1. Introduction

Surface plasmon polaritons (SPPs) are interactions between light and surface electrons in metal and are able to confine light strongly at the nanoscale from the metal-dielectric interface [1]. Various plasmonic nanostructures supporting SPPs have been proposed for light harvesting and manipulation [2]. Among these, we are particularly interested in two-dimensional (2D) metallic nanobowl arrays because of their excellent light trapping properties. When the nanobowl depth is equal to or larger than its radius, localized SPPs can be excited with strong optical field concentrated inside the nanobowls [3,4]. It has been reported that gold (Au) nanobowl arrays are able to greatly enhance Raman scattering [5,6] and are even superior to inverted nanobowls (i.e., Au nanospherical shells) [6]. However, as a typical plasmonic nanostructure, metallic nanobowls have not been applied as widely as their inverted counterparts. The light trapping mechanisms in optoelectronic devices (e.g., photovoltaic solar cells, photodetectors, etc.) have also not been comprehensively investigated. This is mainly due to the challenge of fabricating uniform 2D metallic nanobowl arrays.

There have been many methods developed to fabricate metallic nanobowls. J. Ye et al. proposed to mill synthesized silica (SiO₂)-Au core-shell nanospheres followed by a vapor hydrofluoric (HF) acid etching of the SiO₂ core [7]. In this case, Au nanobowls are randomly distributed on the substrate. However, random nanostructures have been shown to be inferior to the periodic counterparts when employed in photovoltaic solar cells [8,9]. To obtain a 2D uniform nanobowl array, self-assembled SiO₂ or polystyrene (PS) nanospheres usually serve as sacrificial templates. Based on these templates, bowl-shaped metallic nanostructures can be easily obtained with electrochemical deposition [3–5]. However, this requires conductive substrates. L. Johnson et al. took advantage of redox groups on PS nanosphere surfaces to reduce silver (Ag) ions, which preferentially deposited in the pores of the PS templates. The size of the formed Ag nanobowl array, however, was less than 250 μm in diameter [10]. If metals are sputtered or evaporated directly on top of the sacrificial templates, a turn-over and transfer process is necessary. L. Chen et al. did this in HF solution [6]. To avoid using the dangerous HF solution, a tape-assisted stripping process was proposed [11]. In this case, the nanobowl array attached to the tape cannot be easily transferred to other substrates. In our recent work, a large-scale very shallow tantalum nanobowl array has been developed based on a lift-off process assisted with self-assembled PS nanospheres [12].

In contrast to the above techniques, in this work, we introduce a facile approach based on a polymethyl methacrylate (PMMA)-assisted turnover-transfer method, through which Au films deposited on self-assembled PS nanospheres are uniformly turned over and transferred to another substrate, becoming a 2D hexagonal Au nanobowl array. It is known that PMMA has been widely employed to transfer graphene [13,14] or carbon nanotubes [15] onto a target substrate. In those methods, the turnover step is not necessary, which however is very critical in this work. It is also different from our previous work, where PMMA serves as a carrier to uniformly distribute Ag nanowires over a large area [16,17]. When the Au nanobowl array is coated with an amorphous silicon (a-Si) thin film, anomalous light trapping enhancement is observed for the first time and investigated both experimentally and theoretically. Such optical behavior is quite unlike that in the inverted nanobowl array, where the light trapping is suppressed by the a-Si:H coating.

