VO₂ metasurface smart thermal emitter with high visual transparency for passive radiative cooling regulation in space and terrestrial applications

3.1 Atomic layer deposition of the VO₂ thermochromic layer and unpatterned thin-film results

For the proposed smart VO₂ metasurfaces, the thermochromic property of VO₂ is the most critical part. A process based on the atomic layer deposition (ALD) technique was developed and optimized to achieve desirable optical properties, accurate thickness control and large-substrate uniformity (see Supplementary Materials S1 and S5). The vanadium precursor, [V(NEtMe)₄ (TEMAV), from Strem Chemicals], was heated up to 85 °C to achieve a sufficient vapor pressure. Recently, VO₂ processes based on TEMAV were demonstrated by a number of research groups using as oxygen precursor ozone [46], [47], [48], [49] or water [50], [51], [52]. In this work, water was chosen as oxygen precursor as the more promising route to stoichiometric VO₂. Figure 4a provides a summary of optimized key process parameters and growth rate and uniformity. The deposition temperature was set as 150 °C as the precursor risks thermal decomposition above 200 °C. The growth rate is extracted to be 0.37 Å/cycle and 6-inch substrate uniformity is below 2%. The growth rate is lower than some reported ozone process of ∼0.8 Å/cycle [47, 49, 53]. However, there is also some inconsistency in reported growth rates [50, 54]. Therefore, the excellent uniformity of below 2% over 6-inch substrate indicates that the developed process is a self-limiting ALD growth (as the ALD tool, unlike CVD tools, is not hardware optimized for uniformity).

Figure 4:

VO₂ by atomic layer deposition process and the characterizations after different post-anneal treatments. (a) X-ray photoelectron spectroscopy (XPS) result showing key energies corresponding to the VO₂ phase. (b) Raman spectra of VO₂ with 1 h anneal at temperatures of 425 °C, 450 °C, 480 °C and 500 °C. (c and d) FTIR absorption spectra of VO₂/SiO₂/Al stack measured at 20 °C (dashed lines, c) and 90 °C (solid lines, d) for the annealed samples of (b).

A 40 nm VO₂ film as deposited was investigated by X-ray photoelectron spectroscopy (XPS), the binding spectrum is shown in Figure 4a. No argon etch was performed as argon etch was known to preferentially etch oxygen atoms, which then affects the vanadium oxide stoichiometry with extra oxidation states [55]. The binding energies values were charge corrected to the O1s peak at 530 eV due to O–V bonds [56]. The V2p energy level splits into the V2p _1/2 and V2p _3/2 components due to orbital splitting at 523.8 eV and 516.4, respectively, and these values are consistent with those reported for VO₂ [55, 57]. Therefore, XPS analysis shows as-deposited film is mainly VO₂ with V present as V⁴⁺ and similar result was also seen in the work by Peter et al. [51]

As the VO₂ film is amorphous as deposited, a post-anneal treatment is required to crystalize the VO₂ into monoclinic VO₂ (M1) for thermochromic properties. Based on literature knowledge [54], the anneal condition is a trade-off between temperature, oxygen pressure and anneal time. For the anneal studies, VO₂ was deposited on an VO₂/SiO₂/Al OSR stack on 6-inch Si substrate to generate a sufficiently large number of identical samples for anneal test at various conditions. Also the anneal conditions are more easily compared for a high IR emissivity contrast. We tested the performance of VO₂ films annealed in O₂ pressure of 40 mTorr for 1 h, for different temperatures of 425 °C, 450 °C, 480 °C and 500 °C. Figure 4b shows Raman spectra of the annealed VO₂ films with a bulk V₂O₅ spectrum shown as a reference. Known Raman peaks for VO₂ (M1) were also marked at 144 cm⁻¹, 193 cm⁻¹, 223 cm⁻¹, 308 cm⁻¹, 389 cm⁻¹, 497 cm⁻¹ and 613 cm⁻¹ [57], [58], [59]. All four VO₂ films are seen to exhibit Raman peaks highly consistent with known VO₂(M1) peaks and none matches V₂O₅ peaks, indicating that all four anneal conditions achieve VO₂(M1) with little presence of V₂O₅.

