1. Introduction

The optical spectrum, especially in the infrared (IR) region, dominates radiative thermal emission from an object since the total amount of thermal emissive power corresponds to that obtained by integrating the optical absorptivity (=emissivity by Kirchhoff’s law) multiplied by the Planck’s blackbody distribution. As a representative example, passive daytime radiative cooling utilizing the principle has been successfully demonstrated [1], where the optical reflection in the visible to near-IR regions and the optical emissivity in the atmospheric window (ranging from 8 to 13 µm) are simultaneously controlled. The development of passive daytime radiative cooling has pioneered the field of non-electric thermal management based on photonic thermal emission. Hence, optical control of the IR spectrum is essential for the thermal management of an object. Since then, for control of the IR spectrum, various types of photonic structures including metamaterials [2], metasurfaces [3], photonic crystals [4], nanoparticles [5,6], diffusive polymer structures [7], and hierarchical polymer structures [8], have been suggested and demonstrated.

While passive daytime radiative cooling has been successful, most of these photonic structures exhibit “static” IR emissivity once the structures are designed and fabricated. Since these IR emitters continue to emit heat into the space even in, e.g., the cold season and/or nighttime, static IR emitters may even increase the energy consumption if heating is required when implemented in a living space. Therefore, a self-adaptive photonic structure that can enable thermal emission when the environment temperature is above the critical temperature and disable thermal emission when below is desired. To demonstrate self-adaptive regulation of thermal emission, active control of the IR emissivity in the atmospheric window depending on temperature is necessary. Utilizing the metal-insulator transition of phase change materials, e.g., vanadium dioxide [9–13] (VO₂) and Germanium-Antimony-Tellurium [14,15] (GST), modulation of the IR spectrum as a function of temperature has been suggested. Phase change materials dramatically change their optical behavior from that of a dielectric in the low-temperature (LT) phase to metallic in the high-temperature (HT) phase. In particular, VO₂ has a volatile switching feature in that the metallic HT phase returns to the dielectric behavior in the LT phase without any extra power input, unlike GST. Therefore, an active IR emitter based on VO₂ is a promising candidate system for maintaining a comfortable temperature in a living space. However, experimental implementation of such an active IR emitter for self-adaptive radiative cooling has been challenging.

Recently, a few groups have experimentally demonstrated an active IR emitter for self-adaptive radiative cooling with spectral tunability in the atmospheric window [16,17]. Tang et al. demonstrated a temperature-adaptive radiative cooler based on a metamaterial coating on a flexible substrate. The emission peak of the structure in the HT phase lies in the atmospheric window and the emissivity dramatically drops in the LT phase. The critical temperature of VO₂ can also be adjusted to ∼ 22 °C by adequate W-doping of pure VO₂, meaning that this system maintains a comfortable temperature. Xu et al. reported a 1D photonic structure consisting of a thin VO₂/Hafnium dioxide (HfO₂)/Aluminum (Al) reflector, which shows strong peak emission in the atmospheric window due to optical interference depending on the thickness of the spacer (HfO₂ in this case). For active IR emissivity modulation, both of the reported devices require a micro-patterned VO₂ structure and/or high-quality VO₂ thin film deposition by a vacuum process. In other words, cost-effective formation of the active layer for self-adaptive radiative cooling has been challenging.

In this paper, we propose a simple 1D structure in which an active VO₂ layer is formed via spin-coating of a vanadium oxide (VO_x) nanoparticle dispersion for active control of the IR emissivity. The structure shows a clear switching effect for IR emissivity in the atmospheric window. This structure is a potential candidate for a low-cost self-adaptive radiative cooler suitable for large-scale implementation.

2. Concept and structure design

Figure 1 shows the concept behind our self-adaptive IR emitter. As an ideal self-adaptive IR radiative cooler, we consider a device that shows unity (zero) emissivity in the HT condition (LT condition). Such ideal spectral modulation triggered by the volatile phase change of VO₂ enables us to achieve homeostasis, where thermal emission occurs only in the HT condition.

 

Fig. 1. The concept of our self-adaptive IR radiative cooler. The red (blue) line shows the emissivity in high (low) temperature conditions. The orange shade expresses the spectral power density of the sunlight (right axis). The gray shade expresses the atmospheric transmissivity.