2. Fabrication Procedure

Our Au nanobowl array is fabricated by the following procedure illustrated in Fig. 1. A Si substrate was first pretreated with an oxidized hydrophillic surface, on which closely-packed PS nanospheres (average diameter: 500 nm) were self-assembled based on a previously-reported method [18]. On top of the nanospheres, a 68-nm thick Au film was deposited by sputtering, followed by spin-coating of a 300-nm thick PMMA and a 10-min bake at 180 °C. The top PS nanospheres/Au/PMMA three-layer film was peeled off from the Si substrate in a hot sodium hydroxide aqueous solution (1 mol/L, 80 °C), and immediately shrank into a ball. Fortunately, immersing the ball into water made it spread out again and float at the air/water interface due to surface tension. This was the first critical step of the process. With the help of a clean substrate (e.g., polyethylene terephthalate), we carefully turned the floating three-layer film over with the PS nanospheres facing the air. The inverted film was lifted up by a pre-cleaned glass substrate. A 80 °C bake followed until the film tightly adhered to the substrate. The PS nanospheres were completely removed by immersing the sample in cyclohexane for 1 hour. After removing from the solvent and drying in the ambient lab conditions, an intact, closely-packed hexagonal Au nanobowl array was obtained on the glass substrate with PMMA in between. The sample size was 2 cm × 2 cm, which could be made even larger if a larger-scale PS nanosphere array was obtained [19]. The whole process is very simple, safe and repeatable. Without the inversion process, Au nanospherical shells, i.e., inverted Au nanobowls, were fabricated directly on a glass substrate and characterized for comparison. Quite different optical responses were observed when a-Si layers of different thicknesses were sputtered on top of the nanostructured arrays.

 

Fig. 1 Schematic of the fabrication procedure of Au nanobowl arrays: ① Sputtering deposition of a Au film on top of the self-assembled PS nanospheres; ② Spin-coating of a PMMA layer; ③ Disassociation of the top three-layer film from the Si substrate in a hot sodium hydroxide aqueous solution (first critical step); ④ Turning the floating three-layer film over (second critical step); ⑤ Lifting the inverted film up by a pre-cleaned glass substrate; ⑥ Removal of the PS nanospheres with cyclohexane.

Download Full Size | PDF

The SEM (scanning electron microscope) image of the Au nanobowl array [Fig. 2(a)] showed that the Au nanobowls were packed closely into a hexagonal pattern, mainly due to the reproduction process from the well-established self-assembled PS nanospheres. The white triangles were Au, which was sputtered onto the substrate through the gaps of nanospheres and remained through the following steps (top inset of Fig. 2(a)). The bottom morphology of individual nanobowls was inspected from a folding area (bottom inset of Fig. 2(a)), where the hemisphere shapes kept almost perfectly as the Au film covering the nanospheres in the inverted configuration (Fig. 2(d)). This indicated again that our fabrication processes did not cause substantial damages to the Au nanobowls. Looking closer to the inset of Fig. 2(d), one could see larger top hemispheres than the bottoms, meaning that the Au film was mainly deposited on the top part of the PS nanospheres. For these two nanostructures, some defects were inevitable due to the sparse or dense distributions of the sacrificial PS nanospheres. If these defects were neglected, the central distance between adjacent nanobowls (nanospherical shells) was equal to the diameter of the PS nanospheres (500 nm). Deposition of an a-Si film did not affect the hexagonal pattern of Au nanobowls (e.g., Fig. 2(b)) or Au nanospherical shells (e.g., Fig. 2(e)). The a-Si film seemed much rougher on the former than on the latter. For each sample, a small region was milled away by focused ion beam (FIB) and cross-sectional morphology appeared [Figs. 2(c) and 2(f)], which were quite different from our previous assumption of conformal coatings of both Au and a-Si (as reported in [20]). Actually, the Au film did not conformally cover the PS nanospheres but looked like a crescent array. It was thicker at the top of the nanospheres and thinner at the spacing between adjacent nanospheres [Fig. 2(f)]. After turnover, such crescent shape maintained very well [Fig. 2(c)]. The a-Si film was also not a conformal coating over the Au nanobowls or nanospherical shells. For the former, more a-Si was accumulated at the openings of the nanobowls and less was deposited at the bottoms, forming an almost straight air hole in each nanobowl [Fig. 2(c)]; while for the latter, the a-Si film shaped like a crescent on each Au nanospherical shell, similarly to the previously-deposited Au film [Fig. 2(f)]. Since it was very difficult to control the FIB milling to right cut the central Au nanobowls or Au nanospherical shells, the film thicknesses of Au or a-Si obtained from Figs. 2(c) and 2(f) were not very precise. In the following, all the mentioned sputtered film thicknesses were instead determined by profile measurement of spare samples, on which micropatterns were fabricated through successive UV lithography, sputtering deposition under the same condition and lift-off processes.

