Review pubs.acs.org/CR
Cite This: Chem. Rev. 2018, 118, 2994−3031
Probing Nanoparticle Plasmons with Electron Energy Loss Spectroscopy Yueying Wu,† Guoliang Li,*,‡ and Jon P. Camden*,† †
Department of Chemistry and Biochemistry, University of Notre Dame, Notre Dame, Indiana 46556, United States Center for Electron Microscopy, Institute for New Energy Materials and Low-Carbon Technologies, School of Materials Science and Engineering, Tianjin University of Technology, Tianjin 300384, China
‡
ABSTRACT: Electron energy loss spectroscopy (EELS) performed in a scanning transmission electron microscope (STEM) has demonstrated unprecedented power in the characterization of surface plasmons. The subangstrom spatial resolution achieved in EELS and its capability of exciting the full set of localized surface plasmon resonance (LSPR) modes supported by a metallic nanostructure makes STEM/EELS an ideal tool in the study of LSPRs. The plasmonic properties characterized using EELS can be associated with geometric or structural features collected simultaneously in a STEM to achieve a deeper understanding of the plasmonic response. In this review, we provide the reader a thorough experimental description of EELS as a LSPR characterization tool and summarize the exciting recent progress in the field of STEM/EELS plasmon characterization.
CONTENTS 1. Introduction 1.1. Background 1.2. Scope of Review 1.3. Historical Perspective 2. EELS Evolution and EELS Experimental Setup 2.1. Experimental Setup 2.2. Data Analysis 2.2.1. Spectra Normalization 2.2.2. Background Removal 2.2.3. LSPR Peak Analysis 2.2.4. LSPR Mapping 2.3. Sample Preparation 2.4. EELS Simulations 2.4.1. Theoretical Background: Electron Excitation of LSPRs 2.4.2. Numerical Methods 2.5. Progress in STEM/EELS Energy Resolution 3. Probing Surface Plasmons with STEM/EELS 3.1. Spheres 3.2. Nanorods 3.3. Two-Dimensional Nanostructures 3.3.1. Nanodisks 3.3.2. Nanoprisms 3.4. Nanocubes 3.5. Alloys 3.6. Quantum-Size Regime 3.7. Coupled Plasmon Systems 3.7.1. Nanosphere Dimers 3.7.2. Nanorod and Nanowire Dimers 3.7.3. Nanoprism Dimers 3.7.4. Nanocube Dimers © 2017 American Chemical Society
3.7.5. Dimers Approaching the Quantum Regime 3.8. Damping 3.9. Plasmonic Forces 4. Connecting EELS with Far-Field Radiation 4.1. EELS and Cathodoluminescence 4.1.1. Instrumental Aspects of CL 4.1.2. Connecting EELS, CL, and EMLDOS 4.1.3. Surface Plasmon Energy Shifts in EELS and CL 4.2. EELS and Optical Scattering 4.2.1. Comparing Resonance Rayleigh and EELS 4.2.2. Fano Interferences in EELS 4.3. Electron Energy Gain Spectrum and PhotonInduced Near-Field Electron Microscopy 5. Conclusion and Outlook Author Information Corresponding Authors ORCID Notes Biographies Acknowledgments Abbreviations Used References
2995 2995 2996 2996 2997 2997 2998 2998 2998 2999 3000 3000 3000 3000 3001 3002 3003 3003 3004 3005 3005 3006 3007 3008 3009 3010 3010 3011 3012 3012
3013 3014 3015 3016 3016 3016 3016 3018 3018 3018 3018 3020 3020 3020 3020 3020 3020 3020 3021 3021 3021
Special Issue: Plasmonics in Chemistry Received: June 16, 2017 Published: December 7, 2017 2994
DOI: 10.1021/acs.chemrev.7b00354 Chem. Rev. 2018, 118, 2994−3031
Chemical Reviews
Review
1. INTRODUCTION Excitation of the localized surface plasmon resonance (LSPR) in metallic nanostructures (MNs) gives rise to strong optical absorption, scattering, and electromagnetic field confinement, leading to applications ranging from ultrafast communication1−3 and high-density memories4,5 to ultrasensitive detection,6,7 catalysis,8−11 and high-efficiency solar energy harvesting.12,13 In parallel with these impressive applications, remarkable progress in understanding the physics behind LSPRs has also been achieved through decades of research, including tuning the LSPR peaks from the infrared (IR) to ultraviolet (UV);14−18 the dispersion and damping of LSPR modes;19−21 electromagnetic hot spots generated between closely positioned nanoparticles;22−26 enhancement of nonlinear optical responses using nanoparticle plasmons;27,28 Fano resonances resulting from coupling between a broad, optically bright and a narrow, optically dark plasmon mode;29 hotelectron generation and electron transfer correlated to LSPR excitation;30−33 and optical forces in plasmonic nanoparticles,34,35 to name just a few. This fantastic progress has undoubtedly been