1. Introduction

Since the first demonstration by Landy et al [1], metamaterial absorbers (MAs) have stimulated a new scientific research field due to the huge interest in the development of materials for promising extensive applications, such as solar cells [2–7], plasmonic sensors [8–10], detectors [11–14], cloaking devices [15], and thermal emitters [16–19]. To date, tremendous efforts have been devoted to develop MAs at microwave, terahertz and infrared frequencies. The majority of these works mainly focused on the target of limited frequencies absorption, instead limited works were presented to directly achieve broadband perfect MAs at the whole visible light region. For instance, Wang et al. have obtained broadband optical absorption (higher than 95% over the wavelength from 300 to 2000 nm) by tunable Mie resonances in silicon nanocone arrays [20]. Guo et al. reported a nanostructured broadband absorber with flat average absorption greater than 80% from 400 to 700 nm [21]. Furthermore, Guo et al. also demonstrated an ultrathin broadband visible absorber, which obtained 95.5% at a resonance and the absorption can be maintained over a wide angle of incidence up to ± 70° regardless of the incident light polarization [22]. Tun Cao et al. reported a broadband MA composed of an array of thin Au squares separated from a continuous Au film by a phase change material layer and realized polarization-independent, wide-angle near unity absorbance in the visible region [23]. Koray Aydin et al. demonstrated a MA consisting of a metal-insulator-metal stack configuration, which yielded an average absorption of 71% over the visible range 400-700 nm [24]. Unfortunately, complicated fabrication procedures and device configurations still constitute significant problematic to achieve practical device applications due to these recently proposed approaches involves the time-consuming and costly fabrication technology of electron beam lithography or focused ion beam milling. Thereby, the desire to realize high performance broadband MA in the visible regime has long been a research topic of interest for scientists-one particular theme being the construction of a facile device configuration which is compatible with current cost-effective industrial production technology.

Here, a new MA composed of a periodic array of dielectric cylinder sandwiched with non-noble metal of Ni film is proposed and demonstrated. The proposed MA achieves polarization-independent absorption of an average level of above 90% in the whole visible regime (400-700 nm) and nearly perfect absorbance (over 99%) is achieved from the wavelength 500 nm to 560 nm. Furthermore, the absorption is nearly independent of the incident angle below 30° and still remains exceed 80% even when the incident angle reaches 60°. Additionally, the presented fabrication approach is straightly compatible with high-throughput manufacture technology of soft nano-imprinting lithography, making the manufacturing process is promising and cost-effective.

2. Structural design and experimental details

The three-dimensional configuration and the cross-sectional view of the suggested MA are shown in Fig. 1. It was composed of a periodic array of dielectric cylinder sandwiched with non-noble metal of Ni film (Fig. 1(a)). The parameters of the period, the width and the residual layer of cylinder are denoted as p, w, and h₂, respectively. It is noteworthy that the thickness of the bottom Ni layer (h₁) is thick enough to prevent light transmission and therefore the absorption can be calculated by A = 1 - R, where R is the reflection. Noted that, the broadband absorption can be obtained by employing metal Ni, due to the nature of Ni. The thickness of the cylinder and the top Ni layer are h₃ and h₄ respectively. The optimized parameters are: p = 250 nm, w = 100 nm, h₁ = 200 nm, h₂ = 60 nm, h₃ = 85 nm and h₄ = 10 nm. The theoretical optimized procedures will be discussed further in next section.

 

Fig. 1 The schematic of the proposed MA structure. (a) Three-dimensional (3D) topography. (b) The corresponding cross-section configuration of the MA.

Download Full Size | PDF

Schematic of the procedures involved in fabrication of the proposed MA is presented in Fig. 2. The Silicon (Si) template is fabricated by the continuously variable spatial frequency photolithography (SVG Corporation, Nanocrystal) as we previously reported (Fig. 2(a)) [25, 26]. The laser beam (351 nm) is transmitted through the suppressed zero-order diffraction grating and only ± 1 orders of diffracted beams are selected to pass through the plane mirror and objective lens. These two diffracted beams are crossed on the top surface of the photoresist (RZJ-390PG) constructed on the monocrystal silicon substrate for exposure of 15 s and subsequently developed in NaOH solutions (6‰) for 6 s and dried by electric blow drier. Thereafter, the sample is etched through a fluorine-based ICP-RIE process. As a consequence, the Si template with periodic two-dimensional cylinder arrays is successfully produced. To fabricate the cost-effective MA, soft nano-imprinting lithography and thermal evaporation technology are adopted to fabricate the MA and the corresponding schematic procedures are illustrated in Fig. 2(b). Firstly, the thickness of 200 nm Ni film was sputtered onto the flat polyethylene (PET) substrate. Then, the UV resin (D10, PhiChem) was drop-cast onto a pre-fabricated Si template, and the PET substrate was placed onto the UV resin and imprinted under a constant pressure of 1.5 bar for 15s with a UV illumination of 500 mJ/cm² at a wavelength of 395nm. Subsequently, by carefully peeling off the silicon template, the periodic cylinder arrays (UV resin) constructed on the PET substrate coated with flat Ni film was obtained [27]. Finally, ultra-thin Ni film with a thickness of about 10 nm is deposited on the periodic cylinder array by thermal evaporation to give the proposed MA structure.