 

Fig. 2 SEM images of a Au nanobowl array: (a) before a-Si deposition (top inset: a zoomed-in image; bottom inset: a folding area), (b) after deposition of a 60-nm thick a-Si coating, (c) FIB milled cross-section (inset: a zoomed-in image); SEM images of a Au nanospherical shell array: (d) before a-Si deposition (inset: a cross-sectional image), (e) after deposition of a 60-nm thick a-Si coating (A Au film is deposited to improve the conductivity and for clear inspection. It is too thin to be continuous but fortunately does not affect inspection.), (f) FIB milled cross-section.

Download Full Size | PDF

3. Experimental Characterizations

3.1 Anomalous light trapping enhancement

The hemisphere reflection and transmission spectra of both Au nanobowls and Au nanospherical shells were measured by our home-built spectrometer based on an integrating sphere with an incident angle of 7° and demonstrated in Figs. 3(b) and 3(c), respectively, from which the absorption spectra were calculated by subtracting them from one and plotted in Fig. 3(a). For comparison, optical performances of a 68-nm thick flat Au film were also characterized. In Fig. 3(a), the flat Au film demonstrated a high intrinsic absorption in the short wavelength range of < 500 nm, beyond which its absorption dropped drastically to nearly zero due to high reflection [Fig. 3(b)]. With PS nanospheres as the substrate, the reflection of the flat Au thin film was greatly suppressed [Fig. 3(b)], leading to a significant enhancement in absorption, especially in the long range where the flat Au film was highly reflective. For the nanobowl array, its absorption was also much higher than that of the flat film [Fig. 3(a)], due to the reduction in light reflection [Fig. 3(b)]. In comparison with the nanospherical shells, the absorption of the nanobowls was higher in the short wavelength range of 380-620 nm but lower in the long range of 620-880 nm. This is mainly due to the abruptly-changing reflection spectrum, which was lower in the short range but much higher in the long range [Fig. 3(b)]. Instead, we will show later that the nanobowls have a stronger light trapping effect than the inverted structure with an a-Si coating. For the two nanostructures, the transmission spectra were both enhanced compared to that of the flat case [Fig. 3(c)], which was probably due to the plasmon-induced anomalous transmission [21,22].

 

Fig. 3 Measured (a,d) absorption, (b,e) reflection and (c,f) transmission spectra of the fabricated Au nanobowl array (red), Au nanospherical shell array (black), and a flat Au film with the same thickness (blue) before (top row) and after (bottom row) deposition of a 60-nm thick a-Si film. The corresponding optical responses for a glass substrate coated with a 60-nm thick a-Si film (green) are also characterized and plotted in (d-f).

Download Full Size | PDF

Surprisingly, when a 60-nm a-Si film was sputtered onto the Au nanobowl array and its inverted counterpart, anomalous optical behaviors were observed, resulting in quite different light trapping effects. For comparison, a-Si films with the same thickness were also deposited on a flat Au film and a bare glass substrate and characterized. For the a-Si film on top of the flat Au film rather than on the glass substrate, its total absorption was enhanced [Fig. 3(d)]. In the short range of < 450 nm, the incident light was absorbed in the top highly-absorptive a-Si film; while in the long range, the light was absorbed by both the top a-Si and the bottom Au film, because it could reach the highly-absorptive Au and be confined within the a-Si film through constructive interference. When the bottom Au film was constructed into the nanobowl array, the a-Si film could be deposited into the Au nanobowls [Fig. 2(b)], but did not cover the sidewalls conformally [Fig. 2(c)]. The incident light was scattered with extended light path in the hybrid nanostructure, resulting in enhanced absorption in a-Si and Au [Fig. 3(d)]. Both the reflection and transmission spectra were reduced and its transmittance even decreased to nearly zero over the whole wavelength range [Figs. 3(e) and 3(f)]. In contrast, for the sample based on the Au nanosperical shells, its reflection spectrum was enhanced with the top a-Si film and the transmission still remained very high [Figs. 3(e) and 3(f)], leading to an apparent absorption reduction in the whole range [Fig. 3(d)]. A similar absorption spectrum was observed in [20], where an Ag nanospherical shell array coated with a 120-nm thick a-Si film was based on a 300-nm diameter PS nanosphere array. However, detailed explanations were absent.

