We demonstrate strong-to-perfect absorption across a wide range of mid-infrared wavelengths (5-12µm) using a two-layer system consisting of heavily-doped silicon and a thin high-index germanium dielectric layer. We demonstrate spectral control of the absorption resonance by varying the thickness of the dielectric layer. The absorption resonance is shown to be largely polarization-independent and angle-invariant. Upon heating, we observe selective thermal emission from our materials. Experimental data is compared to an analytical model of our structures with strong agreement.
©2013 Optical Society of America
There has been recent interest in a class of composite optical materials capable of strong, or perfect, absorption of light at designed wavelengths. Such materials can be traced back to early radar work, one example of which being a structure known as the Salisbury Screen . In their earliest incarnation, these structures used a lossy semitransparent layer suspended a quarter-wavelength above a reflecting ground plane, designed to act as an anti-reflection coating for the reduction of a surface’s radar cross section. More recent work has attempted to bring perfect absorption to optical frequencies, where in many cases, this light absorption is achieved by means of a patterned metallic top layer consisting of either antenna, plasmonic, or metamaterial resonators, again suspended over a metallic ground plane, but now by a thin (d<λ0/4n) dielectric layer [2–5]. Initial demonstrations of such structures occurred at both the very long wavelength (microwave) and shorter wavelength (near-IR) portions of the electromagnetic spectrum [2,3]. Nonetheless, the fundamental operation of the more recent metamaterial-type structures is very similar, and is typically explained by means of impedance matching or critical coupling [3,6]. The design of the resonant top metallic layer allows for significant control over the thin films’ absorption frequency, and has led to the demonstration of perfect absorbers (PAs) across a broad range of the electromagnetic spectrum [2–5,7,8].
The mid-infrared (mid-IR) portion of the electromagnetic spectrum (~3-30µm) holds particular potential for such structures. The strong field localization that often accompanies the perfect absorption phenomenon could be leveraged to enhance the interaction of mid-IR light with small volumes of molecular and biological species which have strong absorption fingerprints in this spectral region. Early PAs were demonstrated to be effective at sensing relatively small changes in the refractive index above the PA surface at near-IR frequencies , and antenna-based PAs have demonstrated strong coupling to molecular resonances in the dielectric spacer separating the top patterned and bottom ground plane metallic layers . However, the mid-IR is of additional interest as it is the spectral range corresponding to thermal emission for objects across a wide range of temperatures (~200-1400K). Kirchhoff’s Law of thermal radiation links the optical properties of a material to its thermal emission, with the absorptivity of the material A(λ) equivalent to the emissivity ε(λ). While losses in most optical metamaterial and plasmonic structures are considered deleterious to device performance, in the mid-IR these losses can be harnessed and used as a fundamental component of device performance. This is because mid-IR perfect absorbers, by allowing for spectral control of absorption resonances, can spectrally modify and control the material’s blackbody emission upon heating. Thus, such structures hold potential for applications in thermal cloaking, thermal signature mimicry, and frequency selective thermal light sources. Already, PA structures have demonstrated largely angle-independent selective thermal emission from both top-patterned antenna and metamaterial arrays [5,10]. However, such structures require multiple material depositions and at least one patterning step, which can often require high resolution and uniformity. Here we demonstrate that the combination of a new class of engineered mid-IR optical metals, coated with a subwavelength, high-index dielectric layer, can result in a class of polarization-independent and largely angle-insensitive wavelength-flexible PAs and selective thermal emitters. These structures require no lithography, and only two processing steps, one of which is a spin-on and bake process, and the other a standard evaporation, allowing for inexpensive and rapid fabrication of large area samples. Strong absorption is observed, and shown to be controllable across a broad range of mid-IR wavelengths (4.8-12.5µm). Our materials are characterized by mid-IR Fourier transform infrared (FTIR) microscopy, angle-dependent reflection spectroscopy, and thermal emission spectroscopy. Strong selectivity is seen in both absorption and emission spectroscopy, and is measured to be largely angle- and polarization-insensitive. Our results are compared to an analytical model of the system with excellent agreement.
