Electrically driven nanogap antennas and quantum tunneling regime

The optical and electrical characteristics of electrically-driven nanogap antennas are extremely sensitive to the nanogap region where the fields are tightly confined and electrons and photons can interplay. Upon injecting electrons in the nanogap, a conductance channel opens between the metal surfaces modifying the plasmon charge distribution and therefore inducing an electrical tuning of the gap plasmon resonance. Electron tunneling across the nanogap can be harnessed to induce broadband photon emission with boosted quantum efficiency. Under certain conditions, the energy of the emitted photons exceeds the energy of electrons, and this overbias light emission is due to spontaneous emission of the hot electron distribution in the electrode. We conclude with the potential of electrically controlled nanogap antennas for faster on-chip communication.


Introduction
Since the nineties, the field of plasmonics has grown by merging concepts from the fields of material science and RF antennas. This has led to pioneering results where suitably designed metal nanostructures behave like resonant antennas at optical frequencies, thus enabling efficient coupling of visible or near-infrared light to matter in deeply subwavelength volumes [1]. Remarkably, surface plasmons in metal nanostructures enable strongly enhanced electric fields in their nanoscale surrounding, which have proven useful, for example, to locally enhance Raman and fluorescence signals or for photothermal conversion [2].
Since then, achieving plasmon resonances and maximum enhanced fields at selected wavelengths has been the focus of many studies, which have been greatly supported by modern nanofabrication and nanocharacterization techniques to study the optical properties of metal nanostructures, either isolated or grouped as dimers, trimers, oligomers, or supracrystal films [3][4][5]. In particular, it was shown that classical electromagnetic interaction between closely packed metal nanostructures, especially dimers, may lead to a more intense local electric field within the gap, compared with an isolated nanostructure.
This result has motivated researchers to explore the interaction between metal nanostructures separated by a very small, nanometric gap. This exploration has revealed the onset of quantum physics effects that strongly affect the plasmonic and electric field enhancement of the interacting nanostructures. When the distance between nanostructures is sufficiently small, electrons may tunnel between them. Upon applying a bias voltage, this endows them with electroluminescence properties that are appealing for designing electrically driven, nanoscale and ultrafast light emission or detection devices [6,7].
To produce efficient and practical devices, a clear overview of the fundamental optoelectronic properties, and production possibilities and constraints of such nanostructures, also called "nanogap antennas", is needed. With such aim in mind, this article reviews recent works on nanogap antennas.
In Section 2, the plasmonic and local electric field enhancement properties of nanogap antennas are discussed from a theoretical point of view. For a clear understanding, the case of a metal dimer with a gap distance broadly tuned to cover both the classical and quantum regimes is considered.
In Section 3, experimental works showing the measured plasmonic, local electric field, Raman and fluorescence enhancement properties of nanogap antennas in the classical and quantum tunneling regimes are reviewed, and their findings are compared with theoretical expectations.
In Section 4, experimental findings related to the electroluminescence of nanogap antennas, occurring thanks to inelastic electron tunneling, are presented. This section provides an overview over a broad range of reported nanogap antennas, differing by their structure (metal-insulatormetal structures, nanostructure dimers, vertically organized structures) or their chemical nature ("alternative" materials beyond metals). Performance indicators, such as the electroluminescence quantum efficiency, spectrum, and electrical tunability, are systematically reported.
Section 5 makes an emphasis on electroluminescent nanogap antennas harnessing a different light-emission mechanism, which involves hot carrier processes. Such type of antennas promises a broader spectral tuning and higher quantum efficiency than those involving inelastic electron tunneling.
Section 6 briefly discusses recent attempts to produce practical, miniaturized, and tunable light-emission devices based on nanogap antennas.

Theory of nanogap antennas: from classical to quantum regime
Concentrating light into nanoscale dimensions was made possible through structuring sub-wavelength metal objects that can support collective oscillations of electrons when interacting with electromagnetic radiation. This localization effect produces great field enhancements of several orders of magnitude compared to the incident value. Such plasmonic enhancements significantly boost the efficiencies of light-matter interactions at the nanoscale and have impacted applications such as energy harvesting [8], biomedicine [9,10], photo-detection [6,11,12] and plasmon nanolasers [13][14][15][16][17]. Compared to single nanoparticles (NPs), dimer structures -two nanoparticles separated by a nanoscale gap -were proven particularly effective in squeezing light into deep-wavelength volumes [17][18][19]. Gaps formed between metal surfaces control the coupling of localized plasmons, thus allowing gap-tuning targeted to exploit the enhanced optical fields for different applications [20][21][22][23]. Classical electrodynamic approaches are frequently used to describe the plasmonic response of metal gap structures with gaps on the order of few nanometers [24,25]. Using these classical approaches, the different materials are separated by sharp boundaries. The optical properties of materials are described by a frequency-and position-dependent linear dielectric function ( , r) [26].
This theory predicts a monotonic increase in the electric field enhancements with decreasing gap distance, prompting the production of plasmonic structures with subnanometer gaps.
However, as the gap distance enters the nanometer and then the sub-nanometer scale, classical electrodynamics fail to describe the optical response of nanogap structures, where quantum effects become important owing to non-local screening (also called spill-out effect in the literature) and electron tunneling across the gap [28][29][30][31][32][33]. The quantum nature of electrons and the non-local screening strongly alter the optical response and lead to deviations from the classical description as shown in Figure 1. In this quantum regime, surface charges are corroborated with a frequency-dependent distance parameter F , the Feibelman parameter, which defines the actual position of the screening charges with respect to the geometrical boundaries of the metal, leading to an effective gap distance that differs from the geometrical value [27]. Moreover, the spatial profile of the screening charges is described by a where the screening charges with respect to the geometrical boundaries of the metal leads to an effective modification of the metal interface boundaries, producing an effective gap distance that differs from the geometrical value. A tunneling regime, separated from the non-local quantum regime by a threshold tunnel distance d th , where the conductivity of the junction is large enough to strongly perturb the optical response of the structure. Reproduced from ref. [27] with permission; copyright 2016 Nature Publishing.
smooth transition of the electronic densities at the interface. Electron tunneling between the metal surfaces is another key quantum feature that the classical local model fails to capture. Tunneling current across the nanogap induces a conductive contact prior to the direct geometrical contact, which reduces the Coulomb coupling between charges of opposite signs on the two sides of the nanogap and strongly affects the optical response of the nanogap structures [34]. The impact of these quantum mechanical effects on the optical response of nanogap structures is schematically illustrated in Figure 1. This figure shows the different regimes for the plasmon resonances of a pair of particles as a function of the gap distance d. For large gap distances (≥2-5 nm), the structure is in the classical regime, and its optical response can be correctly described using Maxwell equations with gap plasmon modes red-shifting (due to higher effective index of the mode) as the gap is decreased.
As the gap becomes smaller than a few nanometers with a sufficiently small conductance between the particles, the optical response becomes dominated by the non-local screening quantum effect. The non-local screening leads to deviations of plasmon resonances from the classical (dashed line) to the quantum regime (solid line). As the gap distance continues to decrease and becomes smaller than the threshold tunneling-distance d th , the tunneling-induced conductance becomes large enough to allow a fraction of surface charges to tunnel across the gap. The plasmon resonance of the structure is strongly affected by the ac current tunneling across the junction, resulting in a progressive extinction of the red-shifting gap plasmon mode and a gradual emergence of the blue-shifting charge-transfer plasmon (CTP) mode.
In this context, various semi-classical models have been developed to partially account for non-classical features [35]. For example, some of these descriptions illustrate the intrinsic nonlocality of the metals by combining classical electrodynamics and hydrodynamic descriptions of the induced charges [28,30,[36][37][38]. Moreover, other semi-classical approaches account for the spill-out of the induced electron density and Landau damping by incorporating ab initio quantum surface-response corrections at the metal-dielectric boundaries. This can be achieved using Feibelman parameters obtained from quantum calculations [26,[39][40][41][42]. These models have predicted that non-classical phenomena can substantially influence the optical interaction in a nanogap structure. Recently, fully quantum calculations based on the time-dependent density functional theory (TDDFT) have been proposed to capture all the quantum effects affecting the interaction in nanogap antennas and to establish the validity of the different semi-classical models [43].

