1. Introduction

Recent work by M. A. Kats [1] demonstrated a thin-film absorber that operates in the long-wave infrared (LWIR) and is tunable using a phase change material. That seminal work spurred investigations on thin film infrared absorbers by illustrating how low cost absorbers may be developed as an alternative to absorbers that rely on fabrication intensive structured surfaces such as that found in [2]. Two thin film absorber investigations [3,4] took advantage of the doping tunable properties of silicon which may be readily purchased from a vendor and simply etched to achieve LWIR perfect absorption. The Kats [1] and our previous thin film absorber [3] took advantage of specific material properties although both relied on a sapphire substrate and more specifically the large absorption peak that appears in the 11-15 μm range. The Streyer [4] thin film absorber design investigated implementation of a high index Ge thin top layer deposited on doped-Si with an SiO₂ substrate. The high index Ge is a novel idea but is inherently limited to wavelengths above ~2 μm due to its band edge. This layer structure does seem to add complications since the correct Ge thickness now must be chosen to match the doped-Si optical constants and thickness. A more recent work [5] demonstrated a thin film perfect absorber in the visible based on a silicon on aluminum structure. The silicon is thin enough to be optically transparent although multiple reflections with the silicon cavity enable the prefect absorption. Related to that work, we investigate a new thin film absorber structure using a doping-tunable material on a metallic “substrate”. A metallic substrate with high k (imaginary part of index) was sought and chosen based after our previous work [3] indicated that maximizing this value was key to such layered structures. Employing the metallic substrate also enables broader absorption tunability when compared with the aforementioned thin film absorbers that utilized sapphire or SiO₂ as a substrate. One other advantage when using a metallic “substrate” approach will be that a thin metal can be deposited on any substrate as the basis for such thin film absorbers.

In this work we specifically investigate wavelengths in the short-wave infrared (SWIR) and mid-wave infrared (MIR) for applications that will be compatible with telecommunications and infrared detection and imaging systems that fall within the atmospheric transparency window. Our previous work has shown that the thin film perfect absorbers can be designed utilizing doped-Si with its bulk plasmon wavelength, or the wavelength where the real part of the permittivity crosses zero, being just longer than the desired absorber operation wavelength [3]. Thus far, doped-Si results have focused on the LWIR and longer-MIR wavelengths with silicon because of the inability to dope highly enough to push the plasmon wavelength into shorter MIR wavelengths [4, 6,7] with most traditional doped semiconductors giving the same result. The work presented here, while not plasmonic in nature, will take advantage of a tailorable bulk plasmon wavelength and associated material properties which the aforementioned works helped to pave the way for.

The solution for thin film absorbers that reach down to SWIR may lie in doped-transparent conductive oxides. This work will utilize gallium doped zinc oxide (GZO) as the tunable semiconductor layer. GZO has been of interest in recent years for applications to plasmonics [8,9] due to its ability to achieve plasma wavelengths in the telecommunication regime. Those works demonstrated carrier concentrations of ~1.1E21 cm⁻³ which translates into a bulk plasmon wavelength of ~1.4 μm. A more recent material focused investigation [9] predicts possible plasmon wavelengths as low as 1.0 μm which ultimately may be important for pushing the same structures to even shorter wavelengths in the future. It is noted in [9,10] that a large range of GZO mobility and carrier concentration may be obtained based on deposition methods and conditions. Also, the post-deposition carrier concentration can be decreased via air annealing [9], or even slightly increased via annealing face down on Zn foil while in forming gas [11]! These anneal methods provide simple methods to tune the plasmon wavelength. A recent investigation of thin film absorbers utilized a zinc oxide relative, indium tin oxide, for telecommunication perfect absorption albeit at large angles of incidence and for a polarization sensitive case [12] which may not be as practical for detection or other applications. GZO also appears to be the better candidate for such structures due to the rising cost of indium. Based on GZO’s inherent abilities to tune the material parameters, and by association the attainable plasmon wavelengths in the SWIR and longer wavelengths, and low relatively low material cost, GZO is a superior candidate for such thin film absorbers.