IR absorption spectra are presented in Figure 4c and d for VO₂/SiO₂/Al film reflectors using Fourier transform infrared (FTIR) measured respectively at 20 °C (c, denoted as cold) and 90 °C (d, denoted as hot). In the cold state, VO₂ devices annealed at 450 °C, 480 °C and 500 °C shows only little absorption over a broad IR range except for a band of absorption peaks around 10 µm due to the vibrational modes of the underlying SiO₂ layer. In comparison, a significant higher IR absorption is seen for the VO₂ film annealed at 425 °C (purple line in Figure 4c). The low IR absorption in the cold state is attributed to the dielectric properties of VO₂ [26]. In the hot state, VO₂ devices annealed at 425 °C, 450 °C and 480 °C show a broad IR absorption whilst the material annealed at 500 °C gives little IR absorption. The high IR absorption is attributed to the metallic property of VO₂ in the hot state [26]. We see that annealing in the temperature range of 450 °C–480 °C result in a desirably high IR contrast. The huge difference in the IR optical properties observed for these different anneal conditions is of great interest as these films are all identified as VO₂ by Raman spectroscopy. Thus, the VO₂ anneal condition needs to be particularly well optimized for this application. After material optimization, the VO₂ using the promising anneal condition was successfully applied to stacks grown on CaF₂ substrates.

3.2 Fabrication and characterization of visually transparent VO₂ smart metasurface thermal emitters

To study the behavior of the transparent metasurface thermal emitters, we fabricated and evaluated a device stack on a 1 mm thick CaF₂ substrate. Figure 5a shows SEM images of the VO₂ metasurface with feature size L of 1.5 µm, 2.3 µm, 3.5 µm and 4.7 µm. For all three sizes, VO₂ features are well defined in square shape and the gaps are accurately defined as d = 2 µm. Optical spectra of their metasurfaces and film are presented in Figure 5d–i for the spectral range from 0.4 to 20 µm covering visible, NIR and IR, measured at 90 °C (hot state, d–f) and at 30 °C (cold state, g–i). The absorption (A) is obtained from the transmission (T) and reflection (R) using the relation A = 1 − R − T. In our experiments we use a FTIR microscope with numerical aperture of 0.58, corresponding to a range of angles from 0°–35°.

Figure 5:

SEM images and experimental optical spectra of VO₂ metasurfaces as a function of feature size. (a) SEM images for metasurfaces with gap size fixed at d = 2 µm and feature sizes L = 1.5 µm, 2.3 µm, 3.5 µm and 4.7 µm. (b–d) Transmission, reflection and absorption spectra at 30 °C and 90 °C, for the VO₂ metasurfaces with four different feature sizes (colored lines) as well as the unpatterned thin film device (black line).

The VO₂ metasurfaces provide up to 60% transmission in the visible range and almost zero transmission in the thermal IR range (>2.5 µm) in both the hot and cold states (Figure 5d and g). The CaF₂ substrate is completely transparent for wavelengths <10 µm and we find that the transmission of the bare metasurface stack in this range is in good agreement with the numerical simulations of Figure 2. Compared with the planar film device (black curve), metasurfaces offer significantly improved visible/NIR transmission which is related to the reduced VO₂ coverage in the metasurface geometry.

In reflection (Figure 5e and h), VO₂ metasurfaces result in a weakly oscillating reflection around 20% in the spectral range of 0.4–1.5 µm. In the mid-IR range between 2 and 8 µm wavelength, the cold state shows an overall high reflectivity which is associated with the AZO back-reflector. In the hot state, stronger oscillations are seen with an amplitude depending on the metasurface feature size, which is caused by an interplay of the Fabry–Perot fringes in the dielectric spacer and the strong absorption in the VO₂. At odd multiples of a ¼ wavelength, a minimum reflection is seen as the antinodes coincide with the VO₂ layer (Salisbury screen effect), while for even multiples nodes at the location of the VO₂ result in maxima in the reflection.

The absorption spectra (Figure 5f and i) show a relatively flat absorption in the visible and NIR range around 30–35%, where a lower visible absorption is found for the metasurface geometry than for a planar film. Strong IR absorption is seen in the hot state (Figure 5f) both for the planar film as well as for the larger feature metasurfaces, while the IR absorption is reduced for smaller feature sizes. The IR metasurface absorption is influenced both by plasmonic resonance effects and VO₂ area coverage and a balance of both is needed as clearly illustrated by the simulated parameter maps of Figure 3. In the cold state, the IR absorption between 10 and 20 µm (and associated dip in reflectivity) is related to vibrational (LO and TO) modes of the SiO₂ dielectric spacer.