Non-local effects in plasmonic air-nanogap antennas
Savage et al. [49] achieved gap distances in the tunneling regime by bringing two gold-nanoparticle-terminated atomic force microscope (AFM) tips to close proximity using a piezoelectric stage. Gap distances in the sub-nanometer regime were produced, with a dc tunneling conductance G across the gap exceeding the quantum conductance G 0 = 2e 2 ∕h (where e is the elementary charge and h is the Planck's constant) for tips in conductive contact. The dark-field scattering spectra recorded as a function of the reducing gap distance are shown in Figure 2(a). For d > d QR , with d QR the threshold distance at which charges induced by quantum tunneling overcome the near-field capacitive interaction between plasmons, spectra are dominated by the nearfield interaction of the cavity localized charges and plasmons couple according to classical models. As the structure separation decreases below a critical size d ∼ d QR , plasmons interactions enter the quantum regime characterized by a gradual evolution of the bonding modes into CTP modes. Quantum tunneling charge transfer was found to screen the localized surface plasmon charges, therefore decreasing the enhanced fields and reducing the plasmonic Simultaneously measured electrical conductance G∕G 0 and dark-field scattering spectra of the plasmonic cavity with increasing force applied to the tip cavity. Calculated total scattering intensity from a plasmonic cavity system incorporating quantum mechanical tunneling. Reproduced from ref. [49] with permission; copyright 2012 Nature Publishing. (b) Electron energy-loss spectroscopy (EELS) signals of a 9-nm diameter silver dimer collected during particle motion triggered by a STEM-EELS probe. STEM images collected at the beginning and end of each scan (indicated by solid vertical lines) illustrate the separation distances (+0.7, <0.27, −0.3, and −1.0 nm). Image scale bar, 5 nm. Reproduced from ref. [50] with permission; copyright 2012 American Chemical Society. (c) Evolution of Au tip PL spectra in scanning tunneling microscope-based configuration as a function of distance to the Au-surface, z. Pronounced field enhancement is followed by rapid quenching for z < 2 nm. Reprinted with permission from ref. [51]; copyright 2014 American Chemical Society. (d) Two-dimensional histogram of photoefficiency versus distance relative to contact formation for light emission of wavelengths 850 nm and longer. Reprinted with permission from ref. [52]; copyright 2022 American Chemical Society.
coupling. These results agree well with the predictions of the quantum corrected model (QCM, quantum approach incorporating quantum mechanical effects for nanogap plasmons within a classical electrodynamic framework) of Esteban et al. [26,40] and recent experiments by Scholl and colleagues [50]. Scholl et al. observed the plasmon resonances of coupled metal nanoparticles as their gap size was reduced down to sub-nanometer dimensions (Figure 2(b)). As separations were reduced from 7 nm, the dipolar peak exhibited a redshift consistent with classical calculations. However, for gaps smaller than ∼0.5 nm, this mode showed a reduced intensity, implying that field enhancements will saturate around this gap separation, which is again consistent with quantum theories that incorporate electron tunneling. As the particles overlapped, the dipolar mode disappeared and was substituted by a dipolar charge transfer mode [50].
Additionally, Kravtsov et al. explored the photoluminescence (PL) of a coupled plasmonic structure consisting of a sharp AFM gold (Au) tip placed on the top of an ultraflat templated-stripped Au film [51]. Field-enhanced behavior accompanied with a maximum PL intensity dominated until the onset of quantum coupling (gap distance ∼1.5 nm) dramatically reduced field enhancement and emission intensity (Figure 2(c), inset: scheme illustrating the experimental configuration). The PL spectral peaks blue-shifted for smaller gaps, an additional indication of the emergence of electron tunneling.
Recently, the electroluminescence and conductance of Au tunnel junctions have been simultaneously probed [52]. As the gap size was reduced, the plasmonic enhancement increased for junctions biased between 1.4 and 1.8 V, which is consistent with the behavior of charge transfer plasmons. At biases above 1.9 V, the plasmonic enhancement decreased with the decreasing gap, showing quenching due to tunneling as shown in Figure 2(d), which is in remarkable agreement with the trends observed for high energy plasmons in scattering experiments.