driven by the combination of continuously improving nanofabrication techniques, rapid developments in analytical models and numerical algorithms, and relentless advances in characterization techniques. While these three pillars are equivalently important for current and future development of the plasmonics field, we focus here on electron energy loss spectroscopy (EELS) characterization of LSPRs. Given the proliferation of EELS studies in the past decade, this review focuses on measurements performed inside a scanning transmission electron microscope (STEM), whereas energyfiltered transmission electron microscopy (EFTEM)36,37 is not treated in detail here.
optical dark modes have advantages in applications such as sensing,44 low-loss nanoscale waveguiding,45 and plasmonic nanolasing.46,47 One should be careful, however, because this definition of bright and dark is not always uniformly applied in the literature. In the quasistatic limit, referred to as the dipole approximation in molecular spectroscopy,48 the wavelength of the exciting radiation is assumed to be much larger than the particle size and only modes with a nonzero dipole moment are coupled to far-field plane-wave light. In this limit, bright and dark modes can be assigned by determining if a mode has a nonzero or zero dipole moment, respectively. However, as the size of the particle becomes comparable to the wavelength of light, higher-order multipoles can couple the nanoparticle to far-field radiation, thereby allowing excitation of a dipoleforbidden LSPR mode, i.e., a mode with no net dipole moment. For example, the absorption spectrum of triangular nanoprisms displays a prominent quadrupolar resonance,49 and the Miemodes of a large sphere can be coupled to far-field radiation.50,51 It is also important to understand that an optically dark mode can be excited with a fast electron, and, under many circumstances, by altered far-field illumination sources, such as a dipole source,52 an evanescent excitation,53 a radially polarized light,54 non-normal incidence,55 or simply due to retardation effects.50 Even though all plasmons are electron-density waves in nature, LSPRs are of particular interest to researchers because they strongly interact with light and are highly tunable. Excitation of LSPRs requires a time-varying electromagnetic field, and LSPRs were historically probed by illuminating the sample with far-field plane-wave light and studying the optical response by measuring absorption, scattering, or extinction. Despite fantastic progress made by far-field optical-based techniques, the measurement precision and amount of information obtained is limited by diffraction (∼200 nm).56 Furthermore, the selection rules (described in section 2.4.1) governing the photon−LSPR interaction hinders the analysis of optically dark modes.57 In light of these limitations, near-field scanning optical microscopy (NSOM),58 which utilizes a nearfield source formed by focusing far-field light with a subdiffraction sized aperture, was developed. NSOM can probe both optically bright and dark modes with a spatial resolution up to 10 nm, yet it has issues including local heating, imaging artifacts, sensitivity, and tip perturbation which all reduce the interpretability of the experimental results. While optical near-field techniques are not further discussed here, the reader is referred to several reviews58−60 for additional information. Alternatively, EELS, performed in a STEM, exhibits unparalleled power in the study of LSPRs and has greatly deepened our understanding of LSPRs. In contrast to photons, fast or so-called “swift” electrons can be focused to the subangstrom length scale and are, therefore, able to resolve nanostructures at the atomic level. Furthermore, the electromagnetic field induced by a swift electron is spectrally broadbanded and spatially localized. A fast electron is, therefore, a highly localized source of ultrafast, evanescent, white light. These properties relax the common far-field selection rules, and fast electrons can excite the full set of LSPR modes supported by a MN. The subangstrom probe size combined with the unique selection rules of fast electrons makes STEM/EELS an ideal tool in the study of LSPRs. These properties are key features of STEM/EELS-LSPR characterizations and will be explored in detail throughout this review.
