1. Introduction

Since its first experimental demonstration by Landy et al. [1], the metamaterial absorber has undergone a rapid and prosperous development, finding a wide variety of practical applications including selective thermal emitters [2], wavelength-tunable microbolometers [3] and refractive index sensing [4]. The typical structure of the metamaterial absorber is a three-layered architecture, consisting of arrays of patterned metallic sub-wavelength structure on top of a dielectric spacer, backed by a thick metallic ground plane. The top periodic structure, referred to as an electric ring resonator (ERR), is responsible for the electric response of the metamaterial absorber. The metallic ground plane is thicker than the penetration depth of the incident wave to eliminate any transmission. The coupling of two metallic layers and the dielectric spacer determines its magnetic response [5]. Therefore, by altering the geometry of the ERR and changing the thickness of the dielectric layer, the effective permittivity ε and permeability μ can be tuned independently, resulting in an impedance match to the free space, and thus perfect absorption of the incident wave at certain frequencies [6]. By manipulating the electric and magnetic properties, the operation of metamaterial absorbers has been verified from microwave range [1], through terahertz [7], infrared [5] and into the visible [8].

In many applications such as solar energy harvesting and bolometers, broadband absorption is required. Since the mechanism of metamaterial absorbers is based on the strong electromagnetic resonance in the periodic structure, the bandwidth of this resonant absorption is narrow by nature [9]. In 2010, Gu et al. used lumped resistance elements embedded in the metamaterial structure to lower the Q factor [10], but this method is only feasible at low frequencies such as the gigahertz region. As an effective solution to realize broadband absorption, current state of art approaches utilizes the multi-resonance metamaterial absorber, which contains more than one resonant structure in each unit cell. Some designs employ nested elements, such as concentric square rings and achieve double and even triple band absorption [11, 12], but these bands are not close enough to merge to a broader band due to the limitation of element size in the nested configuration. Multilayer structures were demonstrated to expand the bandwidth greatly in microwave [13] and terahertz [14] regime, but such structures are hard to scale down for applications in the infrared range mainly because of the fabrication difficulties, including lithography and alignment between neighbored layers. Numerical studies have demonstrated the multiplexed configuration as a convenient choice to realize multi-resonance and broadband absorption [15]. Experimental works have verified this for gigahertz frequency [16] and more recently for the terahertz range by using three resonating structures of two different sizes in one unit cell [17]. As for the infrared regime, the square resonator is specially favored due to its ease of fabrication, and broadband absorbers were experimentally tested using two [18, 19] or even four [20] multiplexed square patches of different side length.

In this paper, we develop broadband metamaterial absorbers using multiplexed cross resonators in mid-infrared regime. Samples containing a single cross, two multiplexed crosses and four multiplexed crosses are fabricated and tested, and the multiplexed configurations show obvious broadening of the bandwidth by both simulation and measurement results.

2. Design and simulation

Three absorbers, referred as Sample A, B and C, were designed with one, two and four cross resonators located in each unit cell respectively. The schematic diagrams of the three unit cells are shown in Fig. 1. All the samples employ the typical metal/dielectric spacer/metal structure with 100 nm gold ground plane at the bottom, 190 nm SiO₂ dielectric layer in the middle and periodic gold patterns on top. The period of the top metal patterns is set to be 2 μm for all the three samples.

 

Fig. 1 Unit cell schematic diagram of (a) sample A, (b) sample B and (c) sample C. The gold layer is 100nm for ground plane and 60nm for metallic patterns. The SiO₂ dielectric layer is 190nm for all the samples.

Download Full Size | PDF

Cross resonator is a typical ERR [5], which couples strongly to the uniform electric field. By adding a dielectric layer and a metal ground plane, the magnetic component of the incident light can couple to the structure, and make it possible to tune its electric and magnetic response separately, causing an impedance matching to free space and thus near unity absorption. The geometry of the cross resonator have been investigated [21] to suit the metamaterial absorber, showing that the arm width should be increased when reducing the thickness of the spacer in order to maintain an equally high absorption. To optimize our design, computer simulation was carried out using commercially available software CST Microwave Studio 2009. Gold was characterized by a Drude model with a plasma frequency ${\omega }_c=2\pi \times 6.5$ THz and collision frequency ${\omega }_c=2\pi \times 6.5$ [22]. SiO₂ was modeled with dielectric constant of 3.9 and loss tangent of 0.025. Time domain solver was utilized with normal incidence and appropriate boundary conditions to extract the S parameters, from which the reflection can be obtained by$R\left(\omega \right)={\left|S_{11}\left(\omega \right)\right|}^2$. Since transmission was eliminated by the thick metal ground plane, the calculation of absorption was simplified to $A\left(\omega \right)=1-R\left(\omega \right)$. The incident light is perpendicular to the metamaterial surface during the simulation and the electric component is along x axis.