Non-local effects in molecule-loaded-nanogap antennas
Ciraci et al. studied the plasmon resonances of the particleon-film structure, with the gap being formed using a molecular dielectric layer. The peak wavelength of the plasmonic resonance was found to deviate from the predictions of the local classical model for gap distances less than ∼3 nm [53]. These deviations can be accounted for by the hydrodynamic model that incorporates non-locality effects. Zhu et al. also measured the surface enhanced Raman scattering (SERS) enhancement of thiophenol molecules adsorbed on a pair of nanoparticles as a function of the gap distance. The emergence of electron tunneling at optical frequencies for gaps below a threshold tunnel-distance limits the maximum achievable plasmonic enhancement and induces a quenching of the SERS enhancement [54]. QCM calculations were found to be consistent with the experimental data, supporting that quenching of near-field enhancement was due to electron tunneling.
Moreover, Hajisalem et al. probed the onset of the quantum tunneling regime by observing the local field intensity in sub-nanometer self-assembled monolayer gaps of particle-on-film structures [55]. A sudden drop in the third harmonic generation occurred for double the gap size of past studies because of the reduced barrier height of the self-assembled monolayers (Figure 3(a)).
Chen et al. developed a MoS 2 spaced nanoparticle-onmirror plasmonic antenna, with layered MoS 2 as a twodimensional atomic crystal serving as the gap for probing field enhancement within the gap [56]. Measured SERS enhancement factor (EF) showed great agreement with calculations from pure classical electromagnetic theory for gap distances >1.24 nm. For a gap of 0.62 nm, the emergence of quantum mechanical effects yielded a maximum field enhancement of 114-fold, 38.4 % lower than classical predictions, demonstrating that classical model significantly overestimates the field enhancement for sub-nanometer gap distances as shown in Figure 3 Recently, supramolecular systems made of oligoamide sequences were used to reversibly switch the gap plasmons of Au nanoparticles on mirror between classical and quantum tunneling regimes via supramolecular interactions [57]. Plasmonic hot electron tunneling driven by laser excitation was proved to blue shift the dipolar mode of gap plasmons because of the increased conductance in the nanogaps. The plasmon peak switched back to its original value when the laser was turned off as illustrated in Figure 3(c). These plasmon shifts fitted perfectly with both classical and quantumcorrected models, suggesting that the quantum limit is approximately around 0.6 nm.
Also, non-local effects have been investigated in propagating gap surface plasmon modes in ultrathin metaldielectric-metal planar waveguides, exploiting monocrystalline gold flakes separated by atomic-layer-deposited aluminum oxide [58]. Scanning near-field optical microscopy (SNOM) was used to directly access the near-field of such confined gap plasmon modes. Scattering-type-SNOM measurements from samples with different gap thicknesses revealed signatures of gap-dependent non-classical broadening, which was well-described by the generalized hydrodynamic model of plasmonics (Figure 3(d)). Thus, quantum non-local corrections should be taken into account when treating extremely confined gap plasmon modes.

Inelastic electron tunneling
In the previous sections, nanogap antennas with gap distances changing from the nanometer scale to the subnanometer scale were introduced and the difference between classical and quantum tunneling regimes was explained. The case of a metal dimer with a gap distance broadly tuned to cover both the classical and quantum regimes was considered. The plasmonic electric field enhancement properties of nanogap antennas were discussed from a theoretical point of view. Then, some recent key approaches that employed plasmonic structures with sub-nanometer gaps (air-and molecule-loaded) and allowed the experimental observation of tunneling and non-local effects were reviewed. These experimental findings were then compared to theoretical expectations.
In this section, experimental findings related to the electroluminescence of nanogap antennas, occurring based on inelastic electron tunneling, will be presented. Other physical mechanisms including hot-electron injection involved in light emission process will be briefly discussed. Then, an overview over a broad range of reported nanogap emitting-light-antennas, differing by their structure (metal-insulator-metal structures, nanostructure dimers, vertically organized structures) or their chemical nature ("alternative" materials beyond metals) will be provided. Performance indicators, such as the electroluminescence quantum efficiency, spectrum, and electrical tunability, will be systematically reported.

Inelastic electron tunneling, plasmon excitation and outcoupling
When a bias V bias is applied between two metal electrodes separated by an insulating nanogap, the Fermi level of one of the electrodes is raised by an energy eV bias with respect to the other, as shown in Figure 4(a). If the nanogap is thin enough (at most, a few nm), some electrons tunnel through it, a fraction of them elastically and the other fraction inelastically. In the inelastic scheme, energy may be conserved through photons emission from the nanogap structure. The most frequently reported mechanism for such light emission, which is depicted in Figure 4(a), is the following: (i) each of the inelastically tunneling electrons may generate one photon that radiates to the far-field or one surface plasmon at the metal surface. Depending on the morphology of the electrodes, the thereby generated plasmons may propagate at the interface (surface plasmon polariton, SPP) or be localized (localized surface plasmon, LSP). (ii) Plasmons may outcouple into photons released to the far-field. The Tuning of the electroluminescence spectrum with V bias in Au (or silver)/aluminum oxide/aluminum junction. The cutoff at eV bias is clearly seen. Reproduced from ref. [62] with permission; copyright 1976 American Physical Society.
probability of such outcoupling depends on the electrode morphology.
The mechanism involving inelastic electron tunneling and plasmons in a single-electron process has been claimed to be the dominant path for electrically driven light emission when the nanogap conductance G is smaller than the quantum of conductance G 0 (G « G 0 ). However, it has also been observed in some studies where G was close from G 0 . This mechanism endows the nanogap structure with electroluminescent properties that are appealing for devices harnessing the properties of plasmons such as nanoscale light sources and sensors. In this context, it is particularly relevant that the process of inelastic electron tunneling occurs at the femtosecond timescale, i.e. much faster than radiative recombination in semiconductors. This would be promising to achieve ultrafast modulation of the electroluminescence signal.
Other mechanisms for light emission rather than relying on an inherently inelastic tunneling effect in the gap have also been demonstrated. Uskov et al. proposed a light source utilizing light emission by electron passing ballistically through a nanoconstriction [59]. When a voltage is applied across the nanoscale constriction used as the active element of the optical antenna, electrons flow through the contact quasi-ballistically, gain energy and transfer it to excited plasmon modes, provided that both energy and momentum conservation is maintained.
Light emission based on coherent surface plasmon amplification due to the tunneling gain was also discussed [60]. A detailed comparison of the surface plasmon gain due to the inelastic tunneling and loss due to the interband and Drude absorption shows the possibility of the net gain in sufficiently thin tunnel structures. Mechanisms describing photon emission out of nanogap antennas at energies exceeding the applied bias will be discussed in Section 5.