Download Full Size | PDF

On the other hand, for demonstrating radiative cooling under direct sunlight (“daytime” radiative cooling), the emissivity (absorptivity) in the wavelength regime from the visible to near-IR should be minimized simultaneously, as depicted in Fig. 1. To suppress the visible to near-IR absorption, combining an active IR emitter with a porous polyethylene cover [7,18] is effective. The porous polyethylene structures diffuse sunlight while transmitting the mid-infrared thermal radiation from an emitter that lies beneath the cover. Combining with such a porous polyethylene cover, the rough emissivity for the combined structure will be one calculated from multiplying the original emissivity for the structure and the transmittance for the cover. Assuming the usage of the polyethylene cover, we can focus on the design of the active IR emitter without the constraints of visible–near-IR absorption.

To design the active IR emitter fulfilling the requirement of emissivity modulation in the atmospheric window, we adopt a simple stacked layer structure consisting of VO₂/dielectric/Metal reflector similar to that described in Refs. [11,17]. The stacked 1D photonic structure, also known as a Salisbury absorber [18], enables tunability of the peak emissivity via adjusting the optical path length in the dielectric layer. Since there was a problem of crack formation described later when fabricating the stacked structure, we here theoretically screen the layer structure when changing the dielectric material with the optical constants of ideal VO₂, calculated from Ref. [19], as a starting point.

A schematic diagram of the designed stacked layer structure is depicted in Fig. 2(a). Since we aim to realize the VO₂ active layer via spin-coating of a nanoparticle dispersion as discussed later, the morphology is assumed to be connected nanoparticles. Although this assumption results in a lower refractive index/absorption coefficient than that of ideal bulk VO₂, the connected VO₂ layer basically maintains its function as a phase change material [20] as shown in Fig. 2(b).

 

Fig. 2. (a) Schematic diagram of the self-adaptive IR emitter consisting of VO₂ layer/dielectric layer/Al/Si-substrate. The active VO₂ layer is assumed to be composed of necked-VO₂ nanoparticles and air voids. (b) Refractive index and absorption coefficient in the LT phase (left panel) and HT phase (right panel). (c) VO₂ layer-thickness dependence of the simulated absorption spectra in VO₂/MgF₂/Al structures, in the HT phase. (d) Simulated absorption spectra in the stacked structure with various dielectric layers. From left to right, the emissivities for magnesium fluoride (MgF₂), zinc sulfide (ZnS), germanium (Ge), and HfO₂ are shown.

Download Full Size | PDF

Using the optical constants, we discuss how the type of dielectric layer affects the IR emissivity. The optical constants of other materials are also derived from Refs. [21,22]. Here, we assumed that the optical constants of the dielectric materials do not change with its temperature. Figure 2(c) shows the VO₂ layer-thickness dependence for the absorption spectra of VO₂/MgF₂/Al structures in the HT phase, where the MgF₂ thickness(∼2.25 µm) is adjusted to maximize the emissivity in the wavelength ranging from 8 to 13 µm. For the calculation, a 200 nm-thick Al layer is assumed for the back reflector. In terms of the emissivity in the wide wavelength regime, we fixed the VO₂ layer thickness to be 30 nm for the following theoretical part.

Figure 2(d) shows IR emissivities obtained with various dielectric layers. The thickness of the dielectric layer is optimized by the simplex method such that the emissivity difference between the HT phase and LT phase becomes maximum in the atmospheric window. The optimized thicknesses for MgF₂, ZnS, Ge, and HfO₂ are 2251 nm, 1175 nm, 621 nm, and 1453 nm, respectively.

All structures show near unity emissivity for the HT phase in the atmospheric window while maintaining low emissivity for the LT phase, indicating that the designed stacked layer structures fulfill the function of a self-adaptive IR emitter. Among the structures we designed, magnesium fluoride (MgF₂) is theoretically the best dielectric material because of its low refractive index (∼1.4) and low absorption coefficient (∼0.003) around 10 µm. However, the MgF₂ structure requires a large layer thickness to tune the peak emissivity around 10 µm. In addition, the deposition of high-quality MgF₂ requires high temperature (∼300 °C) film deposition [23], which often results in cracking when cooling the stacked film on the Al to room temperature.

