1. Introduction

Surface plasmon polaritons (SPPs) are quasi-two-dimensional electromagnetic excitations propagating along a metal-dielectric interface, so that electromagnetic fields in dielectric are coupled to electron plasma oscillations in metal [1]. SPP propagation characteristics are extremely sensitive to surface properties, such as surface topography and adsorbents. The structure of a metal surface can be controlled by nanofabrication techniques in order to tailor the properties of SPPs and more specifically their interaction with visible and infra-red light, thereby offering the potential for developing new photonic devices and optical interconnects [2]. SPP guiding structures are basic elements of most plasmonic components needed to route optical signals and address sub-wavelength volumes in bio-sensors and data storage devices. Several promising configurations for SPP guiding structures have been reported, including SPP guiding along metal stripes [3–5] and their edges [6], as well as along channels in periodic arrays of scatterers (bumps) placed on metal surfaces [7]. In the latter case, the main principle is similar to that known for photonic crystals [8], i.e., the inhibition of SPP propagation within a certain wavelength range in periodic scattering arrays. Such a band gap (BG) effect for SPPs excited in the wavelength range of 780–820 nm has been shown to result in efficient SPP guiding, bending and splitting (albeit for relatively small angles) by line defects in periodic scattering arrays [9]. Here we consider the SPP excitation and propagation at telecom wavelengths along gold films covered with periodic arrays of gold bumps. Using a scanning near-field optical microscope (SNOM), we demonstrate the inhibition of SPP propagation into the arrays within a certain wavelength range depending on their period and orientation as well as the SPP propagation along straight and bent channels cut through these arrays.

2. Experimental setup

The experimental setup consists of a collection SNOM [10] in which the (near-field) radiation scattered by an uncoated sharp fiber tip into fiber modes is detected, and an arrangement for focusing of a laser beam and SPP excitation in the Kretschmann configuration (Fig. 1(a)). The p-polarized (electric field is parallel to the plane of incidence) light beam from a tunable telecom laser (λ=1425-1640 nm, P~1 mW) is focused (spot size~15 µm) onto the sample attached with immersion oil to the base of a glass prism. Other excitation sources, such as a He-Ne laser (λ=633 nm, spot size ~5 µm) and a Ti:Sapphire laser (λ=740-840 nm, spot size ~10 µm) are also available. The SPP excitation by a weakly focused beam is usually recognized as a minimum in the angular dependence of the reflected light power (attenuated total reflection) [1]. In our case of a strongly focused beam, the SPP excitation is manifested by a dark stripe in the reflected beam spot (Fig. 1(b)).

 

Fig. 1. (a) Schematic of the experimental setup. (b) A reflected beam spot observed on a screen for a 45-nm-thick gold film at λ=633 nm. (c) Microscope dark-field image of a typical sample structure showing periodic arrays of gold bumps with straight and bent channels (line defects) placed along the median line of a 200-µm-wide gold stripe.

Download Full Size | PDF

The SPPBG structures represent 26- and 40-nm-thick gold films deposited on silica substrates and covered with gold bumps having heights of ~80 and 100 nm and diameters of ~250–450 nm and arranged in triangular lattices with periods of 850, 900 and 950 nm. This patterning was produced by electron beam lithography on a resist layer on the thin film, evaporation of a second gold layer and lift-off. Note that the gold films were also structured in 200-µm-wide and 1.5-mm-long stripes hosting the SPPBG structures (Fig. 1(c)). This design allows the usage of two types of the SPP excitation (from below): through the gold film and at the gold edge.

3. Experimental results

The first issue to be elucidated before setting up the investigations of SPPBG structures was to assess the localized SPP excitation and SNOM imaging of SPP propagation with the constructed experimental arrangement. This has been accomplished for three different wavelengths moving from visible light (Fig. 2(a)) to near-IR (Fig. 2(b)) and telecom (Fig. 2(c)) wavelengths, observing a rapid increase in the SPP propagation length (Fig. 2(d)) as expected [1,2]. The cross sections shown should be interpreted as convolutions between (Gaussian-shaped) excitation spots formed by focused laser beams (incident from right towards left) and exponentially decaying SPP intensity profiles. The SPP propagation lengths determined from these cross sections correspond qualitatively to those expected [1].