Electroluminescence spectrum and external quantum efficiency
In the context of such applications, key performance indicators are the electroluminescence spectrum of the nanogap structures and the corresponding external quantum yield ( ext = number of emitted photons/number of tunneling electrons). To give qualitative insights into such indicators, one can relate them to the underlying microscopic mechanisms. The electroluminescence is often modeled as threestep process [61], depicted in Figure 4 (iii) Radiation to the far-field. The power radiated by the dipoles outcouples directly to the far-field, or couples to the electrodes. In the second case, part of it is lost through non-radiative processes, and the other part is outcoupled to the far-field. The efficiency of the outcoupling process is given by the photon energy-dependent radiative efficiency rad (E) of the electrodes.
The electroluminescence spectrum is then given by: From eq. (1), it comes that the electroluminescence spectrum can be tuned by two means. First, in a static way by suitably tailoring the morphology of the electrodes. This enables controlling the plasmonic response of the nanogap structure and thus the energy dependence of gap ∕ 0 and rad . Second, in a dynamic way by adjusting V bias . Since IET = 0 if E > eV bias , photons are emitted only at energies smaller than eV bias . Therefore, increasing V bias enables extending the bandwidth of electroluminescence toward higher photon energies.
The corresponding external quantum yield is obtained by integrating ext (E) over the whole range of photon energy. Achieving a high external quantum yield requires tuning the morphology of the electrodes so that they enable a suitable trade-off between (1) a large fraction of electrons tunneling inelastically, (2) a strong Purcell effect in the nanogap, and (3) an efficient antenna behavior ensuring an optimal outcoupling of the generated optical power to the far-field.
For the case of a scanning tunneling microscope traveling over a metal surface, the physical mechanism governing light emission can be understood as follows. When the tip-sample separation distance is around 0.5-1 nm, the surface plasmons on the two surfaces strongly interact and form a coupled plasmon mode. Light emission is resonantly enhanced due to the formation of the tip-induced plasmon mode in the cavity formed between the tip and sample. The resonance in the emission spectrum occurs at the frequency for which half a wavelength of the coupled mode fits into this cavity. The charge oscillations associated with the coupled plasmon mode have opposite signs on the two electrodes. This has two major consequences: (i) the electromagnetic fluctuations are enhanced since the charge on the first surface give a field that polarizes the second surface, so that the polarizing field at the first surface increases. (ii) The coupled plasmon mode is locally charge neutral which red-shifts the resonance peak relative to the surface plasma frequency. It is worth noticing that the effects of retardation are small for sample materials such as Au or copper and tip materials such as tungsten or iridium [63]. Recently, a STM has been employed to locally modify the surface of a Si/Au film stack via heating, which was enabled by a high-density tunnel current. Using this technique, hybrid Si/Au nanoantennas with a minimum diameter of 60 nm have been formed, resulting in structures that efficiently emit photons in the visible range owing to the inelastic tunneling of electrons through the junction [64].
Considering the theory of photon emission in electron tunneling to metal nanoparticles, one can distinguish between inelastic tunneling where the excitation occurs when the electron is in the vacuum barrier and the hotelectron mechanism, where a tunneling electron injected in the particle excites the surface plasmon. For a particle with a few ten nanometer radius, the hot-electron mechanism is quite ineffective and light emission is governed by inelastic tunneling. But for very small particles, R < 1.5 nm, light emission is dominated by the hot-electron injection mechanism [65].
The properties of surface plasmons excited by electron tunneling near metallic structures have been also examined. A complete calculation is given for surface plasmons localized by particles with curvature near the junction, including the radiation spectrum of the system [66].

Overview of pioneer and recent works 4.2.1 Pioneer works: metal-insulator-metal junctions
Electroluminescence due to inelastic electron tunneling in a nanogap was first observed in the 70s by Lambe and McCarthy in layered metal-insulator-metal nanogap structures [62]. In these structures, the electrodes consisted of noble metals (Au, silver, or aluminum). These early studies showed the potential of such structures as versatile electrically driven light sources emitting in the infrared and visible regions. It was shown that the spectral features of their electroluminescence can be tuned by: (i) Varying the applied voltage, as shown in Figure 4 Increasing the voltage broadens the emission spectrum from the near infrared to the visible, due to the change in IET (E). (ii) Controlling the radiation efficiency of the structure.
This was achieved by several means, such as tailoring the metal roughness [62,67,68], adjunction of a prism [69] or a grating [70]. This also enables tuning the emission pattern of the emitted light.
However, most of these structures presented a relatively large footprint and low values of ext ∼ 10 −6 , far from the theoretical limit estimated to be ∼10 −1 [62].

Recent works: nanogap antennas
The recent progresses of nanofabrication techniques have fostered the design of smaller footprint nanogap structures showing electroluminescence due to inelastic electron tunneling. They combine a strong Purcell enhancement with a more efficient outcoupling of the generated optical power. These structures, which are called 'nanogap antennas', besides presenting a higher external electroluminescence quantum yield than standard layered metal-insulator-metal nanogap structures, also present an excellent potential for controlling the spectrum, pattern and polarization of the emitted light.
Hereafter, the recently reported nanogap antennas are grouped into two classes: in-plane nanogap antennas, in which the metal electrodes are in the same plane and vertical nanogap antennas, which rely on vertical integration. In these two classes, the electrodes consist of noble metals. Finally, nanogap antennas including electrodes based on alternative materials, i.e. others than noble metals, are presented. The advantages and drawbacks of each type of nanogap antennas are discussed, taking into account several aspects: external quantum yield, radiated power, spectral, spatial and polarization properties of the electroluminescence, footprint, integration, and scalability. The related applications are also discussed. The key features of selected nanogap antennas are gathered in Tables 1 and 2.

