.6,18,19 The conducting polymer poly [3,4-ethylenedioxythiophene]:tosylate (PEDOT:Tos) has shown promise by providing strong infrared (IR) electrochromism, 17 which we here exploit to demonstrate electrical regulation of device temperatures at ambient pressure and temperature solely by tuning the radiative cooling of the device. Fiure 1 depicts the general concept and device structure, by which emissivity tuning of the outermost surface of the device is enabled by sandwiching two PEDOT:Tos films by an ionically conductive electrolyte-filled porous cellulose layer.Varying the electrical potential difference between the two PEDOT electrodes controls the redox state of the top (and bottom) PEDOT layer, thereby forming a means to tune the capability of the device to emit thermal radiation.Using an indoor sky simulator, our study shows that this can be used to induce clear variations in device temperature that originate from the electrical modulation of the radiative cooling power of the device.

RESULTS AND DISCUSSION
Spectral IR characterization of free-standing PEDOT:Tos films We prepared around 200-nm-thick PEDOT:Tos films by vapor-phase polymerization, followed by transfer to PET substrates having holes punched in.This enabled determining the IR reflectance and transmittance of free-standing PEDOT:Tos films by integrating sphere spectroscopy (see experimental details in the experimental procedures).The pristine polymerized PEDOT:Tos is in its oxidized and highly conducting state (see chemical structure in Figure 1), resulting in strong IR reflection due to a high density of polaronic charge carriers and optically metallic properties (Figure 2A). 20In this state, the film is low emissive in a broad range around 10 mm, which corresponds to the relevant spectral region for Planck radiation from objects at room temperature.This means that the material is a poor radiative cooler in this state.Partially reducing the film by exposure to vapor from poly(ethylene imine) (PEI) [21][22][23] led to a decrease in the IR reflectance (Figure 2A), and the material instead increased its IR absorptance.The reason is that reduction of PEDOT reduces its polaronic charge carrier density (compare chemical structures in Figure 1), which leads to a low conducting material with correspondingly less efficient metallic reflection.There was still considerable IR reflection for the chemically reduced film, as partly attributed to spontaneous re-oxidation of the polymer before the IR reflection measurements.It is known that PEDOT can spontaneously re-oxidize under normal oxygen atmosphere. 24Importantly, the effect of decreasing the reflectance is that the IR light instead becomes available for absorption by molecular vibrations of the organic material, while the transmittance through the film was close to zero for both states.This means that variations in IR reflectance will directly transform to variations in thermal emissivity of the film, which is promising for tuneable control of thermal radiation.