3.2 The effect of a-Si coating thickness

Based on the above observation, the effect of the a-Si thickness on the absorption spectra for both nanostructures were investigated and shown in Figs. 4(a) and 4(b). The average absorption over the measured wavelength range was calculated for each a-Si thickness and plotted in Fig. 4(c). As the a-Si thickness rose, for the Au nanobowl array, the average absorption increased first and then decreased, peaking at 60 nm with the value up to 82%; while for the Au nanospherical shell array, deposition of an a-Si film of any thickness (even as thin as 39.3 nm) would greatly reduce the average absorption despite a small increasing trend when the thickness became larger than 157.2 nm. Therefore, for a large range of a-Si coating thicknesses (around 30-200 nm), our Au nanobowl array showed a superior light absorption or trapping enhancement over its inverted counterpart [Fig. 4(c)].

 

Fig. 4 The effect of a-Si coating thickness on the absorption spectra for (a) the Au nanobowl array and (b) the Au nanospherical shell array, as well as (c) their average absorption over the measured wavelength range.

Download Full Size | PDF

Specifically, for the Au nanobowl array, the 39.3-nm thick a-Si film enhanced the absorption in the short wavelength range but was too thin to affect the absorption substantially in the long range [Fig. 4(a)]. Therefore, its average absorption was higher than that of the case without a-Si coating but lower than that of the 60-nm case [Fig. 4(c)]. For a-Si films thicker than 60 nm, the obvious change lay in the spectral dip in the long wavelength range, which red shifted and became increasingly deeper with the increasing a-Si thickness [Fig. 4(a)]. Since the sputtered a-Si film was more prone to accumulation at the openings of the nanobowls (e.g., Fig. 2(c)), as the film thickness became thicker, the openings would become smaller. It was thus very difficult for long-wavelength photons to couple into the nanobowls coated with thicker a-Si films. As a result, the measured reflection increased (not shown here). In contrast, for the Au nanospherical shell array, when the thickness increased to 157.2 nm, the average absorption dropped almost linearly to the minimum (except for the occasional value at 60 nm) and then rose to 62% when the thickness was further increased [Fig. 4(c)]. This drastic change mainly resulted from the absorption in the range longer than 550 nm [Fig. 4(b)]. Two competing mechanisms determined the light trapping property. From the morphological point of view, as the a-Si thickness increased, the shell-shaped nanostructured morphology flattened gradually and the light reflection became stronger, leading to lower absorption. On the other hand, thicker a-Si coating would allow more light to be absorbed and lead to enhanced absorption. Figure 4(c) showed that the 196.5 nm a-Si film was sufficiently thick and able to marginally increase the average absorption.