Nanoparticle in a gap and nanorod dimers: tuning the electroluminescence spectrum and pattern
The first demonstration of spectrally tunable electroluminescence by electrically driven in-plane nanogap antennas was reported in 2015 by Kern et al. [71] Each antenna consisted of two single-crystal Au nanoflakes with the same orientation separated by a 25 nm gap, which were connected to the external voltage supplied by metal nanowires as shown in Figure 5(a). These structures were fabricated by focused ion beam lithography. A Au nanoparticle capped by a ligand shell was then brought between the two nanoflakes with an atomic force microscopy tip, until it touched one of the nanoflakes and was separated from the other by a nanogap. Upon applying a voltage, an electroluminescence signal was observed. The corresponding spectrum showed a resonant structure in the near infrared attributed to the LSP of the nanoflakes. Upon increasing the voltage up to 2 V, this band was shown to shift toward higher photon energies. In addition to this dynamic tuning, the electroluminescence spectrum was tuned by controlling the location of the Au nanoparticle within the gap. The electroluminescence of the fabricated structures showed a symmetric dipole-like pattern.
Later, Gurunarayanan et al. fabricated in-plane nanogap antennas based on polycrystalline Au nanorods [72]. Electron beam lithography was used to fabricate a nanorod dimer connected by a nanoconstriction, which was then submitted to an electromigration process to form a nanogap between them. In contrast with the approach by Kern et al., it was not needed to bring any nanoparticle to form the nanogap. Moreover, the nanorods were tilted one with respect to the other to form a "V-shaped antenna", as shown in Figure 5(b). The corresponding electroluminescence spectrum was shown to present a LSP-related resonance which was tuned from the near infrared to the visible by varying the nanorod length. Furthermore, an asymmetric electroluminescence pattern, with most of the power being radiated in a tilted direction, was achieved. Such directional emission was attributed to the far-field interference of the dipolar emission from the gap and the quadrupolar-like LSP scattering from the antenna. The strength of this interference was shown to depend on the antenna structure (length and orientation of the nanorods) and on the applied bias, thus allowing the tuning of the electroluminescence pattern both by design and dynamically.
In a more recent work, the directionality of light emission has been numerically addressed in an electrically driven nano-strip MIM tunnel junctions. The device consisted of an Ag-SiO 2 -Ag stack with SiO 2 of thickness 3 nm, the top electrode of the stack was divided into 16 nano-strips, with two of the tunnel junctions at the center acting as source. The emission was tuned to desired angles by appropriate selection of the periodicity and the excitation source. The mechanism governing this directional emission was the constructive interference of directly emitted photons from the primary and secondary sources, with negligible contribution from the scattered surface plasmons [73].
It is worth noting that, in the three previous works, the electroluminescence spectra presented relatively broad features, with a linewidth greater than 100 nm. Achieving sharp features may be useful for applications such as pure color emission or sensing. In this context, Vardi et al. showed that sharp Fano resonances can be produced in the electroluminescence spectrum of in-plane nanogap antennas [74]. They achieved such a resonance in the near infrared by introducing a colloidal Au nanoparticle capped by an organic layer in the gap of a bowtie nanowire, as shown in Figure 5(c). The bowtie nanowire was fabricated by electron beam lithography and the nanoparticle was introduced by electrostatic trapping. It was shown that the Fano resonance results from the interaction between the narrow LSP resonance of the nanoparticle and the broader LSP resonance of the bowtie.

Parallelepipedal nanostructures dimer: maximizing the external electroluminescence quantum yield
The external electroluminescence quantum yield of the antennas reported in the previous works reached at most values of ext ∼ 10 −4 under few-volt applied bias. This is much higher than with the early layered metal-insulatormetal nanogap structures, but still far from the theoretical limit. In this context, Qian et al. reported a nanogap antenna design enabling ext to reach much higher values [76]. Their design consists of a dimer of single-crystal Ag nanostructures with a parallelepipedal shape. These nanostructures are encapsulated by an ultrathin polymer layer. A polymermediated assembly process is followed to achieve an edgeto-edge orientation of the nanostructures with an ultrathin nanogap. The electroluminescence spectrum of the antenna showed LSP-related resonant features tunable in the visible and near-infrared by controlling the nanostructure shape. By accurately adjusting their three axis lengths, the authors achieved a resonant response for rad and gap ∕ 0 at the same photon energy. Combining these features altogether enabled ext to reach values in the range of 10 −2 under a ∼3 V bias.

Bowtie with nanoconstriction: maximizing the electroluminescence power and operation lifetime
The quantities that are truly relevant for benchmarking the performance of a light source are the optical power it radiates and its operation lifetime. For the nanogap antennas with the highest reported ext ∼ 10 −2 , the radiated optical power was only in the range of 10 pW [76]. This points a weak current of inelastically tunneling electrons, with values typically in the range of tens to hundreds of nA under a few-volt bias for these structures. In addition, the operation lifetime of the antennas was not studied.
Recently, Qin et al. fabricated a nanogap antenna structure aiming at combining a relatively high ext with a strong tunnel current and studied its operation lifetime [75]. The nanogap antenna was produced by shaping a bowtie in a single-crystal Au nanowire by electron beam lithography. An ultrasmall nanogap, shown in Figure 5 More recently, an optical-field-controlled emission of electrons in the ultrafast regime at the metal-vacuum interface was demonstrated [77,78]. When a strong driving pulse illuminates the nanoantenna junction, a large local electric field is generated at the nanoantenna tip. Due to the combined effect of the geometric field enhancement resulting from the sharp radius of curvature and the localized surface plasmon polariton in the antenna, the locally enhanced field can exceed the incident electric field of the driver pulse by more than one order of magnitude depending on the spectral overlap with the plasmonic resonance. If high enough, the local electric field substantially bends the surface potential, and a net tunneling current of electrons flows through the potential barrier resulting in an opticalfield-controlled emission of electrons at the metal-vacuum interface. Owing to the strong nonlinearity of the emission process, the electron bursts generated in the device were on the order of several hundred attoseconds for the case of NIR fields.

Vertical nanogap antennas 4.4.1 Nanohole array: tuning the electroluminescence spectrum and polarization
These advantages were first demonstrated by Parzefall et al., who produced vertical nanogap antennas in which the electroluminescence is outcoupled with nanohole Au structures [79]. In this design, an Au thin film acting as first electrode is first grown. This film is structured by focused ion beam lithography to design a nanohole array. A boron nitride (h-BN) ultrathin film is then deposited onto this structured film. Finally, a continuous Au thin film is grown onto the boron nitride film, to act as second electrode ( Figure 6(a)). In this embodiment, the nanogap is thus the space between the structured and continuous Au films and its thickness is controlled by tuning the boron nitride thickness. Upon applying a few-volt bias between the electrodes, inelastic electron tunneling occurs and the optical power of the generated nanogap dipoles is radiated to the far-field by the nanohole LSPs. Therefore, the electroluminescence spectrum can be tuned from the near-infrared to the visible by changing the nanohole dimensions, in addition to varying the voltage. For structures with rectangular nanoholes, the emitted light was linearly polarized. ext values up to 2 × 10 −5 were achieved.
In a more recent work, the same group has demonstrated twist-controlled resonant light emission from graphene/hexagonal h-BN/graphene tunnel junctions. The tunnel junctions consisted of two twist-angle-controlled single-layer graphene electrodes separated by a h-BN tunnel barrier of 7-12 atomic layers [80]. The device was supported by a transparent glass substrate, and the graphene layers were connected to gold electrodes. For small twist angles, the emission spectrum featured a pronounced resonance, which was also tunable with the applied bias. In nearly aligned devices, approximately 40 % of the photons was emitted on resonance. ext values of only 6 × 10 −7 were recorded, but can be enhanced further by optical antennas.