Electrical tuning of the IR properties of radiative cooling devices
To reveal the possibility to electrically tune the IR reflectance and thermal emissivity, we fabricated electrochemical devices with PEDOT:Tos films as the top layer.The tuneable PEDOT film sits on a porous cellulose layer infused by an electrolyte, which ionically connects the top film with a bottom counter electrode (also made of PEDOT:Tos; see schematic in Figure 1).This configuration allows electrical tuning without the need for an electrolyte on top of the PEDOT film, thereby maintaining the PEDOT film as the outermost tuneable thermal emissive layer of the device.Figure 2B presents the IR reflectance and thermal emissivity at different electrical biases for a device with an around 50-nm-thick PEDOT:Tos top electrode.The results clearly show that the device concept can provide tuneable IR reflectance and thermal emissivity in the mid-IR spectral range that is relevant for thermal emission at ambient temperatures.The spectrally flat reflectance/emissivity could be tuned gradually by varying the voltage applied to the top electrode, using a modest voltage range between À1.5 and 1.5 V.In fact, maximum reflectance was observed already at lower voltages for this device, followed by a slight decrease at 1.5 V.The observed variation in IR reflectance corresponds to variation in the thermal emissivity of the device from around 0.54 at positive voltages (for which the PEDOT is in its oxidized high conducting state) to around 0.77 at negative voltages (at which the PEDOT is reduced to its low conducting state).This agrees with the oxidized PEDOT being low emissive thanks to its high density of charge carriers and metallic properties, while the reduced PEDOT with its drastically lowered charge carrier density becomes a more efficient thermal emitter thanks to its lowered IR reflectance and correspondingly increased emissivity caused by molecular absorption.These results are consistent with the redox switching behavior of the PEDOT material also in the visible and near-IR regimes (Note S1; Figure S1).
Devices with different PEDOT top layer thicknesses provided similar tuning capabilities but with different absolute values.This is shown in Figure 2C, which presents the IR reflectance and thermal emissivity at 10 mm as a function of bias potential for devices with PEDOT top layer thicknesses from around 50 nm to around 200 nm.The wavelength of 10 mm was here chosen as it corresponds to the approximate peak position for Planck radiation from objects at room temperature.The main influence of increasing the PEDOT top layer thickness was larger IR reflectance and lower thermal emissivity.The absolute change in reflectance and thermal emissivity upon electrical tuning was about the same regardless of film thickness, around 0.2-0.25.For example, a device with around 53 nm PEDOT provided emissivity tuning from 0.54 to 0.77, while a device with around 200 nm PEDOT showed tuning from 0.44 to 0.64, corresponding to an emissivity increase of 44%.We note that these changes are larger than those measured for the free-standing pristine and chemically reduced PEDOT films (Figure 2A), highlighting the strength of electrochemically tuning the conducting polymers in situ.The low operating voltages of the concept are an advantage, and applying larger voltages than G1.5 V did not lead to larger emissivity tuning (Note S2; Figure S2, which also presents the emissivity tuning in a wider spectral range).

Temperature tuning of radiative cooling devices
We will now demonstrate that the variations in thermal emissivity provided by the devices can be used to electrically modulate their temperature at ambient conditions through changes in their ability to emit thermal radiation.We first note that nighttime outdoor thermal camera imaging showed large changes in the apparent temperature of around 2.4 C upon switching (Note S6; Figure S6).However, those measurements mostly confirm the emissivity tuning itself, and we need another approach to conclusively prove changes in absolute device temperature.To tackle this challenge, we monitored the device temperature using thermocouples in a sky simulator, as depicted in Figure 3A and detailed in the experimental procedures.In brief, thermal radiation from the device was guided along an IR-reflecting aluminum tower (see reflectance data in Note  S3) toward a liquid nitrogen-cooled IR absorbing black foil acting as ''the sky.'' 25The cooled foil emulates the optical properties of the clear night sky by providing low back reflection of thermal radiation (as emitted from the device) and negligible thermal emission of its own thanks to its cold temperature of $77 K (despite its high emissivity).A comparison suggests that the total back radiation in the atmospheric window is similar for the sky simulator (z33 Wm À2 ) and the real sky (z50 Wm À2 ; see details in Note S7).Although the calculations do not account for the sky simulator providing a larger effective transparency window, we conclude that it enables evaluation of radiative cooling performance in a stable environment that is relevant for but not identical to real outdoor conditions.The net radiative heat flux Q from the device to the simulated sky can be approximated as 16,26

