1. Introduction

The main advantage of nanovoid and nanodome arrays is that due to the periodic nature of the structure, light can be directly coupled to localized (‘Mie’ or ‘cavity’ modes) and delocalized (‘Bragg’ modes) plasmon modes [1]. While the localized modes are restricted to the protrusions of the gold layer, the propagating ones extend over the surface of the structure. The theoretical framework for the description and deeper understanding of the light-matter interaction in these quasi 2D ‘plasmonic crystals’ [2,3 ] paved the way towards their use in many different research areas and suggests a strong incident angle dependence especially for the Bragg modes at different wavelengths. The strong local electromagnetic fields of the localized plasmon modes were successfully utilized for light management in solar cell applications, where enhanced electromagnetic fields and increased optical path due to scattering were shown to increase the efficiency of inorganic as well as bulk-heterojunction organic and dye sensitized solar cells [4–6 ]. Also surface enhanced Raman scattering experiment performed on the nanovoid and nanodome arrays indicate that these structures can be very useful in label free analytical approaches [7–10 ]. A recent demonstration has also shown that despite the inherently low efficiency of the process, these plasmonic nanostructures can allow the generation and extraction of hot charge carriers [11]. Several approaches for the colloidal lithography based preparation of plasmonic nanostructures and nanovoid arrays exist, as reviewed recently by Ai et al. [12], but generally the size of the structured surfaces is restricted by the monolayer fabrication methods to lab-scale areas. Traditionally, metallic nanovoid arrays are fabricated via electrodeposition of noble metal trough a self-assembled monolayer of template particles – an approach that was inspired by the preparation of 3D metallic inverse opals [13], while for the nanodome arrays metal evaporation has been the natural choice.

Herein we present a simple and versatile process for the preparation of metallic nanovoid and nanodome arrays, where a short annealing step of a particle template monolayer on the minute scale and at moderate temperature (below 150° C) is only necessary to develop the geometry. As the main advantage the sample preparation involves only a few technology steps, which are identical for the two different surface structures (void and dome); at the same time the Langmuir-Blodgett technique – although more complicated and time consuming compared to spin-coating - allows the fabrication of macroscopic, large area (up to 3-inch diameter) truly monolayered templates. Additionally, the use of e-beam evaporation results in a well controlled and high quality metal layer. An excellent agreement was found between the experimental data and electromagnetic simulation results, but only when the inherently domain structure of the self-assembled template monolayer and hence the resulting final structure was taken into account.

2. Methods

2.1 Sample preparation

The sample preparation consists of three major steps, involving the template particle preparation, their monolayer formation at the water/air interface and the final structure preparation by heat treatment and metal deposition. An overview of the preparation procedure is provided in Fig. 1 . For all experiments ultrapure water with a resistivity of 18.2 MΩcm was used.

 

Fig. 1 Overview of the sample preparation procedure (see main text for details).

Download Full Size | PDF

First, the template particles have been synthesized according to Stöber’s method with some modification according to the literature [14], using Ludox-S40 silica nanoparticles (Sigma-Aldrich) as seeds for the particle growth. This seeded-growth approach improves the size-distribution of the particles compared to the original Stöber method, which is critical for the directed self-assembly process. Second, a monolayer of the template particles were prepared in a Langmuir film balance and transferred onto a substrate by the Langmuir-Blodgett (LB) method as reported by us earlier [e.g 15.]. The monolayers were transferred onto Si-wafers, coated with a 300 nm thick Shipley S1805 photoresist layer in a class 10-10000 clean room facility. As a last step, the monolayer deposited on the substrate were subjected to heat treatment at 120° C for 60 seconds. During the annealing process, the monolayer submerged into the polymer film to an extent depending on the temperature and duration of the heat treatment process. After completing the annealing step, the template particles were optionally removed from the polymer film by soaking the structure in 2% HF-solution for 20 seconds before e-beam evaporation of 140 nm gold at a rate of 2 Å/s in a ATC ORION 8-E UHV e-beam evaporation system.

