Effects of silicon dioxide as the polar dielectric on the infrared absorption spectrum of the metal-insulator-metal metasurface

Metal-insulator-metal metasurfaces have been widely used as high-performance absorbers in almost all electromagnetic spectral ranges. Their absorption spectra can be engineered by making variations in the geometry of the unit cells and/or by embedding materials with specific optical constants. Including a polar dielectric in their structure is another approach for manipulating their absorption spectra. In this research, we have numerically and experimentally investigated the effect of using silicon dioxide (SiO 2) as a polar dielectric on the absorption spectrum of a metal-insulator-metal metasurface composed of a tri-layer of Ni-SiO 2 -Ni. Our results have shown the presence of absorption peaks in the mid-infrared which are attributed to the excitation of the optical phonons in the SiO 2 spacer layer. Particularly, the excitation of the Berreman mode in the SiO 2 spacer layer was observed and its effect on the total absorption spectrum is studied. The parametric effects of the top patterned Ni layer, the incident angle, and the polarization are also investigated. This study can provide engineering capabilities for the mid-infrared absorbers and reflection filters.


Introduction
Among different high-performance absorber designs, metal-insulator-metal metasurfaces (MIM-M) are of particular importance because they strongly interact with the incident electromagnetic field in the near-field region and can completely absorb the incident radiation. A typical MIM-M consists of a uniform metallic backplane layer, an intermediate uniform insulating layer, and an array of periodic thin metallic structures lithographically patterned atop [1][2][3]. Extensive studies have shown that MIM-Ms are excellent absorbers in the visible (VIS) and the near-infrared (NIR) with the ability to tune the absorption peaks spectrally and spatially [3]. The origin of high absorption in these structures is attributed to the excitation of a variety of electromagnetic resonances which, in turn, results in the strong confinement of the electromagnetic energy in the gap between the top metallic patterned layer and the metallic backplane layer [4]. The absorption wavelength and its bandwidth can be tuned by varying the dimensions of the unit cells as well as the constituent materials with specific optical constants [5][6][7]. Also, external stimuli such as heat [8] can cause considerable nonlinear optical absorption in these structures. MIM-Ms have inspired the realization of devices in the field of photonics and plasmonics including detectors [9,10], novel energy harvesting devices [11], ultrathin flat lenses [12], imaging [13], resonant enhancement of Raman scattering [14], and on-chip integration [1].
MIR is a spectral region of tremendous scientific and technological interest as the vibrational transitions of many important molecules exist in this spectral region making MIR photonic devices crucial for applications in chemical and biomolecular sensing and spectroscopy [10,15,16]. Recently, polar dielectrics such as silicon dioxide (SiO 2 ), silicon carbide (SiC), gallium phosphide (GaP), and gallium nitride (GaN) have attracted much attention as insulating materials platform which can manipulate light-matter interaction in the mid-infrared (MIR) region [16][17][18]. It is possible to excite optical phonon waves in polar dielectrics with their negative real and positive imaginary parts of permittivity in the MIR region. Under specific excitation conditions, the incident radiation can excite surface waves in polar dielectrics known as surface phonon polaritons (SPhPs). This is analogous to its counterpart in metals named the surface plasmon polaritons (SPPs). The excitation of the optical phonon waves in polar dielectrics has been extensively studied both theoretically [19,20] and experimentally [17,21,22]. However, the use of polar dielectrics in the MIM-Ms is not fully explored yet and new and rich physics could be explored by including polar dielectrics MIM-Ms [22,23].
In this paper, we numerically and experimentally studied Nickel (Ni)-based MIM-Ms encompassing SiO 2 as the polar dielectric (Ni-SiO 2 -Ni tri-layer structure). Two different optical phonon resonances known as the Berreman (BE) [23][24][25] and epsilon-near-zero (ENZ) [26][27][28][29] modes are excited, and they are utilized to engineer the absorption spectrum in the MIR region. We have also investigated the interaction between the plasmonic resonances (from the metallic structures) and the BE mode (from the polar dielectric SiO 2 spacer) and found that the BE mode suppresses the plasmonic resonances in a way that they can't be excited simultaneously. In order to fully exploring the engineering capabilities of our proposed MIM-M, the effects of the top Ni pattern dimensions, lattice constant, incident angle, and its polarization on the absorption spectra were also studied.

