2D materials integrated with metallic nanostructures: fundamentals and optoelectronic applications

2.1 Raman scattering of 2D materials enhanced by plasmonic nanostructures

Raman spectroscopy is a nondestructive technique for chemical sensing and mapping of 2D materials, which typically has a very weak signal. Since the discovery of surface-enhanced Raman spectroscopy (SERS) in the early 1970s [86], a variety of plasmonic nanostructured materials have been employed as surface SERS substrates. It is shown that the electromagnetic enhancement dominating in the measured SERS enhancement factor approximately follows the fourth power of the local electric field enhancement [62], [63]. Two-dimensional materials exhibit flat surfaces without dangling bonds, which are thought to be strong candidates for fundamental studies of Raman enhancement effect. For instance, graphene’s electronic structure can be uniquely captured in its Raman spectrum, which allows high-throughput and nondestructive identification of graphene layers [87], and Raman spectroscopy is widely explored to identify the quality and the number of layers in 2D materials.

As an example of enhanced light–matter interaction, the SERS of 2D materials, such as graphene, can be modified tremendously by plasmonic nanostructures. The strong plasmon-induced enhancement of the Raman signal of graphene is due to the enhanced near-field interaction between graphene itself and the plasmonic modes [88], [89], [90], [91]. This interaction results in an increase in absolute light absorption of 30% with up to 700-fold enhancement of the Raman response of the graphene, as shown in Figure 2A [88]. It has been shown that developing a flat 2D surface for Raman enhancement has promising importance for the further applications of SERS [89], [93]. As most 2D materials provide a good choice as an ideal flat substrate, the hybrid 2D materials–plasmonic structure is also with potential for use as Raman enhancement substrates (Figure 2A) [88]. Tip-enhanced Raman spectroscopy, where the enhancement of Raman scattering occurs near the atomically sharp pin [94], has been used for chemical imaging at the single-molecule level [95]. Quantum mechanical effects, such as electron tunneling and nonlocal screening, become extremely important especially when the distance between the tip and material approaches the subnanometer length scale, which are beyond the classical picture for the nanogap plasmons [96]. As shown in Figure 2B and C, Chen et al. [92] used layered MoS₂ as a 2D atomic crystal probe in nanoparticle-on-mirror nanoantennas to measure the plasmonic enhancement in the gap by quantitative surface-enhanced Raman scattering. For a subnanometer gap, the probable emergence of quantum mechanical effects renders an average electric field enhancement lower than classical predictions. The SERS system of 2D nanoplatelets sandwiched between a gold (Au) nanoparticle and an Au surface can also reveal a phonon doublet arising from oscillations perpendicular to and within the platelet plane [97]. The SERS detection from the combination of extreme plasmonic confinement and ultrathin layers opens the potential to excite and probe local light–matter interactions within and in neighboring plasmonic structures, which may provide a new way for further exploring quantum plasmonics.

Figure 2:

Raman scattering of 2D materials enhanced by plasmonic nanostructures.

(A) A hybrid graphene–plasmonic structure for studying light–matter interaction with Au-void nanostructures exhibiting resonances in the visible range. Enhanced coupling of graphene to the plasmon modes of the nanovoid arrays results in significant frequency shifts of the underlying plasmon resonances, enabling 30% enhanced absolute light absorption by adding a monolayer graphene and up to 700-fold enhancement of the Raman response of the graphene. (B) The quantitative probing of the vertical and horizontal SERS of MoS₂ inside plasmonic antennas with a gap distance reaching down to a well-defined subnanometer scale. Layered 2D MoS₂ crystals with strict lattice arrangement were designed as probes to create robust and uniform gaps with intervals of 0.62 nm, indicated by schematic and high-resolution transmission electron microscope images. (C) Atomic displacements of the phonon modes in the unit cell of 1 L MoS₂. The Raman scattering spectra and image (inset) show the plasmon-enhanced Raman signals, scale bar, 2 μm. Reproduced/adapted with permission from Zhu et al. [88] (A) and Chen et al. [92] (B, C).

2.2 Enhancing PL of 2D materials by plasmonic nanostructures

Transition metal dichalcogenides have attracted increasing attention due to their strong light emission, indirect-to-direct bandgap transition, and spin and valley degrees of freedom. The electronic band of TMDCs or their layered stacks is strongly influenced by the atomic compositions, the number of 2D layers, and the rotational or vertical stack of layers. As an example, when bulk MoS₂ is thinned to a monolayer structure, it transforms from an indirect to a direct gap semiconductor, enabling strong exciton PL at room temperature due to the direct electron transition without the loss of momentum [98]. The enhanced PL from TMDCs has been further explored to realize coherent light sources from a TMDC monolayer [15], [99], [100], [101]. In quantum-electrodynamics, local density of photonic states of exciton recombination can be modified by the environmental electromagnetic field. Plasmon–exciton coupling can be used to enhance the emission efficiencies of light-emitting materials and devices. In the surface plasmon resonant energy, the excitonic electromagnetic decay rates are dramatically enhanced, leading to improved quantum yield of PL [61]. Enhanced PL up to 12 times has been observed from the fact that plasmonic resonance couples to both excitation and emission fields and thus boosting the light–matter interaction at the nanoscale [102], see Figure 3A.

Figure 3:

Enhancing PL of 2D materials by plasmonic nanostructures.

