Large-area 1D selective emitter for thermophotovoltaic applications in the mid-infrared

Thermophotovoltaic (TPV) devices enable energy harvesting from (waste) heat.^1–3 In a TPV device, emitters absorb thermal energy from the heat source and radiate photons.⁴ The photons from the emitter are converted into electricity by the photodiode due to the photovoltaic effect.⁴ For a given emission spectrum of the emitter, at a given temperature, the photon flux available for power conversion is proportional to the emitter size. This indicates that a higher output power of a TPV device can be achieved using a larger emitter. To fabricated TPV emitters, subwavelength structures (or metamaterials) are commonly used, where nanometer scale features are two- or three-dimensionally patterned on a substrate.^5–16 One advantage of using metamaterial emitters is that their emission spectrum can be arbitrarily engineered by changing their dimensions,^17,18 which is useful for creating emitters for various source temperatures. However, the difficulties and cost of fabricating metamaterial emitters may make it challenging to realize large-area emitters for practical applications¹⁹ -we note that the TPV power output density is typically on the order of 1W/cm² or below.^20–23 For this reason, emitters with the following advantages will help realize TPV applications on a greater scale: easier fabrication, low fabrication cost, and engineerable emission properties. Compared to metamaterials, structures that use optical interference effects between one-dimensional thin layers have the advantage of easier fabrication.^24–29 The interference effects also provide some tunability of the structure’s optical (or emission) properties. Thus, interference-based emitters that are fabricated with low-cost materials are a good candidate that can help achieve practical deployment of TPV devices. In this work, we demonstrate a selective emitter based on optical interference effects. The emitter is fabricated simply with one-dimensional layers of silicon (Si) and chromium (Cr) on a Si substrate; where all used materials are low-cost. The Si was chosen due to its high refractive index, which leads to a longer optical path length per unit thickness of the material. On optical interference, a material with a longer (or shorter) optical path can lead to a narrower (or wider) bandwidth.³⁰ The intensity contrast between the constructively and destructively interfered wavelengths can be further enlarged by having Cr under the Si due to the optimal reflectance at the Si/Cr interface.³⁰ The higher contrast between maxima and minima of emission intensity is a desired property of a selective TPV emitter. Moreover, the emission spectrum of our emitter can also be engineered by changing the top Si thickness.

The emitter was fabricated by depositing Cr and Si on a Si substrate with a diameter of 100 mm. A few nm thick Ti layer was used between each layer to improve adhesion and prevent delamination upon heating. All materials were deposited in a sputtering chamber (ATC Orion-3-HV, AJA International) at 400 °C to induce densification for bulklike optical properties.³¹ The emitter structure is shown in Fig. 1. A small piece (32 mm × 32 mm) is cleaved out of the 100 mm emitter to test thermal emission. Figures 2(a) and 2(b) show some key components of the custom-built emission measurement setup in this work. Figure 2(c) illustrates the radiation path and the power flow from the sample to the photodetector (PDAVJ10, Thorlabs). The thermal radiation of the emitter is measured at 802 K in air at normal incidence. This temperature is chosen so that the emission peak is in the mid-infrared (mid-IR), which is typically the spectrum of interest in TPV applications.

Prior to the emitter fabrication, a sample of a Cr/substrate was fabricated with the same deposition parameters that were used to fabricate the emitter. The ellipsometric parameters psi (Ψ) and delta (Δ)³² of the single Cr layer were measured at ∼802 K since the optical properties of materials change as a function of temperature.^33,34 The measured data were fit using a Drude, a Lorentz, and a few Gaussian oscillation models. Since the Cr was optically thick, the reflection at the Cr/substrate interface was negligible. Then, the ellipsometric parameters of the emitter in Fig. 1 were measured at 802 K and fitted using the previously obtained dispersion models for the Cr layer and a Cauchy's dispersion model for the Si layer. In the fitting process, the inclusion of the Ti layer, which is only a few nm in thickness, above Cr made a nearly negligible difference. The inclusion of the substrate also did not affect the results because the Cr was optically thick. Then, the absorptivity spectrum of the emitter was plotted based on the measured optical properties of each layer. This will be referred to as the “theoretical” emissivity of the emitter due to Kirchhoff’s law; which states that the emissivity of an object is equal to its absorptivity.³⁵ Also, a highly doped Si substrate, whose room temperature resistivity is 0.008 Ω cm, was used as a reference material to normalize the emitter’s radiation relative to that of a blackbody. The reference sample was annealed in air at 900 K for 30 min to cause oxidation before the emission measurement such that the radiation is stable and is not significantly affected by additional oxidation. After annealing, its ellipsometric parameters were measured at 802 K, which were then used to generate the absorption spectrum. This absorption spectrum was also used as the emissivity spectrum due to Kirchhoff’s law.