The Ge structure shows a good on/off feature as well. However, the emissivity in the LT phase is relatively high compared to the MgF₂ and ZnS structures, indicating that the structure shows relatively high thermal emission even in the LT phase. Since Ge has a high refractive index (∼4) around 10 µm, the origin of loss comes from the large impedance mismatch between the Ge layer and other layers. The HfO₂ structure shows relatively high LT emissivity like that of the Ge structure, but the origin of the loss is different from the Ge case. HfO₂ has multiple vibrational oscillators in mid- to far-infrared regimes [22], resulting in a higher absorption coefficient than that of IR-transparent materials (MgF₂, ZnS, Ge, and so on.). Generally, metal-oxide dielectrics tend to have relatively high absorption coefficients due to vibrational oscillation in the mid- to far-infrared regimes [24]. According to the theoretical design and the experimental insight, we chose the VO₂/ZnS/Al structure. The methods to realize the structure and the experimental results are discussed below.

3. Methods

3.1 Synthesis of VO_x nanoparticles

We started with the preparation of VO_x nanoparticles for formation of the VO₂ active layer by a solution process. The VO_x nanoparticles were synthesized via a previously reported hot-injection method [20,25] with some modification. Briefly, 7.92 g of 1-octadecanol (Alfa Aesar) and 30 mL of oleylamine (Sigma Aldrich, technical grade) were degassed at 125 °C for 1 hour under vacuum. A total of 0.4 g of Vanadium (V) oxytrichloride (99%) was injected into the solution under N₂-flow conditions. After the injection, the N₂-flow was stopped, and the flask was heated up to 250 °C at 10 °C/min. After reaching 250 °C, the reaction temperature was held for 20 min. Then, the solution was cooled down to room temperature. The following purification process was carried out in a N₂-filled glovebox. Forty milliliters of toluene (super dehydrated) was added to the solution and the solution was vortexed until the product dissolved in the solvent. After an excess of methanol (super dehydrated) was added, the solution was centrifuged. The supernatant was discarded, and the resulting precipitate was redispersed in 16 mL of chloroform (super dehydrated).

3.2 Fabrication of the stacked layer structure

Two hundred nanometers of Al as a back reflector was deposited on a Si substrate at 50 °C by electron beam evaporation. A ZnS layer dielectric was deposited on the Al film at 130 °C at a rate of 3 Å/sec until the film reached the intended thickness, also by electron beam evaporation. The VO_x dispersion was spin-coated onto the ZnS/Al film at 1500 rpm for 30 sec. Then, for the transformation of amorphous VO_x to crystalline VO₂, the stacked film was annealed at 400 °C for 60 min under reduced pressure of 120-160 Pa. Although the standard annealing condition of 500 °C for 5 min was reported in Ref. [25], the lower temperature condition was effective for fabricating stacked layer structures without crack formation.

4. Results and discussion

Figure 3(a) shows the optical properties of the 1.2 µm thick-ZnS/metal reflector before depositing the VO₂ active layer. Here, we tested chromium (Cr) and molybdenum (Mo) as metal reflector materials as well. Periodic interference peaks depending on the optical path length of the ZnS were seen at roughly the same wavelength in all structures, while the peak intensities were different. Although Mo and Cr exhibited good adhesion to the ZnS layer and the Si substrate, the mid IR emissivity was found to increase in the IR regime mainly due to low reflectivity. In other words, we experimentally verified that the ZnS/Al structure showed both low loss and high reflectivity over a wide range of the IR region.

 

Fig. 3. (a) IR absorptivity spectra for ZnS/metal reflector/Si structure depicted in the inset. The absorptivity is calculated from 1 – the measured reflectivity. The blue (Mo), red (Al), and green (Cr) lines show the spectra for the stacked layer structures with different metal reflectors, respectively. (b) Temperature-dependent resistivity of solution-processed VO₂ films. The black (red) line shows the resistivity of the film fabricated from hexane (chloroform) solution. (c) XRD pattern for the VO₂ film fabricated from chloroform solution on a Si substrate. The blue and red vertical lines show the PDF data for monoclinic VO₂ crystals.