S3 and corresponding Figure
where s is the Stefan-Boltzmann constant; A device and A sky are the effective areas of the device and sky, respectively; and e device and e sky are the thermal emissivity of the device and the sky, respectively, if approximated as gray bodies (spectrally flat emissivity).The possibility to modulate the radiative heat flux through changes in the thermal emissivity of the device becomes even more clear if approximating the sky material as a perfect black body, which reduces the heat flux to (Equation 2) Figure 3B shows the temperature evolution (top panel) of a tuneable cooling device at ambient conditions during repeated switching of the applied electric potential between G1.5 V. Excitingly, the device temperature (gray and black lines) showed clear and repeatable steps between two consistent temperature levels upon switching, while the chamber reference temperature (green) was not affected by the switching process but maintained steady at a higher temperature.This is also clear from the bottom panel, which presents the difference in temperature between the device and the reference.The results should be understood as the conducting polymer device functioning as a radiative cooler with different cooling powers at different potentials thanks to its varying thermal emissivity.For this device, this led to repeatable tuning of the device temperature of around 0.25 C between the oxidized and reduced states of the top PEDOT coating.Confirmation that such temperature variations can be induced purely by changes in the emissivity of the device is provided in Note S9 and Figure S7 using coatings with static but different emissivity.The magnitude of the temperature tuning originates from the difference in net cooling power and also depends on the thermal load of the system.In turn, the difference in net cooling power upon redox switching is given by DP net = P net;red À P net;ox , for which the net cooling powers for the two states can be obtained from P net = P rad À P back À P nonrad .Here, P rad is the thermal radiation power of the device at the specific state and temperature, as calculated to be around 139 Wm À2 for the reduced state and 96 Wm À2 for the oxidized state in the atmospheric window (7-14 mm).P back is the absorbed back radiation (from the sky simulator), and P nonrad accounts for heat exchange with the surroundings.As detailed in Note S8, DP net could be estimated to around 35 Wm À2 in the atmospheric window upon switching the device.The corresponding calculations for expected tuning at outdoor conditions are around 31 Wm À2 for the same spectral range and temperature difference.Although these values are significant and highly similar, we note that the difference may increase if broadening the spectral range, considering that the sky simulator provides lower back radiation than the real sky outside the atmospheric transparency window.
Besides reversible switching between two temperatures, the added knob for tuning of radiative cooling is expected to also enable continuous temperature regulation by gradually varying the oxidation level of the top PEDOT layer.To confirm this hypothesis, we varied the applied electric potential between multiple different electrical potentials while monitoring the device temperature using the sky simulator, as exemplified for a typical device in Figure 3C. Figure 3D summarizes results obtained from several measurements of different devices.The top panel presents the temperature of the devices compared with that of the ambient as a function of voltage, and the bottom panel presents the same but also setting the value of T device À T reference at À1.5 V to 0 (denoted as DT).The results confirm that temperature could be gradually increased with increasing the oxidation level of the PEDOT film.This trend is consistent between devices and measurements, sometimes combined with nonmonotonic variations during tuning.To ascertain that the effect is due to variations in the devices' ability to radiate heat rather than other effects, we repeated the experiment in the absence of the coolant liquid nitrogen in the sky simulator (red triangular markers in Figure 3D).This effectively turns off ''the sky'' as a heat sink and stops the cooling function.The fact that the sky was turned off is evident from the device temperature becoming close to the ambient temperature in this configuration, compared with more than 1 C device cooling when the sky is on/cold (see top panel in Figure 3D).Having the same temperature for the device and the sky further means that the net radiative heat flux from the device to the sky becomes zero regardless of device emissivity (see Equations 1 and 2), such that temperature tuning is no longer expected based on thermal emission control upon redox tuning.Indeed, the absence of the cold sky led to much smaller variations in DT (DT < 0.04 C) upon voltage tuning.DT remained at around zero until neutral potential (0 V), followed by a slight increase for positive potentials, as can be due to different effects including joule heating.Importantly, these variations are much smaller than those observed for the devices that were exposed to the cold sky (z0.2 C).See Note S5 and Figure S5 for data on recovery of temperature tunability on reintroducing ''the sky.''In a second control measurement, the sky was kept cold, but thermal radiation from the device was instead blocked from reaching it by shading the coolant dewar with an IR-reflecting aluminum (Al) foil.This control experiment resembles the situation of a cloudy night rather than turning the sky off.Also, this control experiment confirms our findings.First, the device temperature was not efficiently cooled, and, more importantly, DT remained rather constant (DT < 0.03 C) and did not show any clear trend when varying the applied potential and oxidation state of the PEDOT.The combined results in Figure 3D confirm that the strong temperature tuning of our devices originates from variations in radiative cooling power.

