Metal nanostructures have enabled rapid advances in numerous interdisciplinary fields such as plasmon enhanced spectroscopy, solar energy conversion, integrated optical nanocircuitry, biosensing, and plasmonic catalysis.^[1–4] All of these distinctive optical properties arise from the localized surface plasmon (LSP) resonance associated with the collective oscillations of free electrons. These features lead to unique optical properties and enable manipulation of light on a nanoscale. Notable features include localized concentration and enhancement of electromagnetic fields, LSP modes that can be tuned by size, shape, and dielectric constant of the surrounding medium, and even the coupling of LSP modes with themselves to yield Fano resonance and plasmonic metamaterials.^[5]

To date, most experimental methods used to investigate singlemetal nanostructures have been based on optical scattering or absorption.^[6] For example, the scattering of surface plasmons has been applied to studies of the antenna characteristics of individual and coupled gold nanoparticles.^[7] Furthermore, optical absorbance and photothermal imaging have been used to determine the orientation of metal nanostructures.^[8] In addition to these features, metallic materials present an intrinsic luminescence which was first reported in 1969 for bulk gold materials and then later on for rough thin films.^[9–11] However, the quantum yield of this luminescence is low with an efficiency of approximately 10⁻¹⁰. Recently, the one-photon luminescence (OPL) efficiency of gold nanorods was found to be in the order of 10⁻⁶.^[12] Although this quantum yield was low, the light emission process was compensated for by the large excitation cross-section of plasmonic nanostructures, thus leading to a large number of emitted photons; hence, this type of fluorescence can appear brighter than single organic fluorophores or semiconducting quantum dots.

The intrinsic photoluminescence generated from metals does not undergo photobleaching or photoblinking and enables stable observations of the spatial distribution of the emitted field of a single antenna.^[13,14] Characterizing plasmonic nanostructures from their intrinsic luminescence provides an alternative flexible and simplified optical method complementary to conventional scattering and absorption techniques. The luminescence of metal nanostructures has attracted considerable interest because of its potential applications in optical recording,^[15] bioimaging tags,^[16,17] biosensing,^[18] orientation probes,^[19] local temperature detection,^[20] and plasmonic resonant mode identification.^[21] In addition, these phenomena are related to the broad continuum background in surface-enhanced Raman scattering^[22,23] and surfaceenhanced fluorescence.^[24]

A deeper understanding of this process would be helpful for understanding and optimizing plasmon enhanced spectroscopy.^[25] However, there remains some controversy concerning the physical mechanism of this effect. Here, we tentatively call this luminescence process photoluminescence (PL). This process was first understood based on interband transitions. Taking gold as an example, a photon excites electrons from the d-band to the sp-band. Then, electrons in the sp-band relax to sp-electronic states near the Fermi level as shown in the inset of Fig. 1(a). At the same time, holes relax to the same k point. Electron thermalization occurs over 100 fs to 1 ps.^[26] The electrons finally recombine with holes to emit photons.^[27] Briefly, when a metal material is illuminated by light with an energy higher than that of the interband transition, electrons relax and emit photons as shown in Fig. 1(a).

Fig. 1.

Figure Option ViewDownloadNew Window

Fig. 1. (color online) (a) PL of bulk gold. Experimental OPL spectrum of gold thin film excited at 405 nm. The inset shows that OPL from bulk gold evolves by a three-step process: a d-band hole is created by 405-nm excitation and scattered within the d-band to a state below the filled sp-band states. The hole subsequently recombines radiatively with an sp-electron (from Ref. [27], Fig. 1). (b) Schematic diagram illustrating surface plasmon polariton (from Ref. [28], Fig. 1(a)). (c) Normalized extinction spectra of as-prepared gold nanorod solutions. Insets show transmission electron microscope images of gold nanorods on a copper grid. The average approximate sizes of nanorods are, from left to right, 37 × 19 nm, 50 × 12 nm, and 50 × 8 nm. Scale bars are 50 nm (from Ref. [15], Fig. 2(b)). (d) Schematic diagram illustrating localized surface plasmon (from Ref. [28], Fig. 1(b)).

In addition to the electronic band structure, surface plasmon resonance is another important factor that influences luminescence processes, particularly in nanostructured metals. Metals possess a negative real and small positive imaginary dielectric constant that enables metals to support surface plasmons, i.e., collective oscillation of free electrons.^[28] Figures 1(b) and 1(d) show two kinds of collective oscillations for a metal film and metal nanoparticle. Figures 1(a) and 1(c) show corresponding representative photoluminescence spectra. There is a strong electric field at the interface of the metal and the dielectric because of surface plasmon resonance. The high density of optical states caused by surface plasmons enhances the rates of both light absorption and emission.