In this work, we investigate gallium-doped ZnO thin film absorbers on a thin metal “substrate”. After selecting an appropriate metal, we will theoretically investigate optical parameters and thicknesses for GZO that enable perfect absorption at particular SWIR and MIR wavelengths. We will also investigate theoretically the effects of the metal thickness. A representative GZO thin film absorber structure is then deposited and characterized via ellipsometry, infrared spectroscopy, and scanning electron microscopy. After the initial characterization, the sample underwent etching to demonstrate tuning of absorption. Those experimental results are then compared with theory. Finally, material parameters and thicknesses are determined that enable greater than 99% absorption for target SWIR and MIR wavelengths. These thin film, polarization insensitive, and perfectly absorbing structures, which are relatively simple and fast to fabricate, will have potential applications as reflection band filters, infrared scene generators, photodetectors and photovoltaics.

2. Structure design for thin film absorber

The simple thin film perfect absorber design discussed here relies on a Ga-doped zinc oxide (GZO) as the 1st layer and a metal as the 2nd layer. Figure 1 depicts the layered structure under investigation. The thin film perfect absorber structure itself is a result of critical coupling between the incident light and a tailorable cavity resonance [1, 5]. More specifically, at the absorption resonance, the partially reflected waves were confirmed by calculation to sum to zero in the highly complex fashion presented as Fig. 1D in [1]. GZO was specifically chosen as the cavity material, layer 1, based on the tuning mechanisms discussed in the introduction. Besides tuning the optical parameters, which will be discussed more in the following section, tuning of the thickness of the film is possible which will be illustrated later.

 

Fig. 1 ZnO on metal structure. Light is normally incident with multiple reflections in the cavity (GZO film) pictorially represented using spatial separation.

Download Full Size | PDF

The goal of this work was to investigate the thin film absorber idea for SWIR and MIR wavelengths. The thin film absorber layer structure is aided by the large k (imaginary part of the refractive index) in layer 2. The large k ensures that light is confined to the cavity and critical coupling is possible. Shown in Fig. 2 are k values for the metals Ag, Au, and Cu [13]. Metals are specifically highlight where the k value is not very wavelength-dependent, as opposed to SiO2 or sapphire which have been used previously [1,3]. This allows a greater tuning range for the structure. All metals show a similar decreasing k trend with increasing wavelength. Of the three metals shown, copper was discounted due to potential oxidation that might occur between deposition steps. All three however could potentially be useful but silver was chosen here due to availability. We also investigate the possibility of an all GZO based system where the “metallic” layer 2 would be very highly conductive material utilizing a doping concentration N = 10²¹ cm⁻³ and mobility μ = 45 cm²/Vs which is the maximum theoretical for that N [14]. The k values for GZO shown Fig. 2 are calculated from the Drude formulation described in the next section. k for GZO is as minimum a factor of 3 smaller than the metals with increasing wavelengths showing an increasing separation. It is noted that a doping concentration near ~2x10²¹ cm⁻³ may be possible for GZO but at a reduced mobility of 35 cm²/Vs. This results in a marginal increase in k although achieving such mobility will be dependent on difficult processing. As will be discussed in section 4, it may be possible to use this highly doped GZO as a substrate but at a cost of thickness. For the wavelength ranges discussed here of 1.5-4.5 μm, silver will be focused on for the metallic layer from this point forward.