Speed of temperature switching
To examine the speed at which the temperature of the device could be tuned, we repeatedly switched the applied potential between G1.75 V at different time intervals, starting at 300 s and decreasing to 210, 120, and 90 s, and finally increasing again to 240 s (see the dashed line in the top panel of Figure 4A).The real-time data in Figure 4A show that the device temperature could be repeatedly tuned for the longer switching time intervals and that it could not keep up when decreasing to the shortest switching time intervals.The temperature switching then recovered when increasing the time interval again.Figure 4B presents the average of the switching-induced change in temperature for repeated electrical switches as a function of switching time duration.The temperature tuning decreased for faster switching, while for longer switching intervals, it tended to saturate at around 0.12 C for this device.In terms of optimizing the performance of future devices, it would be useful to understand if the switching speed is limited by the variations of the optical properties of the device or by heat flows.We therefore measured the optical switching speed of a tuneable radiative cooling device by monitoring the reflectance of a red laser beam incident from the top side while electrically switching the applied voltage at different time intervals (see details in Note S4).As shown in Figure S4, the optical properties could fully switch for all time intervals including the shortest of s.We therefore conclude that the temperature variation of the tuneable radiative cooling devices is not limited by the tuning of the optical emissivity but rather by heat flows within the device and in exchange with the environment.
Practical applications may use this information to optimize performance via, for example, the thermal conductivity of device and contact materials.

Memory effects for the temperature tuning
It would be advantageous if the radiative cooling device could maintain a pre-set state and corresponding temperature after removing the applied voltage and hence without a need for constant energy input.The longer the device can hold a pre-set radiative cooling state, the less frequently it needs to be reset.PEDOT-based systems are promising in that respect as they can show optical memory effects, 17 which for our concept would translate into thermal memory effects.To examine the thermal memory, we first oxidized or reduced a device at +1.5 or À1.5 V, respectively, and then switched it to the opposite state.This initial part of the experiment provided the two base-level temperatures and switching magnitude for the device.We then turned off the applied potential and monitored the sample temperature over time.Figure 5A shows the case for turning off the voltage when the device was in its reduced strong radiative cooling state.First, switching the device from an oxidized to a reduced state led to a decrease in temperature of around 0.15 C thanks to the increased radiative cooling power.Next, turning off the electric power led only to a minor rise in temperature over 25 min, which was much smaller than the switching magnitude.Furthermore, applying the reducing potential again did not fully recover the initially lower temperature, indicating that the slow rise during the period of no applied potential may not be due to the top layer becoming less reduced.Finally, applying a positive potential successfully switched the device back to the oxidized state, and it recovered its higher temperature due to the reduction in radiative cooling power.The experiment thereby shows that the device provided good optical memory when left unplugged in the reduced state with high radiative cooling power.Further improvement and reduction of spontaneous re-oxidation could be expected if encapsulating the device in IR-transparent barrier films like a thin polyethylene film.
Figure 5B shows the corresponding results when instead turning off the voltage after pre-setting a device to its oxidized state.The results show no tendency of decrease in temperature during around 1 h without applied potential, indicating an excellent memory effect also for the device pre-set to its oxidized state with low radiative cooling power.Reducing the device again led to successful switching back to a lower temperature, although to a higher base level possibly due to small drifts or reduced coolant in the simulated sky.
Our demonstration of temperature regulation at ambient conditions through electrical tuning of radiative cooling widens the application areas of this emerging field.One major goal is to contribute to reduced needs for traditional cooling and climate control systems based on vapor-compression refrigeration, which is both energy consuming and uses hazardous materials.Already regulation of around 0.25 C as demonstrated herein is significant and could be further increased through larger emissivity tuning and changes in other properties such as thermal loads and thermal energy exchange with the surroundings.In terms of size, current devices are around 2-4 cm À2 , and upscaling could be aided by recent advances in screen-printing-based deposition of PEDOT. 27,28or outdoor use during daytime, the concept could either use solar-reflective IR-transparent coatings to avoid solar heating or be combined with electrochromism, also of solar light, for synergistic effects.