Fig. 2.

Figure Option ViewDownloadNew Window

Fig. 2. (color online) (a) PL (solid lines) and scattering (dots) of a gold nanorod in different media. (Inset) LSP peak of nanorods in different media and fitting line (from Ref. [18], Fig. 2). (b) Anti-Stokes emission as a function of photon energy. Data are respectively fitted (lines) by a Boltzmann statistics distribution (blue solid line) and Lorentz line shape (purple dashed line) for the anti-Stokes part only. The inset shows the anti-Stokes spectra at different illumination light intensities with corresponding fitting lines (from Ref. [20], Fig. 3). (c) A complete analysis of three components of plasmon-enhanced fluorescence spectrum and coupling-configuration dependent spectra. OPL spectra of free single gold nanorod (black) and free single fluorescent nanodiamond (red) before coupling, plasmon-enhanced spectrum (blue), and scattering spectrum (green) after coupling of a single gold nanorod and a single fluorescent nanodiamond (from Ref. [24], Fig. 2 (a)). (d) Decay dynamics of multiphoton photoluminescence for antenna arrays of L = 140 nm (red curves) and L = 80 nm (black curves). The instrument response function (IRF) is traced with a black dashed line (from Ref. [51], Fig. 3(c)).

A variety of different metals and nanostructures have been extensively studied including gold, silver,^[29] aluminum, and copper.^[10] These materials have been formed into various types of nanoscale structures with different sizes and shapes and in various coupling configurations.^[30,31] In this review, most of the examples discussed concern gold nanostructures and have been widely investigated because of their ambient stability and accessibility. Notably, the same physical mechanism should also operate in other materials supporting the oscillation of free carriers. Here, we review recent advances concerning the luminescence of metal nanostructures excited by continuous wave (CW) light, pulsed lasers, and electrons together with potential applications of this luminescence. The physical origin of the luminescence is also discussed.

Fig. 3.

Figure Option ViewDownloadNew Window

Fig. 3. (color online) (a) In vivo imaging of single gold nanorods in mouse ear blood vessels. Overlay of transmission image (light blue) and single-frame TPL image. Two single nanorods (red spots) superimposed by a line scan (white). Image size is 175 μm × 175 μm (from Ref. [16], Fig. 5(c)). (b) Topographic images of gold nanorods and two-photon-induced photoluminescence images of gold nanorods (from Ref. [74], Fig. 2)). (c) The process of reshaping gold nanorods in the focal volume of focusing objective. Some nanorods that couple with light can be reshaped for optical recording (from Ref. [15], Fig. 2(a)).

Photoluminescence is weak luminescence from metal nanostructures. Here, any secondary light emission at energies different from the excitation laser energy are defined as intrinsic luminescence. When the power-law exponent of the emission intensity as a function of the excitation power is equal to 1, this luminescence can be called OPL. OPL usually takes place after excitation by CW light. OPL from bulk gold excited by CW light was first reported by Mooradian in 1969.^[9] Later, gold thin films with rough surfaces were found to possess considerably higher emission efficiency compared with that of bulk gold by Boyd et al. in 1986.^[10] The effects of LSP on rough surface protrusions has been proposed to explain light emission enhancement. In 2000, Mohamed et al. showed that gold nanorods capped with micelles fluoresced with a quantum yield more than six orders of magnitude higher than that of bulk metals.^[12] The OPL spectrum of gold nanoparticles always closely follows the plasmon resonance band of a particle. Such a spectral coincidence can be tuned from the visible to the infrared in gold nanorods or gap plasmons.^[32,33] The LSP together with the lightning-rod effect have been used to explain the enormous enhancement of interband transitions through local field enhancement. Following these investigations, the OPL of metal nanoparticles has attracted attention because of its many potential applications. Nanostructures such as nanorods, nanoparticles,^[34] nanoflowers, nanocubes,^[35,36] core-shell structures,^[37] and dimers,^[31,38] have been prepared by chemical and lithographical methods,^[33] and their OPL properties had been extensively investigated.

