Reflective and transparent cellulose-based passive radiative coolers

Passive radiative cooling has the potential to reduce the world’s energy-consumption by complementing and replacing traditional active methods to control indoor temperature, such as using air-conditioning. In contrast to traditional methods, radiative cooling requires no external energy to function. Instead, it utilizes the temperature difference between the cold outer space and the earth to passively radiate heat through the atmosphere, which conveniently has a transparency window in the infrared (IR) region that matches thermal radiation at room temperature. The net cooling power density, P_net of a radiative cooler is given by (see details in Supporting Information):where P_rad is the thermal radiation power per area of the cooler, P_atm is the power per area absorbed by the cooler due to incident radiation from the atmosphere, P_nonrad accounts for power lost or gained due to conduction and convection, and P_solar corresponds to incident absorbed power per area due to solar irradiation (Bartoli et al. 1977; Nilsson and Niklasson 1995; Zhao et al. 2019). The last term highlights that efficient radiative cooling is more challenging during daytime than during night time due to solar-induced heating (Bartoli et al. 1977; Nilsson and Niklasson 1995). Daytime radiative coolers therefore need to strongly absorb infrared light (making them strong thermal emitters) while not absorbing light throughout the solar spectrum. Solar radiation must therefore be either transmitted or reflected by the cooling material. Cooling materials that transmit solar radiation are sought for enhancing the performance of solar absorbers (Zhu et al. 2015) and solar cells (Zhu et al. 2014), and have been suggested for cooling objects while preserving their color (Zhu et al. 2013a, b). For many other applications, it is instead critical that solar radiation is reflected by the cooling material to protect underlying objects from absorbing solar light and heating up. Significant effort has been put into creating such reflective radiative coolers, which could cool objects even to sub-ambient temperatures while being exposed to the sun (Rephaeli et al. 2013; Zhai et al. 2017; Li et al. 2019; Mandal et al. 2018; Raman et al. 2014; Chen et al. 2016; Zhou et al. 2019; Kou et al. 2017). These reflective radiative coolers prevent solar transmission using back reflectors (Zhai et al. 2017), non-absorbing visible scatterers (Zhu et al. 2015) or by micro-porous structures (Mandal et al. 2018; Xiang et al. 2020). They are promising as passive techniques to reduce the need for air-conditioning and refrigeration and thereby have potential to decrease energy consumption devoted to cooling (Goldstein et al. 2017). Hence, there is a need for both solar-transparent and solar-reflective radiative coolers, with ideal systems illustrated in Fig. 1a, b, respectively. For transparent coolers we focus on generic systems suitable for different applications, while the transparency window can be further optimized for specific applications, e.g., adopted to the working spectral range of a particular solar cell. Figure 1 shows the ideal reflectance (R), transmittance (T) and absorptance (A) in the UV, visible (VIS) to near IR (NIR) wavelengths and mid-IR (MIR) wavelengths, together with the 1.5 air mass (AM) solar irradiation spectrum (light grey) and the atmospheric IR transparent window (dark grey) from 7 to 14 μm. An ideal transparent radiative cooler (Fig. 1a) is fully transparent (i.e., no reflection or absorption) in the visible spectral range while an ideal reflective cooler instead reflects all visible light (i.e., no transmission or absorption). In the MIR thermal and atmospheric transparency region, both types of materials must possess high absorptance, which via Kirchoff’s law of radiation translates to high thermal emissivity and hence, efficient thermal radiation (Greffet et al. 2018).