Vapor phase polymerization of PEDOT:Tos films and device fabrication
The oxidant for the Tos films consisted of 2 g CB-54 (Fe (III)-Tos in 1-butanol, 12 wt %), 2 g PEG-PPG-PEG (22 wt %), and 5 g 99.5% pure ethanol mixed at 80 C and dropped on cleaned glass slides for spin coating.The oxidant film was annealed at 70 C for 30 s and loaded into a vapor phase polymerization chamber at 60 C, where it reacted with fumes of 3, 4-ethylenedioxythiophene (EDOT) to form the PEDOT:Tos films.
Each glass slide with PEDOT film was carefully put in deionized water and the container was agitated gently to release the PEDOT film to make it float on the water surface.The film was gently transferred to either a PET substrate with holes for freestanding film characterization or to cellulose paper substrates for further device fabrication as follows.The ionic liquid 1-ethyl-3-methylimidazolium ethyl sulfate (EMIM-ES) was mixed with hydroxyethyl cellulose (HEC) in the ratio 10:1 by weight at 100 C for 1 h.A filter paper with the PEDOT:Tos film on one side was soaked in the gel electrolyte and mechanically pressed against another PEDOT:Tos film transferred to the other side of the paper, this being the final step of the device fabrication.

Integrating sphere measurements
The total reflectance, transmittance, and absorptance of PEDOT:Tos free-standing films and multi-layered devices were determined from spectroscopic measurements of the spectral directional hemispherical reflectance (DHR) and directional hemispherical transmittance (DHT) using integrating spheres.For the UV-visible near-IR (UV-vis-NIR) range, a Cary 5000 spectrometer with a Labsphere DRA-2500 integrating sphere was used.It uses two detectors: an R928 PMT for the UV-vis range and a PbS detector for the NIR range.For the mid-IR range from 2 mm, we used a Bruker Vertex 70 FTIR spectrometer with a downward-looking diffuse gold-coated integrating sphere (Labsphere A562) and DTGS detector.The angles of incidence for the two integrating spheres were 8 and 9 in the UV-NIR and mid-IR ranges, respectively.Calibrated Spectralon reflectance standards from Labsphere were used as reference samples in the Cary spectrometer for DHR measurements.For the absolute reflectance measurements in the FTIR, we used Infragold calibration standards to determine the correction factors to be included in the DHR measurements.For each sample measurement in the mid-IR, reference measurements were made with light reflected from the gold-coated inner walls of the sphere the sample was present at the same time.An off-center mirror in the sphere was flipped to direct the incident beam to the sample after the reference measurement and to capture the signals from the sample.The ratio of the sample and reference measurement intensity multiplied by the calibrated correction factor gives an accurate FTIR spectral measurement.A Gamry 1010B potentiostat working electrode was connected to the top PEDOT film and a counter electrode to the PEDOT film at the bottom of the device for voltage switching measurements.

Sky simulator radiative cooling measurements
For the device temperature measurements to be sensitive, repeatable, and accurate, we designed and built a sky simulator.It simulates the cold night sky by having a partially enclosed reflecting Al tower with high IR reflection, as shown in Fig- ure S3A.IR radiation from the sample is guided downward toward the cold sink (the ''sky'') by a gently sloping Al funnel at the bottom end of the square cross sectioned tower.The tower houses an IR-absorbing low-temperature thermal sink near the bottom that serves as the cold sky (schematic in Figure 3A).It is made of black Al-based Cinefoil (Rosco), having high IR absorption and emissivity (see Figure S3B), submerged in liquid nitrogen in a cryogenic flask to maintain a low temperature.The insulated sample area near the top of the tower was sealed by a thin polyethylene sheet to minimize convective heat transfer while allowing IR radiation to pass unhindered.Two sheathed k-type miniature contact thermocouples (model 8103000 from Pentronic AB) were connected to a digital temperature measurement system (Analog Devices DC2608) with 0.001 C temperature resolution (24 bits ADC) that comprises the DC2618, DC2210, and LTC2986-1 for direct computer interfacing and real-time data logging.For most figures showing real-time temperature monitoring, the raw data were plotted together with smoothed results, obtained using LOESS smoothing on the raw data with a window of 30 data points (in Origin).Similar to spectral measurements, for measurements in the sky simulator, a Gamry 1010B potentiostat's working electrode was connected to the top PEDOT film (facing the sky) and a counter electrode to the bottom PEDOT film.