Structure and absorption simulation
A schematic illustration of the proposed MIM-M unit cell is shown in figure 1(a). The MIM-M consists of a stack of three layers: a bottom uniform Ni layer, a SiO 2 spacer layer, and the top Ni micro-disks. The geometrical parameters of the unit cell consist of the diameter of the Ni micro-disk (d) and its thickness (t p ), the thickness of the SiO 2 spacer layer (h) and the lattice constant of the structure (a), i.e., the periodicity. The thickness of the bottom Ni layer (t b = 100 nm) which is much larger than its skin depth (approximately 45 nm for Ni in the spectral range of 5-12 μm) prevents electromagnetic wave transmission. We chose to use Ni as a metal, as it shows a high melting point which is ideal for energy harvesting applications such as solar thermophotovoltaic (STPV ) systems [6]. Ni is also ferromagnetic; thus, it has the potential to couple plasmonic resonance with its magnetic properties [30].
The simulations were carried out in the frequency domain by utilizing the Wave Optics Module of the commercial software COMSOL Multiphysics ® . The periodicities of the unit cell in both x-and y-direction were set by the Floquet boundary conditions. Plane waves impinge upon the unit cell and the reflection and transmission are monitored. A perfectly matched layer (PML) absorbing boundary condition has been employed along the z-direction (the propagation direction).

Sample fabrication and experimental characterization
As previously described, our proposed MIM-M consists of three layers: a bottom Ni layer, a middle SiO 2 layer, and top periodic micro-structured Ni disks. The MIM-M samples have been fabricated by depositing a 100-nmthick Ni layer on top of a Silicon (Si) wafer using electron beam deposition (EBD). The SiO 2 layer was then deposited using plasma-enhanced chemical vapor deposition (PE-CVD). An S1813 resist mask is patterned for the deposition of the top micro-disk array by deep UV lithography and a photomask. The final 50-nm-thick Ni layer was deposited by the EBD followed by lift-off in acetone. A scanning electron microscope (SEM) image of one of the fabricated samples (a = 5 μm and d = 4 μm) is shown in figure 1(b). Since the incident beam size on the samples becomes elongated and large at oblique incidences, we fabricated all samples as large as 1 cm by 1 cm using deep UV lithography. This large size sample helped to focus the light exactly on the sample even at the incident angle θ i = 60°.
The characterization of the samples involves reflection measurements utilizing the Fourier transform infrared (FT-IR) spectroscopy. An external Globar radiation source (550 nm-15 μm) illuminated the samples at normal or oblique incidence. The reflected light from the samples was collected and steered into the FT-IR (Bruker VERTEX 70v). A wire grid polarizer (ThorLabs, KRS-5), which operates in the range of 2-30 μm with an extinction ratio of 300:1, has been mounted in front of the radiation source to distinguish between transverse magnetic (TM)and transverse electric (TE)-polarized modes. It also allows having a mix of TM-and TE-modes by rotating its wire grid axis. All the measurements were normalized to a reference gold mirror (ThorLabs, PF10-03-M01).  . Having an absorption band in the MIR spectral region for the SiO 2 means that the real part of its permittivity becomes zero (ε r = 0) and negative (ε r < 0) whereas the imaginary part becomes positive (ε i > 0) [32]. The crossing point between the positive ε r and the negative ε r where ε r = 0 is known as the ENZ mode. The ENZ mode which is a local electric field confinement occurs because the normal component of the electric field E normal is both enhanced and confined due to the continuity of εE normal at the interfaces [21]. Also, it can be noticed from figure 2(b) that in the wavelength range of 7.8 μm λ 10.9 μm the imaginary part of permittivity of SiO 2 has a positive value which implies an absorption in this range. This is the spectral region where the BE mode (which is a longitudinal optical phonon mode) is excited by the TM-polarized radiation. The point where the imaginary part of the permittivity has a peak is known as the epsilon-near-pole (ENP) [33]. It is worth mentioning the importance of highly absorbing dielectric films which can be prospectively used as ultrathin photodetectors and solar cells [34]. BE and ENZ are two modes of excitations with different physical mechanisms [21]. While the BE mode is a leaky mode which can propagate [25], the ENZ is a non-radiative mode which is confined close to the interface [21]. By depositing a thin layer of SiO 2 on metallic films these excitations become more pronounced [19,20,23].

