Supporting Information: Metasurface Optical Solar Reflectors using AZO Transparent Conducting Oxides for Radiative Cooling of Spacecraft

Transcription

1 Supporting Information: Metasurface Optical Solar Reflectors using AZO Transparent Conducting Oxides for Radiative Cooling of Spacecraft K. Sun 1,2, C. A. Riedel 1,2, Y. Wang 1,2, A. Urbani 3, M. Simeoni 3, S. Mengali 3, M. Zalkovskij 4, B. Bilenberg 4, C.H. de Groot 1 and O.L. Muskens 2,* 1 Electronics and Computer Science, University of Southampton, Southampton, UK. 2 Physics and Astronomy, University of Southampton, Southampton, UK 3 Consorzio CREO, L Aquila, Italy 4 NIL Technology, Kongens Lyngby, Demark * O.Muskens@soton.ac.uk Contents: 18 Pages, 18 Figure, 3 Tables Table of contents Page S1. Simulation of meta-osr design 2 S2. Atomic layer deposition of AZO 8 S3. Electrical and optical characterization 9 S4 Meta-OSR fabrication and characterization 13 S1

2 S1. Simulation of meta-osr design An OSR is a spectrally selective filter, that reflects the ultraviolet (UV), visible (VIS) and near infrared (NIR) parts of the optical spectrum which correspond to the radiation spectrum of the sun. At the same time, an OSR emits the thermal infrared spectrum corresponding to that of a black body at 300 K. Figure S1 shows the typical desired performance, with a sharp transition at around 3 µm wavelength from a perfectly reflecting to perfectly absorbing (emitting) response. Figure S1 Schematic of optical solar spectrum (blue, T= 5777K), thermal blackbody spectrum (orange, T=300K) and desired ideal response function (1-R) for an optical solar reflector. S1.1 Back reflector The proposed structure is a traditional Salisbury screen consisting of a stack of three layers: the metal back reflector, the dielectric spacer layer and the TCO meta-surface. The back reflector can be made of silver or aluminum, both of which are good reflectors in the visible and IR spectral range. While silver has the higher average reflectivity in the visible spectrum, it has a bandgap at around 350 nm, thereby increasing absorption in the ultraviolet. Aluminum on the other hand has a lower average reflectivity and weak bandgap absorption at 800 nm. Simulated absorption spectra for a 100 nm thick layer of the respective metal are shown in Fig. S2(a). The solar absorptance is for Al and for Ag. These values represent the minimum absorptance values possible with the respective reflector. In our experiments, the dielectric spacers tended to peel off of silver due to poor adhesion on native AgO. Furthermore Ag films showed poor stability against oxidation, which is a particular concern in space applications. We therefore chose Al as reflector for our fabricated structures. S2

3 Figure S2 (a) Reflectivity of 100nm thick silver and aluminum films. (b) Real part of the permittivity of different spacer materials. (c) Simulated spectra for different dielectric spacer materials using the full structure device with an insulator thickness of 1130nm, a carrier concentration of 2x10 20 cm -3, a square size of 950 nm and a 250 nm gap. The spectrum for Si 3N 4 is based on extrapolated data (dashed line). (d) Skin depth of transparent conductive oxides as a function of wavelength and its carrier concentration. S1.2 Dielectric Spacer The spacer layer of the structure is designed to create constructive interference of the incoming and the reflected radiation at the position of the AZO meta-surface. This works similar to a λ/4 spacer, however due to the complex permittivity of common dielectrics in the infrared and the coupling of the meta-surface to induced image charges the optimum spacer thickness varies with AZO carrier concentration and generally with changing top material. The permittivity of the spacer material is another important parameter as it allows some freedom in the design of this interference condition through dispersion, which however has to be balanced against strong phonon reflection bands lowering ε. Figure S2(b) shows the real part of the permittivity of several dielectric spacer materials. In the Mid-IR range, most materials show absorption bands due to phonon vibrations, which show as perturbations in the real part of the permittivity. These so-called reststrahlen bands have an increased reflectivity which is undesirable for our high IR absorptive application. Infrared transparent materials like fluorides are of interest for their lack of phonon bands in the thermal infrared. However fabrication is substantially more complicated and they are generally more brittle and prone to damage than oxides. Therefore these are not the materials of choice in this study. To evaluate the influence of dispersion of the dielectric spacer on infrared performance, numerical calculations of the meta-osr design were evaluated using Lumerical FDTD. The resulting spectra from a metareflector with an insulator thickness of 1130 nm, a carrier concentration of 2x10 20 cm -3, a S3