(A) Enhanced light emission from large-area monolayer MoS₂ using plasmonic silver nanodisc arrays, where enhanced PL up to 12 times was measured. (B) The valley-dependent directional coupling of light using a plasmonic nanowire–WS₂ system. It shows that the valley pseudospin in WS₂ couples to transverse optical spin of the same handedness with a directional coupling efficiency of 90±1%. (C) Schematic of the resonant field of the chiral metasurface coupled with valley-polarized excitons. (D) The WoS₂ metacoupling tailors the measured PL spectra in the far field at T=87 K where the PL intensity of MoS₂ monolayer is enhanced by the chiral metasurface, see the inset optical image. Reproduced/adapted with permission from Butun et al. [102] (A), Gong et al. [103] (B), and Li et al. [104] (C, D).

In most monolayer TMDCs, the broken inversion symmetry combined with time-reversal symmetry causes opposite electron spins at the energy-momentum dispersion, so-called valleys [105], [106]. The emergence of 2D TMDCs materials has sparked intense activity in so-called valleytronics, as their valley information can be encoded and detected with the spin angular momentum of light. The valley-dependent directional coupling of light was demonstrated using a plasmonic nanowire–WS₂ system, where the valley pseudospin in WS₂ couples to transverse optical spin of the same handedness with a directional coupling efficiency of 90%±1% (Figure 3B) [103].

In addition, optical metasurfaces [107] can facilitate valley transport and establish an interface between valleytronic and photonic devices [108]. Specifically, plasmonic metasurfaces can evanescently interact with excitons in 2D TMDCs in the near field and at the same time apply a spin-related geometric phase to light via spatially distributed plasmonic nanoresonators. It was reported that valley-polarized PL of MoS₂ can be tailored through near-field interaction with plasmonic chiral metasurface. The resonant field of the chiral metasurface couples with valley-polarized excitons and tailors the measured PL spectra in the far field (Figure 3C and D) [104]. These plasmon-enhanced excitons are promising for application as low-power, ultrafast lasers and modulators and for the study of many-body quantum phenomena [16].

2.3 Nonlinearity enhancement in 2D materials by plasmonic nanostructures

Nonlinear optical processes, including saturable absorption, second-harmonic generation (SHG), third-harmonic generation (THG), and two-photon absorption, enable technology for a large range of applications from light generation, quantum photonics, and attoscience to bioimaging and sensing. Two-dimensional materials also exhibit extraordinary NLO properties, which start with the case of ultrafast and broadband saturable absorption response of graphene in mode-locked lasers [109], [110].

The optical response of a material to an applied optical field can be expressed by expanding the polarization P(t), as a power series in terms of the optical field strength E(t),

where the coefficients χ⁽ⁿ⁾ are the nth-order susceptibilities of the medium, and the presence of such a term is generally referred to as an nth-order nonlinearity, representing both the polarization-dependent nature of the parametric interaction and the symmetries (or lack of) of the material.

The NLO effect is sensitive both to the phase-matching condition and the crystal symmetric. Provided by the in-plain coupling from plasmonic nanostructures, a phase-matching condition is in principle satisfied for 2D materials when both the momentum and energy are conserved simultaneously. The odd-order nonlinearity, such as THG, can be observed in any symmetry, whereas all the even-order susceptibility will be vanished in the centrosymmetric material. As mentioned, 2D materials usually have very short light interaction lengths with the pump laser because they are atomically thin, such that light-harmonic generation is generally inefficient. For TMDCs, the lack of inversion symmetry at the surface can give rise to strong SHG signal under intense optical pump and plasmonic enhancement. Using remotely excited SPPs, axial collimated but transversely divergent SHG in a single silver nanowire–monolayer MoS₂ hybrid system was demonstrated to generate and manipulate SHG emission around subwavelength waveguides, as shown in Figure 4A [59]. In the work done by Wang et al. [111], a 7000-fold SHG enhancement without peak broadening or background in the spectra, was reported (Figure 4B). In addition, SHG amplitude can be dynamically controlled via selective excitation of the lateral gap plasmon by rotating the laser polarization, as shown in the right panel of Figure 4B.

Figure 4:

Nonlinearity enhancement in 2D materials by plasmonic nanostructures.

(A) Left, SHG from silver nanowire–monolayer MoS₂ hybrid structure; Middle, Optical imaging under the excitation wavelength of 796 nm; right, imaging at SH frequency, the white dotted line is the outline of monolayer MoS₂. (B) Left, SHG from monolayer WSe₂ on the Au trench hybrid system; right, enhanced SHG mapping at the transverse polarization resonant excitation. The white dotted line is the outline of monolayer WSe₂. (C) A synthetic structure with a plasmonic metasurface substrate can yield large SHG from the valley band of the WS₂ and split opposite spin components of the second-harmonic valley photons into different directions. (D) A proposed hybrid graphene–plasmonic grating to achieve enhanced nonlinear effects at terahertz frequencies. Reproduced/adapted with permission from Li et al. [59] (A), Wang et al. [111] (B), Hu et al. [112] (C), and Guo et al. [113] (D).

