Introduction

Aluminum (Al) is arguably the only material sustaining surface plasmons (SPs) with moderate loss over the full visible and neighboring spectral ranges. In particular, this applies to the blue and ultraviolet (UV) range, where, e.g., silver and gold suffer significant loss from interband transitions. In addition, with its native oxide, Al micro- and nanostructures are environmentally stable. Accordingly, Al plasmonics has recently been intensely investigated [1,2,3,4]. With a view on applications relying on optical high-energy excitation such as plasmon-mediated chemistry or fluorescence microscopy, Al plasmon-enhanced spectroscopy [5] and coupling to elementary emitters [6] have been demonstrated.

A powerful technique to image and characterize SPs is photoemission electron microscopy (PEEM). With non-linear multiphoton photoemission, SP fields have been probed with high spatial and often with femtosecond time resolution [7, 8]. Propagating SP waves were studied with PEEM [9,10,11,12,13], with advanced setups providing even momentum resolution [14]. Interference patterns on microstructured silver films were successfully interpreted as SP/light interference and analyzed for inclined and normal light incidence with respect to the sample plane [15, 16].

Here, we apply this approach to Al surfaces to derive the SP properties in the UV and blue spectral ranges. A lithographed straight ridge with subwavelength cross section is used for SP excitation with inclined light incidence. This provides SP interference patterns in both forward and backward directions with respect to the exciting light direction. By lithographic fabrication, we vary the ridge orientation with respect to the (experimentally fixed) in-plane angle of the incident light on a single sample. From the observed interference patterns in the experimental PEEM images, we deduce the SP propagation directions and decay lengths of 2.8 and 8.4 \(\mu\)m for light excitation wavelengths of 273 and 410 nm, respectively.

Experimental

Al ridges with nanoscale cross section were fabricated by electron beam lithography on glass slides covered with 70 nm of indium tin oxide. These substrates were first coated with a 0.5-nm thick film of chromium and a 40-nm thick film of aluminum. A 100-nm thick poly(methylmetacrylate) film was then deposited by spin coating, exposed by the electron beam with the lateral ridge geometry (200 nm wide and 40 \(\mu\)m long) and chemically developed. The resulting mask was used to deposit 40 nm of Al by vacuum evaporation. All Al deposition was done at a rate of 3 nm/s and a base pressure of about \(8 \times 10^{-6}\) mbar. The ridge orientation with respect to the incident laser beam (projected on the sample surface) was varied from \(\alpha\) = 0 to 90\(^\circ\) in steps of 5\(^\circ\).

Fig. 1
figure 1

a PEEM setup. A nanosecond laser pulse is incident under an angle \(\Theta\) = 25\(^{\circ }\) with respect to the substrate plane, and k\(_{0}\) and k\(_{SP}\) denote the wave vectors of light and SP, respectively. The in-plane angle \(\alpha\) with respect to the ridge orientation is varied by addressing differently oriented ridges on the sample. Exemplary PEEM images over an area of 12 \(\times\) 12 \(\mu\)m\(^2\) and \(\alpha\) = 90\(^\circ\) are depicted for excitation wavelengths of b 410 nm and c 273 nm

Full size image

PEEM images were recorded using a NanoESCA II instrument (FOCUS GmbH and Scienta Omicron). The second and third harmonic of an Indigo-S Ti:Sapphire laser system (Coherent) tuned to 3.02 eV (410 nm) and 4.54 eV (273 nm) were used as the light source. The laser pulse duration was 30 ns with a repetition rate of 5 kHz, and the laser power was set to 7 mW. The light was incident under an angle of \(\Theta\) = 25\(^\circ\) with respect to the sample surface (see Fig. 1), slightly focused to an elliptical spot with axis lengths of about 50 and 120 \(\mu\)m (full width at half maximum). The polarization state was set to p, i.e., within the plane of incidence. The spatial resolution of the electron optics is about 300 nm.

Exemplary PEEM images are shown in Fig. 1 for \(\alpha\) = 90\(^\circ\) and excitation wavelengths of (b) 410 nm and (c) 273 nm. While the overall appearance of the observed patterns is similar, the wavelength-dependent interference periodicity is evidently different. Further differences are discussed below.