 

Fig. 2 Wavelength dependent k for potential “metallic” substrates [13]

Download Full Size | PDF

3. Analytical absorption properties of GZO films on silver

Before discussing analytical results, we must first discuss the calculations methods. Complex refractive indices are used for all layers with the exception of the final silicon substrate having a refractive index of 3.4. For the calculations, wavelength-dependent complex refractive indices for silver are from [13]. Most importantly we use a Drude formulation for the Ga-doped ZnO layer. The complex refractive indices are determined by

The absorption in the layer structure (Fig. 1) is found analytically via standard Fresnel Equations for conventional thin-film wave propagation in absorbing media [3, 15]. For this work, the light is un-polarized and is incident normal to the surface. We first calculate the transmission, T, and Reflection, R, for the layer structure and then find the absorption by

Figure 3 presents’ contour plots of absorption percentage as a function of μ, and N. Specifically illustrated here at two wavelengths, 1.5 and 3.5 μm. The wavelengths are column specific with the ZnO film thickness increasing trending downwards. In this case, the silver film is 50 nm thick and the silicon substrate is 400 μm. The mobility min and max are greyed out and labeled as “forbidden” regions. Practically speaking, GZO films with μ values far from the forbidden regions are likely easier to fabricate although all non-grey regions remain possible. Drawn on each of the contour plots are the conditions for 90 and 95% absorption as solid black lines for reference.

 

Fig. 3 Absorption contours calculated for Ga-doped ZnO films on Ag and a silicon substrate. (left) and (right) are calculated for 1.5 and 3.5 μm respectively. The ZnO films increase in thickness trending downwards. “Forbidden” regions for Ga-doped ZnO are colored in grey. Black contour lines are included that illustrate the 90 and 95% absorption regions.

Download Full Size | PDF

Examining the wavelength case of 1.5 μm we notice that at 400 nm thick GZO, greater than 99% absorption can be achieved with a doping of ~4.5E20 cm⁻³ although with a μ of ~15 cm²/Vs which borders the lower forbidden region. Upon further examination, increasing the thickness to 600 nm pushes the perfectly absorbing region to higher mobility (μ = 24 cm²/Vs). Interestingly, as the thickness is increased, a 2nd highly absorbing region enters the contour which is achievable with a lower doping. This will be revisited later in conjunction with experimental work. The wavelength of 3.5 μm follows a similar trend although with thicker films. For that wavelength, the 1000 nm thick GZO film with μ = 30 cm²/Vs and N = 8E19 cm⁻³ enables near perfect absorption. For that wavelength, the upper forbidden region is shifted to higher mobility values and exists off the top of each graph. The y-axis is not changed however for better comparison with the 1.5 μm plots.

Utilizing the optical parameters we determined from Fig. 3, we now investigate the absorption spectrum at the wavelengths of 1.5 and 3.5 μm, shown in Fig. 4. In each case greater than 99% absorption is observed at the respective target wavelength. On the left side of and up to the absorption peak are Fabry-Perot fringes due to the transparent ZnO. We confirmed that in each case the zero crossing of the permittivity is on the right side of the peak absorption wavelength. The zero crossing is 1.7 and 4.3 μm for the (left) and (right) of Fig. 4 respectively. It is noted that increasing the film thickness (not shown) results in a reduction of the absorption peak with an eventual increasing absorption peak at slightly shorter wavelengths. Continuing to increase the GZO film thickness, while not in the interest of this thin-film paper, shifts higher order Fabry-Perot fringes into the region of high absorption. As the film approaches a bulk like material, the Fabry-Perot fringes due to the GZO diminish and one begins to observe a single unified absorption peak near the bulk plasmon wavelength. This occurs at thicknesses no smaller than double the original thicknesses shown in Fig. 4. In that thick GZO case, the absorption also decreases due to light not reaching the silver and hence not taking advantage of its large k.

 

Fig. 4 Wavelength-dependent absorption spectra for layered structures with optimized parameters for ~1.5 (left) and ~3.5 μm (right).