Recently, two works by the Capasso group have demonstrated strong absorption from two-layer thin-film systems [11,12]. In the first, a sapphire substrate is coated in the phase-change material VO2, and dynamic control of perfect absorption is demonstrated by heating the VO2 through its phase-change temperature. In the second, Ge thin films of varying thicknesses are deposited upon gold substrates, and strong absorption is observed at visible wavelengths. The physics of both systems are quite similar, relying on the resonant destructive interference between light reflected from the interfaces on either side of the dielectric coating. Ultimately, such structures are a clever new twist on the thin-film interference effect utilized in the Salisbury screen, an effect which also forms the basis of, among many other applications, anti-reflection and high-reflectivity coatings and dielectric Bragg mirrors. Traditionally, the latter structures are designed to give reflection minima coincident with transmission maxima with little to no absorption, as absorption is generally undesirable for most optical applications. In the case of the recent above-referenced works [11,12], the system under investigation is designed for thin-film absorption and uses a two layer system with a high-reflectivity ground plane so that there is no light transmitted through the structure.
For this two-layer system of a dielectric film above a metallic substrate with light incident from air, we can model absorption as a function of dielectric thickness for a fixed wavelength. For a dielectric with small losses and a near-perfectly conducting ground plane, reflection minima (and thus absorption maxima) will occur when the thickness of the film is d = (2m + 1)λ0/4n, where m is an integer, λ0 is the free-space wavelength of light, and n is the refractive index of the thin film material. Thus, the thinnest such structure capable of a reflection minimum has a thickness equal to one fourth of the wavelength of light in the thin film, as shown in Fig. 1(a). This limitation can be significant, especially at longer wavelengths, where a thickness of λ0/4n could prove unwieldy for real-world applications. However, as recently demonstrated [11,12], the picture can change dramatically if the dielectric becomes strongly absorbing and the metal is given a finite conductivity. In such a case, the phase shift for reflection at the metal/dielectric interface is no longer π, and thus reflection minima can be achieved with thickness much less than d = λ0/4n, as shown in Fig. 1(b). It is argued that the combination of lossy dielectrics and finite conductivity metals are capable of the strongest absorption in such a two layer system, which is effectively the case for most traditional metals at visible or near-IR frequencies. Alternatively, if the structure is designed to absorb at or near the plasma frequency of the metal film (where the permittivity of the metal is small), very strong absorption can be achieved with thin, lossless, dielectric films, as shown in Fig. 1(c). However, controlling the optical properties of traditional metals in order to position the plasma frequency appropriately for such structures is not possible.
Recently there has been interest in a new class of engineered metals based on highly doped semiconductors for long-wavelength applications. At high enough doping densities, these semiconductors can, at near-IR and mid-IR wavelengths, mimic the shorter wavelength optical properties of traditional noble metals [13–17]. Such engineered metals are of interest, as they allow for strong confinement and low losses of optical modes in basic plasmonic structures, an effect not possible with traditional plasmonic metals at these wavelengths [18,19]. Moreover, such materials have the potential for integration into semiconductor optoelectronic devices , and have the added benefit of allowing the designers of plasmonic structures to control the optical properties of their metals by simply tuning the materials’ doping density, which cannot be achieved with traditional plasmonic metals at shorter wavelengths. As noted above, when such metals are used as the ground plane for the thin film absorbers, perfect absorption can be achieved at dielectric thickness d<λ0/4n, even with no losses in the dielectric layer, as shown in Fig. 1(c). Small changes in the dielectric thickness allow for tunability of these structures’ absorption across a large part of the mid-IR. As we will show, such structures offer a low cost thin-film material system with no required lithography for the development of spectrally-selective thermal emitters across a broad range of frequencies.
