Probing plasmonic nanostructures by photons and electrons

Exploiting photons and electrons opens up exciting new capabilities to study complex plasmonic nanostructures and related local fields.


Introduction
Frequencies of collective oscillations of the free electrons in metal nanostructures, called surface plasmons, fall in the optical range of the electromagnetic spectrum. 1 In complex metal nanostructures, plasmon hybridization results in a rich plasmon spectrum. 2,3 Plasmon modes can be dipole active, leading to so-called bright modes, but plasmon oscillations can also give rise to dark modes which have a zero net dipole moment and usually do not show up in optical absorption experiments. Resonances between light and the surface plasmons result in enhanced optical elds in the vicinity of metal nanostructures. These high local elds have the potential to revolutionize optical spectroscopy and open up exciting new capabilities for almost all photon-driven processes.
Surface Enhanced Raman Scattering (SERS) 4 is a prominent, example to demonstrate the power of plasmonic approaches. Raman spectra of a single molecule can be measured in the enhanced local elds of metal nanostructures. [5][6][7][8] In particular, single molecule detection can be also achieved using nonresonant excitation in the near-infrared range. 5,8 Giant SERS enhancement has been obtained only when composites of metal nanostructures such as aggregates of silver and gold nanoparticles were used as enhancing plasmonic structures. 5,7-10 Interestingly, extinction spectra of plasmonic structures providing such a high enhancement level do not necessarily show resonances in the NIR, the range which was used for excitation when high enhancement is observed. 5,11 This can be understood in the framework of the plasmonic properties of the enhancing aggregate structures. Theory predicts intense and extremely localized elds, for aggregates formed by metal nanoparticles, fractal structures and semicontinuous random metal lms related to bright and dark plasmonic eigenmodes in these structures. [12][13][14] Also in several other plasmonic structures, coupling between dark and bright modes, which results in the appearance of Fano resonances has been shown to be important for generating high local elds. [15][16][17] Overall, the current experimental and theoretical insight shows that dark modes are of essential importance for the generation of highest local elds and for the spectroscopic performance of a plasmonic nanostructure. Therefore, comprehensive information on the plasmonic spectrum including bright and dark modes is of basic interest for deeper understanding of plasmon supported spectroscopy and for optimizing plasmon supported processes. Experimental tools for a comprehensive characterization of the plasmonic structures also have to address the strong connement of their local elds. Specically, SERS experiments indicate that highest enhancement level is available only for very few molecules 11 residing in extremely small volumes. This means that the local connement of hottest hot spots is in the sub-nanometer range.
In this minireview, we discuss methods for comprehensive probing of the plasmonic spectrum of metal nanostructures and of the related local elds. We start with considering optical experiments using one-and two-photon excitation. Tuned laser excitation enables to study the dependence of local elds on the photon energy. Spectroscopic signals collected from single molecules deliver sensitive information on the near eld at the location of the molecule.
Complementary to photons, we discuss electron-energy-loss spectroscopy (EELS) as an emerging novel characterization tool for plasmonic nanostructures which can provide high spatial resolution at the subnanometer scale. [18][19][20] In addition to excellent spatial resolution, EELS can probe bright and dark modes and, by this way, provides comprehensive information on the plasmonic spectrum. [21][22][23][24][25][26] Probing plasmon resonances and local fields using photons Absorption and elastic scattering of light allow a straightforward probing of plasmon resonances. Beautifully colored glass windows in old cathedrals are related to scattering and absorption of sun light by surface plasmons of small metal nanoparticles incorporated into the glass matrix by old artists and masters hundreds of years ago. Today, hybridized surface plasmon resonances in aluminum nano disks and nanoholes can generate bright reective colors to be used for coloration of plastic consumer products. 27 Simple measurements in commercial spectrometers yield the extinction spectrum (losses due to absorption and scattering) of metal nanostructures. Usually, these measurements are performed on an ensemble of nanoparticles. For probing the plasmon spectrum of individual nanostructures absorption does not provide sufficient sensitivity, and collecting the scattered light using dark-eld illumination is the more efficient experimental choice. Fig. 1a visualizes the plasmon resonances of individual silver nanoparticles of different size and shape which "glow" in different colors in a white light illuminated dark-eld microscope. Combining this microscope with a spectrometer allows us to measure the plasmon spectrum of individual nanostructures. 28 Usually, extinction probes bright plasmon modes. Dark modes become visible when they spectrally interfere with a bright mode and, by this way, borrow some oscillator strength from the bright mode. Such a Fano resonance is visible as an asymmetric dip in the spectrum around the energy of the dark mode. Sophisticated dark eld microscopy which enables illumination with different incidence angle can identify shis of Fano resonance minima in a nanoparticle aggregate in dependence on the angle of incidence of the excitation light. 29 This provides insight into the bright and dark plasmon modes in the structure, and into their excitation mechanisms.
