Wideband mid-infrared thermal emitter based on stacked nanocavity metasurfaces

For a broad series of applications from chemistry [1] and medicine [2] to atmospheric science [3], the efficiency and bandwidth of light sources will ultimately decide the performance of an entire mid-infrared (M-IR) system. For infrared sensing, it is necessary to capture infrared signals in the atmospheric window (i.e. 8–14 μm) [4, 5], where molecular vibrational fingerprints exist. Taking into account that light sources are the crucial part of numerous M-IR detection systems, so far, a few types of M-IR light sources have been developed. They can be classified into three types: light emitting diodes (LEDs) [6], lasers [7], and thermal emitters [8]. Although the laser can produce a very high output power, the working wavelengths are limited and are not always obtainable for the spectra of interest, not to mention the high cost of infrared lasers. LEDs are much cheaper than lasers, yet their working wavelengths are shorter than 5 μm. M-IR thermal emitters, which depend on heating coiled filaments composed by refractory metals to emit light, are the most popularly employed method to produce M-IR light due to the ultrawide operating spectrum and much cheaper price. Yet, the radiation energy from such an M-IR thermal emitter drops pronouncedly at long-wave M-IR spectrum owing to the low emissivity of the heating filament at this spectrum, making them inappropriate for long-wave M-IR sensing over 8–14 μm. Various nanophotonic thermal emitters have been proposed to operate across a wide spectrum from visible to millimeter [9, 10]. So far, wavelength-scale photonic structures such as silicon carbide (SiC) gratings [11], semiconductor photonic crystals [12], optical antennas [13], resonant cavities [14, 15], and plasmonic nanostructures [16] have been successfully used to engineer the absorptivity that, in turn, tailors thermal radiation following Kirchhoff's law. Of these, an approach using nanofabricated metasurfaces presents promise to modulate infrared thermal emission owing to their significant plasmonic resonance-enhanced Joule loss, directionality, and monochromaticity [17–19]. Such a thermal metasurface emitter (meta-emitter) is very thin because the metasurface can produce near-unity emissivity at the nanoscale, enabling the thermal emission to take place in a much smaller volume than the wavelength of emission in free-space. Since the plasmonic resonance of metasurfaces are associated with their geometries, the metasurfaces are limited to a narrow spectral region. However, a resonance near-unity emissivity over a broad spectral regime is highly desired for wide applications from thermal sensing [20] to infrared imaging [21].

A highly efficient thermal emitter is required to eliminate reflection and transmission across a wide spectrum. Therefore, one needs to consider three key components when designing a meta-emitter: robust light coupling for high absorption [3], large density of optical modes [22], and efficient antireflection [23]. Many works in this area concentrate on tapered or multi-resonance metasurfaces to broaden bandwidth [24, 25]. For example, a hyperbolic metasurface waveguide taper array was demonstrated to increase thermal emission in the short-wave M-IR region between 3 and 5 μm [26]. Nevertheless, practical device applications are limited by the complicated nanofabrication requirements of trapezoidal arrays. Another strategy is to unite various kinds of plasmonic resonators within a subwavelength meta-atom, which can achieve broadband or multiband response [16, 27]. However, the combination of resonators within a subwavelength pitch increases the complexity of the fabrication process and decreases the duty ratio and efficiency of emitter cells [28]. Such limitations may hinder the widespread applications of these meta-emitters. Thus, it is important to develop an efficient wideband thermal emitter that has a straightforward structure, is easy to fabricate, and works in the M-IR atmospheric transparency window.