Results and Discussion

Figure 2 shows PEEM images for different Al ridge (on Al film) orientations for excitation with a light wavelength of 410 nm. We observe interference patterns with fringes that are oriented parallel to the ridge for all ridge orientations. For an in-plane incidence angle of \(\alpha\) = 0\(^\circ\), the interference period is identical on both sides of the ridge. With increasing angle, the period at the forward side of the ridge increases, while it decreases on the backward side. For angles above 40\(^\circ\), the backward fringes cannot be resolved due to the resolution limitation of our PEEM system. Figure 3 shows the PEEM images from the same structures as above, however for an excitation wavelength of 273 nm. Qualitatively, identical interference patterns are observed for both wavelengths.

Fig. 2
figure 2

PEEM images for different ridge orientations (and thus light in-plane incidence angles \(\alpha\), as indicated in each panel) for a laser wavelength of 410 nm. The image size is 12 \(\times\) 12 \(\mu\)m

Full size image
Fig. 3
figure 3

PEEM images for different ridge orientations (and thus light in-plane incidence angles \(\alpha\), as indicated in each panel) for a laser wavelength of 273 nm. The image size is 6 x 6 \(\mu\)m

Full size image
Fig. 4
figure 4

SP/light interference. The orientations of the in-plane wave vectors of light (k\(_{0\parallel }\)), its component along the ridge (k\(_{0r}\)), the SP (k\(_{SP\pm }\)), and the interference pattern (\(\Delta\)k\(_{\pm }\)) are plotted for in-plane incidence angles \(\alpha\) of a 0\(^\circ\), b 55\(^\circ\), and c 90\(^\circ\). The solid line marks the ridge, and the dotted circle indicates the modulus of k\(_{SP}\). d–f show the corresponding simulated SP field snaphots, with a light wavelength of 410 nm, a SP wavelength of 402 nm, and an assumed SP decay length of 8.4 \(\mu\)m. g–i show the corresponding simulated interference patterns

Full size image

We first discuss the observed patterns in terms of the interaction of the exciting (inclined) light wave and the SP waves generated at the ridge [10, 15, 16]. As schematically depicted in Fig. 4a–c, the process can be straightforwardly interpreted in terms of wave vectors, corresponding to the momenta of light and SPs and angle-dependent momentum transfer provided by the Al ridge. Figure 4b sketches schematically the case for an incidence angle of 55\(^\circ\), with the in-plane light wave vector k\(_{0\parallel }\) (red horizontal arrow). According to Snell’s law, the component k\(_{0r}\) along the ridge has to be maintained. On the other hand, the wave vector component normal to the ridge is subject to change due to scattering from the ridge. Indeed, the normal vector components \(\Delta\)k\(_{+}\) and \(\Delta\)k\(_{-}\) provide the momentum match to the SP wave vector by pointing to the dotted circle that indicates the length of the SP wave vector following from the SP dispersion relation. Thereby, also, the SP propagation directions are set, indicated in Fig. 4b by the k\(_{SP\pm }\) vectors.

The same arguments apply to the exemplary cases for angles \(\alpha\) of 0\(^\circ\) and 90\(^\circ\), as sketched in Fig. 4a, c. In Fig. 4d–f, simulated snapshots of the SP fields are plotted, with phase fronts normal to their propagation direction as defined by k\(_{SP\pm }\). The excitation wavelength is 410 nm, and the tabulated Al permittivity of \(-\)24.3 + 5.1i (Ref. [17]) was used to calculate a value of 402 nm for the SP wavelength. The SP damping is set by an 1/e propagation length of 8.4 \(\mu\)m (corresponding to the value deduced below from the measurements). We note that while, in principle, the SP wavelength could also be recovered from the experimental interference patterns, the required multiparameter fitting does not allow to do so with meaningful precision.

Finally, Fig. 4g–i depict the simulated resulting interference patterns, considering the interference of an undamped plane wave (representing the incident light) with damped plane waves (representing the SPs). The pattern orientation is perpendicular to \(\Delta\)k\(_{\pm }\). We find that this model excellently reproduces all main features in the experimental images. We quantitatively illustrate this in Fig. 5, where we plot the measured interference periods in forward and backward directions (symbols) as observed in Figs. 2 and 3 together with the values following from the simulations (lines).