The shape of the OPL Stokes spectrum resembles dark field scattering when the OPL spectrum is compared with the dark field scattering spectrum of single nanostructures.^[32] Slight spectral differences between OPL and scattering have been observed in previous reports.^[19,32,39] Recent studies have attributed these differences to the thermalization of excited electrons to form a broad and inhomogeneous population distribution in the conduction band.^[29,38] The difference becomes large when the nanostructures possess multiple or broad LSP modes.^[40] Yin et al. designed a special nanostructure to obtain a blue-shifted OPL spectrum by controlling the polarization of the excited light.^[41] Moreover, there were clear differences between the OPL anti-Stokes spectrum and scattering. These differences are likely associated with the distribution of electrons. Various line shapes have been discussed based on Bose–Einstein-like and Fermi–Dirac-like distributions.^[42] The OPL Stokes spectral shape can be fitted by a Lorentz line shape. Numerical simulations and experimental studies have shown that the peak of the Lorentz line varies with the refractive index. A linear relationship exists between the peak wavelength and refractive index as shown in Fig. 2(a) which suggests potential for applications in sensing.^[18,43] Furthermore, for some asymmetric nanoparticles such as gold nanoflowers, different incident-light polarization leads to different OPL emission patterns.^[40] The polarization dependence of the OPL spectra has potential applications in the configuration determination of plasmonic nanostructures.

Anti-Stokes emission processes have seldom been discussed in previous studies. Anti-Stokes emission was reported by Mooradian in 1969^[9] but was regarded as an emission tail on the high-energy side of a laser attributed to the thermal smearing of the electron and hole distributions. This anti-Stokes emission was considered by Neupane et al. to be an up-converted luminescence process in 2013.^[44] Anti-Stokes emission can be enhanced by over three orders of magnitude compared with that of bulk gold when nanoparticles are excited at LSP resonances. He et al. claimed that electron occupation follows the Fermi–Dirac distribution and the profile of the anti-Stokes emission can in principle be applied to probe the local temperature.^[20,42] Figure 2(b) shows that the temperature can be extracted by fitting the OPL anti-Stoke line shape of a single gold nanorod because the temperature controls the distribution of the electrons near the Fermi level according to Boltzmann statistics. Aquiles et al. showed that the anti-Stokes emission of gold nanorods can be used for imaging applications under biologically relevant conditions.^[45] Gold nanorods are better than fluorescent organic dyes because of their lack of photobleaching effects. Furthermore, nanorods have a large absorption cross-section that enables bright nanoscale probing despite the low photoluminescence quantum yield of nanorods.^[46]

Surface plasmons have been widely applied to enhance molecular fluorescence and Raman scattering because of the strong localized electric field around plasmonic nanostructures.^[22,47–50] Light emission from nanostructured metals can be correlated with the broad background continuum in plasmon enhanced spectroscopy.^[22] Quantitative correlations between the photoluminescence of metals and the SERS background are helpful for recovering native chemical information from surface-enhanced Raman scattering spectrum.^[23] These spectral features contain information on the enhanced absorption and enhanced emission rate of quantum emitters. When placing an emitter around a plasmonic nanostructure, the luminescence of the metal nanostructure is enhanced. Zhao et al. demonstrated that gold nanoparticles coupled with fluorescent nanodiamonds mutually enhanced their light emission in experimental and theoretical investigations.^[24] Figure 2(c) shows the corresponding plasmon enhanced spectra. Similarly, Raman scattering can be enhanced by placing Raman molecules around gold nanoparticles. The mechanism of this background enhancement remains unclear despite extensive investigations over the last 35 years. In recent research, Cheng et al. developed a theoretical model to calculate the background of enhancement.^[47] Within the model, the molecule was simplified as a three-level system in the Λ configuration and the nanoparticle was simplified as a plasmonic nanostructure.^[47] These findings contribute to a deeper understanding and suggest that plasmonic nanostructures and molecules are a hybrid entity for analyzing and optimizing surfaceenhanced spectroscopy.

