Electron energy loss spectroscopy on freestanding perforated gold films

We report on a combined far- and near-field study of surface plasmon polaritons on freestanding perforated gold films. The samples were fabricated by focused ion beam milling of a periodic hole array into a carbon membrane followed by thermal evaporation of gold and plasma ashing of the carbon film. Optical transmission spectra showed a series of characteristic features, which can be attributed to the excitation of surface plasmon modes via the periodic nanohole array. The corresponding near-field distributions were mapped by electron energy loss spectroscopy. In addition to the optically bright surface plasmon modes, we observed in the near-field a dark plasmon mode which is absent in the normal incidence far-field spectra. Our experimental results are in good agreement with numerical computations based on a discontinuous Galerkin time-domain method.

Figure 1

Fabrication process of freestanding perforated gold films: (a) Patterning of a carbon membrane via focused ion beam milling. (b) Thermal evaporation of gold to metallize the structures. (c) Removal of the carbon film by a plasma ashing process. (d) Scanning electron micrograph of the sample after all fabrication steps.

Figure 2

Cross-section of the finite element mesh used for the discontinuous Galerkin time-domain (DGTD) computations. This includes total field (Tf) and scattered field (Sf) regions, both for the surrounding air and gold layer. The swift electron input field is injected on the Tf/Sf contour and the Au Sf region [34]. The computational domain contains a $9\times 9$ hole array and is terminated with perfectly matched layers (PMLs) to emulate a continuous gold film.

Figure 3

Optical transmission spectra of the freestanding hole arrays with different film thicknesses and incident polarizations. The film thicknesses in nanometers are indicated by numbers in the respective color. The electric field is polarized along the horizontal $x$ direction ($800\mathrm{nm}$ array period) in the left panels and along the vertical $y$ direction ($600\mathrm{nm}$ array period) in the right panels. Experimental spectra are shown in the upper panels. Computed spectra are shown in the lower panels. Solid curves: $R=100\mathrm{nm}$, dashed curve: $R=90\mathrm{nm}$.

Figure 4

Comparison of the optical transmission spectrum recorded with horizontal incident polarization to electron energy loss spectroscopy (EELS) data for the $d_z=22\mathrm{nm}$ film. (a) Optical transmission spectrum replicated from the red curve in Fig. 3. (b) EELS data extracted from a horizontal line profile through two holes extended over the spectral range of the optical transmission data. The $(1,0)$ mode at $1.34\mathrm{eV}$, the dark $(1.5,0)$ mode at $1.65\mathrm{eV}$, and the $(1,1)$ mode at $1.84\mathrm{eV}$ are highlighted by gray dashed lines, which indicate their energetic positions in (a) and (b). (c)–(e) EELS maps of a region including four holes at the electron loss energies corresponding to the mode energies marked in (a) and (b). The $y$ position of the profile used to generate the data in (b) is marked by the gray dashed lines in (c)–(e). The scale bar in (c) is $250\mathrm{nm}$ long.

Figure 5

Discontinuous Galerkin time-domain (DGTD) computation results for (a) the optical transmission and (b)–(e) EELS data. The data are presented in the same manner as the experimental results in Fig. 4. The scale bar in (c) is $250\mathrm{nm}$ long.

Figure 6

Summarizing graph of the experimental (triangles) and computational (disks) results in comparison with theoretical calculations (rhombs). The different colors correspond to the three lowest-energy bright modes of the array. The black data points represent the dark $(1.5,0)$ mode.