 

Fig. 2 (a) The experimental setup of the continuously variable spatial frequency photolithography system. (b) The fabrication process for proposed MA.

Download Full Size | PDF

3. Results and discussion

The surface morphology of the photolithographic two-dimensional cylinder arrays was characterized by scanning electron microscopy (SEM) and display in Fig. 3(a). These images clearly reveal that the fabricated device with p = 250 nm, w = 100 nm, indicating that parameters of this sample corresponds very well to our theoretical design as discussed above.

 

Fig. 3 Experimental and theoretical characterization of the proposed MA. (a) SEM images of the photolithographic two-dimensional cylinder arrays. The dimensions of a unit are p = 250 nm, w = 100 nm, h₁ = 200 nm, h₂ = 60 nm, h₃ = 85 nm. (b) The numerical simulated absorption (blue line), reflection (red line), and transmission (green line, negligible) spectra of the MA. The inset depicts the photograph of the fabricated sample with a size of 2.5 × 3 cm² placed on the front of the trees under ambient light. (c) Comparison between the simulated absorption (blue line) and experimental absorption (red line) of the MA sample at the incident angle of 20°.

Download Full Size | PDF

To analyze the optical absorption mechanisms of the MA, optical modeling calculations based on the rigorous coupled analysis (RCWA) method [28] using commercial RSoft 8.1 (RSoft Design Group, Inc.) are carefully performed. Here, the simulated parameters of the Ni were fitted by the Drude-Lorentz model and the refractive index of the dielectric cylinder array was set to be 1.5 with a negligibly small optical loss and the experimental absorption was calculated according to the equation of A = 1- R. It is well known that the absorption is polarization-independent as the symmetrical configuration of the designed MA [23, 24]. Therefore, only TM polarization (the magnetic field of the incident light is kept parallel to the y-axis) is taken as an example to investigate. According to the simulated results (Fig. 3(b)), it can be seen that the strong absorption at an average level of >90% is observed in the spectrum wavelength range 400-700 nm. In particular, a nearly perfect absorbance (over 99%) is achieved from 500 nm to 560 nm. The transmission is zero because of the thick Ni film so that the absorbance is maximal. To investigate the effects and validity of the presented MA, the comparison between the simulated absorption (blue line) and experimental absorption (red line) of the MA sample at the incident angle of 20° was depicted in Fig. 3(c). In this case, the measured reflection is normalized with respect to a Ni plane, while the measured transmission is characterized with respect to the incident wave in free space. The measured absorption spectrum is above 90% in the spectrum range from 400 to 700 nm. This explains why the fabricated sample in Fig. 3(b) appears to be black. The fabricated sample exhibit efficient broadband visible light absorption, resulting in a nearly black image, distinctly demonstrating that almost all incident light photons were absorbed by the MA. As can be observed, the agreement between the experimental and simulated result is very good considering the fabrication imperfections and the measurement error.

Moreover, to investigate the angular dependence properties of the MA, the simulation and the corresponding experimental measurement of the absorption spectra under oblique incident angle for the TM polarized light are performed and displayed in Figs. 4(a) and 4(b), respectively. Noted that, the experimental dates are obtained by employing the spectrophotometer (LAMBDA 750) with the wavelength region from 190 nm to 3300 nm. From Fig. 4(a), it is clearly seen that absorption is above 90% at the incident angle of 30°. Even though there is a drop in absorption up to an angle of 60°, the drop is not so large and the overall absorption still remains around 80%. From Fig. 4(b), it is noteworthy that there is negligible discrepancy between the calculated and measured absorption spectra, which mainly originates from the fabrication error of the sample and the background testing signal. More importantly, even though there is an obvious decrease in absorption curves when the incident angles varying from 30° to 60° in 15° steps, the absorption is nearly independent of the incident angle below 30° and the measured minimal absorption still remains as large as 80% for the incident angle of 60°. In addition, highly flat absorption can be maintained when the incident angles varying from 30° to 60° in 15° steps for TE polarized light (Fig. 4(c)), indicating the polarization-independent absorption.

 

Fig. 4 Simulated (a) and measured (b) angular absorptions of the MA for TM polarized light, measured (c) angular absorption of the MA for TE polarized light. The incident angle is varied from 30° to 60° in 15° steps.