3.3 Performance parameters of visually transparent VO₂ smart metasurface thermal emitters

The experiments of Section 3.2 show that the VO₂ metasurface achieves an improved visible/NIR transparency in both the cold and hot states and an IR emissivity contrast equivalent to the planar film. Similar to their counterparts with Al reflector, the IR absorption enhancement is attributed to the plasmonic effects of the VO₂ metasurface [26]. Table 1 summarizes the obtained values of IR emissivity in term of ɛ _hot, ɛ _cold and emissivity contrast Δɛ, as well as the solar absorption parameters α _hot and α _cold. For IR emissivity and solar absorption calculation, details are available in our previous works [13, 26]. Additionally, visual transmittance (T _{vis, hot} and T _{vis, cold}) in the hot and cold states in the human visual range (T _vis) is obtained from the spectra using the equation [60]:

where T ( λ ) is the measured transmittance and Φ lum is the standard luminous efficiency function for photopic vision in the wavelength range of 400–780 nm [61].

The thermal emissivity contrast Δɛ between hot and cold states is the critical performance metric for smart emitter devices and is given by the difference between hot and cold emissivity (ɛ _cold and ɛ _hot) at 90 °C and 30 °C, respectively. For ɛ _hot, the feature size is seen to have an impact due to the plasmonic resonance enhancement to the IR absorption. Larger feature sizes of 3.5 and 4.7 µm give similar ɛ _hot to the planar film whilst smaller feature size of 1.5 and 2.5 µm give smaller values of ɛ _hot. For ɛ _cold, metasurfaces are not very different from the planar film since the dielectric VO₂ gives no plasmonic enhancement and the cold emissivity is mostly associated with the vibrational modes of the dielectric SiO₂ spacer and the emissivity contribution of the AZO back-reflector. Overall, the emissivity contrast slightly increases with the feature size from 1.5 to 4.7 µm and peaks at Δɛ = 0.26 for a feature size of 4.7 µm. IR emissivity hysteresis is shown in Supplementary Materials Figure S3.

For solar absorption, all metasurfaces show significantly reduced absorption compared to the planar film in both cold and hot states. The metasurface with feature size of 3.5 µm gives the lowest value of α _cold = 0.33, which is 28% lower than the planar film (α _{film, cold} = 0.46). The hot-state solar absorption dips at α _hot = 0.28 also for the 3.5 µm feature size corresponding to a 35% improvement over the planar film (α _{film, hot} = 0.43). Therefore, VO₂ metasurfaces with optimized design provide equivalent IR emissivity with a significantly reduced solar absorptions than the equivalent thin-film stack. This reduction in solar absorption is highly advantageous both for its improved thermal performance (as quantified by the ratio ɛ/α in either state) as well as for the strongly increased visual transparency which overcomes the serious issue of the lack of VO₂ visual transparency even at the reduced thickness of 35 nm [45]. Direct comparison with conventional OSR stacks using a metal back-reflector show a further significant reduction in α owing to the difference in absorption in single-pass transmission through the VO₂ film compared to double-pass reflection in a conventional OSR design [6].

Finally, we evaluate the visual transparency performance of these metasurfaces. All metasurfaces in the hot state give significantly improved visual transmittance over the planar film and the metasurface with feature size of 3.5 µm gives the highest value of T _{vis, hot} = 62%, which is a more than double the value of the corresponding planar thin-film stack (T _{vis, hot} = 29%). The same trend is also seen in the cold state where visual transmittance peaks at 60% for 3.5 µm feature over the planar film which is 71% improvement over the planar film.

4.1 Transition from reflection to transparent OSR and the role of carrier density on the device performance

Compared with OSR designs based on VO₂ metasurface stacks using an Al back-reflector [26], the emissivity contrast of 0.26 is considerably lower. To investigate this, the effect of AZO carrier density has been studied through numerical simulations between 1 × 10²⁰ cm⁻³ and 1 × 10²² cm⁻³. The simulated spectra are presented in Supplementary Materials Figure S4.