In addition to knowing the complete plasmon spectrum including bright and dark modes, information on local optical eld in the vicinity of a plasmonic structure is of basic scientic and practical interest. One-and two-photon excited spectroscopy performed in these local elds is a useful approach for getting this information. Spectroscopic signals collected from single molecules provide sub-nanometer probing of eld amplitudes and polarization direction. For example, single molecule uorescence has been exploited to map the local eld of around a plasmonic structure. 30 In case of one-photon excitation, the uorescence signal quadratically depends on the local eld amplitude. As a scattering process, SERS signals increase with the amplitudes of local optical eld to the power of four. 31 This makes SERS a very sensitive probe for local elds, particularly when also two-photon excited surface enhanced hyper Raman scattering (SEHRS) is applied. In SEHRS, two photons are scattered simultaneously resulting in a Raman signal shied relatively to the second harmonic of the excitation laser 9,32 (see Fig. 2a).
As a two-photon excited scattering process, SEHRS depends on the local eld amplitudes to the power of six. 9,32 Therefore, the ratio between SEHRS and SERS signals P SEHRS /P SERS delivers direct information about local optical eld intensities. Analyzing this ratio for different excitation wavelengths (Fig. 2b) reveals the dependence of local optical elds on the photon energy. The measured increase in intensity of local optical elds with increasing wavelengths as it is shown in Fig. 2b is qualitatively in agreement with computations of local elds for the hottest hot spots. 33 Optical measurements as discussed in Fig. 2 deliver information on local optical elds from a (sub) nanometer probed volume, the location of a single probe molecule in the hot spot. However, due to a spatial resolution of about half of the applied wavelength, these experiments do not provide information where the hot spot is located on a plasmonic structure nor about the eld distribution inside such a "hot area".
In order to get information on the location of hot areas around a plasmonic nanostructure and about distribution of  local elds in these hot areas, sophisticated methods for superresolution imaging and mapping beyond the diffraction limit have been developed based on surface-enhanced uorescence 35,36 and SERS signals. 37 In these experiments, plasmon enhanced spectroscopic signals of single molecules are measured across the diffraction limited spot and localized by a two-dimensional Gaussian t. By this way, emitters can be localized with a precision around 10 nm. Fig. 3 shows not only an intensity map of single uorophores in hot areas, it also identies spectral modulation of the uorescence signal across the hot area.
The spatial distribution of surface enhanced Raman signals at the single molecule level can be used basically in the same way as uorescence for the mapping of plasmonic local elds. Super resolution imaging of SERS hot spots including also the ability to obtain simultaneously spatial and spectral information is discussed in comprehensive recent reviews. 37,38 Another method to measure the distribution of the local eld intensity exploits changes in the decay rate of a uorophore in the vicinity of a metallic nanostructure. This decay rate is proportional to the local density of plasmonic eigenstates which, in turn, is proportional to the intensity of the local eld. Giant uctuations in local elds of a fractal metal surface 39 are identied by enhanced uctuations in the decay rates of molecules. 40 However, even sophisticated optical measurements cannot provide the spatial resolution for mapping out local optical elds with sub-nanometer resolution. As discussed in the next section, new approaches for probing plasmonic structures using fast electrons enable such experiments.