Plasmonic metasurface can entangle the phase and spin of light and simultaneously enhance and manipulate nonlinear emission of the on-top 2D materials. The second-harmonic valley photons in monolayer WS₂ at room temperature were coherently pumped by light, separated, and routed to predetermined directions with a spin-related geometric phase Au metasurface (Figure 4C) [112]. The nonlinear hybrid graphene-covered plasmonic grating was proposed to realize efficient third-order nonlinear terahertz (THz) effects with a limited fabrication complexity (Figure 4D) [113]. These hybrid graphene–plasmonic structures and devices are expected to be useful for nonlinear spectroscopy, noninvasive subwavelength bioimaging, and ultrafast communication applications.

2.4 Ultrastrong light–matter interactions enabled by acoustic plasmons

As mentioned above, the light–matter interaction in 2D materials can be greatly enhanced by polaritons, taking the benefits of their extraordinary field confinement. Relying on the hybridization mechanism, plasmon polaritons with extremely large wavevectors can be achieved when the separation between two layers is decreased. Here we review recent progress of ultraconfined acoustic plasmons existing in hybrid 2D-dielectric-metal systems, particularly focusing on the investigation of the nonlocal response and quantum features of plasmons in 2D materials. As an example, acoustic graphene plasmons (AGPs) exhibiting a linear dispersion [114], [115], [116] are found in a graphene monolayer lying at a small distance d from a metal substrate, (see the top left in Figure 5A). Figure 5A shows plasmon velocity v_p and dispersion curves of AGPs as a function of the thickness d of the dielectric. The solid and dashed lines represent the results obtained by the Drude mode (so-called the local model) and random phase approximation (RPA, i.e. the nonlocal theory), respectively. One can see that the plasmon velocity becomes smaller, and the dispersion curve of the AGPs moves toward larger wavevectors as d is reduced. The classical results obtained by the Drude model agree quite well with those obtained by the RPA model for a large distance, while showing significant quantitative deviations when d is below 10 nm. Especially when the dielectric separation is on the order of a few nanometers, the classical theory totally breaks down, showing that the dispersion diagram falls within the electron–hole continuum and that the plasmon velocities are below the Fermi velocity of carriers in graphene, which are prohibited by the nonlocal theory. The strongly confined AGP supported in the system consisting of hexagonal boron nitride (hBN)–encapsulated graphene deposited onto a metal gate [117] was first experimentally visualized [117] by scattering-type scanning near-field optical microscope (s-SNOM) [120], [121]. A nanoscale-resolved THz photocurrent near-field microscopy was developed (see the top left in Figure 5B), where the near-field AGPs were detected thermoelectrically [122]. The near-field photocurrent (see the bottom-left in Figure 5B), in the graphene was measured by the two metal contacts. By measuring the oscillation period for the photocurrent profiles as a function of frequency, the dispersion relation for the AGPs was experimentally obtained (see the red symbols in the right panel in Figure 5B, showing a nearly linear dispersion at low frequencies). The experimental results were found to be consistent with calculated dispersion (see the blue contour plot in the right panel).

Figure 5:

Ultrastrong light–matter interactions enabled by acoustic plasmons.

(A) Plasmon velocity v_p and dispersion curves of acoustic graphene plasmons supported by a graphene-dielectric-metal system as a function of the thickness d of the dielectric separating a graphene layer from a metal substrate. Both the local and nonlocal responses for AGP are considered, whereas the metal is assumed to be a perfect electric conductor. (B) Left, Schematics of the experimental setup and real-space imaging of acoustic graphene plasmons in a graphene photodetector with split-gate architecture, where the laser-illuminated metal tip of an atomic force microscope serves as a nanoscale near-field light source. Right, The dispersion of AGPs obtained by s-SNOM, see the red symbols, showing a nearly linear dispersion at low frequencies. (C) Left, Schematics of the experimental setup, where terahertz laser illuminates the device, launching the AGP. Right, Photocurrent profiles for three different devices and their corresponding plasmon wavelength λ. (D) Left, Device design for probing ultimate confinement limit for AGP and simulated plasmonic fields for the separations at 10 and 1 nm. Right, Simulated plasmon wavelength as a function of metal–graphene separation under four different models. Reproduced/adapted with permission from Gonçalves et al. [57] (A), Alonso-González et al. [117] (B), Lundeberg et al. [118] (C), and Alcaraz Iranzo et al. [119] (D).

Due to the linear dispersion, AGPs have the potential to achieve extremely large wavevectors, while having their associated plasmon velocity close to the Fermi velocity of carriers in graphene when the graphene–metal separation is largely reduced. In principle, the distance between the graphene and metal can amount to the single-atom thickness if we choose a single-layer of hBN as a spacer. Therefore, the unique 2D-dielectric-metal system allows us to study nonlocal effects and to probe the quantum mechanical features [118], [123], [124]), as discussed in Figure 5A. Lundeberg et al. [118] investigated experimentally the nonlocal response of graphene electron liquid. Notably, Figure 5A shows the concept for enhancing nonlocal effects by slowing down the velocity of graphene plasmons. The left panel of Figure 5C shows the experimental setup, where the graphene is encapsulated by hBN [125] and placed on the top of AuPd metal layer. Three devices with different graphene–metal separation were fabricated and were all electrically contacted by use of one-dimensional electrical contact to the 2D material [126]. The s-SNOM in the photocurrent mode was used to visualize and characterize AGPs. The measured photocurrent profiles (see the right panel in Figure 5C) shows the interference fringes in the dependence of tip position, allowing us to determine the plasmon wavelength λ by fitting to the photocurrent with an appropriately subtracted background. The plasmon phase velocity was extracted from the photocurrent maps for three devices, and it turned out that the local approximation showed a significant discrepancy, while showing excellent agreement for the full nonlocal theory.

