1. Introduction

In solar thermal applications such as solar thermophotovoltaics, the aim is to convert solar energy into heat. The key component in these applications is the selective solar absorber (SSA). Ideally, the SSA would exhibit broadband and near-unity absorption deep into the near infrared (NIR) with an abrupt drop to nearly zero around which the solar intensity is surpassed by blackbody radiation. This allows the SSA to absorb most of the solar energy and suppress losses due to self-emission, thus enabling maximum solar-to-heat conversion efficiency at specific temperatures [1,2]. Moreover, because of the diffusive and unpolarized characteristics of sunlight, it is important for SSAs to operate efficiently under oblique angles of incidence and different polarizations. Recently, metamaterials (MMs) have attracted considerable attention in negative refractive index [3], cloaking [4], and perfect lenses [5] owing to their extraordinary electromagnetic performance absent from natural materials. As the geometric scaling effect upon interaction with electromagnetic waves opens a path for the application of MMs in different frequency ranges, from the microwave [6] and IR [7] to the visible [8], it also provides the possibility of realizing MM-based broadband selective absorbers for solar thermal applications. Previous studies have explored one-dimensional MM absorbers to a large extent, however, the poor directional properties have limited their use in solar applications [9]. In addition, there have been efforts to design some complicated three-dimensional (3D) structures that generate higher dimensions in tuning a broadband electromagnetic response [10,11]; however, to the best of our knowledge, the experimental demonstration has not yet been realized, possibly because of the difficulties in fabrication. As far as we know, SSAs combining the steep cut-off and large angle acceptance have rarely been experimentally demonstrated. In this paper, we focus on the design of an SSA comprising the use of tungsten and a two-dimensional (2D) metal-insulator-metal (MIM) configuration, which is easy to fabricate. Although tungsten exhibits an intrinsic absorption spectrum that has a clear contrast from the UV-vis to NIR, its bulk material is far from ideal as an SSA [10,12,13]. Fortunately, the use of MIM structures adds extra optical loss mechanisms through nano-engineered metasurfaces [14,15].

Here, the as-proposed SSA is comprised of three layers with the SiO₂ dielectric thin film sandwiched by the bottom tungsten thick film and top tungsten layer, where the top layer is patterned as a simple nanodisk array. The use of a simple surface structure can largely prevent fabrication variation resulting from the design of complex nanostructured surfaces, especially in the optical range [16–18]. Owing to the non-polarized characteristics of solar radiation, we use a simple nanodisk array in a square lattice [19,20], for which we will verify that the responses to different polarized light are identical under normal incidence. In addition, the large-angle characterization shows that the solar selectivity remains almost invariable up to 40°, which is advantageous in solar thermal applications. Furthermore, we provide an explanation for the energy-dissipation process from a solar-to-heat perspective, gaining insights into the absorption mechanisms, which have only been partially demonstrated thus far [13,21–23].

2. Structural design and fabrication details

The structure of the proposed solar absorber is illustrated with nine unit cells in Fig. 1(a). The diameter of the tungsten nanodisk and the unit-cell period are denoted by d and a, respectively. The thicknesses of the tungsten nanodisk and SiO₂ layer are represented by h and t, respectively. The absorber is illuminated by a transverse-electromagnetic plane wave with normal incidence. To acquire high absorption in the UV-vis and short NIR regions, we conducted four-parameter sweeps on d, a, h, and t, where they were optimized in sequence using a commercial finite-different time-domain (FDTD) software (FDTD Solutions, Lumerical). The absorption was calculated in a volume of 0.5 μm × 0.5 μm × 6.5 μm, where a plane wave source was set 3 μm above the absorber. We employed a mesh size of 3 nm. A frequency power monitor was set 0.2 μm below the plane wave source, yielding a net power passing through it. The bottom tungsten film was set to 240 nm to avoid any transmission. The optimized parameters were d = 300 nm, a = 500 nm, h = 100 nm, and t = 60 nm. To fabricate this absorber, a 240-nm tungsten film was first deposited using magnetic sputtering on a clean Si substrate, and a 60-nm SiO₂ film was then evaporated using an electron-beam evaporator on the tungsten film. The 100-nm top tungsten film was deposited using magnetic sputtering. A 220-nm ZEP520 E-beam resist was spin-coated onto the top and post-baked for 3 min at 180 °C. Electron beam lithography was employed to pattern the resist, which was thereafter used as an etching mask to shape the top tungsten nanodisk array using ICP etching. The sample was then placed in an RIE chamber under O₂ plasma for 1 min to remove the resist residue. A top-view scanning electron microscope (SEM) image of the fabricated absorber is presented in Fig. 1(b).