2.2 Measurements

The as-synthesized particles were characterized by dynamic light scattering (DLS) (Malvern Nano ZS) to ensure their narrow size distribution. For the structural characterization of the samples scanning electron microscopy (LEO 1540 XB, Philips XL30 FEG) and atomic force microscopy (Bruker Dimension Icon) has been used. The angle-dependent absolute optical reflectance spectra of the samples was measured using a Perkin Elmer Lambda 1050 spectrophotometer with Universal Reflectance Accessory (URA), and allowed the reflectivity spectra to be measured in the wavelength range of 400-800 nm and angular range from 8 to 65 degrees in 13 points.

2.3 Optical simulations

Finite difference time domain (FDTD) calculations were used to evaluate the measured reflectance spectra of the different samples [16]. The optical model of both the nanovoid and the dome array consists of a semi-infinite Si substrate (refractive index from [17]), a photoresist layer of thickness of 300 nm (refractive index from [18]) and a gold layer of thickness of 140 nm (refractive index from [19]). In the simulation setup, both the photoresist and the gold layer were patterned with a hexagonal lattice of spherical voids/domes with vertical position of the spherical void/dome of 175 nm. The lattice period was 570 nm and the sphere diameter was 500 nm, these values were obtained from the AFM and SEM measurements. Bloch boundary conditions were applied in lateral directions and perfectly matched layer (PML) [20] in the vertical one. The structure was excited with a plane wave having variable polar and azimuth angle of incidence with respect to the lattice defined by the simulated unit cell and wavelength between 400 nm and 800 nm. The polar angle (θ) varied between 0 and 65 degree while the azimuth angle (φ) was set between 0 and 30 degree in 5 degree steps. Calculation was only necessary up to 30 degree because of lattice symmetry.

An advantage of the time domain calculation was that the whole wavelength range could be calculated at a single simulation using standard Fourier transform of the time signal. Using Bloch boundary conditions at oblique incidence required the set up of the lateral component of the incident wave vector as constant instead of the polar angle. This led to a wavelength dependent polar angle for each calculation. Finally wavelength and angle dependent spectra were produced in the postprocessing. Due to time limitations we have calculated the reflectance spectra only up to a certain maximum value of the Bloch vector which resulted in incomplete reflectance images at higher polar angles.

Calculating the response of a single structure for all angles, and wavelengths and both transverse electric (TE) and magnetic (TM) polarizations took about a week on a PC with 6 cores and 32 GB of RAM.

3. Results

The polydispersity index of the template particles measured by DLS was less than 0.04. The mean particle size is 500 nm. Figure 2 shows the characteristic SEM images of the two samples. Both for the positive (dome) and negative (void) structures, the top view images (Fig. 2(a), 2(b)) clearly indicate a domain-like structure of the film after the after gold deposition. The typical size of the domains is around 10-20 μm. The domain structure of the film is an inherent property of the template monolayer and is naturally developed for self-assembled mono and multilayers in general. The cross-sectional view of the samples (Fig. 2(c), 2(d)) clearly evidence the partial submerging of the template particles into the polymer film, as well as the dome and void structures after gold deposition. For both types of samples the deposited gold film is continuous.

 

Fig. 2 SEM top (a-b) and cross-section (c-d) images of the final dome (a,c) and void (b,d) structures after gold deposition. The insets show the schematic model setup used during the simulation of the respective samples.

Download Full Size | PDF

The schematic setup of the simulation can be seen on Fig. 3(a) . We have first calculated the specular reflectance of the structures at discrete azimuth angles of the incident wave. As an example Fig. 3(b)-3(e) shows the reflectance spectra for the TM polarization of the void structure at azimuth angles of 0, 10, 20 and 30 degree. A relatively strong absorption line is visible in all spectra at the same position starting at around 600 nm and can be identified as the Bragg plasmon mode of the system [3]. The apparent ‘patchiness’ of the data at this narrow reflection minimum band originates from the limited angular resolution of the simulation and is a direct consequence of data interpolation. The appearance of sharp edges in the spectra with increasing azimuthal angle indicates the change in the diffraction orders and the resulting spectra are markedly different for different orientations.