Figure 6 shows the simulated performance of the VO₂ metasurface (L = 3.5 µm and d = 2 µm) with different AZO carrier densities together with that using an aluminium reflector presented as reference. Hot IR emissivity (ɛ _hot) is little affected and consistent with that of the Al reflector design (dashed line, red). Unlike ɛ _hot, ɛ _cold decreases with AZO carrier density when above 2 × 10²⁰ cm⁻³. The emissivity contrast, as a combined effect, increases with AZO carrier density towards that of the Al reflector. The carrier density of fabricated AZO in this work is estimated to be around 6 × 10²⁰ cm⁻³ and its emissivity contrast is consistent with the performance predicted by the simulation. The infrared emissivity tunability Δɛ is mainly affected by the emissivity of the backreflector itself which is higher for lower carrier density where the skin depth is large and electromagnetic fields are more strongly absorbed inside the conductor. With AZO carrier increasing, solar absorptions (Figure 6b) at hot and cold (α _hot and α _cold) increase and peak at carrier density of around 4 × 10²¹ cm⁻³. This increase is related  to the closing of the visual transparency window in Figure 6c as the plasmon frequency shifts from the infrared, across the visible to the ultraviolet range. For even higher carrier densities, solar absorption losses are reduced as the electromagnetic skin depth reduces penetration of fields inside the lossy metal, eventually reaching the limit of the Al metallic backreflector (dashed lines), where Al is a Drude metal with carrier density of 1.8 × 10²³ cm^–3.

Figure 6:

Effect of AZO carrier density on the VO₂ metasurface performance. (a) IR emissivity, (b) solar absorption and (c) visual transmittance.

Thus, lower AZO carrier density is preferred for lower solar absorption. The visual transmittance (Figure 6c) maintains a value at around 60% over a wide range of low carrier densities, but rapidly decreases when the AZO carrier density exceeds 1 × 10²¹ cm⁻³. This decrease is attributed to the metallic response at shorter wavelengths at higher carrier density. Therefore, the AZO carrier density needs to be chosen as trade-off between emissivity contrast, solar absorption and visual transmittance. Further improvement of Δε may be possible by replacing the SiO₂ with another low-emissivity material, such as CaF₂, MgF₂, or ZnS, to reduce the emissivity background of other parts of the stack.

Our experimental results demonstrate that VO₂ metasurfaces with a transparent backreflector with carrier density around 6 × 10²¹ cm⁻³ offer superior visual transmittance over planar films whilst maintaining an equivalent thermal emissivity contrast, making them potentially superior candidates in see-through applications, e.g., smart windows [39, 60]. We point out that the VO₂ in this work has a transition temperature around 68 °C, which is significantly above the desirable room temperature operation for many temperature regulation applications. However with introduction of dopants, such as tungsten, into VO₂, the transition temperature of the VO₂ can be reduced to room temperature [39, 62], [63], [64], [65].

4.2 Cooling power performance studies

Metasurface based technologies are still at a lower technology level compared to thin-film coatings which can be more easily scaled up. The small scale of current generation of metasurface device demonstrators prevents a direct validation through field studies. We emphasize that IR absorption measurements under the assumption of Kirchhoff’s law are an accepted standard in the characterization of thermal control coatings which is generally highly correlated with the thermal performance in existing literature. We therefore estimate the theoretically expected thermal performance on the basis of the experimentally derived parameters in order to evaluate the cooling power that can be theoretically achieved using our metasurface smart radiator devices (see also Supplementary Materials Section S5). We separately consider here applications in radiative cooling management both in terrestrial and space environments.

To quantify the cooling power for terrestrial applications, three contributing factors have to be considered. The (1) solar absorption P _sun, (2) the thermal radiation from the atmosphere P _a, and (3) the energy radiated from the device itself P _r. The largest heating comes from the absorption of the incident solar flux, which has a total power of 987 W m⁻² incident on the Earth’s surface. Here we use the value corresponds to the standard AM1.5 global standard to provide a meaningful comparison between different devices. More detailed evaluations based on specific geographic locations and daytime variation in solar irradiation should be made for specific applications [48].

Table 2 shows the simulated values for power per unit area in hot and cold states for varying feature sizes and for the planar film devices calculated using the standard radiative cooling equations [6, 66], [67], [68], [69]. The atmospheric effect is taken into account by considering the thermal radiation absorbed by the device from an environment at temperature of 300 K. The difference in P _a between cold and hot states is due to the difference in solar absorption of the VO₂ below and above the critical transition. The net power flux P _r − P _a − P _sun is presented as a standard daytime radiative cooling performance figure showing the total net cooling capability under daytime solar illumination, while for comparison the night-time radiative cooling performance P _r − P _a is also indicated.