4. Numerical Simulations and Discussions

To further investigate the anomalous light trapping behaviors within the two nanostructures before and after the a-Si deposition, we conducted full-vector optical simulations based on a finite-difference time domain method (Lumerical FDTD Solutions). Figures 5(a)-5(f) showed the schematic diagrams of the four nanostructures used in the simulations. To approximate the real hexagonal pattern, the diameter of the element (nanobowl or nanospherical shell) was set to 500 nm and the central distance between adjacent elements was equal to the diameter, i.e., p = 500 nm [Figs. 5(c) and 5(f)]. The Au crescent in Figs. 2(c) and 2(f) was approximated by two spheres (with the same diameter of 500 nm) vertically separated by 68 nm, i.e., t_Au = 68 nm [Figs. 5(c) and 5(f)]. This value was consistent with the sputtered Au film thickness. The bottom sphere was PS, which was kept for the shell-shaped nanostructure [Fig. 5(f)] but was removed after being turned over for the bowl-shaped nanostructure [Fig. 5(c)]. Using the same method, the a-Si coating was created on top the Au nanospherical shells and the disconnection between the bottom Au sphere and the top a-Si sphere was compensated by an a-Si cylinder (indicated by the two horizontal dashed white lines in Fig. 5(f)). The modeling was more complicated for the a-Si coating on the Au nanobowls, which was illustrated in Fig. 5(c). To approximate the real morphology shown in Fig. 2(c), within each Au nanobowl, an a-Si donut with diameter of 500 nm right covered the bowl opening. The cross-sectional radius was set to the deposited a-Si thickness, i.e., t_a-Si, which was varied (as in Fig. 4). The thinner bottom a-Si coating was assumed as a conformal coating, which was bounded by a small air sphere with radius of r_air. The bottom a-Si coating was set to half of the deposited film thickness, i.e., 0.5t_a-Si. The a-Si inner sidewall was assumed to be vertical and created by an air cylinder tangent to both the bottom air sphere and the top donut. In this case, the cylinder height was also 0.5t_a-Si indicated by the two horizontal gray dashed lines and r_air + t_a-Si = 250 nm held [Fig. 5(c)]. x- and y-polarized plane waves were incident normally onto the structures from the top. Perfectly matched layers were treated along the z direction, while to save memory, anti-symmetric and symmetric boundaries were set along the x and y directions for x-polarization, and vice versa. Both reflection and transmission spectra were recorded from 400 nm to 900 nm, based on which the absorption spectrum as well as the average absorptions for each a-Si thickness were calculated.

 

Fig. 5 3D schematic diagrams used in the simulation of (a,b) a Au nanobowl array and (d,e) a Au nanospherical shell array (a,d) before and (b,e) after the a-Si deposition. 2D illustrations of a-Si coatings (c) within adjacent Au nanobowls and (f) on top of Au nanospherical shells.

Download Full Size | PDF

Figures 6(a) and 6(d) showed the average absorption varied with the a-Si coating thickness for both nanostructures for x- and y-polarizations, respectively. Irrespective of polarization, the average absorption for the two nanostructures had similar trends with the a-Si thickness. There was a large gap for the average absorption between the two nanostructures, when the a-Si thickness ranged from around 30 nm to around 150 nm. In this range, the average absorption for the bowl-shaped nanostructure was higher than that for the shell-shaped nanostructure, which was consistent with the experimental results [Fig. 4(c)]. The consistence also included the first increasing and then decreasing trend for the former and the opposite trend for the latter [Figs. 6(a) and 6(d)]. To some extent, it could be believed that the simulation model was accurate enough to reflect the experiments. Nevertheless, it should be admitted that the modeled structures (Fig. 5) were still too ideal. Defects in real samples (Fig. 2) were ignored in the simulations. Typically, for the cases based on the Au nanobowls, the small Au triangles [Fig. 2(a)], the roughness of the sputtered a-Si film [Fig. 2(b)], and the non-vertical a-Si inner sidewalls [Fig. 2(c)] were excluded; while for the Au nanospherical shells, the films deposited onto the glass substrate through the spacing of adjacent PS nanospheres [Fig. 2(d)] were neglected. For the former, due to the accumulation of the a-Si film at each nanobowl opening, the real a-Si thickness was much larger than the nominally sputtered film thickness, which was set in the model. Therefore, the simulated average absorption peaked at a larger a-Si thickness of 78.6 nm [Figs. 6(a) and 6(d)] rather than the experimentally-obtained 60 nm [Fig. 4(c)]. In contrast, the a-Si coatings on the real Au nanospherical shells seemed thinner than the nominally sputtered thickness (also the parameter set in the model). The average absorption valley occurred at a smaller a-Si thickness of 60 nm [Figs. 6(a) and 6(d)] instead of the experimentally-obtained 157.2 nm [Fig. 4(c)]. Apart from these, the measurement setup could also not be perfectly simulated numerically, which also introduced some mismatch between simulation and measurement results. Fortunately, such discrepancies did not affect the following theoretical analysis based on the simulations.