Crossed nanorod-nanostripe antennas: tunable electroluminescence spectrum with narrowband features
To achieve narrowband electroluminescence, He et al. fabricated vertical nanogap antennas by stacking a singlecrystal colloidal silver nanowire on an Au nanostripe with a perpendicular orientation, as shown in Figure 6(b) [81]. The width of the nanogap is determined by the thickness of a molecular layer capping the silver nanowire. In the nanogap, the inelastically tunneling electrons excite gap LSPs, and the power is then radiated to the farfield. The electroluminescence spectrum is thus dominated by resonant features related to the gap LSP modes, which can present a narrow linewidth and can be spectrally tuned by changing the nanowire diameter. Electroluminescence spectra with linewidth of few tens of nm and a 200-nm tunability range in the near infrared were achieved.
In a more recent work, Qian et al. has found a way to enhance the low surface plasmon excitation efficiency by introducing resonant inelastic electron tunneling at the visible/near-infrared frequencies, therefore demonstrating highly-efficient electrically-driven surface plasmon sources [82]. The low efficiency of surface plasmon excitation is due to the fact that elastic tunneling of electrons is much more efficient than inelastic tunneling. In this work, resonant inelastic electron tunneling has been supported by a TiN/Al 2 O 3 metallic quantum well heterostructure, while monocrystalline silver nanorods were used for the surface plasmon generation. The radiated power was clearly boosted owing to the large tunneling current with values up to 2 nW. An EQE up to 30 % was also observed and was claimed to be the highest value demonstrated thus far for an on-chip electrically-driven surface plasmon generation.

Nanoparticle trapping on metal stripes: toward higher throughput fabrication
The two previous antenna designs are not ideal in view of a high-throughput fabrication. Focused ion beam lithography was used to produce the nanohole arrays and positioning reproducibly a nanowire on a nanostripe with a perpendicular orientation is a challenging task. In this context, He et al. demonstrated a scalable method to accurately locate Au nanoparticles forming nanogaps with the top surfaces of two parallel Au nanostripe electrodes [83]. A PMMA layer was first deposited onto the electrodes, and a cavity was engineered in this layer by lithography. The cavity was designed to overlap with the two electrodes, as shown in Figure 6(c). Finally, a solution of nanoparticles was casted onto the structure and an AC voltage was applied between the electrodes. This resulted in trapping nanoparticles within the cavity. With a suitable cavity design, a nanoparticle capped by an organic layer can form nanogaps with the two electrodes. The thickness of the capping layer sets the nanogap width. Antennas fabricated following this method displayed ext values up to 2 × 10 −4 with tunneling currents in the range of few tens of nA. Moreover, by designing adequate cavities, nanoparticle dimers or chains can be located above the electrodes. This enables a broad tuning of the electroluminescence spectrum in the near infrared region.

AC-driven Al-AlO x -Cu tunnel junction: toward longer operational lifetime
Most plasmonic tunnel junctions are studied under DC operation with a relatively large bias (and associated currents) to observe light emission at optical frequencies. Large voltages risk device failure and reduce device lifetimes. This challenge was tackled by Wang et al. [84] in their recent paper by using MIM junctions based on Al and Cu materials. The AlO x layer defined the tunnel barrier thickness, the Al electrode functioned as the bottom electrode, and the Cu electrode formed the top electrode. Light emission mechanism was attributed to inelastic electron tunneling, where the electrons tunnel through the AlO x barrier and lose energy to all available optical modes. Electron-photon conversion efficiencies ext were on the order of 10 −5 -10 −7 . More importantly, the operational lifetime of AC-driven plasmonic tunnel junctions was improved by a factor of three compared to their DC-driven counterparts. Under DC conditions, slow processes that may lead to device failure can readily occur, in particular undesirable electromigration leading to shorts, thus limiting the device operation time to 9.2 h. However, under AC operation, such processes are slow with respect to the voltage changes prolonging the operation time beyond 18.0 h. Recently, the same group has reported a new structure based on plasmonic tunnel junctions and Schottky diodes that improved the device efficiency [85].

Advantages and drawbacks of in-plane nanogap antennas and vertical nanogap antennas
In sum, both in-plane nanogap antennas and vertical nanogap antennas present several advantages that have been described in the previous sections: -They can operate as electrically driven light sources, emitting in the visible and/or near infrared. -Their emission spectrum and pattern can be tuned by properly designing the antenna morphology, and dynamically by adjusting V bias . -Broad and very narrow spectral features can be achieved. -They can efficiently operate under a relatively low bias V bias of few volts. Under such bias, external electroluminescence quantum yields up to ext ∼ 10 −2 and radiated optical powers in the nW range were produced. -Operation lifetimes up to 10 h were reported.
Another important advantage intrinsic to in-plane nanogap antennas is that each antenna can be connected individually to a voltage supply. This is appealing for the design of multiplexed nanoscale light sources.
Beyond these peculiarities, in-plane nanogap still needs to be optimized to produce reliable nanostructures with a high production throughput to be exploited in realworld devices. The current fabrication methods of inplane nanogap antennas suffer in most of the cases from [86]: -A low production throughput, which is especially notorious for electron beam and focused ion beam lithography, atomic force microscopy positioning, and electromigration. It is worth noting that, in many cases, lithography with a nanoscale precision is needed to fabricate the electric contacts connecting the antenna to the voltage supply. -A limited reproducibility, especially in the control of the nanogap width, which will impact the repeatability of the tunnel current, ext and the radiated optical power. This is especially applicable for the approaches that rely on electron beam lithography and/or electromigration for the nanogap design. These methods introduce a surface roughness that is difficult to reproduce exactly from one experiment to another, thus leading to an uncertainty in the nanogap width.
Vertical nanogap antennas enable relieving some of the fabrication drawbacks of in-plane nanogap antennas. In particular, the electrical contact with the voltage supply can be fabricated with no need of high-accuracy lithography tools. Thus, vertical nanogap structures can have a higher production throughout percentage compared to their inplane nanogap counterparts and tackle the main challenge of scalability. Moreover, nanogaps of vertical nanoantennas consist in most cases of a spacer layer whose thickness can be accurately controlled and reproduced.

Nanogap antennas with alternative materials
Nanogap antennas with electrodes consisting of other materials than noble metals (graphene, silicon, eutectic galliumindium) were also shown to exhibit inelastic electron tunneling-induced electroluminescence. Using these alternative materials presented advantages such as the decoupling between electrical and optical properties (graphene), a practical electrical connection (graphene), electroluminescence and sensing on a silicon platform (silicon), or a reversibly tunable electroluminescence driven by photochemical or electrochemical processes.