Download Full Size | PDF

Next, we discuss the electronic transport and crystal structure of the solution-processed VO₂ layer. Figure 3(b) shows the temperature-dependent resistivity of the connected VO₂ nanoparticle layer. We investigated the effect of the solvent used for the VO_x nanoparticles on the electronic transport. In the HT phase, VO₂ films prepared from hexane solution exhibited almost the same resistivity (conductivity σ ∼ 5000 S/m) and phase change temperature (∼70 °C) as those in Ref. [25], whereas VO₂ films prepared from chloroform solution exhibited one to two orders lower resistivity (σ ∼ 120000 S/m) and lower phase change temperature. Here, the phase change temperature of the VO₂ film fabricated from the chloroform solution was estimated to be ∼55 °C from the electrical measurement. The HT conductivity obtained was comparable to films prepared by vacuum deposition processes such as sputtering (σ ∼20000 to ∼60000 S/m [26]) and pulsed laser deposition (σ ∼500000 S/m [27]). This indicates that solution-processed VO₂ films prepared by optimized dispersion conditions exhibit nearly ideal metallic behavior in the HT phase and can be expected to have the high HT emissivity predicted by the design terms above. We also believe that the VO₂ layer prepared from chloroform solution have fewer transport gaps, such as air voids, and/or less residual organic content, which can contribute to the higher HT conductivity and lower phase change temperature. These results suggest that the phase change properties of the connected VO₂ nanoparticles can be tuned by selecting an appropriate dispersion process.

The XRD pattern for the VO₂ film fabricated from the chloroform solution on a Si substrate is shown in Fig. 3(c). Note that the prominent peak at 33 deg is due to the Si substrate. (011), (012), and (022) peaks were relatively significant, indicating that the VO₂ film shows preferential orientation. No peaks from other phases were observed, which means that the spin-coated VO₂ film is single-phase with monoclinic structure as reported in Ref. [25].

Next, we discuss the experimental results for the stacked layer structures consisting of VO₂/ZnS/Al designed above. Figure 4(a) shows a cross-sectional transmission electron microscopy (TEM) image for the stacked layer structure with nominally 1.2 µm thick ZnS, where the active layer is formed from the undiluted VO₂ nanoparticle dispersion. The average layer thickness was 63.9 nm as shown in the inset and no broken areas and/or pinholes were seen from the TEM analysis, meaning that the connected VO₂ nanoparticles cover the dielectric layer well. We also prepared a thin VO₂ structure as a comparison, where the active layer was formed from a VO₂ nanoparticle dispersion diluted by a factor of 2. As shown in Fig. 4(b), the average active layer thickness was 31.2 µm and no broken areas and/or pinholes were apparent as well. In both structures, the dielectric ZnS layer was confirmed by the TEM analysis to have a columnar crystal morphology, as reported in MOCVD-grown ZnS [28], indicating the presence of high crystallinity even in films deposited by electron beam evaporation.

 

Fig. 4. (a)(b) Cross-sectional TEM images of stacked layer structure with nominally 1.2 µm thick ZnS layer. Magnified views centered on the VO₂ layer are shown in the insets. (a) Structure fabricated via spin-coating of undiluted VO₂ nanoparticle dispersion. (b) Structure fabricated via spin-coating VO₂ nanoparticle dispersion diluted by factor of 2. (c) Temperature-dependent emissivity in the thick VO₂ structure with nominally 1.2 µm thick ZnS layer. (d) Temperature-dependent emissivity in the thin VO₂ structure with nominally 1.2 µm thick ZnS layer. (e) Temperature-dependent emissivity in the thick VO₂ structure with nominally 1.0 µm thick ZnS layer. The shade regions in (c), (d), and (e) express the transmissivity of the atmosphere. (f) Temperature-dependent peak emissivity of the thick VO₂ structure with 1.2 µm ZnS layer. The red (blue) line shows the emissivity during the heating-up (cooling-down) process.

Download Full Size | PDF

Figure 4(c) and 4(d) show the temperature-dependent IR absorptivities of the thick VO₂ structure and the thin VO₂ structure, respectively. The thick VO₂ structure shows low emissivity (peak emissivity is 41.1% in the atmospheric window) until the system temperature reaches ∼ 40 °C. Once the system is heated up above 40 °C, the emissivity starts to increase drastically with temperature. The peak emissivity in the atmospheric window reaches over 90% at 70 °C, with the saturation value of 92.8% at temperatures above 80 °C, which results in over 50% modulation of emissivity. On the other hand, the thin VO₂ structure does not show such a significant modulation effect even though the 30 nm thick VO₂ is sufficient for achieving modulation, as discussed in the above theoretical discussion. We inferred that the cause of the discrepancy lies in local imperfections in the electrical percolative conduction in particularly thin regions, since the measured VO₂ thickness was found to fluctuate from 18.8 nm to 42.5 nm in a slice prepared by focused ion milling. It is noted that the experimentally measured emissivity is the directional one and hemispherical emissivity measurement will be required for rigorous calculation of radiative power.