The hybrid 2D-dielectric-metal structure provides a unique platform to probe the ultimate plasmon confinement limits. Alcaraz Iranzo et al. [119] fabricated the 2D-dielectric-metal heterostructure consisting of monolayer of graphene and an array of metallic rods, separated by hBN or Al₂O₃ (see the top left in Figure 5D). Here the presence of the metallic rods leads to efficient screening of graphene plasmons, at the same time facilitating efficient coupling between far-field light and the strongly confined AGPs. The field profiles shown in the left-down in Figure 5D illustrate the strong field confinement between the metal and graphene. Bringing the metal closer to the graphene layer significantly enhances the vertical confinement (see the right panel in Figure 5D), where the simulated results were obtained under four different models. These ultraconfined plasmon modes, discussed in [118], [119], [124], give rise to realize ultrastrong light–matter interactions.

3.1 Hybrid 2D material plasmonic modulators

Being a crucial component in the optical interconnects, electro-optic modulators convert the electrical signal into optical signal in order to take advantage of the high speed and low power consumption in the optical domain. Multiple attributes of light can be employed to achieve light modulation, including amplitude, phase, and polarization [147], [169], [170]. Up to now, most of high-speed integrated electro-optical modulators, especially Si modulators, heavily rely on the plasma dispersion effect [171]. The plasma dispersion effect is related to change the density of free carriers in a semiconductor. Other effects, such as thermo-optic effect [172], Franz–Keldysh effect [173], quantum-confined Stark effect [174], and Kerr effect [175], could also be used for the optical modulators. The key figures of merit used to characterize a modulator are the speed, modulation depth, and operation wavelength range. Although impressive progress has been made in the integrated electro-optic modulators, many reported modulators based on the bulk material still face crucial challenges in realizing ultrahigh-speed modulation, owing to their relatively low carrier mobility.

Because of ultrahigh carrier mobility in 2D materials, they could provide the possibility to realize ultrahigh-speed modulation when combined with conventional optoelectronic devices [176]. Among them, graphene has been extensively explored for the application in electro-optic modulator. Most of graphene-based modulators rely on gate-tunable electroabsorption effect, thus acting as an intensity modulator [32], [177], [178]. It could also perform as a phase modulator by applying a Mach–Zehnder interferometer configuration [179], [180]. The most distinctive performance of the reported graphene modulators lies in its potential to realize extremely high modulation speed. For instance, the all-optical modulator reported by Li et al. shows a response time of 2.2 ps, which is equivalent to a calculated bandwidth of 200 GHz [178]. Another significant advantage originating from the gapless nature of graphene is its broad operation bandwidth, which has been demonstrated to cover visible, infrared, and THz bands [181]. However, for these Si-based 2D material modulators, the optical modes are usually strongly confined in the high-index materials, leading to weak light–matter interaction, as well as poor modulation depth. Besides, the size of the reported modulators is normally limited in tens of microns in order to reach the designed moderate modulation depth [182], which may further impede the miniaturization of the devices.

Because SPPs show the ability to manipulate light on the subwavelength scale, it could be used to direct more energy of light into a 2D material to enhance the performance for a 2D-based device, as well as to shrink the size of a 2D-based modulator. Extensive researches have been conducted to demonstrate graphene-based modulators by taking advantage of the plasmonic nanostructures. Early works were devoted in the flexible control of plasmon excitation [183], [184], plasmon resonance wavelength [185], [186], or plasmon resonance quality factor [187]. Various metallic nanostructures, including Au nanorod, Au strip, Au bowties, and Ag nanowires, were introduced to stimulate plasmonic modes. Although graphene is only a single-atom thick, its effect on altering the plasmonic wave is remarkably strong. For example, in a hybrid graphene–Au nanorod system proposed by Kim et al. [185], a 20-meV shift of resonance frequency, 30% increase in quality factor, and resonance scattering intensity were experimentally observed by applying different gate voltage, as shown in the Figure 6A. Moreover, by combining the graphene sheet with metallic bowties, Emani et al. [187] demonstrated that the transmittance of infrared incident light could also be effectively modulated by changing the gating of graphene, which is displayed in Figure 6B. The decrease in resonance linewidth was experimentally observed to be up to 210 nm by applying external voltage of 35 V. The metallic nanostructure, as a metasurface, could also be utilized to demonstrate perfect absorbing with the aid of the metallic mirror, which is shown in Figure 6C. By switching the absorber in and out of the critical coupling condition via the gate voltage applied on graphene, an impressive modulation depth up to 100% was achieved [188], which could hardly be realized only with graphene. Meanwhile, they also demonstrated a modulation bandwidth up to 20 GHz over a broad wavelength range. Although these early works rely on top illumination and operate within infrared wavelength, which might not be readily applied for the optical interconnects, they could be recognized as endeavor, demonstrating the feasibility for the plasmonic graphene modulator.

Figure 6:

Plasmonically enhanced electro-optic modulation under normal illumination.