 

Fig. 2. Near-field imaging of localized SPP excitation and SPP propagation for different wavelengths. A 60-nm-thick gold film has been used for λ=(a) 633 and (b) 790 nm, whereas a 50-nm-thick gold film was used for (c) λ=1520 nm. (d) The SPP propagation length derived by an exponential fit to the signal decrease along dashed lines indicated in the corresponding SNOM images.

Download Full Size | PDF

The next step was determination of the BG in the fabricated structures. Typical topographical and near-field optical images obtained with the localized SPP excitation and SPP interaction with periodic structures are shown in Fig. 3. It is immediately seen that the detected SNOM signal is considerably stronger at the scattering arrays than at a flat gold film surface. This feature can be explained by the circumstance that the scattered out of the surface plane (i.e., propagating in air away from the surface) field components are detected with the SNOM fiber tip much more efficiently than the evanescent components associated with the SPP fields [11]. This difference in the detection efficiency is further accentuated due to the usage of a bare fiber tip, whose detection of the field components perpendicular to the surface plane is inherently weak [11,12]. One might notice that this feature has not been highlighted in our previous investigations of SPPBG structures [7,9]. We believe that the main reason for the deterioration in the SPP detection efficiency is related to the fact that the SPP mode at telecom wavelengths is much closer to the light line than that at the wavelength of ~750 nm [1]. The latter means that both the out-of-plane SPP scattering is more efficient and relative magnitude of the perpendicular (to the surface) SPP field component is considerably larger for telecom wavelengths. Indeed, the ratio between perpendicular and parallel SPP field components given by |Re(ε)|^0.5 is the metal dielectric constant) increases from ~5 to ~10 for the SPP at a gold surface when the wavelength changes from ~750 nm to ~1500 nm [13]. At the same time, the field components scattered in the direction normal to the surface plane are predominantly polarized parallel to the surface.

 

Fig. 3. SNOM (a) topographical and (b,c) optical images (35×35 µm²) of a triangular 900-nm-period structure with the SPP being excited at the wavelength of (b) 1550 and (c) 1600 nm and incident from the right in the ΓK direction of the irreducible Brillouin zone of the lattice [8]. The gold film thickness is 40 nm, and bump diameter and height are 378 nm and 100 nm, respectively. The structure orientation and an estimate of the position of the exciting focused laser beam are indicated on the topographical image (a).

Download Full Size | PDF

Having in mind the aforementioned circumstances, we analyzed the SNOM images obtained at different wavelengths. It is clear the out-of-plane SPP scattering at a given wavelength is determined by the local SPP field intensity, implying, for example in the case illustrated in Fig. 3, that the incident SPP penetrates into the SPPBG structure over much longer distance at the wavelength of 1550 nm than at that of 1600 nm (cf. Figs. 3(b) and 3(c)). In general, one should simply consider separately the SNOM optical image areas inside and outside the SPPBG structure. Using cross sections of the optical images along the propagation direction (Fig. 4(b)), one can then determine the SPP penetration depth (as a function of the wavelength) from the (exponential) signal decrease inside the structure (Fig. 4(c)). However, at wavelengths outside the band gap the signal decrease is quite weak and in some cases even larger than the length of the SPPBG structure causing a very flat signal profile and weak indication of exponential decrease. This gives increased appearance to other contributions to the signal, such as scattering from the back edge and interference between SPP components and light which propagates parallel to the surface by penetrating the metal layer. Nevertheless, pronounced SPPBG effects, especially for ΓK direction, are observed for different structures. It is seen that the BG in ΓK direction scales with the lattice constant, and that the BG in ΓM direction is markedly shifted towards shorter wavelengths as expected [14]. The cross sections of the optical images in front of the SPPBG structure may be used to determine a wavelength dependence of the structure reflectivity (for ΓK direction). Calculations of the spatial frequency spectra of the oscillatory signal in front of the structure yield peaks whose magnitude is related to the degree of reflectivity (Fig. 4(d)). The reflectivity is seen reaching maximum at a wavelength close to that of the center BG (cf. Figs. 4(c) and 4(d)) in agreement with general considerations of the BG effect [8].