Probing plasmon resonances and local fields using electrons
In general, the energy-momentum mismatch between freely propagating charges and photons prevents the linear coupling of fast electrons and light in vacuum. However, evanescent local optical elds carry the necessary complex momentum to make possible a coupling between freely propagating charges and photons. As an alternative to photons, surface plasmons can also be excited by low-energy 41 and high-energy electrons. 42 This opens up new ways for probing local elds of plasmonic nanostructures by electron energy loss spectroscopy (EELS). 19,20,43 In an EELS experiment, a beam of monoenergetic fast electrons interacts with the plasmonic nanoparticles. The electric eld associated with the fast ying electrons affects the free electrons in the metal nanostructure and generates a charge displacement, leading to a collective oscillation of conduction electrons at their plasmon resonance frequency. Due to this plasmon excitation, the fast electrons are losing the respective amount of energy. The energy spectrum of the electron beam aer passing the plasmonic structure provides information on the plasmonic spectrum.
The capability of fast electrons to lose energy to the collective oscillation of the conduction electrons in a thick metallic foil has been discussed for the rst time more than 50 years ago. 44,45 In 1982, the excitation of localized surface plasmons was observed in EELS experiments on aluminum nanospheres. 42 Surface plasmon resonances of isolated nanospheres and dimers (two-sphere-systems) were measured as low-energy peaks in EELS spectra along with the bulk plasmons of the metal. In particular, it was shown that EELS signals spread outside the nanostructure and that well-dened different plasmonic eigenmodes can be excited depending on the location of the electron beam relative to a nanoparticle dimer. 46 Fast electrons came into the focus of interest as a highly localized excitation source of surface plasmons in 2007. 18,19,47 EELS performed in a scanning transmission electron microscope (STEM) allows imaging of surface plasmons at subnanometer resolution. A general formalism describing the interaction of fast charged particles and plasmonic samples was described. 20 Note that the monochromatic electron beam has energies of hundreds of keV, while electron energy loss appears typically between 1 and 3.5 eV, since plasmon modes typically appear in the optical range of the electromagnetic spectrum between 400 and 1000 nm. Challenges in EELS experiments on plasmonic nanostructures are discussed in a recent comprehensive review article. 43 In general, the same optically active resonances, which determine the optical response of plasmonic structures, can be excited and studied by the electron beam in EELS. Since the electric eld related to the fast electron varies at the scale of the nanostructure, i.e., very rapidly compared to optical eld, the electron beam also excites non-dipolar modes. The EELS spectra in Fig. 4 show this for plasmonic aggregates: in addition to a strong mode around 3.3 eV, which also dominates the absorption spectrum of silver nanoparticles, there exist modes in the NIR range. Excitation wavelengths used in one-and twophoton excited NIR-SERS experiments at extremely high enhancement level match these NIR plasmon resonances identied in the EELS studies. 34 Also EELS spectra of discontinuous aluminum lms show plasmon resonances in the NIR and strong SERS spectra have been obtained for these lms at NIR excitation, even though the dielectric function of this metal supports the UV range. 48 Images based on EELS intensities in the NIR range (see Fig. 4) show that areas of highest local elds correlate with those of lowest electron energy loss intensities. This is obvious by a comparison of EELS intensity maps and computations of local elds for dimers and trimers. 49,50,52 The inverse relation between EELS maps and the local eld distribution has been observed independently on specic geometries of nanoaggregates. 26,34,51 Measurements on dimers with gap sizes between 12 nm and 1 nm (Fig. 4b) show a direct, almost linear relation between the EELS signal measured in the gap and the gap width. In contrast, computations for optical excitation show local eld intensity enhancement proportional to 1/(gap size). 52 The exact relationship between the electron energy loss and the local eld strengths remains a subject of discussion. 53,54 Knowing the relationship between optically excited local elds and electron energy loss signals opens up the exciting possibility of probing local optical elds in the hottest hot spots and for experimental validation of computations that show dramatic spatial variations in local optical elds. 55,56 For example, Fig. 5 provides some experimental evidence of the nanometer-scale connement of plasmonic eigenmodes 39 in random silver lms around the percolation threshold which gives rise to extremely high hot spots on these structures. 57