To understand the process of OPL, the dynamics of OPL emission have also been investigated.^[51–53] Varnavski et al. measured the decay time to be approximately 50 fs by an up-conversion technique.^[54] Hwang et al. reported a decay time of approximately 1–10 ps, which agreed with the results of previous reports. The decay time of metal OPL is considerably faster than that of conventional dynamic fluorescence processes in molecules or semiconductors. A fast dynamic process can be used to extract the instrument response function during lifetime measurements because the decay time is usually shorter than the instrument response time as shown in Fig. 2(d).^[51] For example, Lu et al. used the fast luminescence of gold nanoparticles to calibrate their instrument response function during measurements of the lifetime of dye molecule photoluminescence enhanced with a bowtie nano-aperture.^[52]

Optical probes play an important role in biological mechanisms and soft matter systems. Single fluorescent molecules are widely used as probes. Although single molecules have a high photoluminescence quantum yield, these chromophores are affected by blinking and bleaching. Therefore, gold nanoparticles might replace single molecules because such nanoparticles do not show blinking or bleaching. However, nanoparticles feature a low photoluminescence quantum yield despite their large absorption cross section. Boyd et al. studied the OPL of gold and showed that the quantum yield could be enhanced. Yorulmaz et al. achieved a quantum yield of approximately 10⁻⁶ by photothermal microscopy. After reshaping gold nanorods into nanospheres by melting, they found that the nanoparticles’ quantum yield was related to their shape and resonance energy. Wilcoxin et al. obtained a quantum yield of approximately 10⁻⁵ for 5-nm gold nanospheres.^[32] Mohammed et al. obtained a quantum yield of approximately 10⁻⁴ for gold nanorods. Recently, Park et al. found that plasmonic cube-in-cube nanoparticles with a controllable interior nanogap resulted in enhanced OPL intensity and a high quantum yield of approximately 10⁻³.^[36] The photoluminescence quantum yields of those gold nanostructures essentially depended on the excitation wavelength.^[30] These studies have shown the great potential for applications of this type of luminescence for studying biological mechanisms and soft matter systems.

Nanoparticles applied in biological sensing and imaging can be affected by dynamic processes such as the mobility and aggregation of nanoparticles in a cell. The motion of nanoparticles is typically observed by dark field scattering or absorption spectroscopy. Tcherniak et al. studied the OPL of gold nanorods under different excitation polarizations.^[55] The intensity of the OPL correlated with the angle between the nanorod and the light polarization. Hence, this approach enabled measurement of the orientation of a nanorod. They were able to measure the rotational diffusion constant of gold nanorods in aqueous solutions. Furthermore, the polarization of the excitation light absorption and the farfield angular distribution of OPL also depended on the orientation of the nanostructures. Zhang et al. measured the OPL of a single gold bipyramid, and their results showed that the OPL emission pattern was related to the orientation of the gold bipyramids.^[19] After melting their gold bipyramids into nanoparticles, they showed that the emission pattern was related to the shape of the nanoparticles. Hence, orientation also can be measured through the farfield emission pattern. These studies demonstrate the potential applications of nanoparticles as dynamic probes of polarization and orientation in sensing and imaging.

Now we discuss the physical origin of OPL. For OPL from bulk metals, the shape of the spectrum does not depend on the excitation wavelength. Hence, the origin of this luminescence was first attributed to interband transitions,^[9,10] i.e., radiative recombination between electrons in the conduction sp band and holes in the valence d bands. In terms of the electronic energy band structure of gold materials, the energy separation between the d-band holes and the Fermi surface is roughly 1.8 eV near the X point and 2.4 eV near the L point. Thus, when a 785-nm laser (1.58 eV) is used to excite gold nanostructures, interband transitions cannot be excited.^[56] Moreover, gold nanorods have been used to study OPL because their LSP resonance can be widely tuned to the near infrared, except for regions of overlap with interband transitions.^[39] However, strong infrared emission remains when the photon energy of a CW excitation laser is tuned to the infrared, for example, at 1.58 eV, which is less than the interband energy gap. Controversy remains about the physical origin of OPL based on interband transitions. Beversluis et al. attributed intraband transitions to infrared OPL, which is strongly mediated by LSP modes rather than interband transitions.^[56,57] Alternatively, possible contributions of interband luminescence can be excluded using silver-based material systems because their interband transitions occur at up to 3.9 eV.^[29,57] Suppression of interband transitions can reduce the plasmon dephasing rate.^[58] Recent transient studies of visible photoluminescence have shown that the photoluminescence decay is faster than 50 fs and intraband excitations of surface plasmons provide additional contributions to interband transitions.^[59]