As shown in Fig. 4(c), for the thick VO₂ structure with 1.2 µm ZnS, the emissivity spectra have complicated shapes that are different from the spectra obtained with the ideal VO₂ in Fig. 2(d). We here surmise that the shape of the spectra comes from the superposition of the optical interference and specific optical properties for the spin-coated VO₂ film. In other words, the shape in the wavelength region from 8 to 10 µm corresponds to absorption mainly resulting from optical interference as discussed in Fig. 2(d), while the shape in the wavelength region from 10 to 16 µm corresponds to absorption mainly attributed to the specific optical constants of the spin-coated VO₂ film (the optical constants of the VO₂ are discussed later). To demonstrate spectra suitable for emissivity modulation in the atmospheric window, i.e., to improve the low HT emissivity in the wavelength range from 8 to 10 µm, we tuned the nominal ZnS layer thickness from 1.2 to 1.0 µm. The measured spectra for the structure combined with the thick VO₂ layer are shown in Fig. 4(e). The HT emissivity spectrum covers a wide range of the atmospheric window, where the optical interference peak covers the wavelength region ranging from 8 to 10 µm and the absorption due to the specific material properties covers the rest of the spectrum. This indicates that the spectrum of our stacked-layer structure can be tuned by simply changing the layer thicknesses.

Figure 4(f) shows the temperature-dependent peak emissivity for the thick VO₂ structure with 1.2 µm ZnS in both the heating-up and cooling-down processes. Although ∼15 °C of hysteresis exists, the emissivity returns to the initial value, indicating that the structure acts as a volatile active IR emitter, unlike the GST-based system. In other words, the demonstrated stacked layer structure has the capability of self-adaptive temperature control in the sense that the emissivity changes depending on its temperature.

As mentioned above, the demonstrated structures exhibited complicated spectra that deviate from the one predicted theoretically. At the end of the section, we try to elucidate the discrepancy between the theoretical and experimental results.

To gain experimental insight into the solution-processed VO₂ layer, the optical constants of the VO₂ layer were determined using spectroscopic ellipsometry in the wavelength range from 1.7 µm to 30 µm (IR-Vase Mark II, J. A. Woollam). The sample temperature was controlled using the equipped heating stage. Here, the spectral resolution is about 10 nm in the wavelength range from 1.7 to 10 µm, and the interval subsequently increases with the recorded wavelength, e. g., the resolution is about 200 nm at the 30 µm. The ellipsometric signals were measured with the incident angles of 50, 55, and 60°. For accurate estimation of the optical constants, we prepared reference samples consisting of Al (200 nm)/Si and nominally 1.0 µm thick ZnS/Al/Si-substrate. After step-by-step estimation of the optical constants of the Al and the ZnS layers, the optical constants for the top VO₂ layer in the measured structure were determined.

The obtained refractive index and absorption coefficient for the VO₂ film are shown in Fig. 5(a) and 5(b), respectively. We have confirmed that the optical constants of HT phase return to that of LT phase when cooling the stacked structure down to room temperature.

 

Fig. 5. (a) Refractive index and (b) extinction coefficient for the VO₂ film in the thick VO₂ structure. (c) Refractive index and (d) extinction coefficient for the ZnS film in 1.0 µm ZnS/Al/Si-substrate structure. The red (blue) line shows the optical constants at 353 K (293 K).

Download Full Size | PDF

A feature of the spectra in both the HT and LT phases is a dip ranging from 10 to 14 µm. We assumed Drude-like metallic behavior in the HT phase and low-loss dielectric behavior such that the optical constant stays roughly flat up to 15 µm in the LT phase, in the design section. Therefore, the discrepancy between the theoretical and experimental results is mainly attributed to the peculiar spectral shapes of the optical constants for the solution-processed VO₂ films. Experimental deviation from ideal VO₂ behavior is also reported in sputtered films [29,30], where vibrational oscillations around 10 µm due to V₂O₅ have been seen. On the other hand, no crystalline heteromorphic phase was apparently seen in the XRD pattern for our VO₂ films; thus, we infer that solution-processed VO₂ films contain a small amount of a heteromorphic phase, with V-O bonds having different oxygen valence from the ideal VO₂, in an amorphous form.