2. Fabrication and experimental set-up
Our highly-doped silicon ground plane was fabricated with a spin-doping process, using P509 Filmtronix Spin-on Dopant on a quarter piece of a 6” silicon-on-insulator (SOI) wafer. The SOI had an active layer thickness of 2-3 µm, an oxide layer thickness of 0.9-1.1 µm, and a handle thickness of 610-640 µm. The SOI was first cleaned with a standard acetone/isopropyl alcohol degrease. After a 5 minute bakeout at 125 °C, 0.5 mL of dopant was spun on at 4000 RPM for 40 seconds, and hard-baked at 200 °C for 12 minutes. Finally, the dopant-coated wafer was baked in an 1100 °C furnace for two hours to drive in the dopants, after which the wafer was cleaned in a 1:1 HF:DI solution to remove the baked spin-on dopant layer. The doped wafers were then characterized optically, using mid-IR reflection microscopy, in order to determine the dopant concentration of the sample.
The doped SOI wafers were then degreased and a thin germanium film was deposited by electron beam evaporation from a 99.99% germanium source at a rate of approximately 2Å/s. The film thickness was monitored using a quartz crystal monitor and confirmed with a germanium-selective etch and profilometry. For our Ge-thickness studies, three evaporation runs of 82.5, 165, and 330nm were performed and seven total samples were fabricated with thicknesses ranging from 82.5nm to 577.5nm in 82.5nm steps. For patterned Ge samples standard lithographic and liftoff techniques were used.
Our samples were characterized by a range of optical techniques. Basic reflection spectroscopy measurements were taken on a Bruker IR-II infrared microscope coupled to a Bruker V80V Fourier transform infrared (FTIR) spectrometer. Experimental data was normalized to reflection from a gold surface, which gives near 100% reflection in the mid-IR. The thickness-dependent reflection spectra were collected on the IR microscope set-up. Angle-dependent reflection spectroscopy was performed on a custom-built experimental set-up coupled to a Bruker V70 FTIR. In this case, broadband IR light from the FTIR’s internal source was focused on the sample using a long focal length mid-IR lens, in order to narrow the solid angle of the incident light and thus improve angular resolution. The incident light was passed through a mid-IR wire-grid polarizer to allow for measurement of both s- and p-polarized reflection. The sample was rotated about its vertical axis, and the collection optics and external HgCdTe (MCT) detector were mounted to a rail, which also rotated around the vertical axis of the sample. The set-up was designed to be able to accurately measure polarization dependent reflection at angles of 10-60° from normal incidence across the entire range of the mid-IR. These data were also normalized to reflection from gold at each measurement angle.
Finally, a custom set-up was built to measure thermal emission from our samples. For these measurements, the samples were attached to a copper mount with indium. The mount is designed to hold two cartridge heaters, which are controlled by an external temperature controller with feedback from a thermocouple clipped to the surface of the copper mount. Thermal emission from the sample passes through two apertures for collimation before being focused into the input of the Bruker V80V FTIR, where it is detected by the FTIR’s internal MCT detector. Thermal emission was measured at a block temperature of 190°C for samples of all thicknesses fabricated, which corresponds to a slightly lower surface temperature for our emitters due to the temperature gradient across the Si substrate. Because the background thermal emission from the internal components of the FTIR can be significant, a spectrum was also obtained with the FTIR input blocked, and this spectrum was subtracted from each of the emission spectra from our samples, in order to isolate the thermal emission from the sample surface. Thermal emission was also measured from patterned Ge samples using a FLIR BX-320 thermal imaging camera. These samples were patterned with 425nm of Ge, designed to give selective thermal emission in the middle of the camera’s 7-13µm detection range, and images were taken through a 15x all-reflective microscope objective.