In the setup in Fig. 2, the radiated beam undergoes multiple reflections between the window, quartz, and the sample. In Fig. 2(c), the total power that is reflected from the Au mirror (P_sum) is given by

where it assumes that most power is in the first and second beams transmitting the window. P_rad is the power thermally radiated from the sample, T_w is transmittance of the window, R_w is the reflectance of the window, R_s is the reflectance of the sample, and R_Au is the reflectance of the Au mirror. The power P_sum passes through a focusing lens and a monochromator and then enters the photodetector. Thus, the power that enters the photodetector (P_in) and the power reflected from the mirror (P_sum) are in a linear relationship as below:

We note that the output voltage from our photodetector (V) is linearly proportional to the input power (P_in) at a given wavelength,

From Eqs. (1a)–(1c),

Thus, the relationship between the power radiated from the sample (P_rad) and the output voltage from the detector (V) is

From Eq. (2), the emissivity of the emitter (ɛ_emitter) can be obtained by

where the properties corresponding to the emitter and the reference material are denoted with “emitter” and “ref” in the subscript, respectively. We note that both the emitter and the reference material are optically thick and have near-zero transmittance over the wavelengths investigated. Thus, the radiation from the heater under the sample does not contaminate the emission from the sample’s top surface. Also, a 500 μm thick quartz window was used to help stabilize temperature of the sample by minimizing heat lost to convection. The window was at a temperature lower than 378 K. This is significantly lower than 802 K, and the emissivity of the window is smaller than ∼0.25, obtained from the measured absorptivity, over the wavelengths investigated. Therefore, we argue that the radiation from the window is negligible.

Figure 3(a) shows theoretical emissivity of the emitter as a function of Si thickness at different temperatures, where it assumes no surface oxidation. At a given temperature, in Fig. 3(a), it is seen that the spectrum shifts to longer wavelengths as the Si thickness increases. This is because the wavelengths at which constructive and destructive interference occur shift due to the increased optical path length in the Si layer.³⁰ Similarly, at a given thickness, the spectrum at 802 K (solid line) is located at longer wavelengths than the spectrum at room temperature (dashed line). This shift is because the Si’s refractive index increases as the temperature increases.³⁶ The increased refractive index leads to increased optical path length and, thus, affects the interference in the identical way as above. Figure 3(b) shows measured emissivity of the emitter at 802 K in air. The maximum of the measured emissivity is ∼0.84 at the wavelength of 3.75 μm with a 1.2 μm full-width-at-half-maximum (FWHM). The measured spectrum is also compared with the theoretical spectrum of an emitter model with surface oxidation because the sample was heated in air.^37,38 In the oxidized model, the thickness of Si decreased from the initial 740 nm to 700 nm, producing ∼88 nm of an additional oxide layer. Interestingly, this reduction of Si thickness is much larger than what it would have been in typical (bulk) Si wet oxidation³⁹—we note that the emitter dwelled at 802 K for less than 70 min during the emission measurement. Since Cr and Ti have a thermal expansion coefficient larger than that of Si, this result may be due to the tensile stress in the thin Si layer,^40,41 which opens the Si lattices for oxygen atoms to diffuse more easily. Moreover, from the results of Figs. 3(a) and 3(b), it is evident that the presence of surface silicon oxide due to oxidation does not noticeably affect the spectrum shape or location; i.e., the emissivity spectra at d = 700 nm with and without oxidation are similar. This indicates that the emissivity of our emitter at a given temperature is mostly determined by the Si thickness.

This work demonstrates a large-area selective emitter fabricated with 1D layers of Si and Cr. Due to the simple structure and fabrication ease, this emitter can easily be scaled up to even larger areas, the entire substrate area, for example. This can lead to a higher output power by increasing the photon flux available for power conversion in a TPV device with significantly lower fabrication cost, time, and complexity. Moreover, the emission spectrum of the emitter in this work can be engineered by changing the Si thickness. This indicates that the utility of our large-area emitter can be extended for varying source temperatures as well. Furthermore, the demonstrated emitter also has potential applications as a mid-IR source in molecular sensing or free space communications, for example.⁴² It is reported that the oxidation resistance of Si can be improved using a dielectric coating layer, such as Al₂O₃, that suppresses diffusion of oxygen atoms due to the compressive stress in the coating layer.^43,44 In our future work, we look forward to studying thermal/oxidation stability of the emitter in air for a prolonged time of operation using such layers.

Funding was provided by the National Science Foundation (NSF, Grant No. ECCS-EPMD-2120568). The ellipsometry measurements were carried out at the Tufts Epitaxial Core Facility on equipment supported by the United States Office of Naval Research (No. ONR DURIP N00014-17-1-2591). This work was also completed in part at the Center for Nanoscale Systems (CNS) at Harvard University, a member of the National Nanotechnology Coordinated Infrastructure Network (NNCI), which is supported by the National Science Foundation under NSF Award No. 1541959.

The authors have no conflicts to disclose.

The data that support the findings of this study are available from the corresponding author upon reasonable request.