As a reference, sample A contains only one single cross with arm length of 1000 nm and arm width of 200 nm. The simulated absorption spectrum $A\left(\omega \right)$is shown in Fig. 2(a) and the maximum absorption is 99.7% at 4.5 μm. Figure 2(b) shows a concentrated electric field in the dielectric spacer at the butt of the cross arm along x axis, indicating a strong coupling to the uniform electric field. The magnetic response is evidenced by the antiparallel current in the cross and the ground plane in Fig. 2(c). The majority of energy dissipation is caused by dielectric loss [5], which is confirmed in Fig. 2(d). As the high absorption originates from resonance, the bandwidth of the spectrum is narrow. The full width at half maximum (FWHM) is only 13.45% with respect to the center frequency, and the Q factor is calculated to be 7.43.

 

Fig. 2 (a) Simulated absorption spectrum of sample A. (b) Simulated magnitude of electric field in SiO₂ at resonance frequency; x axis is in horizontal position. (c) Cross-section view of the simulated anti-parallel surface current in the cross arm and ground plane along x axis. (d) Power loss density in the SiO₂ spacer.

Download Full Size | PDF

Since the resonance frequency is mainly determined by the arm length of the cross resonators, the absorption bandwidth of multi-resonance structure can be tuned by the cross size and pattern arrangement. Sample B employs two multiplexed crosses in one unit cell that offset from the center by 360nm in both horizontal and vertical directions. The sizes of the two crosses were optimized with the same arm width of 140 nm and different arm length of 800 nm and 720 nm respectively. The simulated absorption of sample B is illustrated in Fig. 3(a). For comparison, the spectra of two regular non-multiplexed structures with the same cross size as the multiplexed resonators and a lattice constant of 2 μm are demonstrated together, and the absorptions reach 99% at 3.73 μm and 3.43 μm for the absorbers with cross arm length of 800 nm and 720 nm respectively. The multiplexed structure presents two peaks at the same position of the two individual non-multiplexed absorbers with the absorption of 99% and 88.2%. The two peaks merge to form a broader absorption band since they are sufficient close with each other. The power loss densities of the multiplexed structure at the two resonant frequencies are illustrated in Fig. 3(b) and 3(c). The majority of energy loss occurs in the bigger cross for the low frequency resonance, while most energy is dissipated in the smaller cross for the high frequency resonance. This means that the merged two peaks correspond to each single element resonators, and function independently.

 

Fig. 3 (a) Simulated absorption of sample B and two traditional non-multiplexed absorbers. (b) Power loss density at 3.73 μm. (c) Power loss density at 3.43 μm.

Download Full Size | PDF

To further broaden the absorption band, four different size crosses are multiplexed in sample C. In the unit cell, the crosses with arm length of 880 nm, 800 nm, 720 nm, 640 nm and the same width of 140 nm are located in the four quadrants, offset from the center by 500 nm in both directions. Figure 4(a) shows the simulated absorption spectrum of sample C and also four non-multiplexed structures each containing a single cross with the same dimensions of the four crosses as in samples C. Near unity absorptions are obtained at 4.04 μm, 3.73 μm, 3.43 μm and 3.14 μm for the individual crosses. Since the four peaks are designed so close that they merge to a much wider absorption band when multiplexed into one unit cell. The same as sample B, the four peaks of sample C is induced by the independent resonance of the four crosses. However, there is a small red shift of the peaks in multiplexed structure relative to the peaks of individual crosses, which is caused by the coupling effect between the neighboring metallic crosses [12]. The coupling becomes apparent when the distance between neighboring metallic patterns get closer, as the case of sample C, where four crosses are placed in a compact manner in one unit cell. This is also confirmed by the power loss density in Fig. 4(b)-4(e). At each resonance frequency, one cross plays the dominant role in energy dissipation, but the crosses adjacent to it also show an obvious power loss, though much weaker. The electric field concentrates at the ends of the cross bar when resonance occurs, as previously shown in Fig. 2(b). When two cross bars are close enough at the butt, the concentrated electric field at one end will interact with the other, inducing surface current in the neighboring cross and resulting in a small fraction of power loss. This red shift has no big influence on the strength of absorptions with the lowest of the four peaks higher than 90%. The spectrum of sample C shows a good performance of broadband absorption, covering wavelength from 3.07 μm to 4.86 μm with absorption higher than 50%.