4 square size of 950 nm and a 250 nm gap are shown in Fig. S2(c). The perturbations in the real part of the permittivity of the spacer materials lead to heavily distorted absorption spectra in the infrared, that in particular do not resemble the simple mode structure of conventional metamaterial resonators. The absorptance α in the visible and the emissivity ε in the infrared part of the spectrum is obtained using the definition [1] (Eq. S1) with R(λ) being the spectral reflectance, R(λ,T) the black body radiation at temperature T. This way, the reflectivity is spectrally weighted to give the best overlap with the irradiation from a black body. For α, we use the blackbody radiation for the surface temperature of the sun T=5777 K and set the integration limits to λ 1=300 nm and λ 2=2000nm. This includes 90.6% of the total black body energy, with 3.5% being left on the short-wavelength side and 5.9% on the long wavelength side. For ε, we set T=300 K, λ 1=2µm and λ 2=30µm, which includes 89.1% of the total black body energy, with 10-5 % not included on the short-wavelength side and 10.9% on the long wavelength side. Thus, our spectral range of interest spans two orders of magnitude. For all materials under study, the design parameters result in a region of near unity optical absorption in a spectral window around 10$\mu$m.The emissivity values are for SiO 2, for Al 2O 3, and for Ta 2O 5. SiO 2 shows narrow perturbation regions at wavelengths of 10 µm and 21 µm, and supports a particularly broad absorption range, which is attributed to the particular dispersion of the material resulting in an extension of the constructive interference condition. S1.3 Meta-surface plasmonic material Following the choice of back reflector and dielectric spacer, the choice of the meta-surface plasmonic material and its parameters is crucial for the device performance. AZO is a transparent conductive oxide whose optical properties largely depend on the carrier concentration. Its electrical permittivity can be described using a Drude-Lorentz model as (Eq. S2) Here, the plasma frequency ω p, the background permittivity ε b and the scattering constant Γ are material parameters from the Drude model and ω=2πc 0/λ is the optical frequency of the incoming light. Typically, these values are given by ε b =3.8 and Γ =2x10 14,Hz [2,3]. The plasma frequency depends on the carrier concentration N, the elementary charge e, the effective electron mass m* ~ 0.4m e and the permittivity of vacuum ε 0. Parameters f 1, ω 1 and Γ 1 are the amplitude, the spectral position and the damping of the Lorentz oscillator representing the material's UV bandgap at around 300 nm. Connected to the unique properties of AZO are the transition wavelength λ t and the skin depth δ (Eq. S3) S4

5 λ t is the epsilon-near-zero wavelength at which the real part of the permittivity changes sign. For wavelengths shorter than λ t, AZO behaves like a dielectric, while for wavelengths larger than λ t it behaves like a metal and supports plasmonic resonances. As can be seen from Eqs S2 and S3, the lower wavelength limit for the range in which plasmonic resonances are supported is antiproportional to the square root of the carrier concentration. δ is the skin depth of a plasmonic structure, representing the distance that the electromagnetic fields penetrate through the material. The skin depth should not be significantly larger than the thickness of the material, so that absorption of the meta-surface remains strong. Figure S2(d) shows the skin depth for transparent oxides with carrier concentrations from 1x10 20 cm -3 to 5x10 20 cm -3. An optimum transition wavelength and skin depth range is obtained for carrier densities in the range 2-3x10 20 cm -3. S1.4 Combined design optimization To illustrate the scaling surface plasmon resonance with geometry, we calculate the absorption cross sections of the AZO elements suspended in air for size parameters from 550nm to 1950nm, as illustrated by the cartoon in Figure S3a. The resulting spectra in Figure S3b show a red shift of the plasmon resonance with increasing size, associated with a reduction of the absorption factor σ abs/σ geo. The smaller particles have a too small bandwidth and are blueshifted with respect the thermal spectrum. When embedded in the meta-osr, the resonance of the AZO elements is modified by the presence of other materials with a strong dispersion of the dielectric function. Figure S3c shows the dispersion of the dielectric function of SiO 2, while Figure S3d compares the absorption cross-section of an AZO element of L=1350nm in the meta-osr with the same element embedded in air. We see that the strong dispersion of the surrounding dielectric results in strong modification of the resonance profile, as discussed in Figure 1b in the main text. Figure S3 (a) Illustration of simulated geometries of AZO meta-surface in air and on meta-osr device. (b) Absorption factor (absorption cross-section normalized to geometrical particle cross-section) of a single AZO square in the array versus size L, for meta-surface in air. (c) Dispersion of SiO 2 and (d) comparison of the absolute absorption cross sections of the square in air and on the meta-osr. S5