Fig. 5
figure 5

SP/light interference periods \(\Delta \lambda _B\), data extracted from Figs. 2 and 3 in forward (red) and backward direction (blue) for light wavelengths of 273 nm (open circles) and 410 nm (dots). The symbols depict measured data, and the lines depict the simulated values

Full size image
Fig. 6
figure 6

SP decay derived from data extracted from the PEEM images in Fig. 2 (410 nm, dots) and Fig. 3 (273 nm, open circles) for \(\alpha\) = 0\(^\circ\) (orange) and 90\(^\circ\) (blue). The solid and dashed lines show exponential fits to the like-colored data symbols, yielding 8.7 ± 1.1 \(\mu\)m (410 nm, \(\alpha\) = 0\(^\circ\)), 8.0±0.9 \(\mu\)m (410 nm, \(\alpha\) = 90\(^\circ\)), 2.7 ± 0.3 \(\mu\)m (273 nm, \(\alpha\) = 0\(^\circ\)), and 2.9 ± 0.4 \(\mu\)m (273 nm, \(\alpha\) = 90\(^\circ\))

Full size image

Importantly, the SP propagation direction is not directly observed in the PEEM images as the standing wave fringes are always parallel to the ridge. By understanding the image formation, however, the SP propagation direction can be determined, which allows us to deduce the SP propagation lengths from the PEEM images. We tested various approaches to do so, including fitting of the full mathematical interference model. However, spatial photoemission inhomogeneity (presumably in part due to the focused excitation), defects, and noise limit the validity of approaches requiring the adaption of multiple parameters. We thus chose to directly fit the intensity maxima of the interference patterns in the direction of SP propagation to infer the SP decay. This approach yields, for example, the 1/e decay length for the SP excited with a wavelength of 410 nm under \(\alpha\) = 0\(^\circ\) as 8.7 ± 1.1 \(\mu\)m (Fig. 6). The uncertainty gives the \(68\%\) confidence interval of the exponential fit. Rather than averaging over all data, we chose to exemplarily analyze individual images which gives us a measure of the reproducibility of our experimental approach (including sample fabrication and PEEM measurements) in view of the above-mentioned parasitic contributions. Accordingly, we find 8.0 ± 0.9 \(\mu\)m (410 nm, \(\alpha\) = 90\(^\circ\)), 2.7 ± 0.3 \(\mu\)m (273 nm, \(\alpha\) = 0\(^\circ\)), and 2.9 ± 0.4 \(\mu\)m (273 nm, \(\alpha\) = 90\(^\circ\)); see Fig. 6. We find that all individual results for a given excitation wavelength coincide within experimental error. For the sake of simplicity, we use averaged values for the SP decay length of 2.8 \(\mu\)m and 8.4 \(\mu\)m for excitation wavelengths of 273 nm and 410 nm, respectively, in the summary parts of this work.

It is noteworthy that the propagation lengths we find here are consistently higher than the values around 5 \(\mu\)m reported in the literature for plasmon-based measurements with light wavelengths of 375 nm [18] and around 400 nm [19, 20]. Our values fit however quite well to the values of 2.9 \(\mu\)m and 7.4 \(\mu\)m as derived from the tabulated data of Ref. [17]. We see differences in aluminum surface quality and roughness as the most likely reasons for any discrepancies.

Comparing the PEEM images for an excitation wavelength of 410 nm (Fig. 2) and 273 nm (Fig. 3), a few differences catch the eye. In the former case, the Al ridge shows a high photoelectron yield that is absent for the excitation with 273 nm. The same is true for some randomly distributed (assumably fabrication-related) defects on the sample that are only present for the excitation wavelength of 410 nm. These observations could be related to the difference between a two-photon emission process for the excitation with 410 nm (corresponding to 3.02 eV) and a one-photon process at 273 nm (corresponding to 4.54 eV) as these values lie below and above, respectively, the work function of a polycrystalline Al film of around 4.3 eV [21]. However, further work including the dependence of the photoemission yield on the excitation intensity and the role of the alumina film expected to be present on the aluminum surfaces is needed to clarify this point.

Summary

In summary, we have used PEEM to image SP/light interference patterns generated upon SP excitation on Al with light wavelengths of 273 nm and 410 nm. By comparing the measured data to a simple interference model, we have identified the SP propagation direction, which allowed us to deduce SP propagation lengths of 2.8 \(\mu\)m and 8.4 \(\mu\)m for excitation wavelengths of 273 nm and 410 nm, respectively. We thus demonstrate plasmonics with moderate loss in the UV and blue spectral region that on one hand could further benefit the understanding of plasmon/emitter interactions in this spectral region. On the other hand, PEEM work could proceed from here by including ultrafast schemes to probe the femtosecond dynamics of UV aluminum plasmons.