Herein, we experimentally develop a high efficiency and wideband thermal meta-emitter for long-wave M-IR light production, which has emissivity of ∼80% ranging from λ = 8 to 14 μm. The meta-emitter consists of two pairs of circular–patterned dielectric silicon nitride (Si₃N₄)–gold (Au) stacks on top of a Si₃N₄/Au dual-layered structure. We attribute the broadband high emissivity to the combination of propagation surface plasmon resonance (PSPR) and multiple gap plasmon resonance (GPR). The PSPR is strongly related to the period of metasurface. The GPR is supported by the metal-dielectric-metal (MDM) cavity structures with different dielectric thicknesses. In our proposed stacked nanocavity meta-emitter, the Au cylinders and the Au bottom reflector act as mirrors, while the Si₃N₄ cylinders and Si₃N₄ dielectric layer represent the dielectric spacers. Note that the near-unit emissivity can be upheld over a broad range of observation angles (θ) from 0° to 60° and is independent of the polarization of light. In contrast to the current commercial M-IR thermal emitter based upon refractory metals, our proposed stacked nanocavity meta-emitter can pronouncedly boost thermal emissivity over the entire long-wavelength atmospheric transparency window of 8–14 μm. The rather simple meta-emitter can be fabricated via a magnetron sputtering system and standard optical lithography techniques, opening possibilities to obtain cost-effective, compact, and wafer-scale fabrication of an M-IR light source.

The broadband M-IR meta-emitter residing on a silicon (Si) substrate was composed by arrays of circular-patterned Au–Si₃N₄ stacked nanocavities on a Au–Si₃N₄ dual-layered film, shown in figure 1(a). The inset presents a single period of meta-atoms with designed geometrical parameters. The parameters were optimized to achieve a high emissivity in the spectral range of 8–14 μm. The resonator relies upon three stacked MDM cavities with different cavity lengths that can excite not only the PSPR mode but also three GPR modes in the M-IR region. A hybrid interference between the PSPR and multiple GPR modes leads to a high emissivity spanning the spectrum ranging from 8 to 14 μm. In figure 1(b), we present a top scanning electron microscope (SEM) image of the meta-emitter. We pattern the cylinder array using optical lithography (see 'fabrication section' for details). The schematic of the fabrication process is shown in supplementary figure S1 (available online at stacks.iop.org/IJEM/4/015402/mmedia). The thicknesses of each layer of two pairs of circular-shaped Si₃N₄–Au stacks, a Si₃N₄ spacer, and an Au reflector are t₁ = 100 nm, t₂ = 500 nm, t₃ = 300 nm, t₄ = 700 nm and t₅ = 200 nm, respectively (inset of figure 1(a)).

Zoom In Zoom Out Reset image size

Si₃N₄ is a crucial material employed in bolometer applications due to its advantageous mechanical and thermal features [29]. For this study, Si₃N₄ was selected as a dielectric film because its dielectric property possesses a large imaginary part over a broadband M-IR regime, indicating a high absorption coefficient. In supplementary figure S2, we experimentally presented the complex refractive index (N_SiN = n_SiN + i × k_SiN) of a 100 nm thick Si₃N₄ layer on a Si substrate. Variable angle spectroscopic ellipsometry variable angle spectroscopic ellipsometer (VASE) was used to measure the N_SiN. A Gaussian model was employed for the fitting. The imaginary part (k_SiN, red line) is above 1 across the M-IR region ranging from 9.6 to 13 μm.