 

Fig. 3 Schematic setup of the optical simulations (a) and the simulated specular reflectance spectra of the void structure for TM polarization at azimuth angles of 0 (b), 10 (c), 20 (d) and 30 degree (e).

Download Full Size | PDF

The predicted angle-dispersive plasmon band can be also observed in the optical reflection measurements (Fig. 4 – top row), which is consistent with the coupling mechanism of the incident light to the propagating plasmon mode by the regularly spaced scatterer centers in a hexagonal lattice [1]. The domain structure of the film has important implication on the optical behavior of the samples. Due to the different orientation of the domains with respect to the plane of incidence and the macroscopic size of the light beam, all lattice directions are represented randomly in the reflected beam, since reflected electromagnetic waves arising from each domain are summed as they reach the detector. Since the domain size is larger than the wavelength and the propagating plasmon decay length (which is well below 20 microns for a smooth gold film in this wavelength range [21]) the reflection for each domain can be accounted for independently. Hence despite the lateral 'multicrystalline' nature of the sample surface, the Bragg mode is clearly developed in such a macroscopic ensemble measurement.

 

Fig. 4 Measured (a-d) and simulated (e-h) specular reflectance spectra of the dome (c, d, g, h) and the void (a, b, e, f) structure for TM (a, c, e, g) and TE (b, d, f, h) polarization. Simulated spectra are incoherently averaged over the azimuth angles.

Download Full Size | PDF

Moreover, due to the uncorrelated domain position, summation of calculated spectra can be done incoherently, leading to an average over the lattice orientations, allowing a direct comparison with the experimental spectra. In order to calculate the signal measured on the detector, we have carried out an incoherent averaging of the reflected intensity over the different lattice orientations (azimuthal angles). After averaging, an excellent agreement can be seen between the simulated and the measured spectra. The lower row in Fig. 4 shows the simulated reflectance spectra for both the nanovoid and the nanodome structures for both polarizations. The good agreement indicates a very good quality of the structures (i.e. high local order within the domains), and implies indirectly, that the optical response of each domain is well described by the theoretically calculated spectrum at a particular lattice orientation. The absorption lines on Fig. 3(b)-3(e) at given incident angles correspond to plasmon modes, suggesting strong near fields at the structure surface. This behavior makes the structure useful for potential applications, where strong near fields and local plasmonic modes are advantageous.

The measured reflectance values are about 10 percentage point lower than the calculated ones. This difference can be attributed to losses associated with enhanced scattering from lattice defects and domain boundaries. It is important to note, that the reflectance spectra for the void and dome structure are very similar for TE and TM polarizations respectively. This indicates that in these relatively shallow structures the spectrum depends mainly on the lattice geometry, not on the shape of the particular scatterer.

4. Summary

In this work we have presented a novel and simple method to fabricate metallic nanodome and nanovoid arrays, based on a short annealing of a polymer layer supported particle monolayer template. As a result of the annealing process the template particles partially submerge into the polymer layer and after an optional removal of the template the structured surface can be coated with a uniform gold film using electron beam evaporation. AFM and SEM measurements show that the prepared structures have a domain-like texture with a typical feature size of 10-20 μm and high local order of the two-dimensional hexagonal lattice within the domains. The measured specular reflectance spectra show a strong absorption band for TM polarizations and a weaker one for TE polarization, which can be attributed to a propagating plasmon mode. Calculated reflectance spectra are in very good agreement with the measured one for all configurations provided incoherent averaging of the theoretical data is performed over all lattice orientations. The good agreement indicates that locally each domain reflects light close to the theoretically predicted reflectance. The results underline the importance of the inherently domain-like structure of self-assembled template structures in general. It also demonstrates that local order at the length scale of a few tens of microns is sufficient to allow the development of spectral features associated with the Bragg plasmon modes. This less strict structural requirement can allow the design and realization of large area optimized surfaces for enhanced light-matter interaction in sensing, energy harvesting and light management applications.