Our structures were modeled using a 1D transmission matrix (T-matrix) formalism. The samples are broken into layers of thickness di, with a complex refractive index, ñi and a complex transmission angle (calculated from Snell’s Law) assigned to each layer. The transfer matrix Tij is determined for each interface of the structure (between layers i and j), as well as the transfer matrix Ti for propagation through the ith layer. The total reflection and transmission through the system can be extracted from matrix product of all of the layers’ and interfaces’ T-matrices, as detailed in most any fundamental text on Optics . The doping profile of our samples is estimated assuming a constant surface concentration of dopants (provided by the spin-on dopant coating) and a diffusion coefficient for phosphorus atoms in Si. From this, the distribution of dopants in our material is expected to follow a complementary error function (erfc) profile as a function of depth (z). The Drude model can be applied to these films , resulting in only the surface doping concentration and scattering time as adjustable parameters. Using a doping-independent scattering rate in our material allows for the determination of the material’s plasma frequency, and thus the material’s complex permittivity as a function of depth. For the samples used in this work, little variation in plasma wavelength is expected over the thickness of the active Si layer, an expectation verified by reflection measurements on doped active Si layers etched to varying depths. This sample was modeled with a surface plasma wavelength of 4.2μm. At the longer wavelengths (λ0>3μm) we are working with, the doped active Si is either highly absorbing or highly reflective, and little to no light reaches the buried oxide, making the SOI wafer functionally equivalent to a heavily doped Si wafer. Thus we can treat our system as an infinitely thick doped-Si wafer with an erfc(z) doping profile, as shown schematically in Fig. 2(b,c). Choice of only the surface dopant concentration and a constant scattering time allows for accurate fitting to our experimental data.
3. Results and discussion
The normal-incidence reflection data from one of our doped SOI wafers, before Ge deposition, is shown in Fig. 2(a) along with the modeled reflection using the T-matrix method. Good agreement is observed between model and experiment, and the parameters used to fit the experimental data for the uncoated sample are then held constant throughout our modeling of the subsequently Ge-coated samples. Our model suggests we are able to highly dope our Si material, with an average plasma wavelength for the active Si layer of λp<4.4μm. For the Ge-coated samples, the Ge was treated as a lossless dielectric with nGe = 4.
Figure 3 shows experimental reflection as well as modeled reflection and absorption from a heavily-doped silicon sample coated with 165nm of germanium. Strong absorption is seen at a free-space wavelength λ0 = 5.58μm. The Ge thickness of this sample is d = (λ0/4n)/2.11, at resonant absorption, approximately half the thickness required for strong absorption from a near-PEC/low-loss dielectric structure. As shown in Fig. 3(a), our model matches the data nicely. There is a slight shift between the spectral position of the experimental and modeled data, which we attribute to uncertainty in model parameters such as the germanium index of refraction, as well as to possible small variations in germanium coating thickness from our intended thickness, due to spatial variations on our large area substrates or quartz crystal monitor uncertainty during the deposition process. However, given these experimental uncertainties, the model reproduces the experimental data very well, in both reflection amplitude and the position of spectral features.
Reflection data was taken on the seven germanium-coated samples of thicknesses from d = 82.5-577.5nm, as well as an uncoated, doped silicon wafer (d = 0), using the mid-IR microscope experimental set-up described above, and is shown in Fig. 4(a). Figure 4(b) shows the predicted reflection as function of Ge thickness using our T-matrix model. The experimental results match the model extremely well. For thin Ge films (on the order of 100nm), the absorption resonance is at a free-space wavelength of approximately λ0≈4.6μm, while thick samples show resonances at wavelengths as long as 12.5μm. For thick films, a second absorption resonance appears at shorter wavelengths. This is a higher-order resonance, observed at a shorter wavelength (λ1), which for a near-perfect electrical conductor and low-loss dielectric, would correspond to the d≈3λ1/4n counterpart to the primary absorption resonance. At the thinnest Ge coating (d = 82.5nm), the measured reflection is ~0% at resonance, to within the noise of our spectrometer, demonstrating near-perfect absorption in our thin film structures. At all other thicknesses, reflection is less than 10% on resonance.