1.1. Background
To aid newcomers, we first introduce several important concepts in plasmonics. A plasmon is quasiparticle that quantizes the free-electron density oscillations in a metal, analogous to a phonon that quantizes the atomic vibrations in a crystalline material.38,39 The bulk plasmon is the quantum of longitudinal electron-density waves within the bulk of a metal, and it cannot be excited by transverse electromagnetic plane waves. The surface plasmon polariton (SPP) is the quantum of longitudinal electron-density waves confined to the metal− dielectric interface and it is often excited using the Kretschmann or Otto configuration,40 which involves the coupling between a dielectric prism and a thin metal film to match the photon and SPP wavevectors. By creating total internal reflection at the dielectric surface, an evanescent wave is induced which then excites the SPP at the metal surface. In contrast, the LSPR is a nonpropagating electron-density wave that is confined at the surface of a metallic nanoparticle. Unlike SPPs, which can be described by a continuous dispersion curve, LSPRs are characterized by a family of solutions to the Maxwell equations that depends critically on the boundary conditions imposed by the particle geometry. Each solution is labeled as a particular LSPR mode and possesses a unique energy, line width, magnitude, multipolar moment, and charge distribution. The LSPR modes are often further categorized into optically bright and dark modes, depending on whether the mode is coupled to far-field radiation or not.41−43 Since optically dark modes cannot radiate into the far-field, an important plasmon decay pathway is inhibited; therefore, they generally have longer lifetimes relative to optically bright modes. As a result, 2995
DOI: 10.1021/acs.chemrev.7b00354 Chem. Rev. 2018, 118, 2994−3031
Chemical Reviews
Review
Figure 1. (a) Schematic of EEL spectrum acquisition in a STEM equipped with a Wien-type monochromator and an aberration corrector. The schematic in the dashed square illustrates the working principal of an Omega-type monochromator. (b) An EEL spectrum of a silver dimer revealing different LSPR modes supported by the system. Also presented are the HAADF image and LSPR mode maps generated at loss energies of 1.55, 2.00, 2.70, and 3.60 eV. The green dot in the HAADF image is the beam position used for the spectrum acquisition.
1.2. Scope of Review
changing effect. It would be many centuries before the underlying mechanism was understood. The first theoretical studies date to the early 1900s when a series of papers were published to explain the scattering of light by small spherical particles.61,69−73 The first rigorous theoretical description was by Gustav Mie,61 who solved the scattering of a monochromatic plane wave by a homogeneous dielectric sphere of arbitrary size in a homogeneous medium. By analyzing the solutions to Maxwell’s equations, he related the color changing of gold colloids to the size of gold spheres changes. The theory of plane-wave scattering by a homogeneous isotropic sphere is also called Lorenz−Mie theory74 or Lorenz−Mie−Debye theory,75 due to the contributions of Ludvig Lorenz and Peter Debye to solving this problem. Despite the usefulness of Mie theory in understanding key properties of the LSPR, i.e., scattering and extinction, the physics underlying these properties was not recognized until the mid-20th century, when collective electron oscillations in thin films were probed by fast-electron techniques.76−86 Among the early EELS experiments was the first observation of the bulk plasmon by Ruthemann in thin aluminum and beryllium foils82 and its interpretation by Bohm−Pines’s plasma oscillation theory.83−85 In 1957, the existence of surface plasmons in a thin metal film was theoretically predicted by Ritchie78 and soon experimentally demonstrated in 1959 by Powell and Swan79 using reflection of incident fast electrons. Following this work, Stern and Ferrell, in 1960, used transmission experiments to probe the dispersion of the electron’s energy loss as a function of scattering angle and its dependence on the thickness of oxide coatings.87 These pioneering studies established a solid foundation for understanding LSPR phenomena. The experimental observation of LSPRs in metal spheres excited via bombardment with 50 keV electrons was reported by Creuzburg et al.88 in 1966 and Fujimoto et al.89 in 1967. The phenomenon was theoretically treated by Fujimoto and Komaki using a hydrodynamical classical model90 and by Crowell and Ritchie, who calculated the radiative decay rate in
