1. Introduction

Using Surface Plasmon Polaritons (SPPs), confined electromagnetic waves at metal/dielectric interfaces, to transfer information, has the potential to bridge the gap between microelectronics and photonics, provided their intrinsically lossy character can be mitigated [1]. Indeed, as SPPs can transport optical signals on waveguides of truly nanoscale dimensions [2], they would allow for the use of the bandwidth and speed provided by optical systems, but on nanometric scales so far accessible only to semiconductor electronics [3]. In a context of information transfer and processing, it is important to design means to convert optical signals into SPPs to link long distance transfers, using standard low loss photonic devices, with miniaturised functions that would benefit from the use of SPPs. Such a converter should be compact and able to excite SPPs in a specific direction, towards the region of information processing.

SPPs have an in-plane momentum k_SPP larger than the momentum of free space propagating light, which means that converting light into SPPs requires the use of methods providing extra in-plane momentum to the photons. Many methods have been used to achieve this goal, but the ones relevant for nanoscale integration all use corrugations in the metal film sustaining the SPPs. Periodic arrangements of holes for instance allow for the excitation of SPPs and directionality of the excitation can be achieved when the pump beam impinges on the grating at oblique incidence [4–6]. Preferential SPP launch can be obtained under illumination at normal incidence of a grating provided the symmetry of the grating is broken [7–9]. A very efficient asymmetric grating coupler using concepts similar to the Yagi Uda antenna has been demonstrated recently [10]. Unfortunately, efficient gratings have a footprint of many wavelengths on the metal film: more compact approaches are desirable.

Schemes based on the use of one or two slits in the metal film are much more compact. A variety of such schemes have been proposed and studied theoretically [11, 12]. When a single slit is used to excite SPPs, the excitation efficiency depends on the width of the slit compared to the SPP wavelength [11–13], and directionality of the SPP excitation is possible for excitations at oblique incidence [14]. Similarly to the gratings, breaking the symmetry of a single slit can lead to a directional coupling of SPPs at normal incidence [15,16]. Furthermore, adding a Bragg reflector next to the slit allows for a directional coupling at normal illumination, but of course has the same drawback as for a grating: the footprint of the whole system represents many wavelengths [17]. Systems using two parallel nanoslits have been discussed as well and lead to directionality of SPP excitation at oblique incidence [18]. Filling each slit with dielectrics of different refractive index is expected to allow for directionality with excitation at normal incidence, although this has not been demonstrated experimentally so far [19, 20]. More schemes for directional coupling under normal incidence have been discussed theoretically [21–23].

To date, there is a lack of experimental demonstrations of all the schemes proposed. One report has shown that the SPP excitation efficiency of a single nanoslit depends on its width [24]. Another one demonstrated directional launch of SPP for one slit with illumination at 75° [25]. Both of these studies are performed using Scanning Near-field Optical Microscopy (SNOM), which is expensive, complex to implement, and SNOM images are long to acquire. The role of SPPs in the optical transmission of a pair of slits has been observed in the far-field, for slits separated by many wavelengths [26]. Eventually, the directional coupling of SPP under oblique incidence via a pair of nanoslits has been observed using gratings to out-couple the SPPs generated at the nanoslits – which allows for observations in a narrow wavelength range only [18].

In this article, we use leakage radiation microscopy (LRM) [27–31] to study comprehensively how light is coupled to SPPs via the use of arrangements of slits. LRM is extremely flexible and allows for a fast exploration of the full parameter space of the problem. In contrast to SNOM, it is cheap to implement, simple to operate and measurements are fast. Moreover, no out-coupling gratings are required to observe the SPPs: thanks to this LRM allows for the use of any wavelength without modification of the setup or sample. We show that single and double slits under oblique illumination allow for directional coupling of SPP, probing for the first time the whole parameter space of interest: angles up to 20°, slit widths/distances in the range 150 to 950 nm (λ₀/4 to 1.5λ₀, with λ₀ = 633 nm). The experimental findings are well described by a simple model accounting for the diffraction of the incoming beam by the slits. Eventually, our approach allows for a direct comparison between the performances of single and double slit-based devices.