Download Full Size | PDF

One important question is that of the effect of silver thickness on the nearly perfect absorption. To investigate this we first decrease the silver film thickness for purposes of comparison with Fig. 3, or more specifically we determine how this decrease affects the peak absorption in the μ and N parameter space at 1.5 μm. Beginning with 50 nm, we decreased the silver to thicknesses. At 30 nm, the 90 and 95% regions show only slight compression inwards indicating the highly absorbing area is minimally changed. At 20 nm, a more noticeable decrease in the size of the absorption region begins at 20 nm followed by a dramatic reduction at even smaller silver thicknesses. By the time that 5 and 0 nm are reached, the absorption does not surpass 95 and 90% respectively. Using the measured k for silver, we approximate the optical skin depth to be ~25 nm which corresponds well with the thickness threshold above for near perfect absorption. One other noticeable change besides the absorption decrease is the increase in fast Fabry-Perot fringes due to the silicon substrate. This is evidence that the metal layer is imperative for critically coupling to the cavity. The implication of this thick metal film approach is that the GZO-Ag bi-layer can be deposited upon any substrate, such as plastic, and perfect absorption will nevertheless be attained. Of course it is more costly and more material intensive to deposit a thick metal film, and thus in practice the less expensive thin film approach would usually be used.

4. Experimental characterization of a GZO on Ag layered structure

For fabrication of the films, first a 50 nm silver film was e-beam deposited on top of a clean, highly resistive 3” Si wafer. The Ga-doped ZnO films were then deposited onto the silver coated silicon via pulsed laser deposition (PLD). A GZO target was used that had 0.5% Ga to 99.5% Zn. Based on previous measurements with samples made from the same target, the achieved nominal N and μ are ~2.2E20 cm⁻³ and 35 cm²/Vs respectively. Scanning electron microscopy (SEM) was used to characterize the film thickness and overall film quality (see Fig. 5). The film is visibly rough with the top down surface feature sizes being ~100 nm and the film thickness ranging from ~1.3-1.45 μm. The roughness is more clearly observed in Fig. 5 (lower) from the SEM image captured at a 45° angle. In both images, the thin silver film is visible in between the GZO and Si substrate. J. A. Wollam IR-VASE and V-VASE ellipsometers were used to measure the optical constants for both the silver and GZO. The thicknesses measured via the SEM were used in the fittings to determine the complex permittivity in the range of 1 – 5 μm. It is noted that the zero crossing of the real part of the permittivity, or again where the film becomes “metallic”, is at 2.4 μm. This complex permittivity will be used in subsequent thin film calculations when fitting to measured data except where noted otherwise.

 

Fig. 5 SEM images of the deposited film layer. (upper) shows the cross section while (lower) shows the sample held at a 45° angle to visualize the roughness.

Download Full Size | PDF

Reflection was completed on the deposited layer structure using a Bruker Vertex 80v FTIR with coupled Hyperion microscope. The measured reflection data was normally incident but may have a small a (few degrees) light cone associated with it. A gold mirror was used as a reference with the reference being scaled by 97% to account for absorption losses in the gold at the MIR wavelengths of interest. This allows the reflection percentage values measured to be as accurate as possible when compared to incident intensity. The experimental transmission was confirmed to be zero, which is as expected with the 50 nm thick silver layer. The absorption thus is calculated using Eq. (5) using only R.

Figure 6 shows the measured absorption data and a calculated curve using the measured permittivities. The film thickness was used as a fit manual parameter. The fitted thickness of 1340 nm gave a reasonably good fit to the measured absorption peak and was in agreement with the SEM image. The long wavelength side of the curves does not fit well likely due to the roughness on the surface which impacted both the FTIR and ellipsometry measurements. This however is not important since the highly absorbing region is the focus. The largest absorption peak observed experimentally is 98% at a wavelength of 1.9 μm. A broad shoulder on the high wavelength side though indicates another peak is present at 2.1 μm with an absorption value of ~93.5%.

 

Fig. 6 Empirical absorption data compared with a calculation using measured permittivity values.