Our purpose in this review is 2-fold: first, we wish to provide the reader with a practical experimental description of EELS as a LSPR characterization tool, and second, we wish to summarize the exciting recent progress enabled by EELS studies. While the background material is available elsewhere, we endeavor to make our presentation self-contained so that newcomers can easily follow the discussion, and a list of commonly used acronyms (Abbreviations Used) is provided. We begin with a short historical perspective and theoretical background before turning to a detailed description of the instruments and experimental procedures. Special attention is paid to data acquisition and analysis procedures, along with sample preparation. The remainder of the review highlights EELS characterization of fundamental LSPR properties. After an in-depth discussion of isolated nanospheres, nanorods, and nanocubes, as these representative geometries trace their lineage back to the well-known Mie theory.61 We continue with studies of LSPRs in alloys, quantum plasmons, plasmon resonances in coupled-particle systems, LSPR damping measurements, and plasmon force studies illustrating in detail how EELS has been utilized to explore the rich physics of the LSPR. Throughout the review, we strive to make comparisons between EELS measurements and the often more familiar optical properties. We conclude with an outlook on potential future developments in the field. The topic of EELS characterization of plasmons is a diverse and expanding field, with new studies appearing at an everincreasing rate. Interested readers are also directed to previous excellent reviews by Garciá de Abajo,62 Hohenester,63 Kociak et al.,64 Kociak and Stephan,65 Roth et al.,66 Colliex et al.,67 and Cherqui et al.68 for additional information and viewpoints. 1.3. Historical Perspective
The use of metallic nanostructure LSPRs dates back to ancient times when, for example, gold nanoparticles were incorporated into the famous Roman Lycurgus cup to create its color2996
DOI: 10.1021/acs.chemrev.7b00354 Chem. Rev. 2018, 118, 2994−3031
Chemical Reviews
Review
Figure 2. EELS spectrum processing (a) the raw (middle panel) and normalized (bottom panel) EEL spectra obtained from a 45 nm silver sphere sitting on a 30-nm-thick Si3N4 substrate. The red and blue traces are acquired via aloof and inner beam positions, respectively (top panel). The green rectangle is the region of interest for the acquisition of the spectrum images (SIs) in panel d. (b) Background removal via subtraction of a reflected tail model (top panel), a logarithm-tail model (middle panel), and a background spectrum obtained far from the particle (bottom panel). (c) Lorentzian fitting of a residual spectrum. The residual spectrum used for fitting is the solid-black trace in the middle panel in part b. (d) HAADF image and two LSPR maps generated from the raw data and normalized data. These maps are generated by plotting the spectral intensity at 3.54 eV over the SI.
the dipole approximation.91 Shortly afterward, the first studies employing silver and gold particles were reported by Kreibig and Zacharias in 1970.92 These early EELS experiments were not able to reveal the spatial profiles of LSPRs, and it was not until the emergence of the energy-filtered TEM mode (EFTEM) and spectrum-image (SI) mode in the STEM in the 1980s that mapping studies became possible.93
not until the mid-2000s that the SI method was employed to map LSPRs. This is because early EELS experiments were limited by long acquisition times and generally had no access to features in the visible region (400−700 nm,