2. Leakage radiation microscopy

To conduct our studies, we use the fact that on a thin metal film (a few tens of nanometer thick) deposited on a glass substrate, the SPPs bound to the metal/air interface ”leak” on the glass side [29], i.e. they are coupled back to a propagating light field on the glass side. This emission occurs at an angle larger than the total internal reflection angle of a glass/air interface. Therefore, collecting the leakage radiation requires the use of an oil immersion objective. Figure 1(a) presents the schematic of the leakage radiation microscope. Its design is similar to the one described in Ref. [28]. The beam from a polarised HeNe laser is focused on the sample by a 20×, NA 0.4 microscope objective (O₁). A polariser and half waveplate are inserted in the beam path before the objective, to align the polarisation of the beam perpendicularly to the slits (cf. Fig 1(c)). The width of the beam is much smaller than the pupil of the objective. This allows to change the angle of incidence of the beam onto the sample by translating the laser perpendicularly to the optical axis, keeping it parallel to it, as shown on Fig. 1(a) and 1(c). The spot size at the sample level is about 10 μm in diameter. The light then goes through a 100×, NA 1.49, oil immersion microscope objective (O₂), which allows for the collection of light leaking from the SPPs excited at the gold/air interface. The lens L₁ projects the back focal plane of O₂ (which is contained inside O₂) onto the lens L₂. Placing masks in front of L₂ could allow for a physical filtering in the Fourier space of the image from the sample’s surface (we do not use this in the present study). The lens L₃ then images on a CCD camera either the surface of the sample, in position A, or the back focal plane of O₂, in position B.

 

Fig. 1 Experimental configuration. (a) artistic and sketched views of the leakage radiation system. (b) SEM images of slit pairs and single slit etched in the gold film. The scale bars represent 200 nm. (c) SPPs are launched on the film by focusing an HeNe laser on the slits at an angle θ from the normal to the sample, from the gold side. The light is polarised perpendicularly to the slits. Leakage radiation is collected on the sample side by an oil immersion objective.

Download Full Size | PDF

The samples are produced by thermal evaporation of a 3 nm thick germanium layer on Corning glass coverslips, followed by the evaporation of a 36 nm gold layer. The germanium serves as sticking layer for the gold, and its influence on the optical properties of the system are negligible. Slits of 20 μm length are then produced by Focused Ion Beam milling (Helios Nanolab 50 DualBeam). Figure 1(b) presents SEM micrographs of some of the slits used in this article.

The measurements of the SPP intensity are done in Fourier space, projected on the detector by an appropriate arrangement of the lenses in the leakage microscope. Figure 2 presents a typical measurement in the Fourier plane: in the example presented, the laser is focused on a slit pair separated by 350 nm. In the Fourier plane, one can observe three elements: a disc of intense light, where the detector is saturated, a broad bright line, and two spots of light symmetrically positioned compared to the center of the image (Fig. 2(a) and 2(b)). The saturated spot corresponds to the pump light directly transmitted through the sample, which is partly transparent. The position of the spot indicates the angle of incidence. The broad line is the light diffracted by the slit. Its alignment is perpendicular to the slits. The twin spots placed symmetrically from the center of the image are due to the light leaking from the SPPs. The intensity of the left- and right-propagating plasmons are measured thanks to profiles taken along the Fourier images, as shown in Fig. 2(c).

 

Fig. 2 Observations in the Fourier plane. (a) (resp. (b)) image of the Fourier plane when the laser is focused on a slit pair separated by 350 nm, at quasi-normal incidence (resp. at an angle θ = −18°). Both images have been obtained under the same conditions. (c) Profiles taken along the images (a) black line, and (b) red line. The peaks at ±1.66.10⁶m⁻¹ correspond to the SPPs. No preferential direction of coupling is observed at normal incidence, whereas at 18° mainly right-propagating SPPs are excited. Further analysis of this picture can be found in section 5.

Download Full Size | PDF

3. Theoretical model

The experimental results are compared to a simple analytical model to describe the efficiency of coupling of the incident light into SPPs. The underlying principle is that the pump light is diffracted by the slits, and the angular spectrum of the diffracted light can be calculated by a Fourier Transform of the field at the slit’s position. We assume that the intensity of the SPPs excited is proportional to the intensity of the component of this angular spectrum matching the SPPs’ wavevector [13, 14].

