Selective absorbers and thermal emitters for far-infrared wavelengths

The far-infrared (far-IR) region of the electromagnetic spectrum can be considered to fill the space from the edge of the long-wave infrared (∼13 μm) all the way through to the mm-wave portion of the spectrum. The long-wavelength side of the far-IR is most frequently referred to as the THz wavelength range, which is of great interest for a wide variety of optoelectronic, sensing, and imaging applications. The shorter wavelength side of the far-IR (∼20–100 μm, which we will refer to as the Reststrahlen band), unlike the THz, has received little attention thus far, despite being a wavelength range of potential interest for explosives detection,¹ chemical and biological spectroscopy,^2,3 and astrophysics applications. The exploration of any wavelength range requires an optical toolkit, consisting of sources, detectors, and optical components such as lenses, beamsplitters, polarizers, and filters. In the ∼20–100 μm wavelength range, such components are few and far between. In large part, this is due to the strong phonon absorption of semiconductor materials, which form the basis of most all optoelectronic devices and in the infrared, most optical materials. We refer to the wavelength range of strong phonon absorption as the semiconductor's Reststrahlen band, which varies between materials and depends on the optical phonon energy of the semiconductor. For wide bandgap materials (SiC, diamond, III-nitrides), the Reststrahlen band lies at high energies (corresponding to free-space wavelengths between 10–20 μm). But, for the III–V materials that make up the majority of our optoelectronic infrastructure, phonon energies range across optical wavelengths from 20–40 μm, though phonon absorption dominates the total optical losses of these materials for an even broader wavelength range. For this reason, we will refer to a general Reststrahlen band, covering the wavelength range where typical III–V materials become highly lossy (⁠$λo=20–60 μm$⁠). These losses have largely prevented the development of any significant optical infrastructure for the Reststrahlen band, as evidenced by the quantum cascade laser (QCL), which serves as arguably the mid-infrared (mid-IR) source of choice, as well as a promising THz source, but whose performance decays significantly for devices even approaching the Reststrahlen band,⁴ a result of phonon absorption in the QCL's constituent materials.⁵ 

Thus, the Reststrahlen band is something of an optical frontier, with little in the way of optical infrastructure and, admittedly, little in the way of motivating applications (though one could argue as to which of the above is the “cause” and which the “effect”). It would not be unreasonable to suggest that the distinct lack of a Reststrahlen band optical and optoelectronic toolkit has limited the exploration of this wavelength range, which in turn has prevented the demonstration of potential applications to drive further exploration. The development of emitters based on electronic transitions in traditional optoelectronic III–V materials seems unlikely, and while efforts to utilize III-nitrides for quantum cascade-like emitters at long wavelengths are exciting,⁶ efficient emission from such devices may not be realized for some time. In fact, in the short run, the best option for Reststrahlen band sources might be thermal emitters.

Thermal emitters are ubiquitous in mid-IR optics, serving as cost-effective broadband IR sources in most Fourier transform infrared (FTIR) spectrometers, as well as in a range of IR sensor and illumination systems. Such emitters work reasonably well at mid-IR, or even near-IR, wavelengths, which correspond to peak thermal emission for a blackbody across a range of hot, but achievable, temperatures. At longer wavelengths, however, thermal sources become more problematic, simply due to their weak far-IR spectral emittance, even at high temperatures (Figure 1(a)). At temperatures where emitted power from a $1×1$ cm² blackbody surface, across the entire 30–40 μm wavelength range, approaches the ∼10mW range, the fraction of total emitted power in this wavelength ranges is well below 1%, meaning that the vast majority of the system's energy is being lost at wavelengths outside of this band. Thus, while cost-effective, thermal emitters are highly inefficient, especially for applications requiring light emission in only limited wavelength bands, or at long wavelengths.