 

Fig. 4 (a) Simulated absorption of sample C and four traditional non-multiplexed absorbers. (b), (c), (d), (e) Power loss density at the four absorption peaks from long wavelength to short wavelength in the spectrum.

Download Full Size | PDF

3. Experimental results and discussion

The fabrication process of the absorbers began with a 525 μm thick polished silicon wafer. First, a 20 nm chromium adhesion layer was sputtered on the substrate followed by an 80 nm gold film, which serves as the ground plane of the metamaterial absorber. Then a 190 nm thick SiO₂ layer was deposited on the gold layer by Plasma Enhanced Chemical Vapor Deposition (PECVD). A layer of e-beam resist of PMMA was spin coated on the SiO₂ dielectric spacer and e-beam lithography was used to define the patterns of resonating structure. The patterns of sample A, B and C were transferred to the resist after development. Next, an adhesion layer of 10 nm titanium and a 50 nm gold film were evaporated onto the patterned resist. Finally, a standard lift-off process was carried out in acetone to remove the resist and the unwanted gold. Figure 5 displays the SEM image of the fabricated metamaterial absorbers and the inset is the pattern of a single unit cell.

 

Fig. 5 SEM image of (a) sample A, (b) sample B and (c) sample C.

Download Full Size | PDF

We measured the power reflection of the three samples using a microscope coupled Fourier transform infrared (FTIR) spectrometer (magna-IR 750), and the microscope field of view was reduced to 200 μm × 200 μm to guarantee a more uniform area for measurement. Reflection of a gold mirror was taken as the normalization reference to obtain the reflectivity of the sample in percentage, from which absorption can be calculated. The measured absorptions for sample A, B and C are plotted in Fig. 6.

 

Fig. 6 Measured absorption for (a) sample A, (b) sample B and (c) sample C.

Download Full Size | PDF

As shown in Fig. 6(a), sample A displays a single absorption peak of 86.7% at 4.44 μm, and has more than 50% absorption from 4.21 μm to 4.85 μm, with a range of 0.64 μm. The FWHM is 15.68% and Q factor is 6.39, which shows the bandwidth is a little bit wider than the simulated results. Figure 6(b) shows the absorption of sample B, which has two peaks at 3.58 μm of 91% and 3.78 μm of 89.5%. The two peaks are so close to each other and form a relatively wider peak. The band range with higher than 50% absorption is from 3.23 μm to 4.09 μm. The FWHM and Q factor are calculated to be 26.47% and 3.78 while we take the central frequency at the wavelength of 3.66 μm.

The measured absorption of sample C is plotted in Fig. 6(c). Similar as the simulated results, the measured spectrum displays two main absorption peaks of 93.06% and 95.35% at the wavelength of 3.62 μm and 4.28 μm respectively, and the other two less obvious peaks at 3.33 μm and 4.62 μm. The two less obvious peaks are so close to the two main peaks that they almost merge into the spectrum, but their function to broaden the bandwidth remains unaffected. The band range is from 2.98 μm to 4.84 μm with the absorption more than 50%, which almost covers the full mid-infrared band. Taking the central frequency at 3.91 μm, the FWHM is calculated to be 54.1% and the Q factor is 1.85. More noticeable, the absorption reaches beyond 80% from 3.40 μm to 4.50 μm, corresponding to a bandwidth of 28.1%, which is even wider than the FWHM of sample B, indicating an excellent effect on bandwidth expansion of the four-cross multiplexed structure. The measured performances of the three samples are summarized in Table 1.

Table 1. Performance of the Fabricated Metamaterial Absorbers

View Table

We can see from Fig. 6 that the measured spectra show a higher degree of mergence than simulated both for sample B and C. We attribute this discrepancy to fabrication errors, which include the accurate size controlling of the patterns and surface roughness. Since in the multiplexed configuration, the crosses are very close in size and are densely arranged, a little deviation in the cross size or the distance between neighboring crosses during fabrication will bring a measurable influence on the absorption spectrum.

4. Conclusions

In summary, broadband metamaterial absorbers in the mid-infrared regime using multiplexed cross resonators have been investigated theoretically and experimentally. Three samples were fabricated and measured to give a systematic demonstration of multi-resonance metamaterial as a broadband absorber. FDTD simulation was carried out to give an explanation of the absorption mechanism for the multi-resonance structure. The Q factor drops from 6.39, 3.78 to 1.85 by employing one, two and four crosses in a single unit cell, which agrees well with the simulation results. The absorber with four crosses multiplexed achieves absorption above 50% at the band range from 2.98 μm to 4.84 μm, almost covering the full band of mid-infrared. The demonstrated broadband absorber shows promising potential applications such as energy harvesting and IR detection.