Thermal radiation is the emission of a light wave by an item at a temperature greater than absolute zero. The spectrum and intensity of thermal emission depend on the frequency (ω). According to the Kirchhoff's law of thermal radiation, the emissivity (ε(ω)) of a blackbody is the same as its absorptivity (A(ω)) in thermodynamic equilibrium [12, 30]; hence, the A(ω) can be engineered to obtain the desired ε(ω). The structure is designed to possess an extremely low reflectance (R(ω)) in the spectral range from 8 to 14 μm, producing a near-unity A(ω). Thus, our proposed structure acts like a blackbody with the emissivity ε(ω) = A(ω). Kirchhoff's law can be fully validated since the area of our fabricated nanocavity metasurfaces (0.9 cm × 0.7 cm) is much larger than the M-IR wavelength. Since the metasurface is not transparent in the M-IRregion (i.e. the transmitted power is zero), the reflectance is directly measured via a FTIR spectrometer integrated with an infrared microscope (see 'experimental section') [31]. We can then derive the thermal emissivity via ε(ω) = 1 − R(ω). The reflectance measurement was performed under normal incidence. The FTIR measured ε(ω) (red line) are shown in figure 1(c) together with the numerically calculated one (black line). We experimentally obtain a high emissivity of over 0.8 across the spectral range of 9–13 μm that corresponds to thermal radiation. It is clear that a broadband near-perfect emissivity is achieved with the coupling of four independent emissivity peaks, with peak values of 0.71, 0.81, 0.96, and 0.84 at the wavelengths of 8.0, 9.2, 10.4 and 13.2 μm accordingly. These four resonance peaks are expected to account for this broadband high emissivity [32, 33]. Compared to commonly used refractory metal thermal materials (e.g. Ni, Cr, and W) that produce high emissivity in the visible and near-infrared regions and low emissivity at the M-IR region, our proposed meta-emitter can provide a near-unity emissivity over a broad M-IR range. As was seen, the ε(ω) of our device over 8–14 μm is as large as ∼0.8, which is ∼10 to ∼50 times larger than that of Cr and W thermal emitters, respectively. Such a feature makes our proposed structure a very efficient M-IR thermal light source for detecting. The physical mechanism of our meta-emitter depends on not only the PSPR mode of the stacked nanocavity array but the multiple GPR modes supported by stacked MDM nanocavities. The pitch of the stacked nanocavity array and the geometry of the stacked nanocavity resonator (i.e. the radius of the cylinder and the thickness of each layer) that significantly affect the resonances were optimized to allow for both impedance matching and high emissivity, and the cylinder resonator offers polarization independent excitation of resonances [34]. The simulated spectrum of ε(ω) has an excellent agreement with the measured one. As was seen in figure 1(c), we calculated the ε(ω) spectrum using the FDTD method solver. The geometrical parameters of the metasurface were set to those measured using the SEM pictures shown in figure 1(b). The refractive indices of all materials are obtained from experimentally measured data [35]. The details of the numerical model are shown in the 'simulation section'. We further use a thermal camera to study the ε(ω) of the meta-emitter. We can achieve ε(ω) of the meta-emitter by heating up the structure with a reference material having known ε(ω). A method was then developed to solve the ε(ω) of an unknown sample, that is, at thermal equilibrium, the temperature recorded by the thermal camera should be identical to both materials. To precisely measure the meta-emitter temperature, we subtracted the external contributions (e.g. the thermal emission produced by imaging optics and the ambient thermal emission reflected from meta-emitter) from the total thermal emission imposed on camera sensor. (see supplementary figure S3(a)). In the left column of figure 1(d), we present the thermal picture of the meta-emitter heated to 150 °C (the measured thermal image of the meta-emitter under room temperature, T_r = 20 °C, is shown in supplementary figure S3(b)) where the red region presents the meta-emitter. A reference temperature (T_ref) of 110 °C is given by the thermal image of the reference black electrical tape (see supplementary figure S3(c)). In the right column of figure 1(d), we calculated ε(ω) dependent temperature for both meta-emitter (T_D) and the Si₃N₄–Au dual-layered structure surrounding it (T_b). The meta-emitter's ε(ω) is larger than ε(ω) > 0.8 and ε(ω) in the unpatterned dual-layered structure is only ε(ω) ∼ 0.3. These values match the qualitative ε(ω) values obtained from FTIR measurement.