While the majority of radiative coolers have been based on inorganic materials, organic materials are emerging as promising alternatives. Not least, cellulose and its derivatives have shown promise by providing both high thermal emissivity and low visible absorption (Li et al. 2019; Xiang et al. 2020; Gamage et al. 2020). Cellulose is an abundant forest-based material with attractive sustainable properties such as biodegradability, biocompatibility and nontoxicity. This makes cellulose-based materials suitable also for passive radiative cooling in human body applications (Wei et al. 2020; Zhang et al. 2019; Peng et al. 2018; Xiao et al. 2019). Furthermore, the visible properties of cellulose materials can be manipulated by additives or via changes in the structural arrangement at the nanoscale and microscale (Vasileva et al. 2017; Koivurova et al. 2018). For example, highly transparent or translucent materials have been reported based on nanofibrillated cellulose (NFC) (Zhu et al. 2013a, b; Fang et al. 2014; Wu et al. 2015; Mahpeykar et al. 2017; Gamage et al. 2020) as well as by delignified and polymer-impregnated wood (Li et al. 2016, 2017a, b, 2018). Translucency can be controlled using particle scatterers (Gamage et al. 2020) or by introducing microstructure, which also enables materials to become reflective instead of transparent in the visible range (Xiang et al. 2020). Self-assembly of cellulose nanocrystals into ordered helical structures further allows for materials with structural coloration and preferential reflection of left-handed circularly polarized light (Hewson et al. 2017; Fernandes et al. 2019).

Our previous work (Gamage et al. 2020) showed that transparent NFC films can be used for sub-ambient daytime radiative cooling. We also embedded silica microparticles as resonant IR absorbers and visible scatterers to control translucency without compromising cooling performance. Such transparent systems are suitable for coatings on solar cells and other optoelectronic devices, but not to cool objects that need protection from solar radiation and corresponding solar-induced heating. Here, we demonstrate the possibility of creating both reflective and transparent radiative coolers based exclusively on cellulose, without any active additives or back reflectors. The prime difference between the two coolers is the structure of the cellulose at the nano- and microscales. The transparent cooler is homogeneous while the reflective cooler is made porous using an electrospinning process. We characterize the two types of materials in detail using integrating sphere measurements in the UV–VIS-NIR and MIR ranges and compare them in terms of daytime radiative cooling when used as coatings on solar-absorptive silicon substrates.

Figure 2a and b show the processes and SEM images of cellulose films made by electrospinning and casting, respectively. The two fabrication methods result in highly different structure at the microscale. The electrospun cellulose film is highly porous, composed of a scaffold network of cellulose fibers with micro- and nanoscale dimensions. By contrast, the casted NFC film is highly homogenous. The microporous structure of the electrospun film leads to broadband scattering in the visible region, which makes the film white, in stark contrast to the transparent casted NFC film (see inset photographs in Fig. 2a and b). Successful conversion of the cellulose acetate to cellulose was verified via changes in the vibrational peaks of FTIR spectra before and after conversion (Huang et al. 2015), as shown in Fig. S1. Examining SEM images for converted cellulose samples indicates a unimodal fiber diameter distribution with a peak around 175 nm (Fig. S2).

Detailed studies of the reflectance, transmittance and absorptance of the two different cellulose films confirm our initial indications based on visual appearance. As shown in Fig. 3a (blue line), the porous cellulose material system is exceptionally reflective in the visible, with average 94% reflectance in the range from 0.4 to 0.7 μm for a 275 μm thick sample. The reflectance decreases in the UV but remains above 75% for wavelengths down to 0.25 μm. Likewise, the reflectance slowly decreases but remains fairly high at longer wavelengths all the way to 2.5 μm. The film shows less than 10% transmittance in the whole UV–Vis-NIR region (0.25–2.5 μm, Fig. 3b), thereby efficiently preventing light absorption by underlying objects when used as a coating. The broadband high reflection is due to scattering by the wide distribution of nano- and microstructures in the porous material. By contrast, a casted homogenous cellulose film (green line in Fig. 3b, 65 μm thick) provides high transmittance and less than 10% reflectance (Fig. 3a) in the whole UV–Vis-NIR region (0.25–2.5 μm). Based on this marked difference, we now denote the electrospun porous cellulose films as reflective coolers and the homogenous casted films as transparent coolers. Importantly, both types of coolers show very low absorptance from 0.3 to 1.3 μm (Fig. 3c), which is a promising feature to avoid solar heating of the cooling materials. The absorptance increases towards the UV region and at longer wavelengths in the NIR, but in terms of solar heating we note that solar irradiation is also lower in these regions (indicated by the light-grey shaded spectrum). The increase in absorptance at longer wavelengths is more pronounced for the reflective cooler, which may be due to the high microporosity and differences in film thickness.