Download Full Size | PDF

In investigating the difference between the 98% absorption observed and the near perfect absorption possible, we calculated the effects of decreasing ZnO film thickness using the techniques shown earlier. These calculations showed the 1.9 μm absorption peak decreasing and the 2.1 μm peak increasing to near perfect absorption. To corroborate the calculations, the sample underwent successive wet etching. Wet etching was chosen as a controlled demonstration technique as opposed to dry etching due to known heating during such processes depleting carriers in GZO [16]. Anneal effects during etching would complicate the analysis and would require much further material characterization which is out of the scope of this paper. The sample was etched in 1:900 HCl: H₂O with 30 second etch times expected to etch 100 nm [17]. The etch times were initially 30 second intervals with the last two etch steps being 15 seconds each. FTIR measurements were completed at each etch step.

Figure 7 (upper) shows the measured absorption data corresponding to specific cumulative etch times on the same sample. Figure 7 (lower) shows calculated curves using thicknesses that best qualitatively match the experimental absorption peak and lineshape. The overall trends are in qualitative agreement. As the sample undergoes more etching, i.e. decreasing thickness, the absorption peak near 1.9 μm decreases in strength until not apparent anymore which occurs near the 1 μm thick level. Simultaneously, the experimental peak near 2.1 μm grows in absorption strength to an observed maximum of 99%. Subsequent etching then reduces this absorption peak along with the overall absorption which is trending to that of bulk silver. The absorption peak then observed at 2.1 μm is the peak observable with the thinnest film thickness for the deposited GZO used. The shown 99% percent may not have been the maximum experimental absorption possible with this deposited film. A simulated curve with a thickness of 900 nm is specifically shown in Fig. 7 (lower) to show that near perfect (~100%) absorption is possible. We ran a simple Lumerical FDTD calculation to confirm that with a 900 nm thickness, the GZO film itself absorbs 99.5% of the incident light. With the large etching steps taken, it is likely that we etched through a thickness that would have given an absorption value larger than 99%.

 

Fig. 7 (upper) Experimental absorption measured as a function of etch time. (lower) Calculated absorption using measured permittivities for varying ZnO film thicknesses.

Download Full Size | PDF

We characterized the film post-etching and observed that the film now has a very rough surface with structures that modulate almost 250 nm in height which is now roughly 1.4x the base height of 600 nm. The increasing roughness during etching may be responsible for the increasing discrepancy between the line shape of the experimental and calculated curves in Fig. 7. An etch rate of ~370 nm / minute was determined from the pre- and post-etching SEM images and matches well with calculations (Fig. 7), with both however differing notably from the etch recipe. Lumerical FDTD simulations were used here to analyze the effects of the etched surface roughness. In this case the power absorbed was observed with a monitor over the absorption region as opposed to utilizing Eq. (5) with T and R. The implementation of a rough surface similar to that seen in the SEM for the case of 900 nm slightly decreased the 99.9% absorption (by~1%) and caused a broadening of the long-wavelength side of the absorption similar to that seen in the Fig. 6 and in Fig. 7 (upper) at later etch steps. In these aspects, the simulated roughness matches the 90s etch sample measurement fairly well. The fact that we use an absorption monitor however means that if any light is scattered some of it is also still being transmitted, or more likely focused, into the absorbing region. Ultimately, however this means the experimental peaks observed could be reduced from a higher potential perfect absorption, increasingly so beyond the 90s etch time, due to the surface roughness. This point however does not detract from the overall goal of the perfect thin film absorber as the experimental etch measurements and analytical thickness based calculations qualitatively agree and ultimately such structures would have smooth surfaces. Investigations of better etch techniques, non-heated dry etching techniques that do not affect the MIR optical properties, will be helpful in long term development of such GZO based layered structures. The experimental and matched calculation results shown here provide proof of principal that a perfectly absorber is achievable through this simple thin film layered structure.