Beyond classical and ultrafast
Usually, plasmonic nanostructures for creating high local elds are in dimensions between 10 and 100 nm, i.e., in the case of metals, they are bulk materials that can be treated by classical theory. However, for very small particles below 5 nm or for small interparticle gaps in sub-nanometer dimension, plasmonic structures can exhibit also quantum nature. 25,[58][59][60][61][62][63][64] New experimental methods for probing plasmonic systems at high energy and spatial resolution enable us to observe such quantum effects. In individual metal nanoparticles, quantum effects result in characteristic blue shis of the dipolar plasmon resonances. 58,59 For nanoparticle dimers with narrow gaps, theory predicts the appearance of new charge transfer plasmon modes, as well as a breakdown of high local elds in the gap as hallmarks of quantum nature. 65 EELS spectra collected from individual dimers with decreasing gap sizes down to atomic dimension allow to monitor the evolution of these modes. 25,60 They are related to the onset of tunneling and, when the two spheres start to touch, to an oscillating current through a nanobridge between them. The classical bright and dark modes of the dimer should disappear with decreasing gap widths. Interestingly, silver dimers with atomic scale gaps can exhibit a regime, in which charge transfer modes and classical modes coexist. 25 The preservation of classical dipolar gap modes in the quantum regime when tunneling occurs can have important implications for local elds of plasmonic nanostructures in subnanometer gaps. In optical experiments, while SERS enhancement for a "classical" plasmonic dimer increases with decreasing gap widths, 10 a decrease of the enhancement has been reported for gaps below 6.7 Angstrom when quantum effects emerge. 66 Particularly exciting capabilities for probing plasmonic structures are opened up by experiments where fast electrons interact with surface plasmon elds which are pumped by incident photons. Optical pumping of plasmons gives rise also to an electron energy gain signals. 67 Such experiments using loss and gain allow to combine the spatial resolution of electron microscopy with energy resolution as it can be achieved in optical excitation and probing. Moreover, today, not only light pulses but also electron pulses can be generated at femtosecond duration. This allows to probe photon excited plasmonic elds by electrons with temporal resolution at the time scale of the plasmonic elds. 68 By exploiting all these capabilities, a new type of electron spectroscopy so-called photon-induced near-eld electron microscopy (PINEM) has been invented. 69,70 Experiments using PINEM enable space-time imaging of localized surface plasmon elds aer excitation by incident photon pulses 71 and provide basic insight into the nature of surface plasmons.

Summary and outlook
Comprehensive information on plasmonic nanostructures and related local optical elds can be achieved by exploiting both photons and electrons. Their complementary use enables to access a new level of information by combining the high energy selectivity of laser radiation with the atomic scale spatial resolution of electron microscopy. Also, in the meantime, nanotechnology has allowed us to prepare tailored complex metal nanostructures such as nanoparticles at controlled size and  shape or aggregates and composites with well characterized interparticle gaps down to atomic dimension. 10,66 These developments in spectroscopy and nanotechnology provide us with tools and targets for comprehensive experimental studies of plasmonic excitations and related local elds. New capabilities to characterize plasmonic properties might become of particular interest for structures where quantum effects play a role, such as metal nanoaggregates with atomic scale interparticle gaps. For example, for dimers with subnanometer gaps, EELS and optical spectroscopy can indicate the onset of quantum effects. 25,60,66 Composites of nanoparticles with very narrow gaps at the transition from the classical to the quantum regime seem to provide the highest local elds.
In the future, exploiting photons and electrons to study complex plasmonic structures will allow us to probe their complete plasmon resonance spectrum, to generate maps of local elds and hot spots at subnanometer resolution, and to determine local eld intensities as a function of photon energy. These studies are of basic interest for the deeper understanding of the physics behind plasmonics and for enhancing and optimizing plasmon supported spectroscopy and other photondriven processes.