 

Fig. 1 (a) 3D structure of as-designed absorber in nine unit cells and (b) SEM top-view of the fabricated absorber.

Download Full Size | PDF

3. Results and analysis

The measurement of reflectance was carried out using a spectrophotometer (Lambda 950, PerkinElmer) with an integrating sphere (IS, with an oblique angle 8° away from the normal to avoid direct reflectance out through the beam entrance) and universal-reflectance-accessory (URA, for measuring specular-like surfaces) module. Here, the near-normal reflectance from 300 to 2500 nm was taken from the IS (for measurement under 600 nm) and URA (for measurement above 600 nm) measurements, respectively, as the reflectance of light with sub-period wavelengths has larger diffuse components, while that of light with longer wavelengths is more specular [22]. The absorptance spectrum below 600 nm was normalized with the spectrum taken from a specular reflectance standard, while that above 600 nm was normalized with the spectrum taken from a diffuse reflectance standard. Data from the reflectance measurement were used to calculate the absorptance using the relation A = 1 – R. Figure 2 presents the simulated and measured data from 0.3 to 2.5 μm, from which we see a good agreement over a wide span of wavelengths. The inset shows the calculated spectra of the MIM geometry and the flat tungsten slab from 0.3 to 8 μm. Notice that there are some absorptance dips that do not match with the simulation result under 400 nm. We attribute this to the surface roughness at the W/SiO₂ interface caused during the fabrication process because sub-wavelength light is more easily scattered by surface fluctuations than its longer counterpart.

 

Fig. 2 Measured and simulated absorptance from 0.3 to 2.5 μm. (inset) Calculated spectra of the MIM geometry and flat tungsten slab from 0.3 to 8 μm.

Download Full Size | PDF

The calculation of the absorption described above is an indirect method. However, any absorption of light power can be directly obtained by calculating the energy loss within a specific material, which is a better approximation to reality. Therefore, to validate the simulation results obtained above, we perform another calculation on energy loss using the relation q = 0.5ωIm(ε)|E|². From this equation, we observe that the energy loss is proportional to the angular frequency (ω), the imaginary part of the permittivity (Im(ε)), and the square of the electric field amplitude (|E|²). There is negligible energy loss in the SiO₂ layer owing to the close-to-zero imaginary part of the permittivity (ε) within the considered wavelength range. Two FDTD analysis groups were used to record the optical constants and field data of the top tungsten nanodisk and bottom tungsten film. The calculated energy loss within different layers and the previous absorption results under normal incidence are presented in Fig. 3(a). It is observed in the plot that there is a negligible difference between the two methods of calculation, which is a validation of the previous simulation results. For simplicity of calculation, all the subsequent results are based on the 2D power monitors except where indicated.

 

Fig. 3 Calculated absorptance from FDTD 2D power monitors (blue solid), the energy absorbed in the tungsten bottom layer (red area) and in the tungsten nanodisk array (orange area), and the reflected energy by the absorber.

Download Full Size | PDF

Until now, we have acquired the absorptance from two perspectives, where the consistency of both yields evidence of the conservation of energy within the simulation volume. By assuming the imaginary part of the permittivity of SiO₂ to be negligible, it is shown that the dissipation of electromagnetic energy is largely due to the ohmic loss in the top tungsten nanodisk array and bottom tungsten film. The total dissipated energy is decomposed into two parts, which are shown in Fig. 3 in different colors. Surprisingly, it is the top tungsten nanodisk array rather than the bottom tungsten film that comprises most of the energy dissipated in the absorber, although the top tungsten layer has a thickness less than half of that of the bottom tungsten film. Tungsten itself has a high absorption in the UV-vis and low emissivity in the NIR, indicating good selectivity. However, the absorption spectrum of tungsten still needs to be tailored to achieve a high energy conversion efficiency [10]. By using a nanostructured surface and dielectric insertion layer in our absorber, we enhance the absorptance in the UV-vis and short NIR without doubling the thickness of the structure.