Graphene-based antennas: decoupling electrical and optical properties, practical electrical connection
Graphene is an electrical conductor with an excellent optical transparency in the visible and near infrared. Therefore, it was used as a transparent electrode in vertical nanogap antennas consisting of a stack including an Au thin film as bottom electrode, a thin aluminum oxide [87] or boron nitride insulating layer [88] and a single-layer graphene sheet as top electrode. Finally, Au or silver nanostructures were deposited onto the graphene sheet to form an antenna structure as in the example depicted in Figure 7(a).
The major contrast between graphene-based structures and the other metal-insulator-metal structures is that the former allows to partially decouple the electrical and optical properties of the antenna. While the graphene/insulating layer/Au stack drives the electron tunneling properties, the nanogap LDOS gap ∕ 0 and the radiative efficiency of the antenna rad can be tuned by adjusting the morphology of the Au or silver nanostructures. Electroluminescence spectra with LSP-related resonances were achieved with Au nanorods, and these resonances were tuned in the visible and near infrared by adjusting the nanorod aspect ratio. This also enabled controlling the polarization of the emitted light [87]. Narrow band electroluminescence (linewidth ∼50 nm) was achieved in the near infrared with silver nanocubes [88]. Decoupling the optical and electrical properties thus brings an additional degree of freedom for optimizing the antenna's performance compared with standard metal-insulator-metal structures. Furthermore, using a graphene sheet as top electrode enables a more practical electrical connection of the antennas to the voltage supply because its implementation does not require high accuracy lithography.

Silicon-based antennas: plasmonic sensing with an electroluminescent silicon platform
The reported silicon-based nanogap antennas present a vertical geometry with a silicon electrode at the bottom and a metal electrode at the top. The outcoupling of the radiated optical power and thus the electroluminescence spectrum and external quantum yield can be tuned by structuring the metal electrode [89,91] in the same way as with metal-insulator-metal antennas. As shown by Dathe et al., planar Au/insulating layer/silicon structures supporting Au nanoparticles capped by a thin organic layer present a resonant electroluminescence spectrum peaking in the near infrared. This is depicted in Figure 7(b). The resonance energy is driven by the LSP of the Au nanoparticle-Au electrode structure and is thus sensitive to the thickness of the organic layer. Therefore, this structure enabled the selective detection of molecules (citrate, BSA, BSA + IgG) capping the nanoparticles by tracking the electroluminescence signal.

Eutectic gallium indium-based antennas: electrochemical and photochemical electroluminescence tuning
By using conducting eutectic gallium indium as electrode in nanogap antenna structures, tunable inelastic electron The electroluminescence spectrum shows a NIR plasmonic resonance band that is sensitive to the nature of the molecules capping the nanoparticles. This is useful for molecular detection. Reproduced from ref. [89] with permission; copyright 2016 American Chemical Society. (c) Eutectic-gallium-indium based antennas. Tunneling occurs between Au nanorods and a eutectic gallium indium electrode, separated by a redox polymer. Partial and reversible oxidation and reduction of this polymer is achieved electrochemically or photochemically and enables the tuning of the NIR electroluminescence intensity in a synapse-like way. Reproduced from ref. [90] with permission; copyright 2020 American Chemical Society.
tunneling-induced electroluminescence properties were achieved. Static tuning was achieved by filling the nanogap between eutectic gallium indium and Au electrodes with a self-assembled molecular junction. By using a suitable molecular material, an asymmetric energy barrier was set across the nanogap. By such means, a bias-selective electroluminescence behavior -similar to the one of a diode -was achieved [92]. By tuning the orientation of the molecules, SPPs could be launched directionally [93].
Dynamic electrochemical and photochemical tuning of the electroluminescence power was achieved by Wang et al. in antennas with the structure displayed in Figure 7(c). Each antenna consisted of a vertical Au nanorod separated from the eutectic gallium indium electrode by a nanogap filled with a redox polymer (PLH) [90,94]. The optical power generated by inelastic electron tunneling was outcoupled to the far-field through the near infrared plasmon modes of the nanorods. The intensities of the tunnel current and electroluminescence were tuned by controlling the oxidation state of the polymer. This control was achieved by introducing the antennas in an oxidizing gas (air) or a reducing gas (hydrogen), while applying a bias or shining light onto them to generate hot electrons from elastic tunneling across the nanogap. The electrically (resp. optically) generated hot electrons enabled the electrochemical (resp. photochemical) oxidation or reduction of the polymer. Remarkably, the polymer can be left in a tunable partial oxidation state and then fully recover its fully oxidized or reduced state. This enables a reversible multilevel tuning of the electroluminescence signal, with a memory effect like that of a synapse [90].

Overbias emission
In contrast with the works reviewed in the previous section, some other studies have reported nanogap antennas featuring photon emission at energies above the bias (E > eV bias ). The origin of this "overbias emission" is still under discussion. Several mechanisms accounting for overbias emission have been already proposed, depending on the experimental configuration considered: excitation of a plasmon by multiple inelastic tunneling electrons followed by outcoupling of the plasmon [95], simultaneous electrical and optical excitation of a plasmon followed by outcoupling of the plasmon [96,97], hot-hole formation and electronelectron interaction at the junction [98,99], generation of a hot electron gas of elastically tunneling electrons followed by blackbody radiation from this gas [100][101][102][103], excitation of plasmons by inelastic tunneling electrons, followed by relaxation of these plasmons into hot electron-hole pairs that recombine radiatively with the assistance of plasmons [104]. While the three former mechanisms were proposed to account for the trends of scanning tunneling microscopy experiments, the two latter were proposed in the case of in-plane nanogap antennas.

Blackbody emission by hot electrons
Buret et al. reported overbias emission in Au nanogap antennas produced by direct electromigration of a nanowire fabricated by electron beam lithography. An example of the geometry is shown in Figure 8(a). This allowed the conductance to reach values as high as G = 0.8G 0 [100]. For V bias in the range of several hundreds of mV, while the electroluminescence peak was in the near infrared, a significant light intensity was measured at overbias energies in the visible. In contrast with antennas for which light emission is driven by inelastic electron tunneling followed by plasmon excitation and outcoupling (previous section), the shape of the emission spectrum was not changed when V bias was increased.
Based on a qualitative model, the following mechanism was proposed to account for the experimental trends. Due to the high conductance of the nanogap and its small volume, the density of electrons tunneling elastically through it is remarkably high. Therefore, electron-electron collisions occur at a much faster rate than bulk electron-phonon interactions, which are less likely because of the high surface/volume ratio of the antenna in the nanoconstriction region. This leads to the formation of an electron gas in the metal around the nanogap, with temperatures reaching 2000 K. Its energy is lost radiatively by collisions at the metal surfaces, following a Boltzman curve that extends up to the visible region at energies E > eV bias . Therefore, the overbias emission was attributed in this case to blackbody radiation from a gas of elastically tunneling hot electrons.
Light emission by this mechanism was shown to be compatible with photonic integration. It was coupled to plasmonic waveguides [101] as shown in Figure 8(b), and to dielectric waveguides [103]. Therefore, it is appealing for the design of integrated light sources. Such light sources could present fast modulation capabilities, because the hot electron gas deexcitation processes can occur in times as short as few ps.