However, the spectral range of emission for a thermal source can by controlled using surfaces and structures with engineered absorption resonances. By Kirchoff's Law, which relates the emissivity and absorptivity of a surface [$ε(λ)=1−a(λ)$], these resonances will, upon heating of the surface, result in spectrally selective thermal emission. Numerous approaches have been used to engineer the IR absorptivity, and therefore the emissivity, of such thin films, leveraging a variety of layered, patterned, and/or engineered surfaces and materials.^7–17 Much of the effort towards the development of selective thermal emitters has focused on the mid-infrared (reasonably efficient for temperatures up to ∼1000 K) and, more recently, the near-IR (T > 1000 K) for thermophotovoltaic applications.¹⁸ In such frequency selective surfaces, absorption can be engineered via thin-film interference,^19,20 antenna or plasmonic resonators,^10,13,14,18 structured dielectric waveguides,²¹ propagating surface modes,^7,11 or even surface phonon polariton modes. The majority of the work utilizing phonon resonances have leveraged the high energy (10.6–12 μm) phonons of SiC,^22–28 though thermal emission has been observed via outcoupling of surface phonon modes on materials with lower energy phonon resonances.²⁹ However, these surfaces have operating wavelengths limited to discrete bands determined by the material's phonon energies, and suffer from significant emission outside the spectral band of interest. Thus, the ability to design and engineer selective thermal emitters at long wavelengths, away from phonon resonances, may provide cost-effective thermal light sources with improved efficiency and greater wavelength flexibility, even if the total power emitted in a given wavelength band is still limited by Planck's radiation law. In this work, we demonstrate an engineered surface of subwavelength thickness with designed absorption resonances across the ∼30–40 μm wavelength range. We also demonstrate that such surfaces, when heated, can serve as reasonably narrow-band thermal sources at far-IR wavelengths.

The design of our surface is shown in Figure 2, and consists of a patterned metal (Ti/Au) top layer, separated from a solid metal ground plane (Mo) by a dielectric spacer (AlN). Here, we utilize AlN due to its high energy Reststrahlen band (∼11–15 μm),³⁰ which gives the AlN nearly constant refractive index at wavelengths much longer than its optical phonon resonances. Samples with varying thickness of AlN (1200, 1400, 1600, and 1800 nm) deposited above a layer of Molybdenum (100 nm) were purchased from OEM Group, Inc. A layer of Au (∼65 nm) was then deposited via e-beam evaporation (with a 5 nm Ti layer for adhesion) over the AlN. The Ti/Au top surface was patterned using a KI/I₂-based gold etchant (Transene Gold Etchant TFA) through a lithographically defined photoresist etch-mask. Grating patterns with periods $Λ=20$ μm were chosen in an attempt to ensure that effects from diffraction would occur away from the wavelength range of interest. One-dimensional (1D) grating patterned samples with a range of grating widths were fabricated on wafers of each AlN thickness. In addition, two-dimensional (2D) grating patterns were fabricated for emission experiments.

The structures utilized in this work are, to a certain extent, long-wavelength variations of the metal/dielectric/patterned metal structures used for selective thermal emission and/or absorption at shorter wavelengths.^10,13,14 At frequencies far below the metal's plasma frequency, the top patterned metal layer acts as an array of optical antennas, coupling light into the structure at a wavelength dependent on the geometry of the patterned metal. For the 1D pattern, only TM-polarized light is coupled into the structure (as the antennas are infinite in the electric field direction for TE-polarized light). For the 2D pattern, both TE- and TM-polarized light can couple to the now square patch antennas. A magnetic moment, created by the anti-parallel surface currents induced on the top metal and ground plane, couples strongly to the incident radiation. The thickness of the dielectric spacer effectively determines the strength of the absorption resonance for a given patterned metal geometry. For such a structure, near-perfect absorption can be achieved even with loss-less dielectrics, indicating that light is primarily absorbed by the metal. Translating such structures to the far-IR brings a host of challenges, primarily in the choice of dielectric, which must be low-loss in order to achieve narrow-band selectivity. In this work, we use AlN as our dielectric, as the high phonon energies of the III-nitrides move the high-absorption AlN Reststrahlen band far from our operational wavelengths.

Reflection spectra were collected in a Bruker V80v Fourier transform infrared (FTIR) vacuum spectrum using a Pike Technologies 10Spec specular reflectance accessory, allowing for reflection spectra at 10° from the sample normal. A KRS-5 wire-grid polarizer was placed in the beam path in order obtain both transverse electric (TE) and transverse magnetic (TM) polarized reflection spectra. For emission experiments, the sample was mounted on a hot-plate external to the FTIR, and the thermal emission from the sample, spatially filtered by an aperture, was collected and focused into the FTIR (with the input window removed) using reflective optics. Both the sample and collection optics were enclosed with a plastic “tent” structure, and the FTIR and “tent” kept at nitrogen gas overpressure, in order to mitigate atmospheric absorption and the significant transmission losses of the FTIR window at longer wavelengths (⁠$λo>35$ μm). In order to isolate the sample emission from the significant thermal background, emission from the system, with the aperture blocked, was collected. We also collected emission from a “black soot” reference sample with near-uniform broadband emissivity (in effect acting as a calibrated “greybody”). This sample was fabricated by placing our Si/Mo substrate (with AlN etched off) in the flame of a candle until the surface is optically black.^31,32 2D arrays (as shown in Fig. 2(b)) were used for the emission experiments in order to maximize the sample signal. Our structures were all modeled by 3D rigorous coupled wave analysis (RCWA).³³ This technique, which was introduced in Ref. 34 and further expanded in Ref. 35, is ideal for modeling the optical properties of periodic, layered material systems such as our far-IR absorber structures.