Results and discussion
MIM-M is a structure used for studying and manipulating the gap-surface plasmon (GSP) modes [1,4]. By embedding a polar dielectric such as SiO 2 as a spacer in a MIM-M, it is possible to study the GSP and the BE modes together and probe the interaction between these two excitations. In fact, we have used a patterned structure (micro-disk on top of the SiO 2 as a polar dielectric) to explore the excitation of the BE mode alongside probing its interaction with the GSP mode.
For the sake of comparison, the simulation and experimental results for the reflectance spectra of a layer of SiO 2 on Ni layer without any patterning on top of the SiO 2 layer is shown in figures 2(c) and (d). The effects of the incident angle and polarization on the reflectance spectra of SiO 2 /Ni structure (the thicknesses of SiO 2 and Ni are fixed at 300 nm and 100 nm, respectively) are shown in these figures. Figure 2(c) shows a comparison between simulated and measured reflectance spectra for TM polarization for θ i = 45°and θ i = 60°. In addition to the reflectance dip at the λ ≈ 8 μm (BE mode), there is another dip centered at the λ ≈ 9.5 μm. The latter reflectance dip is the ENP point [see figure 2 Figure 2(d) depicts the reflectance spectra as the incident polarization changes from the TM-to the TE-mode. As the incident polarization varies towards the TE-mode, the amplitude of the reflectance dip associated with the BE mode reduces till there is no dip for the TE-mode. The polarizer rotation angle is also shown for each case.
For MIM-M samples, the thickness of the SiO 2 spacer layer has been fixed to h = 300 nm in all structures which is the same as for the unpatterned structure. Based on our simulations, this thickness causes the best impedance match between air and the MIM-M. The lattice constant was fixed at a = 5 μm and different top micro-disks' diameter (d 1 = 3 μm, d 2 = 3.5 μm, and d 3 = 4 μm) were fabricated. Measured and simulated reflectance spectra at normal incidence are shown in figure 3(a). From this figure, it is clear that as the diameter of the micro-disk increases, the GSP mode shifts to longer wavelengths. The reflectance dip varies from λ 1 ≈ 7.6 μm to λ 3 ≈ 7.8 μm as the top micro-disk diameter was varied from 3 μm to 4 μm, respectively (red-shift) [4]. Measured reflectance spectra for the sample with a = 5 μm and d = 3.0, 3.5, 4.0 μm at θ i = 45°and for TM-mode are illustrated in figure 3(b). The reflectance dips due to the BE mode at λ ≈ 8 μm is the dominant feature in the TM-polarized reflectance spectra for all samples. As the diameter of the micro-disk increases, the reflection dips associated with the GSP mode shifts to longer wavelengths and becomes closer to the BE mode. Figure 3(c) shows a comparison between simulated and measured reflectance spectra for one the samples (a = 5 μm and Permittivity of SiO 2 which shows the ENZ at 8.2 μm and the ENP at 9.5 μm. The absorption peak starts from 7.6 μm end ends at 10.6 μm (shaded area) [22]. (c) Simulated and measured reflectance spectra of SiO 2 /Ni structure at θ i = 45°and θ i = 60°. (d) Experimental reflectance measurement for SiO 2 /Ni structure as the incident light polarization changes from TM to TE (the direction of the arrow). The polarizer angle j for each case is also shown. d = 3.5 μm) for both TM-and TE-modes. Simulated and measured reflectance spectra for the sample with a = 5 μm and d = 3.5 at θ i = 60°for TM-mode are illustrated in figure 4(d). As it is observed, the amplitude of the reflectance dip due to the BE mode enhances more compared with θ i = 45°. From the experimental measurements, the amplitude of the reflectance dip due to the BE mode reaches its maximum at λ ≈ 8 μm which is 8% reflectance (92% absorptance). Figure 3(e) depicts the change of reflectance as the incident polarization changes from the TM-to the TE-mode for the aforementioned sample. Similar to the unpatterned structure, by changing the polarization from the TM-mode to the TE-mode, the amplitude of the reflectance dip associated with the BE mode reduces. By comparing the incident wavelength range (5 μm to 12 μm) for the samples with a = 5 μm, it turns out that the diffraction losses become significant at high oblique incident angles. One possible approach to minimize the influence of the diffraction losses is to implement the random structure. In this type of structure, the top metallic patches are the same size, but they are arranged in random positions on top of the dielectric spacer layer [35]. Another approach is to decrease the lattice constant in a way that it becomes quite shorter than the minimum incident light wavelength. Figures 4(a) and (b) shows the reflectance measurements and simulation results for the structure with a = 4 μm and d = 3.5 μm at θ i = 45°and θ i = 60°, respectively. Compared with the previous cases (a = 5 μm), the diffraction orders from the structure are less accentuated and the reflection dips become wider, particularly for θ i = 60°. The observed differences between simulated and experimental spectra occur mostly due to the sample fabrication process which leads to deviations from the original design. Additionally, comparing the TM-and TE-mode reflection spectra for the MIM-Ms shows that for TM-mode, the reflection dips associated with the GSP mode decrease and are also suppressed compared with the TE-mode. This occurs due to the emergence of the BE mode at the λ ≈ 8 μm [20][21][22]. In addition to the fabrication issues, another possible source of error can be associated with the wavelength dispersion of the permittivity used in our simulations which might slightly differ from those of the fabricated samples.