We have also confirmed the optical constants for ZnS under temperature control, as shown in Fig. 5(c) and 5(d). The ZnS optical constants showed little temperature dependence, as expected. The absorption coefficient for the deposited ZnS film was negligible up to ∼25 µm, indicating that ZnS can be an ideal dielectric spacer for an active radiative cooler.

Figure 6 shows simulated IR spectra based on these optical constants along with the experimental spectra. Here, we use the optical path length of 136 nm for the VO₂ film and 930 nm for the nominally 1.0 µm thick ZnS film determined by IR ellipsometry. In the LT phase, the simulated IR spectra reproduce the experimental spectra well, including the intensity. The simulated HT phase spectra reproduce the spectral shapes as well, especially including the fine structure ranging from 10 to 14 µm. In other words, this means that the complex optical constants in the solution-processed VO₂ film are unraveled, which enables us to predict more realistic emissive performance.

 

Fig. 6. Comparisons between measured and simulated spectra for (a) nominally 1.2 µm thick ZnS structure in LT phase, (b) nominally 1.0 µm thick ZnS structure in LT phase, (c) nominally 1.2 µm thick ZnS structure in HT phase, and (d) nominally 1.0 µm thick ZnS structure in HT phase.

Download Full Size | PDF

On the other hand, the experimentally measured emissivities are slightly larger than predicted in the HT cases. We believe that the intensity difference results from the spot size difference between FTIR for the emissivity measurement and IR ellipsometry for determining the optical constants. About 3 mmϕ of incident light is used for characterizing emissive performance in the stacked layer structure, while about 10 mmϕ of incident light is used for the IR ellipsometry. Since the device size is 10 mm x 10 mm, the ellipsometry measurement may unintentionally average the thickened areas resulting from the spin-coating process and/or amorphous areas in which the crystallization does not proceed as expected. In other words, the obtained n, k change between HT and LT phases in the ellipsometry may be underestimated, compared with that obtained in the emissivity measurement for the actual device with the spot size of ∼3mmϕ. Therefore, the modulation depth simulated from the optical constants determined by ellipsometry may be decreased compared to the actual value (the measured emissivity). This also implies that local imperfections exist in the solution-processed VO₂ film. Thus, to achieve ideal on/off switching as depicted in Fig. 1, further improvement of film quality and uniformity will be necessary via optimizing the nanoparticle synthesis, purification and redispersion, and annealing conditions.

5. Concluding remarks

We designed and demonstrated a simple 1D photonic structure consisting of VO₂/ZnS/Al, where the active VO₂ layer is formed via a solution process. Such a solution process does not require a costly vacuum process and/or lithography process, and have a possibility of large scale implementation by a large-scale coating process. Based on screening by theoretical simulation in the ideal stacked structures and optimizing the device fabrication process, we found the layer structure and the process that suppress the crack formation during the layer stacking via the solution-processed VO₂ formation. The VO₂ films formed via the optimized condition exhibited one to two order higher conductivity than that of films prepared by the conventional solution process [25], indicating high film quality comparable to one fabricated via vacuum deposition processes.

Based on the volatile phase change of the VO₂, over 50% modulation depth of peak emissivity has been achieved in the atmospheric window ranging from 8 to 13 µm. To elucidate the complicated emissivity change seen in the experimental demonstration, we also unraveled the optical constants of the solution-processed VO₂ films up to 30 µm, using IR ellipsometry under temperature control. The determined optical constants reproduce the measured IR emissivity well and enable us to predict precise and realistic emissive performance. Although the phase change temperature of pure VO₂ used in the study is ∼60 °C, the phase change temperature of nanocrystal-based VO₂ is controllable via adequate doping [25] as well as by using a sputtered VO₂ film [16,31,32], indicating the possibility for the active IR emitter to maintain its temperature in a desirable range. This can be a promising platform enabling the cost-effective and large-scale implementation of a self-adaptive radiative cooler.