Figure 3 shows the RCWA-modeled and experimental TE- and TM-polarized reflection data for a 1D grating structure with $w=13$ μm, $hAlN=1400$ nm, and $Λ=20$ μm. As expected, strong absorption features are observed for TM, but not TE polarized light. Because the Mo groundplane is optically thick, and the periodicity of the grating structures is subwavelength, the strong TM reflection dip at $λ=32.5$ μm cannot correspond to either transmission or diffraction, and must therefore be evidence of strong, spectrally selective absorption in the three-layer system. The inset shows the magnetic field amplitude (⁠$H→y$⁠) for TM-polarized reflection at resonance, indicating that we are coupling to a higher-order antenna mode in the top patterned Au layer. Strong absorption at similar wavelengths can be simulated with grating structures having $w∼4$ μm, $hAlN∼600$ nm, and $Λ∼6$ μm, but these lower order resonances have spectral widths approximately twice those of the structures with the wider stripes presented here, and thus we chose to focus on the larger stripes, for their improved spectral selectivity. Figure 4 shows the (a) simulated and (b) experimental TM polarized reflection spectra from a series of samples with 1D grating structures with $hAlN=1400$ nm and $Λ=20$ μm, and with fabricated stripe widths of $w=9.7, 11.7, 13, 14$⁠, and $15$μm (undercutting during the metal wet-etch decreases the stripe widths from the designed by $∼2$ μm). The experimental data show a good fit to our RCWA simulations. All samples show peak absorption of ∼90%, with off-resonant reflection >90%, with resonances across the $λ=27–37$ μm wavelength range.

For selective emitter applications, achieving strong off-resonant reflection is as important as the strength of the resonant absorption feature, as thermal emission away from resonance is integrated over a broad range of wavelengths, and even weak broadband, off-resonance emission can overwhelm the designed selective emission, especially for emitters designed to operate at long wavelengths. Figure 5 shows the experimental emissivity at 200 °C (and room temperature reflectivity) from a sample with $hAlN=1400$ nm and a top patterned layer consisting of a 2D array of patches, with $Λx=Λx=20$ μm and $wx=wy=15$ μm. A 2D grating pattern was used to ensure polarization independent emissivity, allowing us to capture thermal emission without a polarizer, thus improving our signal to noise. As can be seen from Figure 5, a clear peak in the emissivity is observed at the designed resonance of the surface, and the spectral features observed in the room-temperature, 10° incident angle, reflectivity are generally reproduced in the experimental emissivity (despite the slight difference in angle). In addition, we used our RCWA simulations to calculate the fraction of the samples' total thermal emission contained within the spectral range defined by the FWHM of our far-IR emission resonance. For the sample in Figure 5, we note that 2.4% of the total thermal emission sits within this wavelength band. For comparison, a perfect blackbody at the same temperature emits only 0.42% of its power in this same band. Thus, in addition to providing spectrally selective emission in the far-IR, these structures also improve the efficiency of emission into the resonant wavelength band by over a factor of 5.

In summary, we have demonstrated strong selective absorption and emission at far-IR wavelengths from subwavelength thickness structures with designed absorption/emission resonances. We have shown both polarized and unpolarized resonances, and demonstrated control of the resonances across the $λ=27–37$ μm wavelength range. Samples were characterized by FTIR reflection and emission spectroscopy, and modeled using 3D RCWA, with excellent agreement between the simulated and experimental response of the structures. While the far-IR, in particular, the portion of the far-IR overlapping with the Reststrahlen band of traditional optoelectronic semiconductor materials, remains an extremely challenging wavelength range for the development of optical and optoelectronic structures and devices, the work presented here offers a small step towards the development of an optical infrastructure for this largely unexplored region of the electromagnetic spectrum.

The authors would like to acknowledge funding from the National Science Foundation, Award Nos. ECCS 1420952 (D.W. and W.S.) and ECCS 1420176 (A.H. and K.F.). The authors gratefully acknowledge useful advice and discussion with V. Podolskiy and C. Roberts (UMass Lowell) regarding RCWA simulations.