Download Full Size | PDF

To observe the behaviors of light absorption within the fabricated MA, the electromagnetic field distributions of TM polarized light at various resonant (450 nm, 550 nm and 650 nm) wavelengths at normal incidence are calculated and depicted in Fig. 5. As shown in Fig. 5 (a), (c) and (e), the magnetic field at 450nm, 550nm and 650nm are always trapped in the air-slots and the magnetic fields located at the air/Ni interface decreases exponentially with the distance from the Ni surface, which suggested that the roles of surface plasmon polaritons (SPPs) and cavity modes supported inside the air-slots results in the strongly broadband absorption in the visible regime. The light funneling effect with narrow slots much smaller than the incident wavelength occurs and the light is trapped in the air-slots [29]. Looking at the fields in Fig. 5 (b), (d) and (f), strong localized electric fields at the corners of the air-slots can be seen, which further indicates supported SPPs [30, 31]. As we know, the bulk Ni layer can achieve broad absorption in the optical regime. Therefore, the roles of surface plasmon polaritons (SPPs) and cavity modes combined with the strongly attenuating properties of the Ni leads to a strong absorption at these wavelengths [32, 33].

 

Fig. 5 Calculated electromagnetic field distributions at some wavelengths at normal incidence. (a), (c), and (e) are for the magnetic amplitude at 450nm, 550nm, and 650nm, respectively. (b), (d), and (f) are for the magnetic amplitude at 450nm, 550nm, and 650nm, respectively.

Download Full Size | PDF

The Poynting vector distribution at 550 nm wavelength is shown in Fig. 6, it shows that the incident flow bens when reaching the Ni surface, and propagates along the surface toward the air-slot, which can be attributed to the magneto electric interference of the incident wave with the scattered evanescent field. The fundamental TM mode guided between two neighboring Ni sidewalls does not have a cut-off frequency, and then the energy flow squeezes into the air-slots which size is much smaller than the wavelength. Most of the energy is found to be trapped in the air-slots, which gives a solid proof that the funneling effect of the cavity mode plays a key role in the absorptive behavior in the structure.

 

Fig. 6 Calculated Poynting vector distribution at 550 nm wavelength at normal incidence.

Download Full Size | PDF

Noted that, the excitation of SPPs needs additional momentum provided by the proposed nanostructures which resulting in phase matching between the incident field and SPPs. Therefore, the absorption peak is affected by the change of wavelength and incident angle as described by the momentum matching condition [34], which is supported in the Fig. 4(a). At the wavelength further away from the diffracted order of the grating, the magnetic field supported along the surface of the Ni grating get weaker, and the supported cavity mode in air-slots dominate the absorption response at longer wavelengths, as indicated clearly in Fig. 5 (a), (c) and (e). This leads to an absorption peak at longer wavelengths and a more broadband response as shown in Fig. 3.It is suggested that the excitation of cavity modes in the air-slots can couple to one another leading to a supported resonant surface wave [35], which facilitates efficient coupling of the incident photons to be funneled into the air-slots and is almost completely absorbed by the MAs.

To gain further insights into the physical origin of these unique absorptivity, the magnitude of the simulated S parameters for TM polarized light and the corresponding real part and image part of the effective impedance through the inversion of the S parameters are calculated and are displayed in Fig. 7. Here, transmission T(w) = |S₂₁(w)|² and reflection R(w) = |S₁₁(w)|² and the absorption is calculated as A(w) = 1- T(w)- R(w) where only few incident photons are scattered by the absorber [36]. We calculate the effective impedance according to the formula [37]

 

Fig. 7 Magnitude of the simulated S parameters for TM polarized light and calculated real and image part of the effective impedance

Download Full Size | PDF

Finally, the absorption spectra on the normally incident TM polarized light for different period (p) and the thickness of the cylinder (h3) are shown in Fig. 8. The absorption peak blue shift as the grating period increases for the effective index obtained from the metal-insulator-metal plasmon waveguide reduce due to the air-slot width. On the other hand, the resonant wavelength of the absorption peak red shifts as the period continues increasing, for the SPP coupling into the metal-insulator-metal plasmon waveguide becomes less significant as the slot width increases. As shown in Fig. 8 (b), the broadband absorption in the visible region can be obtained when the thicknesses of the cylinder range from 50 nm to 150 nm. When the thicknesses of the cylinder is less than 50nm or larger than 150 nm, the light funneling effect becomes weak, which results the trend of absorption decreasing.

 

Fig. 8 Demonstration of geometric effects on the normally incident TM polarized light: (a) the period, (b) the thickness of the cylinder.

Download Full Size | PDF

4. Conclusions

In conclusion, a kind of broadband, polarization-independent absorber in the whole visible regime has been proposed, fabricated and characterized, whose absorption can exceed 90%. Moreover, the absorption is nearly independent of the incident angle below 30° and still remains greater than 80% at the incident angle of 60°. The proposed MA can be easily obtained without the need for ion or electrochemical etching of metal, which can be easily integrated into other optoelectronic devices. The suggested method helps to show that it is worth reflecting on the potential opportunities that the discussed MA may bring.