- Abstract•
- Introduction•
- Results•
- Discussion•
- Methods•
- Additional information•
- References•
- Acknowledgements•
- Author information•
- Supplementary information
SPP imaging
When irradiated by light, isolated metallic nanowires have been shown to behave as quasi-1D plasmonic nanoantennas, whose radiation patterns are governed by the properties of the excitation (wavelength, incident and polarization angles) and their geometry with respect to the orientation and aspect ratio of the wire22, 23. The incident light photoexcites propagating SPPs in the metallic nanowire whose back-and-forth reflections from the wire ends give rise to an SPP standing wave, making the straight nanowire the plasmonic equivalent of an FP nanoresonator14. In general, the confined electromagnetic fields of such SPP standing waves, whether electron- or photoinduced, are captured through scanning-based techniques such as combined STEM-EELS5, 18, 24 or scanning near-field optical microscopy (SNOM)25, which probe the electric near-field component perpendicular to the sample plane. Here we alternatively employ a fundamentally different, field-of-view approach, based on ultrafast transmission electron microscopy (UTEM).
UTEM is typically performed by modifying a conventional TEM such that ultrashort bunches of imaging electrons, containing at most one particle each26, 27, can be photoemitted from the cathode by fs laser pulses. Optical access is also provided for photoexcitation of the specimen, and the delay between the two pulse trains is controlled via an optical delay line, allowing for time-resolved optical-pump/electron probe experiments (see Fig. 1a and Methods)27, 28. When the specimen being imaged is a (metallic) nanostructure, the temporal and spatial evolution of a photoinduced SPP standing wave can be visualized via the PINEM imaging technique19, 20. PINEM relies on the inelastic exchange of energy quanta between the photoinduced electromagnetic SPP wave and the relativistic imaging electrons, which probe the SPP electric field component along the electron propagation direction20, 29, 30, 31, 32, 33. In our experiments, nanowires (~50 nm radius, few-μm length) are isolated and dispersed on a graphene-covered TEM grid and photoexcited using a pulsed 800-nm laser beam at a 5-mJ cm−2 fluence, corresponding to a peak excitation energy density of ≃10 GW cm−2. Under these experimental conditions, excitation of SPPs by the electron probe beam is a much weaker effect, which can be considered entirely negligible32. The few-layer graphene substrate is used to efficiently dissipate the laser-induced heat.
Figure 1b shows the energy spectra of the probing electrons before and after interaction with an isolated, photoexcited nanowire. The spectrum at negative delay (Δt=−1.6 ps) shows the initial energy distribution of the electron bunches, that is, the zero-loss peak (ZLP), whose full-width half-maximum (FWHM) determines the spectral electron energy resolution as better than 1.1 eV. By contrast, when the optical pump and electron probe are overlapped (Δt=0 ps), the interaction with the photoinduced SPP electric field leads to acceleration (deceleration) of the probing electrons, and the corresponding quantized gain (loss) of energy (peaks found at ΔE=±n · ℏω). In the resulting electron energy spectrum, the (net) exchange of up to nine energy quanta can be observed. The panel further depicts the continuous temporal evolution of this electron energy spectrum as a function of Δt, showing a 1.5-ps FWHM cross-correlation of the optical pump and electron probe pulses. As was previously shown, when optimally configured for time resolution, the same system can readily achieve a sub-ps temporal cross-correlation27.
The electromagnetic field of the photoexcited SPP in the nanowire can be captured by using an imaging energy filter to select out only electrons that have gained energy (see white arrow in Fig. 1b), and subsequently reforming an image. Repeating this procedure at different time delays produces a series of images (Fig. 1c–g) of the temporal evolution of the SPP field, showing its interferometric standing wave pattern in the silver FP nanoresonator. This type of modal interference pattern reveals the wave character of the electromagnetic SPP field, and is typically observed when the properties of light excitation (wavelength and polarization) are close to a resonance condition of the excited nanowire.