 

Fig. 6 Simulated average absorption as a function of a-Si coating thickness for the hybrid nanostructures based on Au nanobowl array (red circle) and Au nanospherical shell array (black square) for (a) x- and (b) y-polarizations. The simulated absorption spectra of the Au nanobowl array (red) and Au nanospherical shell array (black) (b,e) without a-Si coating and (c,f) with a 78.6-nm thick a-Si coating for both (b,c) x- and (e,f) y-polarizations.

Download Full Size | PDF

When the a-Si thickness was 78.6 nm, the greatest contrast was observed in average absorption for the Au nanobowls compared to the inverted Au nanospherical shells [Figs. 6(a) and 6(d)]. The simulated absorption spectra in Figs. 6(b), 6(c), 6(e), and 6(f) showed an obvious enhancement (suppression) in absorption in the long wavelength range for the Au nanobowls (Au nanospherical shells) after the a-Si deposition for both polarizations. Such anomalous optical behaviors were also similar to the experiments (Fig. 4), confirming again the reliability of our simulations. In these simulated absorption spectra, some ripples and sharp peaks appeared, which were however absent in the measured spectra (Figs. 3 and 4) due to the sample defects as well as the low spectral resolution of our spectrometer. Despite this, we could still get some insight on the light trapping behaviors from numerical simulations of these nanostructures. The spectra for both x- and y-polarizations were almost the same, indicating that the hexagonal symmetry did not exert too much sensitivity to light polarization.

In order to clearly see the light confinement within the two nanostructures before and after the deposition of a 78.6-nm thick a-Si coating, electric field distributions of all the nanostructures at four typical wavelengths (i.e., 500, 608, 659 and 782 nm) were calculated and plotted in Fig. 7. Here, only distributions for the x-polarization were presented due to the weak polarization dependence as analyzed above. The four wavelengths actually corresponded to the main peaks in the absorption spectrum of the a-Si coated Au nanobowl array (which are the main reason behind the extended absorption spectrum and thus the greatly enhanced average absorption), while for other nanostructures, no special features appeared at these wavelengths [Figs. 6(b) and 6(c)]. But it was to be shown below that the electric field distributions at these wavelengths could best reflect the changes of the optical responses for all the nanostructures.

 

Fig. 7 Simulated electric field distributions at wavelengths of 500, 608, 659, 782 nm in the xz plane for x-polarization of a Au nanobowl array (a) without and (b) with a 78.6-nm thick a-Si coating; as well as a Au nanospherical shell array (c) without and (d) with a 78.6-nm thick a-Si.

Download Full Size | PDF

Before a-Si deposition, in the short wavelength range below about 550 nm, the light could be well trapped into the Au nanobowls [3,4], while for the Au nanospherical shells, most light was confined at the top surfaces of the Au crescents (e.g., the electric field distributions at 500 nm shown in Figs. 7(a1) and 7(c1)). They showed similar absorptions in this range [Fig. 6(b)]. Compared with the convex nanospherical shells, the concave nanobowls had better light concentration ability and thus a little bit higher absorption (especially at even shorter wavelengths in Fig. 6(b)). As the wavelength increased, the trapped electric field within the nanobowls moved up [Fig. 7(a)], meaning increasing reflection and thus decreasing absorption [Fig. 6(b)]; while for the nanospherical shells, SPPs were excited and became increasingly stronger [Fig. 7(c)]. The SPPs, propagating along the convex surface, resonated between the periodic boundaries [Fig. 7(c)], leading to apparent absorption peaks [Fig. 6(b)]. The cross-section of each Au nanobowl or nanospherical shell looked like a nanocrescent, whose broadband light harvesting behaviors had been extensively investigated theoretically based on transmission optics in [23–26]. Because the sizes of our Au nanostructures were comparable to the light wavelength, radiation loss and retardation effects occurred, making the analytical model inaccurate to some extent [23–26]. Nevertheless, some similar optical field features could still be captured and the optical behaviors in our cases could be re-interpreted further based on the main conclusions from [23–26]. SPPs, which were excited in the fat parts of the Au nanobowls or nanospherical shells, propagated along the inner and outer surfaces to the openings. During the propagation, their wavelengths shortened and velocity decreased due to the increasing effective refractive index [27]. At the blunt openings, SPPs were reflected and interfered with the ones propagating towards them. In this case, resonances of SPPs occurred, leading to multiple ripples in the simulated absorption spectra, especially in the short wavelength range [Fig. 6(b)]. Like the kissing cylinders [23,26], at the touching points between adjacent Au nanobowls or nanospherical shells, strong optical field was concentrated, particularly at long wavelengths [Figs. 7(a) and 7(c)]. In comparison with the Au nanobowls, whose concave metallic surfaces were reflective, the convex surfaces of the Au nanospherical shells were more favorable for light coupling due to the gradually-changing geometry. Therefore, the light, shining directly onto the fat convex part, was more easily to excite SPPs. With much stronger optical field concentration [Fig. 7(c)], higher absorption was obtained for the Au nanospherical shell array [Fig. 6(b)].