(A) The hybrid graphene–Au nanorod system exhibits a 20-meV shift of resonance frequency and a 30% increase in quality factor upon external voltage. (B) Metallic bowties are employed to enhanced light–matter interaction within the channel region of FET, resulting in an effective modulation with a resonance wavelength decrease of 210 nm. (C) Metallic nanostructures are placed on top of the graphene, which combine with the metal reflector on the bottom of the dielectric interlayer. The designed structure forms a perfect absorber, exhibiting a modulation depth up to 100%. Reproduced/adapted with permission from Kim et al. [185] (A), Emani et al. [187] (B), and Yao et al. [188] (C).

Compared to the device with top illumination, waveguide-type modulators are much more favorable due to the compatibility with other devices in terms of the large-scale photonic integration. By taking the benefits of SPPs, plasmonically enhanced waveguide-type modulators have attracted more and more attention in recent years. The hybrid graphene–plasmonic waveguide modulator was first demonstrated by Ansell et al. [189]. In their design, the gold strip supporting SPPs is covered by an hBN, which acts as a dielectric spacer. Three different plasmonic modes, that is, flat plasmons (FPs), corrugated plasmons (CPs), and wedge plasmons (WPs), can be excited separately by focusing the incident light beam with the wavelength of 1.5 μm to different parts of the coupler, as illustrated in Figure 7A. The modulation was achieved by applying gating to the graphene sheet, thus changing the light absorption rate within the waveguide. Among these three modes, WPs exhibited highest modulation depth of 0.03 dB/μm, which is 30 and 230 times larger than the performance of CPs and FPs modes, thanks to its strong interaction with graphene. Although the proposed modulator might not hold best modulation performance, it paves a promising way toward practical realization of very compact and efficient, potentially ultrafast, and broadband hybrid graphene–plasmonic optical modulator at 1550 nm.

Figure 7:

Plasmonic waveguide optical modulators.

(A) Three different SPP modes supported by different structures, which are flat plasmons, corrugated plasmons, and wedge plasmons, respectively. The highest modulation depth of 0.03 dB/μm is reached for the wedge plasmons. (B) An efficient modulator based on the leaky mode of plasmonic slot waveguides. The leaky mode is achieved by optimization of the hybrid waveguide to ensure a low insertion loss of 0.68 dB/μm. The modulation depth is 0.13 dB/μm, and modulation bandwidth is 200 kHz. (C) Plasmonic modulator based on monolayer WSe₂ encapsulated by hBN. The SPP is excited either by the metallic grating or by the metallic surface. The modulation could be realized by either plasmonic control or optical control, which has a modulation depth of 0.004 dB/μm and 0.04 dB/μm, respectively. Reproduced/adapted with permission from Ansell et al. [189] (A), Ding et al. [34] (B), and Klein et al. [190] (C).

In order to improve the modulation depth, Ding et al. [34] proposed an efficient modulator based on leaky-mode plasmonic slot waveguides. The plasmonic slot waveguide (Figure 7B) was formed by introducing a small slot in a thin Au sheet. Two layers of graphene with a thin layer of Al2O3 form an effective capacitor, thus enabling modulation of the light propagation in the waveguide through tuning the optical absorption of graphene sheet. They experimentally demonstrated the modulation depth of 0.13 dB/μm with relatively low propagation loss of 0.68 dB/μm for the slot width of 120 nm. More importantly, the demonstrated modulator was fully integrated with Si technology, where the light was coupled in/out the plasmonic waveguide by use of a tapered Si waveguide. The high modulation depth corresponds to a considerable extinction ration of 2.1 dB in the dynamic performance characterization. Although the modulation bandwidth is approximately only 200 kHz, which is limited by the large RC constant of the device, the proposed device holds the highest modulation depth of the graphene modulator based on plasmonic waveguides. Further optimization of the response time can be achieved by using optimized structure, such as metal–insulator–metal waveguide, which has an ultrafast response time of 2.2 ps.

Besides graphene, TMDCs also exhibit impressive optoelectronic properties including large third-order NLO susceptibilities and strong light–matter interaction [191], [192]. Most recently, Klein et al. [190] proposed a plasmonic modulator based on WSe2 monolayer integrated on top of a metallic waveguide. As shown in Figure 7C, WSe2 monolayer is electrically isolated from the metal by encapsulating with hBN. Two different methods to control the SPP propagation were experimentally demonstrated in their work, which were optical control and plasmonic control, respectively. The modulation was realized by the strong interaction between the SPPs and excitons in the WSe2 monolayer. In both schemes, the SPPs were excited by focusing the light onto the metal grating coupler. For light control scheme, another free-space laser was focused on the hBN-WSe2-hBN structure to act as optical pump. The modulator had a normalized transmission change of 4.1×10⁻³ with the pump energy of 1.739 eV, equivalent to a modulation depth of 0.004 dB/μm. For the plasmonic control, however, both the pump and probe laser were coupled into the input grating coupler. The normalized transmission change is 10-fold larger than that of the optical pump scheme, which reaches 4.1×10⁻² with the pump energy of 1.739 eV. This value corresponds to a modulation depth of 0.004 dB/μm. Moreover, the response time was experimentally verified to be 290 fs, equivalent to a modulation bandwidth of 1.5 THz. The performance of this plasmonic modulator with 2D material is quite impressive, including high modulation depth of 73% and fast response time comparable to the previously reported plasmonic modulator [34], as well as ultralow power consumption of 650 fJ pump energy. Although these performances were achieved under low temperature (around 10 K) in order to prevent thermal broadening effect, it still could be recognized as a milestone in the area of the plasmonic modulator based on the 2D material. As another important member in the 2D materials family, black phosphorus (BP) and its analog have also been utilized in realizing optical modulation, as well as the Q-switched laser [193], [194], [195], [196], [197]. One of the representative works was reported by Peng et al. in 2018 [198]. When the BP is integrated with Si waveguide, a 5-dB modulation depth and 1-dB insertion loss could be expected based on the measured results of absorption coefficient. However, the integration of BP and metallic nanostructures is yet to be investigated up to now.