 

Fig. 4. (a) Orientation of the SPPBG structure in our experiments for ΓK orientation and (b) the typical cross section of the SNOM optical image obtained. Similar cross sections are used to determine wavelength dependencies of (c) the SPP penetration depth and (d) the peak in spatial frequency spectra (of the signal variations in front of the SPPBG structure). (Movie 920 KB)

Download Full Size | PDF

It appears that in the investigated SPPBG structures the BGs for two main orientations do not overlap. Nevertheless, we attempted to realize the SPP propagation along a 30⁰ bend having the input channel of ΓM orientation and the output channel of ΓK orientation in a 950-nm-period lattice. The SNOM optical images recorded at the wavelengths near 1520 nm are shown in Fig. 5.

 

Fig. 5. SNOM (a) topographical and (b,c,d) optical images (52×26 µm²) of a triangular 950-nm-period structure with the SPP being excited at the wavelength of (b) 1500, (c) 1520 and (c) 1540 nm and incident from the right in ΓM direction. The gold film thickness is 23 nm, and bump diameter and height are 438 nm and 80 nm, respectively.

Download Full Size | PDF

It is clearly seen that the SPP propagation along the bent channel is strongly dependent on the wavelength and that the SPP confinement within the input channel is best at 1520 nm, which is inside the BG for ΓM orientation (Fig. 4 (c)). It should also be borne in mind when comparing the images obtained that the positioning of a focused beam spot at telecom wavelengths is rather difficult and that its position is wavelength dependent due to dispersion of a focusing fiber and a prism. Ideally, one should realize the BG simultaneously for both input and output channels and match the mode propagation constants in these channels. These conditions were apparently not met in our experiment explaining poor transmission of the bend. The best optical images were obtained at the wavelength of 1515 nm (Fig. 6) that show efficient bending of the SPP field along with its spreading to both sides of the channel, since this wavelength is outside the BG for ΓK orientation expected to be centered at much longer wavelengths (Fig. 4(c)).

 

Fig. 6. SNOM (a) topographical and (b) optical image (35×35 µm²) of a channel bend in a triangular 950-nm-period structure with the SPP being excited at the wavelength of 1515 nm and incident from the right in ΓM direction. The gold film thickness is 23 nm, and bump diameter and height are 438 nm and 80 nm, respectively.

Download Full Size | PDF

4. Conclusion

To summarize, we have presented first results on the SPP propagation in and scattering by periodic arrays of surface scatterers at telecom wavelengths. Using a collection SNOM, we have directly observed the SPP local excitation and interaction with periodic triangular arrays of gold bumps having different lattice constants and orientations. It should be pointed out that the SNOM imaging of SPP propagation and scattering at telecom wavelengths is quite a challenge because the radiation wavelength is considerably longer than that in visible, causing difficulties in its visualization and detection. Furthermore, understanding of the corresponding SNOM images is rather cumbersome because the SPPs are very close to the light line, resulting in an increase of the SPP out-of-plane scattering and decrease in the SPP field component parallel to the surface plane. We have characterized the inhibition of SPP propagation into the periodic structures due to the BG effect as well as the SPP reflection by the SPPBG structures of ΓK orientation and qualitatively explained the effects observed. We have also reported preliminary results on the SPP propagation along 30⁰ bent channels cut through these arrays. It appears that the main problem is the difference in the locations of the BGs for two main orientations (viz., ΓK and ΓM) observed for the investigated structures. It should be noted though that their positions are expected to strongly depend on the dimensions of surface scatterers [14,15], implying thereby the possibility for further improvement of SPP guiding and bending by use of the SPPBG structures. We conduct further research in this area.