Plasmonic nanoresonators
For symmetry reasons, light at normal incidence that is polarized parallel to the wire long axis exclusively excites odd-order SPP modes, that is, modes that have an odd number of SPP field nodes m (ref. 14). By contrast, electron-excitation of SPPs involves no such symmetry-based selection rules, such that in STEM-EELS both odd- and even-order SPP modes can be excited. To be able to photoexcite even-order SPP modes as well, one requires an excitation geometry where the light is incident at an oblique angle and the azimuthal angle between the light polarization and the wire long axis is nonzero14. In general, under such photoexcitation conditions (in s-configuration), different SPP modes can be excited at the same time. This can result in non-trivial photoinduced SPP field distributions, which require numerical simulations to reproduce and understand them13. As shown in Fig. 2, these general selection rules are valid for the photoexcitation of SPPs in PINEM as well. First, we image a 3.4 μm long, ≃45 nm radius nanowire, illuminated by s-polarized light with an azimuthal angle of ϕ=0° with respect to the wire long axis, see Fig. 2a,b. The resulting photoinduced SPP standing wave corresponds to an odd-order mode (m=11), in excellent agreement with preceding reports13, 14, 23, and with our own finite-element simulation (see Methods) shown in Fig. 2c. The experimentally estimated SPP wavelength ( 
) and wavevector ( 
) for this mode are ≃615 nm and ≃10.2 μm−1, respectively, from the average antinode distance dav≃308 nm in the standing wave pattern (see the spatial profile in Fig. 2a). At an energy of 1.55 eV, this is in excellent agreement with both the calculated and experimental dispersion curves of resonant SPP waves in nanoantennae5, 13, 18, 34. In general, the resonance condition of an order-m SPP mode in a 1D FP resonator of length L can be written as14, 18:
where δθ is the SPP phase shift on reflection from the resonator ends, which is often negligible for higher-order modes13, 18. Calculating the expected SPP wavelength for a resonant mode of order m=11 in an idealized 1D FP resonator of length L=3.4 μm (δθ=0) yields a value of 618.2 nm, which is in good agreement with the estimated 
. The slight difference, disregarding the estimation error margin, would imply a negative reflection phase shift δθ according to equation (1). However, it is more intuitively interpreted in terms of a shortened effective wire length Leff., resulting from the fact that the hemispherical wire caps were included in the determination of L14, 18.
Next we illuminate a 5.7 μm long nanowire using light polarized at a ϕ=45° azimuthal angle, obtaining a non-trivial SPP field distribution in which antinodes of opposite phase are concentrated on alternating sides of the wire (Fig. 2d). Under these excitation conditions, the nodal lines are clearly at an angle with respect to the long axis of the nanowire. As shown in Fig. 2e, the features of this transversally asymmetric SPP mode (m=17) are numerically accounted for by our finite-element simulations. The continuous dependence of the simulated SPP field distribution on the azimuthal polarization angle of the light excitation is illustrated in Supplementary Movie 1. To our knowledge, the experimental observation of such an asymmetric distribution of the SPP field using either SNOM or STEM-EELS techniques has never been reported. In principle, these features could also be observed in SNOM experiments14. The absence of such asymmetric field distributions in STEM-EELS experiments is likely related to the differences between electron- and photoexcitation of SPPs in terms of mechanism and selection rules20, 24, 31, 32, 33. For photoexcited SPP modes, Dorfmüller et al.14, 23 showed that while odd-order modes emit the strongest field in the direction perpendicular to the nanowire, thus maximizing the PINEM effect, the even-order modes have a minimum of radiation in this direction, resulting in a more difficult detection. However, at the same time, the relative dipole coupling strength of odd-order modes was shown to quickly decrease with m, while staying fairly constant for even-order modes. In our experiments, two different nanowires were used to maximize the strength of the photoinduced SPP field in the different excitation geometries. The excellent agreement between the experimental and simulated SPP field distributions demonstrates the potential of controlling SPPs using an external light field.
Energy-space SPP mapping
After interacting with the photoinduced SPP field, the imaging electrons carry all the information about the exchange encoded in their spatial and energy distributions. In EELS and 2D energy-filtered imaging, one typically collapses one or more of these coordinates to obtain either a 1D energy spectrum (collapsing both spatial coordinates) or a 2D image (collapsing the electron energy). Instead, to simultaneously observe both the quantized spectrum and its spatial distribution, here we resonantly excite an odd-order SPP mode in an isolated nanowire and align the corresponding image such that the wire long axis is parallel with the vertical detector axis (see Fig. 3a). By then collapsing only the perpendicular (horizontal) spatial coordinate, we obtain an image that contains spectroscopic information along the horizontal detector axis and spatial information along the vertical detector axis. The experimental energy-space map is shown in Fig. 3b. To optimally resolve the inelastic exchange process, we zoomed in on a selected section of the nanowire (4.6 μm length, ≃61 nm radius, ϕ =0°). As is clear from the experimental image, taking a horizontal cut (horizontal dashed line) yields the quantized spectrum of the interaction between the SPP field and the imaging electrons. At the same time, by taking a vertical cut at an energy corresponding to one of the peaks in the energy spectrum (vertical dashed line), the spatial distribution of the interaction between single electrons and a discrete number of photons is obtained, Fig. 3c, displaying the typical interference fringes of the resonant SPP standing wave. Though both the wave and particle character of SPPs were already observed separately in individual, tailored experiments35, 36, 37, here we obtain a very direct and illustrative view of both aspects of the SPP field simultaneously in a single experiment.
Комментарии
Глядишь, и я такое намалюю.
http://www.nature.com/ncomms/2015/150302/ncomms7407/full/ncomms7407.html
Там картиночки несколько другие
Цветом обозначена не амплитуда, а абсолютная высота точки поверхности.
Но не сильно переврал - в оригинале это не напряженность поля, а число электронов, то есть заряд.