In addition, the OPL spectral shape coincidence of metal nanostructures confirms a strong correlation between OPL and their LSP resonances. A microscopic mechanism based on the radiative decay of surface plasmons has been proposed to explain orders of magnitude increases of the luminescence quantum yield.^[60] In this model, photon excited hot electron–hole pairs recombine nonradiatively. The excitation energy is then transferred to the plasmon, and radiative decay of the excited plasmon yields luminescence. This spectral coincidence has been verified from the Stoke emission of nanoparticles excited by a laser at a wavelength much shorter than the LSP resonance wavelength. Mechanisms involving interband and intraband transitions cannot sufficiently explain ultrafast processes, their dipole transition forbidden nature, or conservation of momentum. An electronic Raman scattering mechanism was proposed to match the free electron dispersion and ultrafast speed of transitions.^[61–64] Several recent studies have suggested that the breakdown of symmetry in nanostructures, i.e., phonon coupling and Landau damping, allows sufficient momentum change for intraband transitions.^[27,29,56] In particular, spectrally resolved nonlinear analysis of metal luminescence tends to suggest that intraband transitions are more compatible with the observed photoluminescence exponent of nonlinearity.^[65] In addition, it has recently been proposed that light emission has a universal nature and the same mechanism operates across the entire spectrum based on hot-electron intraband luminescence.^[57] Moreover, we have found that the OPL process is related to both the sp-band free electron distribution and localized surface plasmon resonance. Thus, we developed a phenomenological theoretical model based on a quantized optical cavity concept to simulate interactions of an electromagnetic field and a plasmonic nanoresonator.^[20,66] An understanding based on the resonator concept provides an intuitive alternative way of describing these phenomena simply and quantitatively. This model helps us understand the excitation wavelengthdependent photoluminescence quantum yield of gold nanoparticles.^[30] By correlating a two-temperature model and electron distributions, we demonstrated that both OPL and two-photon luminescence (TPL) from metal nanostructures involve the same mechanism.^[67] To date, the physical mechanism of light emission from plasmonic nanostructures has been the subject of debate, and quantitative models of this phenomenon have yet to be developed.

Intense anti-Stokes broadband luminescence of metal nanostructures can be observed under pulsed-laser illumination.^[68,69] When the power-law exponent of emission intensity as a function of excitation power is equal to or greater than 2, the luminescence phenomena can involve a two-photon or multiple photon process.^[70,71] For simplicity, only TPL is mentioned and discussed in the following. Similar to OPL, the exact physical process of TPL is still under debate. The actual mechanism of TPL might not be related to two-photon absorption excitation as initially believed.^[61] Luminescence excited by a CW laser might also involve TPL.^[72] For simplicity, we use the term TPL to describe the luminescence process occurring after pulse-laser excitation.

The TPL of metal nanostructures has many interesting properties and potential applications.^[16,17,73] The quantum efficiency of OPL can be enhanced by a factor of one million owing to the LSP effect. In the same way, TPL is also considerably enhanced compared with that of bulk materials. In general, metal nanoparticles are a good choice for biological imaging. The absorption of water and biological molecules is low in the near-infrared band. Although fluorescent semiconducting quantum dots have good nonlinear characteristics, these materials are often toxic. Hence, the TPL of gold nanoparticles has potential applications in the nonlinear optical imaging of biological samples. For example, Wang et al. applied an 830-nm laser to excite TPL in nanorods and obtained in vivo scans. Figure 3(a) shows images of nanorods in mouse ear blood vessels. There is potential for these materials to be used in detecting cancers and in biological tagging.

Furthermore, because of the important role of LSP resonance in the TPL process, we are able to connect the spatial distribution intensity of TPL with the LSP mode.^[21,74] Ghenuche et al. systematically studied the TPL of gold nanoantennas. Their results showed that the TPL signal spatial distribution agreed with the convoluted distribution of the fourth power of the local electric field calculated by numerical simulations. Hence, with the use of a scanning near-field microscope, plasmon modes of the nanorods were distinguished by TPL imaging. Imura et al. presented near-field images of nanorods of different sizes by TPL imaging.^[74] The dotted line in Fig. 3(b) shows the shape of the nanorod, and the broadened image was attributed to the near-field probe tip. Nanorods of different sizes supporting different LSP modes are indicated by different spatial distributions in the TPL imaging.^[75]

For LSP resonances involved in the TPL process, light absorption is polarization dependent for excited laser pulses and sensitive to the wavelength of the excited laser pulse. Owing to these features, Zijlstra et al. demonstrated a five-dimensional optical recording which contained information on the wavelength and polarization of an excited laser pulse and three spatial degrees of freedom.^[15] The photothermal effect can also be applied to reshape nanorods to enable the recording of information.^[19] Data can be recorded in three-dimensional space because photothermal melting occurs within a focal volume. Figure 3(c) shows the recording process and the high-density recording device. The TPL process also enabled read out of the data because TPL is sensitive to the polarization and wavelength of light. This work suggests the potential for applications of TPL in metal nanostructure memory devices.