5. Optimized absorption parameters for target wavelengths

We now investigate optimization parameters for four target wavelengths. The four target wavelengths were selected to highlight parameters that are relevant for particular applications. The 1.5 and 2.0 μm wavelengths are relevant for current and future telecommunication systems, while 3.5 and 4.7 μm are relevant for detection systems that operate in the atmospheric transparency window. For each of the four wavelengths, we investigate three thicknesses, and GZO μ and N which enable 99% absorption. The thicknesses for each wavelength were specifically chosen to highlight three μ regions that enable 99% absorption: mobility’s at least 10% higher than the minimum μ, at the center of the μ range, and at least 10% lower than the maximum μ.. In principal the 99% absorption possibilities will likely extend beyond 5 μm in the long-wave infrared. As noted earlier, shorter wavelengths may also be possible with utilizing very highly doped GZO that has bulk plasmon wavelengths of ~1 μm.

A grid of μ and N contour plots, like those seen in Fig. 2, were calculated for the selected λ and thicknesses. From each of these a 99% absorption ring was observed and the ranges of corresponding μ and N values were extracted and are shown in Table 1. For each of the λ cases, the middle thickness value shown will likely be the most practical deposition value although the upper and lower thicknesses are likely obtainable. Some may be skeptical of the increasing μ required for the wavelength of 4.7 μm. As pointed out earlier, a large range of μ and N may be obtained [9,10] based on varying deposition methods and conditions. Specifically relevant to 4.7 μm is the case of [18] where GZO was sputtered to achieve a sharply varying μ that peaks at ~70 cm²/Vs with N ~0.3-1.3 x 10E20 cm⁻³. With fine tuning, it is likely that such a sample can be accomplished through the means mentioned earlier (slight variation in fabrication technique, annealing either in air or forming gas, or thickness etching). Implementation of these means to improve absorption will require knowledge of where your parameters (μ, N, d₁) lie with respect to the potential absorption maximum. Table 1 is presented here as a rule of thumb guide to obtain near perfect absorption using the GZO/Ag thin film layered system.

Table 1. Criteria for 99% absorption in the GZO/Ag layer structure for four target MIR wavelengths. The GZO thickness has 3 values for each wavelength while Ag is 50 nm thick. For the specific λ and thickness, the range of N and μ give the range of each that allows 99% absorption.

View Table

6. Summary

Presented in this work are designs and experimental demonstrations for the SWIR and MIR pertaining to a GZO/Ag thin film perfect absorber that operates based on critical coupling between incident light and a cavity resonance. All calculations and experiments were completed for un-polarized lights, so all results are relevant for polarization insensitive absorbers. First shown were calculations that describe the method to determine material parameters necessary to achieve near perfect absorption using with specific regard to GZO mobility and carrier concentration, as well as both GZO and Ag thickness. A fabricated GZO/Ag structure was shown to be able to achieve at least 99% absorption at a wavelength of 2.1 μm. Shown also were specific designs for achieving greater than 99% absorption at target SWIR and MIR wavelengths. Notably, a 99% absorber intended for operation at 1.5 μm only requires a moderate carrier concentration of ~5e20 cm⁻³. It is likely that with even higher doping, which is possible with GZO, operational wavelengths for this system could be pushed to 1 μm. The other wavelengths investigated indicate that perfect absorbers based on this structure can be fabricated easily for virtually any wavelength in the span of 1.5-4.7 μm. Also likely is that this structure can be pushed to operation wavelengths greater than this range. To the best of our knowledge, no other 99% thin film absorbers have been demonstrated at these SWIR and MIR wavelengths. In addition this work is the best of our knowledge the first to leverage doping tunable ZnO, or any doped conductive transparent oxide, for such a polarization insensitive thin film absorber. GZO has advantages such as (1) direct tuning of carrier concentration and mobility based on deposition methods and conditions, and (2) the extra and relatively simple methods of air annealing and annealing in forming gas to fine tune optical constants as needed. Those techniques combined with the ability to etch the GZO film offer an additional tuning “knob”, to reach perfect absorption. This shows the benefit of the presented thin film absorber. Our simple and low cost perfect absorber system has potential applications in reflection band filters, infrared scene generators, photodetectors and photovoltaics.