An ideal thermal emitter needs to maintain polarization independent high ε(ω) over a broad observation angle (θ) [36]. In figure 2, we experimentally (figure 2(a)) and numerically (figure 2(b)) present ε(ω) as a function of wavelength and θ in the stacked nanocavity metasurfaces on Si substrate, with the left and right columns showing the ε(ω) for p- and s-polarized incident lights, respectively. The observation angle dependent ε(ω) was characterized using an M-IR J. A. Woollam variable angle spectroscopic ellipsometer (VASE). The θ varied from 31° to 81° at a step of 1°. The meta-emitter possesses omnidirectional large ε(ω) for both p- and s-polarization states. As shown in figure 2(a), for the p-polarization state, the ε(ω) is larger than 0.85 for all θ and wavelengths (left column). For the s–polarization state, the ε(ω) decreases at the wider θ; yet, the ε(ω) larger than 0.8 is upheld for θ < 60° (right column). This is due to fact that the direction of the H-field of p-polarized light does not vary pronouncedly with θ and it can produce looped currents at all incident angles. On the other hand, for the s-polarization, the H-field less effectively drives the circulating currents at wide angles [37]. However, a good spectral overlap occurs between the ε(ω) for p- and s- polarization states. We conclude that the meta-emitter is not only polarization-insensitive but omnidirectional. As was seen in figure 2(b), the numerical simulation matches the performance of the fabricated device, and we attribute the difference in the locations of some resonances to radiation scattering from fabrication errors [38]. On the other hand, several factors that may cause the difference are not considered in the numerical model. These factors include the fabrication imperfections, such as the thickness of the native oxide layer and grain boundaries.

Zoom In Zoom Out Reset image size

In figure 3(a), we numerically present ε(ω) of the meta-emitter and Si₃N₄/Au dual-layered structure under normal illumination. The multilayered metasurface has a high emissivity over a broad M-IR region (8–14 μm) due to the hybrid mode interference of PSPR and GSPR [39]. Whilst the Si₃N₄/Au dual-layered structure exhibited a much smaller emissivity due to the absence of the plasmon resonance, its emissivity peak around λ = 13.2 μm is caused by an asymmetric Fabry–Pérot cavity composed of a 700 nm thick Si₃N₄ dielectric film sandwiched between a 200 nm thick Au reflector and air [40, 41]. It suggests that the design of the stacked nanocavity array on the Si₃N₄/Au film is an extremely efficient device for radiating thermal energy in the M-IR region. To explore the underlying mechanism of the broadband high emissivity, it is instructive to study the magnetic (H)-field distribution for the various resonance modes in the stacked nanocavity metasurface, along the cross-section plane, x–z presented in figure 1. Figures 3(b)–(e) presents the distributions of total magnetic fields ($H = \sqrt {H_x^{\,2} + H_y^{\,2} + H_z^{\,2}}$ ) and the displacement current (J_D) at the PSPR mode (λ₁ = 10.4 μm) [42] and three GPR modes (λ₂ = 8.0 μm, λ₃ = 9.2 μm, and λ₄ = 13.2 μm), respectively, where the color and arrows represent the magnitude of the H-field intensities and J_D, respectively. In the meta-emitter, the PSPR mode is associated with delocalized plasmon (or photonic mode coupling) and surface plasmon (SP) that strongly depends on the pitch of meta-emitter [43]. The PSPR can be produced when the pitch is on the order of the working wavelength [44]. As was seen in figure 3(b), the H-field at λ₁ = 10.4 μm can be strongly confined between two neighboring resonators, confirming that the emissivity stems from the metasurface induced PSPR. On the other hand, the three GPR induced emissivity peaks at λ₂ = 8.0 μm, λ₃ = 9.2 μm, and λ₄ = 13.2 μm are supported by the cavity between the Au reflectors as presented in similar MDM structures [45]. Herein, the GPR modes are strongly dependent upon the radius of the circular stacked nanocavity and the thickness of each layer. For the GPR modes, the dipolar oscillation of the free electrons in the upper Au cylinders causes an anti-parallel dipolar oscillation in the bottom film, which forms J_D loops to effectively produce magnetic dipolar moments for the MDM nanocavity structures [46]. Such magnetic moments can strongly interact with the incident light beam, and finally, the magnetic resonance can be produced in the dielectric space layers [47]. As observed in figures 3(c)–(e), the H-fields at the three GPR wavelength of λ₂ = 8.0 μm, λ₃ = 9.2 μm, and λ₄ = 13.2 μm are concentrated in three Si₃N₄ dielectric layers with different thicknesses of t₂ = 500 nm, t₃ = 300 nm, and t₄ = 700 nm. The GPR at λ₂ = 8.0 nm occurs in the first Si₃N₄ dielectric layer (figure 3(c)), where the H-field can be efficiently confined in the top Si₃N₄ layer, whilst the confinements of H-fields are relatively weak in the middle and bottom Si₃N₄ layers. The GPR at λ₃ = 9.2 nm mainly occurs in the middle Si₃N₄ layer (figure 3(d)), while the H-field is intensely confined in the bottom Si₃N₄ layer at λ₄ = 13.2 μm (figure 3(e)). This is owing to the hybridized mode of the magnetic resonance exciting in the individual Si₃N₄ layers [48]. In our stacked nanocavity metasurface, both the top and sandwiched Au layers are very thin, which enables the light to propagate through them. Thus, the magnetic resonance moments in the three Si₃N₄ layers can interfere strongly with each other. Namely, the three GPR modes exciting in the stacked nanocavity metasurface are related to the hybrid magnetic resonances coupled in the three Si₃N₄ layers. Thereby, the broadband high emissivity in the meta-emitter owes to the hybrid interference of PSPR and multiple GPR modes.