While both types of materials show low absorptance in the visible, they instead show high absorptance in the mid-IR region (Fig. 3c). As discussed above, these combined absorptance characteristics makes the materials suitable for daytime passive radiative cooling. From 7 to 10 μm, both coolers are almost opaque. The transmittance of the reflective cooler further remains very low at longer MIR wavelengths while the transparent cooler shows a broad peak at around 12 μm with about 20% transmittance. Comparison between additional samples reveal that this difference in MIR transmittance between the samples is related to differences in film thickness and porosity, as investigated next.

Figure 3d–f summarizes the visible and MIR properties for three reflective cellulose coolers of different thicknesses (61, 137 and 275 μm) and for the 65 μm thick transparent cellulose cooler. Since the visible reflection, transmission and absorption were relatively flat, we present the data at a representative wavelength of 550 nm (Fig. 3d). All three reflective coolers show very low absorption (< 3%) in the visible region. In terms of reflection, even the thinnest reflective cooler provides values higher than 80%. This value increases with thickness, reaching above 93% for the thickest reflective cooler. The gradual increase in reflection with thickness can be explained by additional scattering sites for thicker films and corresponding increased probability for light to back scatter instead of escaping the film in the forward direction. Further increasing the electrospun thickness while maintaining film uniformity is challenging because the high voltage applied between needle and collector becomes less effective when more material is deposited on the collector. Figure 3e presents the equivalent summary of the MIR properties of the same samples. We first use 10 μm as representative wavelength at which the thermal emission is peaking at room temperature. All four coolers are essentially opaque at this wavelength, with around 90% absorption and around 10% reflection. Since the MIR atmospheric transparent window spans from about 8 to 13 μm and thermal radiation is spectrally broad, we also present the average MIR properties for the four coolers in this wavelength range (Fig. 3f). This analysis reveals larger variations between the samples. We first note that the thinnest reflective cooler shows lower absorptance (i.e., lower emissivity) and higher transmittance in the MIR compared with the transparent cooler despite very similar thicknesses, which can be explained by lower density and less material in the beam path for the porous material. We further observe that increasing the thickness of the reflective cellulose cooler helps to improve the MIR absorptance and thermal emissivity to around 90%. The merely 65 μm thick transparent cooler also shows high (> 85%) averaged MIR absorption in the atmospheric window, verifying that both types of materials can provide high thermal emissivity at room temperature.