6 The combined effect of spacer dispersion and meta-osr design results in a broad absorption spectrum which can be further optimized by tuning the parameters of the AZO carrier density, spacer thickness and feature size. Figures S4(a) and (b) show color maps of IR spectral reflectivity versus meta-osr square size, for a choice of SiO 2 as the dielectric spacer and Al as the back reflector. While the lower carrier concentration gives broader resonances, the spectra for the higher carrier concentrations have a higher amplitude. The corresponding values of epsilon are plotted in Figs. S4c and d. The maximum emissivity values are 0.80 for 2x10 20 cm -3 and 0.78 for 3x10 20 cm -3, which are reached over a range of 150 nm of the insulator thicknesses and 400 nm of the square sizes, showing the tolerance of the optical response to dimensional imperfections. Figure S4(a,b) Spectral maps showing absorption (1-R) of the full meta-osr structure as a function of square size, using 1130nm of SiO 2 and AZO carrier concentration of 2x10 20 cm -3 (a) and 3x10 20 cm -3 (b). (c, d) Emissivity values ε for a 2Dsweep of square size and insulator thickness with AZO carrier concentrations of 2x10 20 cm -3 (c) and 3x10 20 cm -3 (d). S1.5 Angle dependence of the meta-osr performance As the meta-osr design so far has been optimized for perpendicular incidence, an important question is how the structure will perform for other angles of incidence (AOI). In particular, the angle-integrated ε will put a larger weight on the higher AOI of around 45 as these contribute to the majority of the solid angle. Higher AOI are less important because of the decrease of the total spectral radiance with increasing angle (Lambert s cosine law). Figure S5a shows the calculated absorption spectrum for the meta-osr with 950 nm feature size, 250 nm gap, 1200nm SiO 2 spacer and carrier density of 2x10 20 cm -3, for AOIs between 0 and 60. For the infrared spectral range, a shift of the meta-osr resonance toward longer wavelengths is observed with increasing AOI, which results in an increased epsilon. Thus, the meta-osr performance is improved with increasing AOI. This increase can be attributed to an increased S6

7 effective optical path length in the dielectric spacer as the AOI is increased, resulting in a red-shift of the effective λ/4 interference condition of the Salisbury screen. Resulting values for α and ε are shown in Figure S5c and Table S1, and indicate an improvement of the emissivity with angle, while simultaneously the UV-visible losses are slightly increased by around 0.05 in α. Similar results are obtained for the L=1350 nm meta-surface as shown in Figure S5b and S5c. The additional increase with angle is less strong for the larger feature sizes which were already optimized to have a maximum emissivity at zero angle. Figure S5 Calculated absorption versus angle of incidence of the meta-osr with square size of 950 nm and gap of 250 nm. AOI ( ) α ε, L=950 nm ε, L=1350 nm S7

8 Table S1. Values for α and ε extracted from the angle-dependent simulations S2. Atomic layer deposition of AZO The ALD AZO process is schematically shown in Figure S6 and is a combination of ZnO process (referred as ZnO cycle or DEZ cycle) using Diethylzinc (DEZ) and H2O precursors and Al2O3 process (referred as Al2O3 cycle or TMA cycle) using Trimethylaluminium (TMA) and H2O precursors. Both ZnO and Al2O3 processes were set at 175 C. By interleaving Al2O3 cycles with several ZnO cycles, referred as super cycle, the Al content can be tuned and accurately controlled in the deposited film. In the very beginning and ending part of the deposition, several DEZ cycles were used to form a pure ZnO on the very top and bottom of the formed AZO. In this work, we investigated AZO with Al cycle ratio in range of 0-5 % as shown in Table S2. This range of doping levels is commonly used in the literature as further Al content gives detrimental effect to film properties. The total DEZ cycle is set as 632 while the TMA cycle is adjusted to meet the desired cycle ratio and therefore the measured AZO thicknesses are very consistent. Figure S6 Schematic of atomic layer deposition process of Aluminium doped Zinc Oxide with different Al cycle ratios. S8