Table 2:

Calculated terrestrial performance in radiation cooling and solar power through at room temperature for VO₂ in cold (blue, 30 °C), and hot (orange, 90 °C) states.

Calculations were also performed for space applications where radiation contributions by the atmosphere are negligible and the incident power from the sun is much greater, which can reach up to 1345 W m⁻² for continuous direct exposure, while in-orbit the total sun hours depend on the inclination of the plane of orbit known as the β-angle. Typical results for full direct exposure in the space environment are presented in Table 3, again for VO₂ in the cold and hot phases.

Table 3:

Calculated performance for radiation cooling in the space environment, at room temperature for VO₂ in cold (blue, 30 °C), and hot (orange, 90 °C) states.

The net power P _r − P _a − P _sun radiated by the devices in a terrestrial environment reaches a maximum for feature sizes of 3.5 µm for both the cold and hot states. The 3.5 µm feature size appears to be the optimized value for IR terrestrial cooling with a hot state cooling power of 296 W m⁻². The radiated power from the device itself, P _r, increases by a factor of three between the cold and hot states, of which a factor two results from the fourth-power scaling of total radiation power with temperature, while a factor 1.5 is attributed to the thermochromic emissivity change Δɛ. We see that the optimized metasurface reflector outperforms the equivalent thin-film device by 60% in daytime radiative cooling above the critical phase transition, while simultaneously providing 70% higher visual transmission as shown in Table 2. This performance enhancement is directly related to the reduced solar absorption for the metasurface structure.

In the cold phase, all devices under study provide negative radiative cooling under solar illumination. The VO₂ thermochromic transition therefore provides the ability to switch between heating and cooling of the material depending upon its temperature. We note that these calculations do not take into account any on-board energy dissipation of the spacecraft itself which will result in additional thermal loading of the device.

Similar results are obtained for the space environment, here the effects of changes in solar absorption are even more critical and result in larger variations in net cooling power during solar irradiation between different devices. The space environment benefits from the absence of an atmospheric feedback effect, thus facilitating more effective radiation cooling. In absence of direct illumination by sunlight, the values of P _r for space and P _r − P _a for terrestrial applications show only a weak benefit from the metasurface device, indicating that the purely radiative performance is more or less the same for structured and unstructured. Here the benefit of high visual transparency remains main benefit of the metasurface device geometry.

6.1 Material characterizations

VO₂ was characterized in a Thermo Scientific Theta Probe X-ray photoelectron spectroscopy (XPS) system in ultrahigh vacuum conditions (base pressure P ∼ 1 × 10⁻⁷ Pa). The XPS data was analyzed with Themo Avantage software. Crystallized VO₂ (through post-deposition anneal) was characterized in a Renishaw inVia laser Raman spectrometer using 633 nm laser with the exposure power less than 1 mW to avoid film overheat [59].

6.2 Optical characterizations

IR reflectance was measured over the range of 1.6–20 µm using a Fourier transform infrared microscopy (FTIR, Thermo-Nicolet Nexus 670, Continuum microscope) using a ×15 optical objective and MCT-A detector. The KBr beam splitter and IR source were used and the transmittance and reflectance were normalized with air and aluminum mirror, respectively. For the visible and near-IR spectra range (0.4–1.6 µm), reflection and transmission were measured using a separate home-made using a pair of Si and InGaAs spectrometers with a supercontinuum light source (Fianium SC400-1). Two dimensional scans of the surface were performed using motorized stages (Thorlabs) to obtain spectra for the individual arrays (120 × 120 µm).

6.3 Numerical modeling

The simulations of the metasurface thermal emitters were done using the finite difference time domain (FDTD) method implemented in Lumerical software. Spectra from 0.4–20 µm wavelength were obtained using a broadband short pulse source. The source was a plane wave incident normal to the surface. Symmetric and antisymmetric boundary conditions were used to reduce the computation volume. The refractive index and extinction coefficients of AZO and VO₂ as well as other materials including SiO₂ and CaF₂ are from tabulated references [71] and are presented in Supplementary Materials Figure S6.