Plasmon-assisted radiative recombination of hot electron-hole pairs generated by inelastically-induced plasmons
More recently, Cui et al. studied the overbias emission in nanogap antennas fabricated by electromigration of a nanowire with a bowtie designed by electron beam lithography [104]. By such means the nanogap was located accurately at the center of the bowtie. The typical geometry is shown in Figure 8(c). Many devices were fabricated, showing G values up to 0.25G 0 . Different materials were used to fabricate the antennas: Au, palladium, or an Au-palladium alloy. In contrast with previous works on electromigrated antennas where a chromium adhesion layer was always introduced under the Au, Cui et al. produced antennas with and without a chromium adhesion layer. Au antennas without chromium showed the strongest overbias emission whereas those with the highest G emitted only at photon energies in the visible above eV bias and, surprisingly, those with G as small as 10 −3 G 0 still showed some overbias emission. Their light emission spectrum showed features characteristic of blackbody radiation together with antenna LSPs. However, the differences in emission intensity between the antennas made of different materials could not be explained by the related changes in gap ∕ 0 .
To account for these trends, it was proposed that light emission in the Au antennas without chromium wetting layer involves hot carriers and plasmonic effects, the latter coming two times into play. In the proposed mechanism, inelastically tunneling electrons excite plasmons at the nanogap surface. These plasmons then generate electron-hole pairs with a high density. These electron-hole pairs thermalize to generate a gas of hot carriers, which recombine radiatively with the assistance of plasmons.

Single-and multi-electron processes for light emission mechanism
Light emission as a function of conductance in a scanning tunneling microscope operating at room temperature and ambient conditions was examined by Fung et al. [105] A spherical Au single-crystal formed by melting 250-nm diameter Au wire and a cut Au wire were used as the two electrodes. It has been shown that overbias emission was due to a mixture of single-and multi-electron processes rather than being due to blackbody radiation (Figure 8(d)). In particular, the power law of the overbias emission implied America. (c) Au bowtie nanowire with nanoconstriction with no chromium wetting layer under the Au film. The whole light emission spectrum appears at photon energies in the NIR together with plasmonic resonance bands. Reproduced from ref. [104] with permission; copyright 2020 American Chemical Society. (d) Illustration of the mechanism for overbias emission from multi-electron processes and 2D photovoltage-conductance histogram of the overbias emission with the mean fitted power law overlaid in black. Reproduced from ref. [105] with permission; copyright 2020 American Chemical Society.
an increase of electronic temperature of approximately 200 K which cannot be accounted for using blackbody radiation theory, whereas multi-electron processes can produce blackbody-like phenomena at room temperature. Recently, Zhu et al. proposed to tune the dominant light emission mechanism through single-electron or higherorder multi-electron inelastic tunneling to recombination from a steady-state population of hot carriers by progressively altering the tunneling conductance of an aluminum junction [106]. It has been shown that the dominant emission mechanism was set by a combination of tunneling rate, hot carrier relaxation timescales, and junction plasmonic properties, which could potentially resolve the longstanding debate regarding the relative roles of these candidate mechanisms.

Toward miniaturized and tunable nano-optoelectronic devices
Direct electrical excitation of plasmonic modes integrates plasmons and electric nano-circuitry which is highly advantageous as it dramatically reduces the overall footprint of the experiment resulting in novel nano-optoelectronic devices. In this context, inelastic electron tunneling offers an alternative nanoscale light source with switching times at the femtoscale, without the need of additional semiconductor materials as for plasmonic organic and inorganic LEDs. In this context, Ochs et al. demonstrated a device for the selective electrical excitation of two distinct plasmonic modes, which makes it the smallest electrically driven light source with switchable polarization states [107]. Mode selectivity is realized by precisely positioning nanoscale excitation sources within the respective modal field distribution. A nanoscale organic light-emitting antenna was also recently presented as a color-and directionality-switchable point source. The device consisted of laterally arranged electrically-contacted gold nanoantennas with their gap filled by the organic semiconductor zinc phthalocyanine (ZnPc) [108]. Since ZnPc shows preferred hole conduction in combination with gold, the recombination zone relocates depending on the polarity of the applied voltage and couples selectively to either of the two antennas. Thereby, the emission characteristics of the device also depend on polarity.
The emission of graphene can be extended toward telecom and visible spectral range by stacking two graphene layers and twisting one with respect to the other, which results in altering the band structure of the system and, hence, further tuning the output emission. Kuzmina et al. reported visible light emission from tunnel junctions consisting of a twist between two single-layer graphene flakes separated by a thin layer of heaxagonal boron nitride, and excited via inelastic electron tunneling [80]. A twist between the two flakes of graphene alters the band structure of the system, as demonstrated by a visible photoluminescence produced by strongly bound excitons. The emission spectrum was shown to be strongly dependent on the twist angle . For close to 0 • , the emission spectrum exhibited a resonant spectral peak in the near-infrared range, which broadens with increasing and eventually disappears.
An additional class of devices exploiting metallic nanogap antennas and molecular junctions was recently demonstrated. The molecular junction geometry consisted of a monolayer of molecules trapped between a single Au nanoparticle and a flat Au film incorporated within an electrode structure [109]. This robust lithography-free molecular optoelectronic device geometry enabled creating molecular junctions with reproducible morphology and electrical response. The well-defined geometry allowed generating optical-frequency voltages ∼30 mV within the molecular junction from 100 μW incident light, and producing photocurrent by optical rectification (>10 μA/W) from only a few hundred molecules.

Conclusions
In this review, we presented nanogap antennas with gaps in the nanometer and sub-nanometer scale and displayed the impact of quantum tunneling effects on the plasmon resonances of these structures. We investigated light emission in electrically driven nanogap antennas based on inelastic tunneling effect or hot-electron mechanism and demonstrated a spectrally tunable electroluminescence with the applied bias. Such nanogap structures have been proven as a dominant tool for controlling and manipulating light at the nanoscale level, and enhancing the efficiency of photo-detection, light emission and sensing. We then presented the potential applications of electrically driven nanoantennas and discussed some future challenges and opportunities.
Research funding: This work was partly funded by l'Agence Nationale de la Recherche (ANR), project ANR-22-CE24-0011-03. Author contributions: All authors have accepted responsibility for the entire content of this manuscript and approved its submission.

Conflict of interest: Authors state no conflicts of interest.
Data availability: Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.