3.2 Hybrid 2D materials plasmonic photodetectors

As an opposite building block of optical modulator, photodetectors can convert the light signal into the electrical signal for the subsequent processing. Several bulk materials have been widely used in the integrated photodetectors, such as Si [199], [200], [201], [202], germanium (Ge) [203], [204], [205], [206], [207], [208], and III/V materials [209], [210], [211]. To date, most commercialized integrated photodetectors for optical communication wavelength are based on III/V materials. However, they hold several disadvantages impeding its practical applications including high cost and incompatibility with the compatible metal-oxide semiconductor (CMOS) technology. For example, Ge is an appealing absorbing material for optical communication wavelength and has been widely investigated. However, the large dark current and limited operation bandwidth requires to be addressed before its practical application. Silicon-based photonic components are especially attractive for realizing low-cost photonic integrated circuits using high-volume manufacturing processes, thanks to its high compatibility with the mature CMOS industry. Due to its transparency in the telecommunications wavelength bands near 1310 and 1550 nm, Si is an excellent material for realizing low-loss passive optical components [212]. For the same reason, however, it is not a strong candidate for detectors, which requires material to have strong light absorption. Therefore, most integrated photodetectors employ either III/V materials or Ge as absorption medium. Although impressive performance based on these two material systems has been demonstrated, both of them hold distinct drawbacks including the incompatibility with Si photonics and large dark current and difficulty in reaching high responsivity and large operation bandwidth simultaneously. Thanks to its gapless nature, graphene exhibits a wavelength-independent absorption property, ranging from ultraviolet to THz regime. Moreover, graphene also holds great potential in reaching ultrahigh operation bandwidth. Based on all these reasons mentioned above, graphene has been widely explored for the photodetector integrated with plasmonic devices, in order to reach high-speed photodetection.

The first generation of the plasmonic photodetector with 2D materials is based on the free-space scheme, which directly projects light onto the absorption region. Among them, Xia et al. [213] proposed one of the earliest graphene photodetectors. In their work, although plasmonic structures are not directly utilized to increase the light absorption, the enhancement near the metal–graphene interface was observed. In 2011, Echtermeyer et al. [77] increased the responsivity of the photodetector by up to 20 times, thanks to the combination of graphene and plasmonic nanostructures. Various types of nanostructures were fabricated and investigated, and they found that the grating with 110-nm finger-width and 300-nm pitch showed the best performance, where the scanning electron microscope image of the structure is shown in Figure 8A. The photovoltage mapping shown in Figure 8A exhibits clear responsivity enhancement near the tips of the nanostructures, which could be ascribed to both large electron band bending and the strong enhancement of the optical field caused by the plasmonic effect. Among all the different polarization states and excitation wavelengths for normal light incidence, the maximum enhancement was observed at 514 nm with a transverse polarization. Similar conceptual demonstration was also reported in nanogap, nanoparticles, or nanoscale antennas. Unlike other structures, Fang et al. [214] proposed a novel graphene-based structure by sandwiching nanoscale antennas between two graphene monolayers (Figure 8B). Each nanoantenna is formed by an Au heptamer with seven individual components with diameter of 130 nm and gap distance of 15 nm. The main physical mechanism of the light absorption enhancement is ascribed to the excitation of tunable Fano resonances. It strongly suppresses the scattering near-transparency window and contributes to hot electron–hole pair generation, thus enhancing the direct excitation of electron–hole pairs in a single-atom-thick graphene layer and leading to higher photocurrent. The photoresponsivity of 13 mA/W was experimentally obtained, and the photocurrent also increased when having larger heptamer sizes. Compared to the structure without antennas, an 800% enhancement of the photocurrent was observed. Although the maximum photoresponsivity remains low, approximately only 10 mA/W, it directly proved the viability of the concept of using plasmonic nanostructures for light harvesting in 2D material–based photodetectors. Last but not least, the response time of these devices mentioned above was not studied, which is another crucial parameter of the photodetection.

Figure 8:

Two-dimensional materials plasmonic hybrid photodetectors with top illumination.

(A) Interdigit metallic strip is deposited on top of the graphene to induce plasmonic enhancement. The finger width is 110-nm finger-width, and the pitch is 300 nm. The experiment data exhibit a 20-times higher photoresponsivity compared to the device without metallic nanostructure. (B) Au heptamer is sandwiched between top layer graphene and bottom layer graphene to enhance photodetection effect. The measured data show an 800% enhancement of the photocurrent. (C) Graphene photodetector integrated with QDs. The I-V curve of the 405-nm incident light and the temporal response of 1550 nm is shown. (D) Plasmonic MoS₂ photodetector with a subwavelength size of detection region. The SPP-mediated responsivity is up to 18 mA/W, with the detection region of 0.03 μm². Reproduced/adapted with permission from Echtermeyer et al. (A), Fang et al. [214] (B), Sun et al. [215] (C), and Blauth et al. [216] (D).