The ability to obtain near perfect absorption from effectively lossless dielectric films of thickness d<λ0/4nGe relies on our ability to operate near the plasma frequency of the underlying doped Si, our engineered mid-IR metal. Near the doped-Si plasma frequency, we can obtain phase shifts at the ‘metal’-dielectric interface well-removed from the π phase shift obtained at a typical metal-dielectric interface. This can be seen in Fig. 5 (red line), which shows the modeled phase shift for reflected light at the doped Si-Ge interface as a function of wavelength. In addition, Fig. 5 also plots the scaling factor (SF) by which our Ge thickness is reduced from the λ0/4n thickness associated with typical thin-film absorption from traditional metals and low-loss dielectric coatings (see Fig. 1(a)). We calculate this by taking the ratio SF = (λ0/4nGe)/d, where d is the thickness of the sample’s Ge coating for each sample at that sample’s absorption resonance λ0. As can be seen in Fig. 5, for thin Ge layers, which give strong absorption near the doped-Si plasma frequency, we obtain experimental scaling factors of >3.8. However, as our resonance moves to longer wavelengths with the thicker Ge films, this factor approaches 1. This is because, at longer wavelengths, our doped Si is behaving more like the traditional metals of Fig. 1(a,b), and thus the phase shift at the Ge-Si interface approaches π. Because our metals are effectively engineered metals, strong absorption could be achieved across the mid-IR, with d = λ0/4n, by engineering the Si plasma frequency to track with the spectral position of the desired absorption resonance.
Figure 6 shows experimental reflection data for the d = 412.5nm thick Ge sample as a function of angle for incidence angles of 10-60 degrees, for s-polarized light (a) and p-polarized light (c). The corresponding modeled data for s- and p-polarized light are shown in Figs. 4(b) and 4(d), respectively. As shown in these Figs., the spectral position of the reflection resonance remains constant across the broad range of angles investigated. In addition, the polarization dependence of our samples is minimal across the same angular range, with only a slight decrease in the strength of the reflection resonances for p-polarized light at large (θi>45°) angles. The experimental and modeled data presented in Fig. 6 demonstrates the largely angle- and polarization-insensitive nature of the observed resonances, as well as the good agreement of our experimental results with the modeled data.
Thermal emission measurements were also taken on all seven germanium-coated films using the setup detailed previously. The background-subtracted emission spectra from all seven samples are shown in Fig. 7(a), for a mounting block temperature of 463K. A clear red-shift in the thermal emission spectra as a function of increasing Ge thickness is demonstrated, corresponding to the observed red-shift in our reflection data for the same set of samples. Emission from our samples was modeled by calculating the sample emissivity ε(λ) = 1-T(λ)-R(λ) and multiplying the emissivity by Planck’s blackbody emission expression for a blackbody of temperature T = 450K, in order to account for the lower temperature of the sample surface when compared to the copper mounting block. Finally, the modeled emission was scaled by the spectral response of the FTIR’s internal MCT (obtained from the datasheet supplied with the detector), in order to most accurately model the signal detected by our system. The modeled emission is shown in Fig. 7(b) for all seven samples, as well as the blackbody emission curve for a perfect blackbody, modulated by our detector response. The modeled data again agree nicely with the experimental results.
Lastly, a block “I” (the University of Illinois Urbana Champaign logo) was fabricated on the doped silicon by standard photolithography, Ge deposition, and lift-off. The Ge thickness for our patterned “I” was determined by profilometry to be approximately 425nm, corresponding to a thermal emission peak near λ = 10μm, directly in the center of our thermal imaging camera’s 7-13μm spectral window. This sample was then heated on a hot plate and a picture was taken using a FLIR BX320 thermal imaging camera, shown in the inset of Fig. 7. The image shows up clearly against the background, demonstrating the potential for such films in thermal emission applications.
In conclusion, we have demonstrated strong-to-perfect absorption from subwavelength germanium films of varying thicknesses, deposited on heavily-doped silicon engineered metal substrates. The ability to control the spectral position of the absorption resonance with germanium thickness across a wide range of mid-IR wavelengths was demonstrated. In addition, the spectral position and the strength of the observed resonances were shown to be largely angle- and polarization-insensitive. Thermal emission from these films was measured and shown to be spectrally selective, and patterned films were demonstrated to give spatially selective thermal emission. Our structures were modeled using a transmission matrix approach, and the model results agreed nicely with the experimental data, for both reflection and emission. Our experimental and modeling results indicate that the strong absorption observed results from the unique properties of our engineered semiconductor metals, which allow for control over the metal substrate’s optical properties across a broad range of wavelengths and strong absorption with effectively lossless dielectric coatings. The films demonstrated here are low-cost, with a straightforward fabrication process, and thus hold potential for use as low-cost thermal light sources or emissivity controlling coatings.