The slits are represented by the rectangular function $\text{rect}(\frac{x-x_0}{w})$, with x₀ the position of the center of the slit considered and w its width. By definition:

The pump beam is a plane wave of wavelength λ₀ impinging on the sample at an angle θ. The field is polarized perpendicular to the slits, as illustrated on Fig. 1(c). The component of the electric field along the samples’ surface at the surface of the glass substrate in the absence of the slits is $E\left(x\right)=\text{exp}(i2\pi x\frac{sin\hspace{0.17em}\theta }{{\lambda }_0})$. When the slits are considered, we make the approximation, derived from the scalar diffraction theory of Fourier optics, that the metal film prevents the radiation from reaching the substrate, whereas it is not perturbed at the position of the slits. Once this assumption is made, at the sample surface and in the presence of the slits the component of the field along the sample surface becomes:

The angular spectrum of E(x) at the slits position is obtained via a Fourier transform of the field Ẽ(α), where α is the coordinate of the angular spectrum corresponding to the x coordinate in real space. Using the properties of scaling, shifting in space and frequency domain of the Fourier transforms, it can be shown that:

To estimate the SPP generation efficiency, we then calculate the intensity of the angular spectrum for α = α_SPP, where α_SPP is derived from the dispersion relation of the leaky SPPs mode in our system. This dispersion relation is derived as described in Ref. [29], and the values of the dielectric function for gold ε_Au(λ) are taken from Ref. [32]. With a thickness of gold of 36 nm, λ₀=633 nm and (ε_Au(633 nm) = −11.43 + 1.19i), we find an effective index of n_SPP = 1.048. Right-propagating SPPs and left-propagating SPPs are given by I_right = |Ẽ(α_SPP)|² and I_left = |Ẽ(–α_SPP)|², respectively.

From this expression, we can derive the maximum coupling efficiency for single slit width w = w_Max. They happen to correspond to the conditions for distance between slits giving the weakest coupling in a double slit system, x₂ = d_Min:

The directionality D of the SPP generation is then defined as:

As will be seen further, this model gives a qualitative description of the phenomenon, and is useful to identify simply the regions of high coupling efficiency as well as high directionality. We direct the reader to Ref. [11–13, 33–35] for more in-depth descriptions of the interplay between the scattering of slits and the generation of SPPs.

4. Single slits

Let us first turn our attention to the case of a single slits used to excite SPPs. In the experiment, slits of increasing widths have been milled into the gold film, by steps of 50 nm (100 nm, 150 nm,..., 950 nm). The experimental plots are generated by making a linear extrapolation of the values between measured values. Fig. 3(a) presents the intensity measured for the left propagating SPPs, normalised to the highest intensity, alongside with the results of Eq. (4) in Fig. 3(b). The agreement between experiment and the simple model is remarkable.

 

Fig. 3 Single slits, at λ₀ = 633 nm. (a) Experimental coupling efficiency into left propagating SPPs. (b) Theoretical results obtained from Eq. (4). Both intensity plots are normalized to their maximum intensity. (c) Measured directionality. (d) directionality calculated with the model explained in the text. Directionality=1 means that only left propagating SPPs are excited, 0: only right propagating, and 0.5 means equal excitation in both directions. The experimental plots are obtained by making a linear extrapolation between the points actually measured. In the theoretical plots (b) and (d), the black and green lines indicate $w_{\text{Max}}^{\text{Left}}$ and $w_{\text{Min}}^{\text{Right}}$, respectively.

Download Full Size | PDF

The directionality of SPP generation by single slits is presented in Fig. 3(c) and 3(d). As can be seen on these figures, regions showing directionality in excess of 90% can be found: more than 90% of the generated SPPs are launched preferentially in one direction. For the case presented here, such high directionality are obtained with slits of widths between 450 and 550 nm, and angles close to 20°.

Optimal directionality/excitation efficiency of SPP occur at the intersection of the curves of maximum coupling efficiency into one direction w_Max and the curves of minimum coupling efficiency into the other direction w_Min obtained thanks to Eq. (5) and (6). In Fig. 3(b) and 3(d), we highlight $w_{\text{Max}}^{\text{Left}}$ in solid black lines and the $w_{\text{Min}}^{\text{Right}}$ in solid green lines. From these considerations, it appears that there is a general trade-off to be found between angle of incidence and width of slits for optimal operation. To operate at small angles, one has to use wide slits (for instance in this case: operation for 10° is optimal for a slit width of 750 nm). Conversely, the optimal point with narrowest slit (455 nm width here) occurs at an angle of operation of −20.7°. A more complete picture of $w_{\text{Max}}^{\text{Left}}$ and $w_{\text{Min}}^{\text{Right}}$ can be found in Fig. 5(a).

 

Fig. 4 Double slits – for λ₀ = 633 nm. (a) Experimental measurement of coupling efficiency into left-propagating SPPs. (b) Theoretical results obtained from Eq. (4). (c) Experimental directionality. (d) Theoretical results obtained from Eq. (7) for the directionality. In (b) and (d), the black and green lines indicate $d_{\text{Max}}^{\text{Left}}$ and $d_{\text{Min}}^{\text{Right}}$, respectively.