Thermal camera measurements
Thermal images were captured with an IR camera (FLIR Systems, ThermoVision A320G).All the images and characterizations were conducted outdoors at night.A highly reflective metal plate and a highly absorptive rubber foam were used as references for IR camera measurement range calibration.The devices were connected to a potentiostat with different applied biases during the measurements, and video mode was used to record the time response of the apparent temperature variation of the devices.

Figure 1 .
Figure 1.Device concept Schematic illustration depicting the concept of reversible electrically tuneable radiative cooling based on the tuneable thermal emissivity of PEDOT:Tos.The chemical structures of PEDOT:Tos are shown for the reduced (left) and oxidized (right) states.

Figure 2 .
Figure 2. Thermal emissivity tuning of PEDOT:Tos films and devices (A) Total IR reflectance (R), transmittance (T), and absorptance (A) of free-standing pristine (Ox.) and chemically reduced (Red.)PEDOT:Tos films as obtained by integrating sphere measurements.(B) IR reflectance (left y axis) and corresponding emissivity (right y axis) at different electrochemical potentials for a tuneable radiative cooling device with an around 50-nm-thick top PEDOT:Tos electrode (schematic in the inset).(C) Reflectance (left y axis) and emissivity (right y axis) at a wavelength of 10 mm at different applied potentials for tuneable radiative cooling devices with different thicknesses of their top PEDOT layer.

Figure 3 .
Figure 3. Demonstration of temperature regulation at ambient pressure and temperature via electrically tuneable radiative cooling (A) Schematic illustration of the sky simulator, with thermal radiation from a device indicated as orange arrows.Liquid nitrogen is abbreviated as Liq.N2 and LN.(B) Tuneable radiative cooling measurement for switching the applied voltage between two fixed levels (G1.5 V, see dashed black line).The top panel shows the evolution of the device temperature (black, T device ) along with the ambient reference temperature (green, T reference ).Raw data and locally weighted smoothed data (LOESS, points of window 30) are shown as lighter and darker colors, respectively.The bottom panel presents the corresponding difference in temperature between the device and the ambient.(C) Same as in (B) but for the applied potential being varied between different values, as indicated by the dashed black line.(D)Collected results from several measurements like the one presented in (C), measured for different devices (blue and purple) and repeated for the same device (purple squares and triangles).Shown are also results for control measurements conducted by either not cooling ''the sky'' (red triangles) or by shading the cold sky (orange diamonds).The top panel shows results as the difference in temperature between the device and the ambient reference temperature in the chamber, and the lower panel presents DT, for which the value of T device À T reference was set to zero at À1.5 V.

Figure 4 .
Figure 4. Investigation of temperature switching speed via tuneable radiative cooling (A) Temporal variation of device temperature (top gray/black in top panel) and degree of radiative cooling (bottom panel), respectively, as a function of the speed of switching of the electrochemical bias potential (indicated by the dashed line in the top panel).The applied potential interval was varied stepwise from 300, down to 90, and finally up to 240 s. (B) Temperature switching magnitude as a function of switching time (duration between switching steps), extracted from (A) as the average results from for multiple switches for each time interval together with the standard deviation between those multiple switches.The dashed line is a guide to the eye.

Figure 5 .
Figure 5. Memory effects for tuneable radiative cooling devices (A) Evolution in temperature and degree of radiative cooling for a device pre-set to its reduced (high cooling) state followed by removing the applied voltage and leaving it undisturbed.(B) Same as in (A) but instead for a device that was pre-set to its oxidized (low cooling) state before removing the applied electrical stimulus.In both (A) and (B), the measurement included several additional steps to measure base levels and ensure functioning devices (see voltage indications above the top panels).Times for refilling liquid nitrogen in the sky simulator are marked as LN refill.