The authors would like to acknowledge funding from the AFOSR Young Investigator Program (WS, DW, Award #FA9550-10-1-0226) and the NSF REU program (GR, Award #ECCS-1157933).
References and links
1. W. W. Salisbury, “Absorbent body for electromagnetic waves,” U. S. Patent 2599944 (1952).
3. Y. Avitzour, Y. A. Urzhumov, and G. Shvets, “Wide-angle infrared absorber based on a negative-index plasmonic metamaterial,” Phys. Rev. B 79(4), 045131 (2009). [CrossRef]
5. J. A. Mason, S. Smith, and D. Wasserman, “Strong absorption and selective thermal emission from a midinfrared metamaterial,” Appl. Phys. Lett. 98(24), 241105 (2011). [CrossRef]
6. C. Wu, B. Neuner, G. Shvets, J. John, A. Milder, B. Zollars, and S. Savoy, “Large-area wide-angle spectrally selective plasmonic absorber,” Phys. Rev. B 84(7), 075102 (2011). [CrossRef]
8. K. Aydin, V. E. Ferry, R. M. Briggs, and H. A. Atwater, “Broadband polarization-independent resonant light absorption using ultrathin plasmonic super absorbers,” Nat. Commun. 2, 517 (2011). [CrossRef] [PubMed]
9. J. A. Mason, G. Allen, V. A. Podolskiy, and D. Wasserman, “Strong coupling of molecular and mid-infrared perfect absorber resonances,” IEEE Photon. Technol. Lett. 24(1), 31–33 (2012). [CrossRef]
10. X. Liu, T. Tyler, T. Starr, A. F. Starr, N. M. Jokerst, and W. J. Padilla, “Taming the blackbody with infrared metamaterials as selective thermal emitters,” Phys. Rev. Lett. 107(4), 045901 (2011). [CrossRef] [PubMed]
11. M. A. Kats, D. Sharma, J. Lin, P. Genevet, R. Blanchard, Z. Yang, M. M. Qazilbash, D. N. Basov, S. Ramanathan, and F. Capasso, “Ultra-thin perfect absorber employing a tunable phase change material,” Appl. Phys. Lett. 101(22), 221101 (2012). [CrossRef]
12. M. A. Kats, R. Blanchard, P. Genevet, and F. Capasso, “Nanometre optical coatings based on strong interference effects in highly absorbing media,” Nat. Mater. 12(1), 20–24 (2012). [CrossRef] [PubMed]
13. J. C. Ginn, R. L. Jarecki, E. A. Shaner, and P. S. Davids, “Infrared plasmons on heavily-doped silicon,” J. Appl. Phys. 110(4), 043110 (2011). [CrossRef]
14. M. Shahzad, G. Medhi, R. E. Peale, W. R. Buchwald, J. W. Cleary, R. Soref, G. D. Boreman, and O. Edwards, “Infrared surface plasmons on heavily-doped silicon,” J. Appl. Phys. 110(12), 123105 (2011). [CrossRef]
15. G. V. Naik, J. Kim, and A. Boltasseva, “Oxides and nitrides as alternative plasmonic materials in the optical range,” Opt. Mater. Express 1(6), 1090–1099 (2011). [CrossRef]
16. R. Soref, J. Hendrickson, and J. W. Cleary, “Mid- to long-wavelength infrared plasmonic-photonics using heavily doped n-Ge/Ge and n-GeSn/GeSn heterostructures,” Opt. Express 20(4), 3814–3824 (2012). [CrossRef] [PubMed]
18. J. B. Khurgin and A. Boltasseva, “Reflecting upon losses in plasmonics and metamaterials,” MRS Bull. 37(08), 768–779 (2012). [CrossRef]
19. S. Law, V. Podolskiy, and D. Wasserman, “Towards nano-scale photonics with micro-scale photons: the opportunities and challenges of mid-infrared plasmonics,” Nanophotonics (to be published).
21. B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics, 2nd ed. (John Wiley & Sons, 2007).