9 Designed Thickness TMA/(TMA+DEZ) Cycle Set ratio (nm) 0% 112 0% 632 DEZ 1% % 16 DEZ+ 6x(1 TMA+100 DEZ)+1 TMA+ 16 DEZ 2% % 16 DEZ+ 12x(1 TMA+50 DEZ)+1 TMA+ 16 DEZ 3% % 16 DEZ+ 20x(1 TMA+30 DEZ)+1 TMA+ 16 DEZ 4% % 16 DEZ+ 25x(1 TMA+24 DEZ)+1 TMA+ 16 DEZ 5% % 12 DEZ+ 32x(1 TMA+19 DEZ)+1 TMA+ 12 DEZ Table S2. Cycle sets for different Al-cycle ratios S3. Electrical and optical characterization S3.1 Hall Effect Measurement The deposited films were initially characterised by Hall Effect Measurement and the extracted carrier concentration, resistivity and mobility are shown in Figure S7. The sheet resistance decreases from 286 to 74 Ohm/Sqr with the increase of Al cycle ratio from 0-5%. The carrier concentration increases from cm -3 for pure ZnO to cm -3 with the increase of Al cycle ratio, showing that the carrier concentration can be tuned through the Al cycle ratio control. This well match the literature that the Al can be effective dopant to increase carrier concentration. These results show that we can grow AZO films in a wide range of carrier concentrations to meet the design requirements x Sheet Resistance ( Ω/Sqr) Carrier concentration (cm -3 ) TMA /(TMA+DEZ) % TMA /(TMA+DEZ) % Figure S7 Sheet resistance and carrier concentration as a function of Al cycle ratio from Hall Effect Measurement on AZO films with different Al cycle ratio. S3.2 Ellipsometry To optically characterize the films, Ellipsometry was used to investigate the UV-Vis range optical properties. The dielectric function was modelled using a Drude-Lorentz model with a Drude dispersion describing the infrared response and Lorentz oscillators to model the UV absorption. Figure S8 shows S9

10 the real and imaginary permittivities fitted from Ellipsometry measurement. For all AZO films, an absorption in wavelength below 400 nm is seen and the absorption range shifts blue with the increase of Al cycle ratio, indicating that the 5% AZO has the highest carrier concentration and would be preferred to achieve a lower α value for its lower UV absorption. Figure S9 shows a carrier comparison between Hall Effect, Ellipsometry and FTIR measurement. The carrier concentrations by Ellipsometry and FTIR are found to be consistent, whilst that by Hall Effect measurement is more than one order higher. This discrepancy matches other results in the literature [4]. As for meta-osr application, the optical response is our main concern and therefore the consistent extracted carrier concentrations from Ellipsometry and FTIR gives us a guideline for the behavior of the AZO films, particularly beneficial for transferring the optimised design from simulation to experiments. In particular the cross-over wavelength from dielectric to metallic response matches well with the numerical calculations based on the Drude model presented in the main text. To illustrate the metallicity of the AZO in the spectral range under study, we present in Figure S10 the calculated dielectric function using the material parameters extracted from ellipsometry and for three free carrier densities from 1 3x10 20 cm -3. For the AZO under study, the material is metallic in the infrared range, with negative values of the real permittivity of between -10 and -30 in the thermal infrared from 5 to 20 µm 7 Real permitivity ( ε 1 ) AZO 0% AZO 1% AZO 2% AZO 3% AZO 4% AZO 5% Wavelenth (nm) 2.5 Imaginary permitivity ( ε 2 ) AZO 0% AZO 1% AZO 2% AZO 3% AZO 4% AZO 5% Wavelenth (nm) Figure S8 (Left) Real and (Right) Imaginary permittivities fitted from Ellipsometry measurements on fabricated AZO films with different Al cycle ratio. S10