The map of energy loss already indicates where the absorption arises from. However, from a microscopic perspective, we still need to determine the underlying mechanisms of how electromagnetic waves are dissipated within specific regions. In Fig. 4, we present the field intensity and energy loss in the x–z plane with the largest cross-section of the tungsten nanodisk at wavelengths of 503, 671, and 1569 nm, corresponding to the three peaks indicated by arrows in Fig. 3. According to the dispersion relation of surface plasmon polaritons (SPPs), the wavelength at which first-order SPPs appear has the same value as the period. Therefore, what appears in Figs. 4(a) and 4(d) at 503 nm is a typical characteristic of SPPs. The SPPs will propagate along the tungsten–air interface until most of the energy is lost in tungsten. As a result, as shown in Fig. 4(g), the energy loss is largely conserved in the central area of the tungsten nanodisk near the metal–air interface and the intensity is two orders of magnitude larger than that in the bottom tungsten film. Compared to the resonance at 500 nm, the magnetic field at a resonance of 671 nm shown in Fig. 4(b) is largely confined to the SiO₂ layer, especially near the bottom tungsten/SiO₂ interface between adjacent tungsten nanodisks. The electric fields induced by magnetic fields with a relatively strong intensity and smaller integrating volume of energy loss (q) near such regions contribute to an almost equal energy dissipation in the top and bottom tungsten layers at a resonance around 671 nm. Notice that the bright contour indicating an enhanced localized magnetic field between the tungsten nanodisk and the bottom tungsten film is weaker than that in the SiO₂ layer between the neighboring nanodisks. However, starting from a resonance of about 671 nm, the magnetic field intensity between the tungsten nanodisk and the bottom tungsten film strengthens, while that between the nanodisks weakens as the wavelength increases. At 1569 nm, where magnetic polaritons (MPs) appear, the magnetic field intensity reaches its maximum, which is two orders of magnitude higher than that at 671 nm, while the bright field contours between the nanodisks disappear, as shown in Fig. 4(c). When MPs appear, the magnetic components are parallel to the dielectric layer. According to Lenz’s law, strong electric fields will surround the center of the magnetic dipoles. Although the dielectric plays no role in the absorption, the current flux, which is driven by the strong electric fields in the tungsten layers, creates extra loss mechanisms for the SSA. The characteristics of the field distribution at 1569 nm clearly indicate the excitation of MPs, of which the magnetic field is mostly confined in the dielectric layer, resulting in a circulating current around the magnetic contours. Consequently, the conduction current will cause energy loss in the metal layers, as can be seen in Fig. 4(i). Moreover, the absorption band is further expanded to the NIR because of the excitation of MPs.

 

Fig. 4 Field intensity and energy loss at 503, 671, and 1569 nm in the x–z plane.

Download Full Size | PDF

Furthermore, we studied the influence of the polarization angle on the absorptance, and the result is presented in Fig. 5(a). Regarding the symmetry of the structure, the simulation was only conducted with a polarization angle up to 45° with a 5° step under normal incidence. The result indicates that the as-designed absorber is insensitive to the polarization angle under normal incidence. Moreover, this criterion of polarization independence is very crucial for absorber applications in devices where unpolarized waves are applied. Thus, we believe our absorber would have potential applications in solar devices.

 

Fig. 5 (a) Simulated absorptance of different polarization angles up to 45° under normal incidence, (b)(c) simulated absorptance of different angles of incidence under TE and TM polarization, respectively, and (d) measured absorptance under 8, 20, 40, and 60°.

Download Full Size | PDF

For a solar absorber, one would expect the absorptive behavior under oblique angles of incidence to be omnidirectional, i.e., the absorber should be capable of absorbing light even for large angles of incidence. Therefore, we estimate the absorptance under different angles of incidence up to 65° with a 5° step under both transverse electric (TE) and transverse magnetic (TM) polarizations in our study, as shown in Figs. 5(b) and 5(c), respectively. In these plots, we see several bright bands demonstrating high absorption while they shift slightly when the angles of incidence increase because of their dispersive characteristics. Detailed explanations of such behaviors have been reported in several previous papers [21,23,24]. A broadband absorption can be realized with large angles of incidence under both polarizations. The large-angle acceptance can be further confirmed by measurements with unpolarized light, which is shown in Fig. 5(d). The results show that the high broadband absorptance is maintained up to an angle of incidence of 40°. In addition, the solar selectivity remains invariable up to 40°, demonstrating the promise of the SSA as a key element in solar thermal applications.

4. Conclusions

In this study, we designed, fabricated, and characterized a tungsten-based broadband SSA. The average measured absorptance is more than 90% from 0.5 to 1.75 μm, which matches well with the numerical calculation. The predicted absorptance beyond 2.5 μm is less than 12.6%. We’ve shown that the SSA is polarization-independent and that the broadband solar selectivity remains invariable up to 40°. Although the fabrication has led to some undesirable variance, given more technical rigor, we have the confidence to show excellent agreement between experiment and simulation results. The demonstration and the overall performance of the SSA will facilitate the development of solar thermal applications with more innovative solar devices.