One of the earliest works investigating the temporal property of 2D materials plasmonic photodetectors based on the free-space scheme was reported by Sun et al. [215] in 2017. Unlike conventional plasmonic structures that usually utilize bulk metals as nanostructures, they used self-doped colloidal copper phosphide quantum dots (QDs) both as plasmonic enhancement structure and as tunable light absorption medium. The schematic of the photodetector is displayed in Figure 8C. Except for the plasmonic enhancement effect, the copper phosphide nanocrystals could also induce strong doping in graphene, which enables the formation of a p–n junction by partially covering graphene, thus leading to efficient photodetection. They experimentally demonstrated an impressive responsivity of 105 A/W at visible wavelength (405 nm) and a high responsivity of 9.34 A/W at 1550 nm. The excellent photoresponsivity is attributed to these three aspects: strong light absorption in the copper phosphide nanocrystals, the high carrier mobility in the graphene film, and the plasmon-enhanced NLO properties. Although the hybrid system of copper phosphide and graphene exhibits very high responsivity, the measured rise and decay times of the graphene–copper phosphide hybrid photodetector are only 0.42 and 1 s, respectively. The response time is far below the requirement of today’s high-speed optical network, which could severely limit the potential application of the proposed device. Similar concept of combining QDs with graphene as the photodetection medium was also proposed by Ni et al. [217], with the similar photoresponsivity performance.

Besides graphene, monolayer MoS₂ has also been widely explored as the photodetection material due to its direct bandgap at the visible wavelength as well as near-unity radiative internal quantum efficiency [98], [191], [218]. More recently, Blauth et al. [216] experimentally realized an ultracompact photodetector by integrating monolayer MoS₂ with a plasmonic waveguide. Figure 8D shows the schematic of the proposed device. The plasmonic waveguide was formed by two Au metallic strips, located on top of the MoS₂ layer. When the light is focused on the device, SPPs are excited and propagate toward the nanoscale detection region, which is centralized around the metal-MoS₂ interface with a subwavelength size of only 0.03 μm². Moreover, one can see the distinctively enhanced electrical field at the edge of the Au and MoS₂, leading to strong light–matter interactions, as well as enhanced photoresponsivity. The SPPs-mediated light absorption could yield a responsivity up to 18 mA/W. Despite that the value is not very impressive, the ultrasmall detection region demonstrated in this work could be a potential solution toward the dense on-chip integration of photonic circuit in the future.

Although plasmonic photodetectors based on free-space scheme could provide more convenience in the potential applications, they hold several intrinsic drawbacks including low responsivity and narrow bandwidth owing to large size and poor light confinement. Guiding the light into the waveguide is considered as a promising solution to boost the performance of the photodetector. Because of intrinsic loss of the metal, the metallic nanostructures may induce strong excess loss when integrated with optical waveguides. Careful designs should be taken in terms of utilizing the plasmonic enhancement effect without sacrificing the insertion losses. Therefore, both the design and the fabrication could be more complex than the case for the free-space scheme. At an early stage, researchers proposed to use the graphene strip for the waveguide and graphene ribbons for the photodetector [219]. But the insertion loss of the graphene waveguide is extremely high, thus rendering the photoresponsivity of the device. Then people started to use low-loss platform, such as Si or SiN waveguide to be integrated with the plasmonic waveguide. As one of the representative works, Ding et al. [35] reported an ultracompact graphene plasmonic photodetector on the Si platform. The schematic of the device is shown in the inset of Figure 9A. The plasmonic mode is excited by the optical mode in a tapered Si waveguide. Here the metal is considered as the material for the waveguide, as well as for the contact. The subwavelength confinement of the plasmonic mode gives rise to strong light–graphene interaction, and the narrow plasmonic slot enables short drift paths for the photogenerated carriers. In order to break the inversion symmetry, two different metals (Ti and Pd) were employed, which could induce potential difference for the contacts. The light absorption was increased by one-order of magnitude higher than the device without metal slot, reaching 1 dB/μm. By applying an external voltage of 2.2 V and increasing the length of the device to 19 μm, an impressive responsivity of 360 mA/W was experimentally reached. Another crucial benefit resulting from the narrow slot is the ultrahigh-speed photodetection. The ultrawide bandwidth larger than 110 GHz was experimentally demonstrated, confirmed by several characterization methods. The underlying mechanism could be ascribed to both the ultrafast transit of photogenerated carriers and the strong built-in electrical field within the narrow gap. Another graphene photodetector based on the metal slot was reported by Ma et al. [220] in 2019 (Figure 9B). Although the structure is to some extent similar to [35], significant differences could still be found in their work. One of the major differences is that the gap size is further reduced to only 15 nm, resulting in fundamental change of the photodetection mechanism within the device. It turned out that both photovoltaic (PV) and photobolometric (PB) played the dominant role. The highest photoresponsivity based on these two effects are 0.35 and 0.17 A/W, respectively, with the waveguide length of only 5 μm. Considering the ultracompact footprint, the photoresponsivity of the device was quite impressive. However, the bandwidth of the proposed device was not mentioned. Very recently, Muench et al. [223] reported a plasmonically enhanced graphene photodetector employing the photothermoelectric effect. The metallic slot remains to be the key structure, although the input light was converted to the voltage signal rather than current signal in PB or PV effect. Based on their measurement results, the plasmonically enhanced structure generates the external responsivity of 12.2 V/W and a 3-dB bandwidth of 42 GHz. Considering that the subsequent electrical circuit would prefer voltage signal from the photodetector rather than current signal, this design holds considerable application potential in the future, if the fabrication process could be simplified.