By comparing the results from the unpatterend (SiO 2 /Ni) and patterned (Ni/SiO 2 /Ni) structures, it is concluded that the BE mode is always excited even with having the periodic Ni pattern above the SiO 2 layer. On the other hand, the top Ni pattern causes the two distinct modes (BE and ENP) to merge into each other and makes the reflectance dip wider. This dip (37% absorptance at θ i = 60°for the TM-mode) is merged in the dip originated from the BE mode for the MIM-M reflectance spectra.
For any MIM-M, there are three different contributions to the total absorption: (1) the continuous metallic backplane layer, (2) the periodic metallic patterns, and (3) the dielectric layer [23]. Previous studies have only investigated the total absorption without any study of the contribution of the different layers. In order to calculate each individual absorption contribution, we computed the integrated absorbed power for each region of the MIM-M. Figures 5(a) and (b) illustrate the contribution of each layer to the total reflection dip for the sample with a = 5 μm and d = 3.5 μm at θ i = 60°for TM-and TE-modes, respectively. From these figures, it is clearly seen that for both TM-and TE-modes at the wavelength range where the imaginary part of the permittivity of the SiO 2 becomes positive the contribution from the GSP mode falls to zero. While the BE mode is excited in this range for the TM-mode and the absorption from the SiO 2 spacer layer becomes dominant, the TE-mode shows low absorption (less than 10%). Other geometries also exhibited very similar trends. This is a direct consequence of the fact that as the real part of the permittivity of the SiO 2 becomes negative [see figure 2(b)], any absorption due to plasmonic excitation falls to zero and only the absorption from the SiO 2 layer becomes dominant. The condition for having any plasmonic resonance is that the metallic layer has a negative real permittivity whereas the insulator layer in contact with it should have a positive permittivity [36]. Hence, the two aforementioned absorption modes cannot be excited at the same wavelength (there is no interaction between them) although by appropriate design they can be very close to each other [37,38]. For the TE-mode, any absorption related to the SiO 2 spacer layer is only associated to the ENZ mode. Figures 5(c) and (d) illustrates the TM-mode magnetic field intensity (H-field) distribution inside the MIM-M at the two main reflection dips for the structure with a = 5 μm and d = 3.5 μm at θ i = 60°. For λ ≈ 7.3 μm the magnetic field is mainly concentrated within the gap between the top Ni micro-disks and the bottom Ni layer. This is direct evidence that these reflection dips correspond to the GSP excitation, and the absorption occurs only due to the top Ni microdisks and the bottom Ni film [4]. However, at λ ≈ 8 μm, no magnetic field enhancement is observed which indicates that at this wavelength the GSP excitation does not happen and the absorption inside the SiO 2 spacer layer occurs merely due to the BE mode excitation.

Conclusions
In this research, the reflectance spectra of MIM-Ms encompassing a polar dielectric (SiO 2 ) have been investigated both numerically and experimentally in the MIR region where the optical phonon waves in the SiO 2 spacer are excited. The effects of varying the size of the top metallic layer, lattice constant, and the radiation incident angle and its polarization on the reflection spectra have been studied in detail. It has been demonstrated that all the three layers of the MIM-Ms contribute differently to the reflection spectra at normal and oblique incidences. However, in a wavelength range at which the real part of the permittivity of the SiO 2 spacer layer is negative, no plasmonic mode is excited whereas the BE mode comes into play, and this means that the plasmonic and the BE modes cannot be excited simultaneously. This research is important in the fields of thermophotovoltaics (TPV ) employing a MIM-M as the active element. By embedding a polar spacer dielectric and an appropriate architectural design, it is possible to enhance the light-matter interactions at high oblique incidences for MIM-M-based devices.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).