After coating a 78.6-nm thick a-Si film, which was highly absorptive at wavelengths shorter than about 550 nm, the absorptions of both Au nanobowls and Au nanospherical shells were enhanced with a peak at around 500 nm. At this wavelength, strong electric field was trapped at the nanobowl opening, where most a-Si was accumulated [Fig. 7(b1)], while for the Au nanospherical shells, the electric field was mainly concentrated at the top convex surface of the a-Si crescent [Fig. 7(d1)]. Therefore, the former showed a little bit higher absorption [Fig. 6(c)]. At a longer wavelength, e.g., 608 nm, there was still a strong electric field concentrated at the opening of the bowl-shaped nanostructure. Meanwhile, through the top a-Si donut, the incident light was partly coupled into the a-Si coating, going through the sidewall and being trapped at the bottom, and partly localized under the connections of adjacent Au crescents [Fig. 7(b2)]. As the wavelength increased, such light confinement became increasingly stronger with SPPs propagating along the curved a-Si/Au and Au/PMMA interfaces. Especially at the concave a-Si/Au interface, the electric field seemed stronger [Figs. 7(b3) and 7(b4)]. The resonant SPPs within each period led to greatly enhanced absorption in both a-Si and Au and thus several high absorption peaks in the long wavelength range [Fig. 6(c)]. This trapping behavior was in great contrast to that in the case without the a-Si coating [Figs. 6(b) and 7(a)]. It was also quite different from the case based on the Au nanospherical shells, where increasing light reflection and decreasing light trapping with weak SPP-featured electric field distributions were observed at the wavelengths ranging from 608 nm to 782 nm [Figs. 7(d2)-7(d4)]. Therefore, with the a-Si coating, the Au nanospherical shell array demonstrated greatly reduced absorption [Fig. 6(c)]. Generally speaking, the a-Si coating assisted (resisted) the excitation of SPPs in the Au nanobowls (the inverted Au nanospherical shells) and allowed the opposite light trapping behaviors in the two nanostructures.

5. Conclusion

In conclusion, we have developed a facile PMMA-assisted turnover-transfer approach to fabricate a uniform 2D hexagonal Au nanobowl array, which shows inferior light trapping ability to its inverted counterpart mainly due to the higher reflection induced lower absorption in the long wavelength range. Surprisingly, after being coated with a 60-nm thick a-Si film, anomalous light trapping enhancement was observed for the Au nanobowl array with greatly enhanced average absorption (82%), while for the Au nanospherical shells, the light trapping was significantly reduced with average absorption of only 66%. Systematic experimental and theoretical results show that the top a-Si coating plays an important role in mediating the excitation of SPPs and the electric field distributions in both nanostructures, leading to such different light trapping behaviors. Based on these findings, new designs of high-efficiency optoelectronic devices (e.g., solar cells, photodetectors, etc.) are promising.