Figure 9:

Plasmonically enhanced photodetectors based on the waveguide.

(A) Ultracompact graphene plasmonic photodetector on Si platform. The gap of the plasmonic slot is 120 nm. Pd and Ti are employed as contact with graphene to induce effective band bending. The maximum measured responsivity and bandwidth are 360 mA/W and 110 GHz, respectively. (B) Plasmonic photodetector with the metal slot width of 15 nm. The responsivity based on PV and PB effect are 0.35 A/W and 0.17 A/W. (C) Plasmonic photodetector based on bowtie nanostructure. When the input light power is 80 μW, the responsivity could reach up to 0.5 A/W. The bandwidth is larger than 110 GHz. (D) Black phosphorous photodetector with three-dimensional integration scheme. An intrinsic responsivity as high as 10 A/W and a 3-dB roll-off frequency of 150 MHz are experimentally demonstrated. Reproduced/adapted with permission from Ding et al. [35] (A), Ma et al. [220] (B), Ma et al. [221] (C), and Chen et al. [222] (D).

Besides the metallic slot structure, arrayed metallic bowtie was also utilized to improve the performance of the 2D-based photodetector [221]. As shown in Figure 9C, the light was guided by the SiN waveguide, and the nanostructured bowties were positioned on top of the graphene. The strong field enhancement arisen by the bowtie structure can be seen in the inset of the schematic. Except for the well-known plasmonically enhanced light–graphene interaction, the arrayed bowtie structure could also induce Bloch mode of the SPPs, thus decreasing the group velocity and further contributing to the enhancement effect. For the monolayer graphene, the responsivity could reach 0.5 A/W for the input power of 80 μW. When the input power was further increased, both Joule heating and Pauli blocking significantly reduced the responsivity of the device. On the other hand, the double-layer graphene device could handle high input power without the saturation effect, maintaining the high responsivity of 0.4 A/W with the input power of 1.3 mW. The measured frequency response of devices was fairly good, showing a flat response covering from 100 kHz to 110 GHz. Moreover, the wavelength-dependent responsivity of the device was checked, exhibiting the flat response covering from S-band, C-band, to L-band of the optical communications. Considering the responsivity, bandwidth, and the footprint, this work could be one of the best photodetectors based on the plasmonic nanostructure integrated with 2D materials so far.

Black phosphorus could be another promising material in terms of photodetection owing to its direct and tunable bandgap, as well as the high sensitivity in the near- and mid-infrared regimes [224], [225], [226]. Black phosphorus analogs, such as Bi, Te, and SnS, have also been applied in the photodetection, but with moderate responsivity and response speed [196], [227], [228]. Multiple works of the BP photodetector based on both free-space and waveguide-integrated schemes were reported. Similar to graphene, these devices still suffer from low responsivity as well as large footprint. To address these issues, Chen et al. [222] proposed a hybrid plasmonic BP photodetector consisting of three different layers with different functionalities (Figure 9D). The bottom layer is a standard silicon-on-insulator platform, which transmits the light into the detection region with ultralow propagation loss. The plasmonic layer in the middle converts the photonic signal into the SPPs wave, concentrating light into the nanogap. The top layer is formed by an exfoliated BP flake as well as the top gating. Based on simulation results, up to 1% photonic emission from the waveguide was transmitted through the nanogap. The high responsivity of 10 A/W was measured in the device without the top gate when the applied bias was 1.5 V. The extremely high photoresponsivity could attribute to both the plasmonically enhanced light absorption, as well as the large gain factor owing to the charge trapping and short transit time within the structure. However, the RC constant severely limits the frequency response of the device, showing 3-dB bandwidth of only 150 MHz. Further optimization regarding the contact resistance, parasite capacitance could be valuable, but also challenging, in order to boost the frequency response while maintaining high responsivity of the proposed device.

Compared with the top-illumination photodetectors, the waveguide-type photodetectors show better performance in terms of large bandwidth, and they are also favorable for the integration with other on-chip optoelectronic devices. Here we summarize recent progress with respect to 2D-based waveguide-type photodetectors (Table 1). Thanks to metallic nanostructures, it is clearly seen, for example, in [35], [221], that photodetectors with plasmonic nanostructures hold much larger bandwidth as well as higher responsivity when compared to pure silicon waveguides. The better performance of these 2D material plasmonic hybrid photodetectors really takes advantages of plasmonic-enhanced light–matter interactions, and the miniaturization of plasmonic waveguide also gives rise to short drift paths for carriers in the photodetection region, resulting in large bandwidth. The chemical vapor deposition (CVD) graphene is widely exploited for integrated devices, thus allowing for mass production. Moreover, various 2D materials besides graphene, such as BP and vdWs heterostructure, have also been exploited with impressive performance. We believe that these developments will further the 2D materials toward practical applications, for example, in optical interconnects, high-speed optical communication, and so on.