11 10 22 No anneal Carrier concentration (cm -3 ) Extracted from Ellipsometry Extracted from FTIR Hall effect measurement Al cycle ratio (%) Figure S9 Carrier concentration comparison using different measurement techniques, Hall Effect measurement, Ellipsometry and FTIR. Figure S10 Calculated Drude mode permittivity in the infrared range, using parameters for AZO obtained from ellipsometry Figure S8 and carrier densities N e of 1x10 20 cm -3, 2x10 20 cm -3 and 3x10 20 cm -3. S3.3 UV-Visible-NIR measurements Spectroscopic investigation of large-area samples (>1 cm 2 ) was done using a commercial large-area reflectivity setup (Bentham PVE) allowing total reflectance measurements from 200nm 2500nm. For investigation of small-area (100x100µm 2 ) arrays obtained using e-beam lithography, we used a home-built setup illustrated in Figure S11. A broadband supercontinuum light source spanning a wavelength range 400nm-2,400nm was used as the illumination source and was focused onto the sample at an angle of 8 with the normal. The supercontinuum laser provides a collimated beam of light that can be focused down to an illumination spot area within the individual arrays. S11

12 Figure S11 UV-visible measurement setup for small-area arrays. (Left) arrangement for aligning laser beam onto arrays using a 10x microscope objective. (Right) Microscope is replaced by integrating sphere to measure total reflectivity. Input and specular ports are at 8 angles with the sample surface. In order to accurately position the arrays in the incident beam, a microscope with a magnification of 10x was positioned at an angle to the sample, allowing inspection of the sample while leaving the illumination path unobstructed (Figure S11(left)). This configuration allowed precise alignment of the illumination spot onto the arrays. Following alignment, the microscope was replaced by an integrating sphere without affecting the alignment, as illustrated in Figure S11(right). Spectroscopic measurements of the total reflectance were done by scanning the wavelength of the supercontinuum source using a double-prism monochromator subtractive mode. The total reflectance was measured using a combination of silicon and InGaAs photodiodes with lock-in detection. To extrapolate the performance of the small area devices further into the UV range, we used the large-area measurements of the thin films as an estimate for the material performance. It was found that thin-film performance and metamaterial performance in the measured range are very comparable, justifying this assumption. S3.4 FTIR microscopy Measurements of reflectance in the range 2,000nm-30,000nm wavelength were taken using an FTIR setup consisting of a Thermo-Nicolet Nexus 670 with a Continuum microscope. The instrument is equipped with a far-ir light source, KBr beam splitter and nitrogen-cooled DTGS detector. The microscope makes use of a reflective cassegrain objective with a numerical aperture of 0.2. Reflection from individual arrays was obtained by selecting a square aperture of 100x100 µm 2 using a mechanical slit assembly in the microscope path S4 Meta-OSR fabrication and characterization S4.1 Sample design and fabrication The square meta-osr structures are designed with square size varying from 750 nm to 1350 nm. The spacings between the neighboring structures are set to be 250 and 150 nm from the simulation design. Then an array of this structure is created periodically to form a 120 µm 120 µm square S12

13 array and several arrays form a chip layout. Arrays in this chip layouts are detailed in Table S3. The arrays are named by their structure and dimensions. For example, S950S150 stands for structure of 950 nm square and a 150 nm spacing. Design Cell Name Square Size (nm) Gap (nm) Period (nm) Array size (nm) Number of units / direction 1 S750S , S950S , S1150S , S1350S , S750S , S950S , S1150S , S1350S , Table S3 Overview of design parameters. Square arrays have a total period of (square+gap).. Figure S12 shows the optical micrographs of the fabricated AZO meta-reflectors after the NMP rinse but before ultra-sonic bath. The patterned AZO is seen in dark color (Figure S12 (Left)). This indicates the resist is not fully removed after IBE etch and NMP rinse. A higher magnification image (Figure S12 (Right)) shows that AZO is patterned into an array of squares. However, squares can be seen in two different colors, indicating that resist is only partially removed and resist residues left on patterned AZO. It needs to be mentioned that the resist residue is hard to be identified by SEM as the geometrical structures are the same. This resist residue explains the darkness of the patterned AZO. To remove the resist residue, the samples were treated in ultra-sonic bath with heated NMP for 5 mins. Figure S13 (Left) shows an optical micrograph of the same chip after NMP bath. The patterned structures are seen in lighter colors than Figure S11 (Left), indicating a significantly less visible absorption. It should be mentioned that the survival of the whole patterned structure shows excellent adhesions between all layers and this is very important for the final applications. The higher magnification image (Figure S13 (Right)) shows all squares are in a uniform color, indicating the resist residues are removed. S13