Zoom In Zoom Out Reset image size

The robustness of the thermal meta-emitter to the observation angle (θ) over a broad range from 0° to 80° is examined in figure 4. Figure 4(a) presents the infrared image at the observation angle of 0° where a high radiation temperature of T_D = 40.0 °C occurs due to the near-unit thermal emissivity. As the observation angle increases from 0° to 80° with a step of 15°, the measured T_D of the meta-emitter remains almost the identical to that of the observation angle of 0°, indicating that the thermal meta-emitter works perfectly even at broad observation angles (figures 4(b)–(e)). It is due to the omnidirectional wideband high emissivity shown in figure 2. The M-IR light source needs to be efficient from a broad range of observation angles, and, thus, the robustness of the thermal meta-emitter to various observation angle is of utmost importance. In figure S4, we numerically studied the effect of the number of layers on thermal emissivity. The peak emissivity increases with the number of layers However, the high emissivity bandwidth (i.e. the range of wavelength where the emissivity is above 0.8) is narrower as the number of layers increases from 3 to 6. Moreover, the fabrication processing is much more difficult for a larger number of layers owing to its greater depth-to-width ratio.

Zoom In Zoom Out Reset image size

We have experimentally and theoretically explored how to combine the propagating SP resonance mode with multiple GPR modes to create a polarization independent, omnidirectional, broadband, and high thermal emissivity metadevice in the M-IR region. Fabricating an array of stacked nanocavities consisting of two pairs of circular Si₃N₄–Au stacks on a Si₃N₄–Au dual-layered structure offers a nearly perfect thermal emission of ∼0.8 over a broad spectrum ranging from 8 to 14 μm. The meta-emitter also possesses broadband emissivity even at a large observation angle (θ ∼ 60°) that is polarization independent. Our theoretical analysis shows that such a broadband high emissivity stems from overlapping the propagating SP resonance and the GPR excited at the different parts of the M-IR region. The rather simple geometry of the structure provides tolerant fabrication processing. Our strategies pave an avenue towards realizing active M-IR meta-devices including thermal radiators and imaging sensors that can be manufactured using complementary metal oxide semiconductor compatible techniques.

4.1. Sample fabrication

4.2. Sample measurements

4.3. FDTD simulations

T C and M L contributed equally to this work. T C acknowledges support from the National Key Research and Development Program of China (2019YFA0709100, 2020YFA0714504), Fundamental Research Funds for the Central Universities (Nos. DUT20GF108, DUT20RC(3)007, DUT20RC(3)062, DUT19RC(3)010), and the Program for Liaoning excellent Talents in University (Grant No. LJQ2015021).