A theoretical understanding of the TPL of metal nanostructures was first based on interband transitions which operate by a mechanism similar to that of the OPL process, except for the involvement of two- or multiple-photon absorption processes.^[76–79] Recent work has proposed the universal nature of light emission and considered the same mechanism to operate across the entire spectrum.^[57] Work by Huang et al. presented an alternate description of TPL as a resonant electronic Raman scattering process.^[61] A new understanding of the resonator concept has provided an alternative description of both OPL and TPL phenomena simply and quantitatively.^[80] In fact, oscillator models combined with a temperature dependent Fermi–Dirac distribution^[26,81,82] provide a unified understanding of OPL and TPL processes.^[80] The decay of excited electrons after photon excitation can be considered as a photoelectric effect or thermionic emission because laser pulses heat the electrons of the metal. We can measure a large thermionic emission and small photoelectric effect. A laser pulse can heat electrons to several thousand degrees instantaneously because the pulse duration is so short that the energy of the electrons cannot be transferred to the lattice.^[26,81,82] Agranat et al. calculated the intensity of the emission, and these results provided support for thermionic emission.^[68,69] The features of the phenomenon might be useful for studying the relaxation kinetics of electrons and the lattice and the noninertial conversion of picosecond IR laser pulses into visible light. Although the resonator concept provides a general and intuitive model for understanding intrinsic light emission phenomena of metallic nanostructures, a complete microscopic model has yet to be determined. The electronic band structure, dynamic processes of electron and hole distribution, interactions of electrons and phonons, and coupling between plasmon modes should also be considered.^[83–85]

In addition to photon excitation, intrinsic luminescence of metal nanostructures also occurs under electron excitation such as cathodoluminescence (CL) in electron microscopes, tunneling electrons through the probe apex in scanning tunneling microscopes, and tunneling electrons within the nano-gap of electrodes. These excitation processes are very different; however, the shapes of emission spectra of gold nanostructures show a strong correlation with their LSP modes. Hence, there should be a common physical mechanism operating for these emission processes regardless of the excitation process. This correlation might be helpful for determining the physical mechanism of luminescence of metal nanostructures. In terms of the high similarity of the emission features for different excitation processes, an understanding of the luminescence process based on an oscillator model could provide a unified model.^[80] Here, we briefly review three ways of electronic excitation that can generate luminescence of metal nanostructures.

At first, electron microscopy involves several kinds of processes depending on the signal collection mode, such as CL, secondary electron, and electron energy inelastic losses. When metallic nanoparticles (e.g., silver particles^[86] or gold nanoclusters^[87]) are exposed to fast electron beams, the localized plasmon mode can be excited and decay radiatively and contribute to light emission.^[88,89] Electrons also excite e-h pairs of semiconductors (or molecules) which decay radiatively and contribute to the luminescence. Here, CL is restricted to luminescence from metal nanostructures after electron beam excitation in electron microscopy. We will not review the mechanism of CL since several comprehensive review articles already exist on this subject.^[88] CL techniques can be used for imaging at the spatial resolution of the LSP modes because of the high-spatial resolution of electrons compared with that of visible light. For instance, Atre et al. applied CL tomography to image metal structures in three dimensions on a nanoscale. They demonstrated CL reconstruction imaging of a three-dimensional metaldielectric crescent. Vesseur et al. imaged a nanowire by CL image spectroscopy. Figure 4(a) shows the plasmon behavior and the eigenmodes with distinct spatial profiles of gold nanowire.^[90,91] Coenen et al. studied the spectral and angular CL patterns of nanoholes. They also showed a directional emission of a linear array of Au nanoparticles by angle-resolved CL spectroscopy when a high-energy electron beam irradiated the first Au nanoparticle.^[92]

Fig. 4.