14 Figure S12 (Left) Optical micrographs of the fabricated AZO meta-reflector chip after IBE etch and NMP rinse and (Right) high magnification image of square array. Figure S13 (Left) Optical micrographs of the fabricated AZO meta-reflector chip after ultra-sonic NMP stripe and (Right) high magnification image of square array. S4.2 SEM images Figure S14 shows plan SEM micrographs for the square meta-osr structures. For all patterns, the structures are in good agreements with the designed structures. This unambiguously demonstrates that the developed fabrication process can achieve the proposed meta-reflectors. S14

15 S750S150 S950S150 S1150S150 S1350S150 S750S250 S950S250 S1150S250 S1350S250 Figure S14 Plan SEM micrographs of designed 8 different pattern structures of AZO meta-reflectors. S15

16 S4.3 Optical characterization The complete experimental dataset corresponding to the results of Figure 3 in the main text is presented in Figure S15a and b, corresponding to gaps of 150nm and 250nm, respectively. Solid lines are the experimental data, dashed lines represent numerical model calculations of the meta-osr devices using Lumerical FDTD. Figure S15 Experimental (solid) and simulated (dashed) spectra of AZO meta-reflectors with different feature sizes for 150 nm gap (a) and 250 nm gap (b). The spectra in UV/Vis/NIR and Infrared ranges were measured using Quantum emission and FTIR systems, respectively. S5 UV reflector design and characterisation The UV reflector (UVR) used in this work is a multilayer stack consisting of 17 layers of alternating Ta 2O 5 and SiO 2. Figure S16 shows the full stack and the corresponding reflection and transmission spectra calculated using Essential MacLeod software for the model stack on a transparent substrate. The UVR coating results in strong reflectivity in the wavelength range of 250 nm 450 nm. The primary aim of this coating was to reduce the high UV reflectivity of the AZO layer in the UV spectral range. The performance of the small-area meta-osr devices in the presence of the UVR coating was tested using the FTIR microscope. Figure S17 shows the resulting spectra of the meta-osrs (solid lines), together with that of the thin-film planar device reported in Figure 5a of the main text (dashed line), all in the presence of the UVR coating. Compared to the thin-film device, the meta-osrs show much higher absorption in the spectral range µm. However the performance at longer wavelength is somewhat reduced to below that of the planar thin-film. Still, the emissivity for the structured meta- OSRs exceed the thin-film planar device, with values of ε up to S16

17 The role of the UVR coating on the total reflectance in the UV-visible-NIR range was explored for the thin-film reflector as shown in Figure 5.5a of the main text. For the small-area devices, total reflectance measurements are shown in Figure S18. In the accessible experimental range, very little difference is found in the reflectance with UVR as compared to the case of the meta-osr without the UVR coating. This is consistent with the results from Figure 5.5a showing that apart from the functionality in the UV range, the UVR coating does have very little impact on the reflectance in the visible and NIR parts of the spectrum. Figure S16 (left) Schematic of multilayer stack for UV-reflector labelled UVR3. (right) Reflectivity (dash) and transmission (solid) of UVR3 coating. Figure S17 Infrared absorption (1-R) for meta-osrs covered with UVR coating, for gaps of 150nm (left) and 250nm (right). S17

18 Figure S18 UV-Visible absorption (1-R) for meta-osrs covered with UVR coating, for gaps of 150nm (left) and 250nm (right). References [1] R. A. Breuch and K. N. Marshall, "Optical solar reflector - A highly stable, low alpha sub S/epsilon spacecraft thermal control surface," Journal of Spacecraft and Rockets 5, (1968) [2] Losego, M. D.; Efremenko, A. Y.; Rhodes, C. L.; Cerruti, M. G.; Franzen, S.; Maria, J.-P. Journal of Applied Physics 2009, 106 [3] Liu, X.; Park, J.; Kang, J.-H.; Yuan, H.; Cui, Y.; Hwang, H. Y.; Brongersma, M. L. Applied Physics Letters 2014, 105. [4] A. K. Pradhan, R. M. Mundle, Kevin Santiago, J. R. Skuza, Bo Xiao, K. D. Song, M. Bahoura, Ramez Cheaito & Patrick E. Hopkins, Extreme tunability in aluminum doped Zinc Oxide plasmonic materials for near-infrared applications, Scientific Reports 4, 6415 (2014) S18