From the data presented in Fig. 3, we conclude that both types of cellulose materials possess very low absorptance in the visible range while being highly absorptive in the MIR region and thereby efficient thermal emitters. These results suggest that both types of materials are suitable for daytime radiative cooling. We performed real-time outdoor radiative cooling measurements during daytime with reflective and transparent coolers. The sky was clear with no clouds, with effective solar irradiance of around 0.8 sun during the measurement and relative humidity of around 27%. The coolers were placed in a cardboard box with bottom and side surfaces covered with reflective metalized mylar sheets (see Fig. 4a). Convective heat transfer was lowered using an IR-transparent polyethylene film as wind barrier. We placed foam blocks wrapped with metalized mylar sheets at the bottom of the box and silicon wafers on top of these blocks representing objects to be cooled (and protected from solar heating, see Fig. 4b). The optical properties of the silicon substrate are given in Fig. S3. The reflective and transparent cellulose radiative coolers were then used as coatings on these silicon wafers while monitoring their temperatures during exposure to the sun and the sky. The top panel of Fig. 4c shows the temperature variation during such measurements for silicon wafers coated with a reflective cellulose cooler (275 μm thick, blue), a transparent cellulose cooler (65 μm thick, green), a non-coated silicon substrate (dark grey) and the ambient temperature in the measurement box (light grey). The ambient temperature is presented as the average between two shaded thermocouples in the box. The temperature of the bare silicon wafer reached around 68 °C while the wafers coated with the cellulose coolers both maintained lower temperatures (around 63 °C and 53 °C for the transparent and reflective coolers, respectively). The elevated temperature in the box compared with the outdoor temperature (up to around 36 °C) is due to solar absorption and heat generation of components in the box, including the bare silicon substrate. The reduction in temperature by the cellulose coolers (ΔT = T_Si− T_cooler) is presented in the bottom panel of Fig. 4c. The transparent radiative cooler showed ΔT ≈ 5 °C, despite allowing most of the solar radiation to reach (and heat up) the wafer. The absolute temperatures varied from day to day depending on, for example, weather conditions, and we observed transparent cooler with temperature reduction even up to 15 °C (Fig. S4). Similar cooling performance of 14 °C relative to bare silicon has been reported for a cooler based on SiO₂ microspheres on glass (Jaramillo-Fernandez et al. 2019) and the temperature reductions are significant compared to the theoretically suggested value of 18.3 K for transparent coolers (Zhu et al. 2014). The reflective cooler reduced the temperature more than the transparent cooler and continuously managed to maintain temperatures several degrees Celsius below the ambient air temperature inside the box. The reason for the more efficient temperature reduction (around 15 °C for the measurement in Fig. 4) is that the reflective cellulose cooler also suppresses solar-induced heating of the underlying silicon wafer. Such sub-ambient daytime cooling highlights the dual ability of the reflective cooler to both prevent solar heating and also remove heat by radiative cooling. While absolute temperatures varied between measurements, the qualitative differences between the two types of coolers were consistent. To further understand the functions of the transparent and reflective coolers (both on silicon) we estimated their net cooling power density by calculating the different terms in Eq. 1 (see Supporting Information for details). Figure S7 present the results for different ambient temperatures, sample temperatures and nonradiative heat transfer conditions. The results predict the transparent cooler to provide negative net cooling power density at T_cooler = T_ambient for all investigated conditions, which explains why the transparent cooler only shows reduction in temperature from the bare silicon substrate reduction but not sub-ambient cooling. By contrast, the reflective cooler provides positive net cooling power density at T_cooler = T_ambient, corroborating its ability for subambient cooling thanks to additional suppression of solar absorption.

To avoid effects from solar heating, we also performed measurements during nighttime using the same coolers. As expected, these results present subambient cooling also for the transparent cooler (Fig. S5). Furthermore, we note that also the ambient temperature in the box decreased significantly (≈ 4 °C) when the setup was exposed to the sky. This interesting aspect may be due to the coolers also cooling the region where the ambient temperature was measured. In turn, this means that the cooling performance may be underestimated when the temperatures of the coolers are compared with the measured ambient temperature in the box. We finally present results for a daytime measurement without the IR transparent wind barrier (Fig. S6). Also here, both coolers could significantly reduce the temperature of the underlying silicon wafers compared with the non-coated silicon, with the reflective cooler being most effective. These measurements without the wind barrier also showed overall lower temperatures compared with corresponding measurements with wind barrier, as attributed to additional convective heat removal for the open system.

Our study shows that cellulose is suitable for passive radiative cooling and that different types of coolers can be made by varying the preparation method. We formed homogenous cellulose films by casting and porous cellulose fiber network films by electrospinning. Even though both systems are cellulose-based, they demonstrate opposite visible reflection/transmission behavior due their structural differences at the nano- and microscales. They both further show very low solar absorptance combined with high thermal emissivity in the MIR. The average MIR absorptance, and hence thermal emissivity, of both types of materials was greater than 74% for all investigated sample thicknesses and reached up to 90% for the thickest reflective cooler. Daytime radiative cooling measurements demonstrate that the reflective and transparent coolers could reduce the temperature of coated silicon wafers by around 15 °C and 5 °C, respectively. Moreover, the reflective cooler reached a few degrees Celsius even below the ambient temperature. These results suggest that transparent cellulose may be used as coating materials for optoelectronic and solar harvesting devices to reduce heat-induced performance drops, while reflective cellulose shows promise as coatings for cooling other underlying objects, such as water masses for building space cooling systems. To ensure durability and weather resistance of the coolers without compromising their optical characteristics, future work may also employ thin transparent lamination layers, for example, of polyethylene as is now used to cover the measurement box.