Figure Option ViewDownloadNew Window

Fig. 4. (color online) (a) Upper part: SEM image of 725nmlong Au nanowire on Si substrate (scale bar is 250 nm). Lower part: CL images measured at wavelengths of 592, 640, and 730 nm from the nanowire (from Ref. [90], Fig. 1). (b) Left: schematic of experiment. Angular distribution of emission occurs when the tip is moved to a different position on the gold nanoparticle. Right: Energy transfer schematic of Ref. [96]. (c) Electron microscope image of lateral tunnel structure: electrically connected single crystalline gold nanoantenna loaded with coated gold nanoparticles on glass substrate (V + : applied voltage; e−: electron flow; hν: light emission). Second: tunnel gap with applied voltage (EF, Fermi energy, and electrochemical potential of Ref. [99], Fig. 1).

For the tunneling-electron induced luminescence of metals, luminescence from metal-oxide-metal sandwich junctions (Au or Ag) has been studied and discussed.^[93] However, a theoretical explanation of this phenomenon has yet to be agreed on. It is not clear how inelastic electron tunneling or hot electron excitation of radiative plasmon modes contribute to the emission. When fluorescent molecules are introduced between a metal tip and substrate, the luminescence process becomes more complex.^[94] Tunneling electrons can excite plasmonic nano-cavities, and the cavity can excite the electrons of molecules to higher vibrational states. The higher vibrational states are induced by a plasmon-assisted multistep inelastic scattering scheme. For simplicity, Kravtsov et al. measured light emission from a tip-film gap structure at different distances.^[95] The peak energy of the luminescence correlated with the distance, which indicated that the luminescence was from plasmon resonance rather than tunneling electrons because they used a low bias voltage. Le Moal et al. used a similar method to investigate light emission by changing the position of the tip with respect to the gold nanoparticle as shown in Fig. 4(b).^[96] They demonstrated that inelastic electron tunneling between the tip and gold nanoparticle can result in directional emission, which is related to the interference of the LSP modes.

It has already been shown that visible light can be generated through electron tunneling. Lambe et al. discovered broadband light emission when applying a voltage to a metal-insulator-metal tunneling junction.^[97] The emission with a high-frequency linear cutoff can be described by the expression L(υ) = P(ν, V)(|V| − hν/e) θ (|V| − hν/e). Here, P(ν,V) is a packaging function and θ(x) is a cutoff function. The luminescence process is described by two steps: (i) electrons tunnel through the insulator from occupied states to unoccupied states in the other electrode without energy loss, and (ii) these higher energy electrons are thermalized. In this process, the surface plasmon modes in the tunnel junction can enhance inelastic tunneling. Parzefall et al. performed a control experiment to show that surface plasmon modes can enhance emission and that the cutoff voltage influences the emission spectrum.^[98] Furthermore, Kern et al. assembled a gold nanoparticle between two electrodes to form an electrically driven optical antenna, as shown in Fig. 4(c).^[99,100] Gurunarayanan et al. placed an antenna in a V-electrode and achieved broadband unidirectional emission owing to the enhancement of the dipolar emission at the tunnel junction and the fundamental quadrupole-like resonance of the antenna.^[101] Ward et al. used a typical nanogap consisting of two gold nanojunctions to demonstrate surfaceenhanced Raman scattering.^[102] Dathe et al. used a similar structure in label-free detection of small biomolecules. When biomolecules were adsorbed around the nanoparticles, the peak of the emission spectral shifted.^[103] The luminescence of metal nanostructures caused by the tunneling electric current includes various interesting phenomena. Applications of these materials include biosensing, high-frequency modulation, and basic components for transmission and reflection spectroscopy.

In this review, we discussed several aspects of the most exciting and promising research directions in the field of intrinsic luminescence of metal nanostructures. Surface plasmons can be excited by CW light, pulsed lasers, electron beams, or tunneling electric current. Then, the plasmon decays radiatively resulting in luminescence for various applications. This luminescence can be excited in other ways, such as by chemical reactions^[104] or by highenergy photons. For example, the luminescence of gold nanoparticles induced by X-rays has been reported.^[105] The physical mechanism has yet to be confirmed in this research field. Although many theoretical models have been proposed to explain these phenomena, there is a lack of a theoretical consensus in this field. Owing to the losses of the metal, the quantum yields of this luminescence are low, which limits its applications. Hence, efforts are still needed to further improve the quantum yield of intrinsic luminescence from plasmonic nanostructures. Since there is a strong correlation between surface plasmon modes and luminescence features which can be controlled through the shape, size, and coupling configuration of the nanomaterial, optimizing these properties could improve luminescence for various functions and different devices. A considerable scope remains for exploring potential applications based on metal luminescence.