Download Full Size | PDF

 

Fig. 5 Comparison of figures of merit (FOM) for single (a) and double slits (b). The log of the FOM is plotted (see text for the definition of the FOM). Here, λ₀ = 633 nm, n_SPP = 1.048, and the FOM is calculated for left propagating SPPs. The purple regions correspond to D ≤ 0.5, i.e. a majority of right-propagating SPPs. In (a), the black and green lines indicate $w_{\text{Max}}^{\text{Left}}$ and $w_{\text{Min}}^{\text{Right}}$ respectively. In (b) they indicate $d_{\text{Max}}^{\text{Left}}$ and $d_{\text{Min}}^{\text{Right}}$.

Download Full Size | PDF

5. Double slits

We now consider the case of parallel sub-wavelength slit pairs. The slits milled by FIB are 90 nm wide, and the distance between slits is varied between 150 nm and 700 nm by steps of 50 nm. Again here the experimental plots are generated by using a linear extrapolation between the values actually measured. Figure 4 presents the excitation efficiency of left propagating SPPs, as well as directionality, depending on slit distance and angle of incidence. Similarly to Fig. 3, the agreement between experiment and model is good, although the measurements here are slightly more noisy. The slits milled here have a width approaching the resolution limit of the fabrication method employed, which could explain the increased noise compared to the single slits. Indeed the quality of the milling varies more from one slit to the next when we operate close to the limits of the instrument, which can lead to variations in the optical transmission of the slits, and thus variations along the horizontal axis in Fig. 4(a). The directionality is not affected by this as it is a ratio of intensities, thus insensitive to variations in absolute coupling to SPPs. Indeed, Fig. 4(c) does not show the sharp variations visible in Fig. 4(a). The directionality depends little on the slit width (data not shown, but can be calculated using Eq. (4) and (7)), which means that the key parameter is the distance between the slits.

The specific situation shown in Fig. 2(c) (slit distance 350 nm, θ = −18° and 0°) can be analyzed further in light of these considerations. The large directionality observed exists due to the fact that the coupling efficiency towards the left direction changes little between θ = −18° and 0° (from 0.08 to 0.06 normalized units, Fig. 4(b)), whereas the coupling efficiency to right propagating SPP increases from 0.06 to 0.44.

To better compare the performances of single and double slits, we now introduce a Figure Of Merit (FOM). We define the FOM for left propagating SPPs as the product of the directionality D, rescaled between −1 and 1, and the left-propagating SPP intensity I_left (from the theoretical model). In this definition a majority of right-propagating SPPs are represented by a negative FOM:

The base 10 logarithm of the FOM is plotted on Fig. 5. The purple regions represent regions where right-propagating SPPs are preferentially excited, or extremely inefficient coupling to a majority of left propagating SPPs. Figure 5 highlights the fact that high directionality combined with high SPP excitation occurs on different regions for single and double slits. These plots can be used to determine what system is best to achieve a specific goal. For instance, one can be interested to have a directional coupler with the smallest footprint possible. The footprint of a slit pair is the distance between slits, plus the width of one individual slit (in the case presented here, 90 nm). With these considerations, the point of optimal coupling for smallest device occurs for a footprint of ≈ 455 nm for single slits, and ≈ 545 nm for double slits (cf. Fig. 5, points A and B) – a single slit will always provide smaller footprint. Moreover, if one can sacrifice some coupling efficiency, single slits allow to work at smaller angles, as shown by the region lying between point A and the line θ = 0.

6. Conclusions

In this study, we directly compare the directionality of SPP launch by single and double slits on a thin metal film. The measurements, using leakage radiation microscopy, allow for a rapid probing of the parameter space of this problem: slit width, distance, and angle of incidence of the pump beam. Moreover, the wavelength of illumination can be changed without modifications to the sample. The experimental findings are well described by a model using the angular spectrum of the light diffracted by the slits to estimate the SPP generation efficiency. Single as well as double slits allow for directional SPP excitation provided that the illumination is at oblique incidence. There is a general trade-off between size of the devices and the angle of incidence required for optimal coupling. Overall, single slits allow for better performances than double slits and have a smaller footprint. The approach developed here should allow for extensive studies of most of the other schemes proposed for unidirectional SPP coupling, and performs better than the ones used in all previous reports.