Metasurface-tuned light-matter interactions for high-performance photodetectors

Abstract

Photodetection, the conversion of optical signals to electrical signals, plays a pivotal role in optical communication, sensing, and computing. However, with the advent of novel applications such as quantum computing, intelligent driving, high-speed communication, and augmented reality, traditional photodetectors encounter challenges associated with miniaturization, the assurance of high performance, and the need for multifunctionality. These demands have prompted the development of novel photodetectors. One particularly promising avenue involves integrating conventional photodetection with metasurfaces. Metasurfaces facilitate highly localized field manipulation and versatile modes control at the subwavelength scale via plasmon resonance or Mie scattering, thereby enabling superior performance and multifunctionality detection. Consequently, metasurface-enhanced photodetectors have garnered significant attention in recent years. This review provides an overview of the unique roles played by metasurfaces in photodetection and summarizes notable breakthroughs in recent years.

1. Introduction

A photodetector is a crucial optoelectronic device that transforms incident light or electromagnetic radiation into electrical signals, finding extensive applications in information communication [1,2], image sensing [3,4], biomedical research [5], and quantum computing [6]. The photodetectors have evolved into one of the fundamental technologies. With the increasing demand for higher computing speeds and enhanced communication capabilities, photoelectric detection plays a progressively important role. Traditional photodetectors operating in the visible spectrum typically rely on indirect bandgap semiconductors - silicon [3]. Notably, silicon boasts a well-established fabrication process and seamless integration with electronic chips. Meanwhile, direct-bandgap semiconductors, such as III–V compounds [7], find use in specialized domains such as aerospace, where high responsivity is necessary and high costs can be accommodated. These semiconductor materials leverage p–n junctions to efficiently separate photogenerated carriers through a built-in electric field bias-free manner, minimizing dark current and achieving a high on-off ratio. InGaAs and HgCdTe semiconductors [8,9], with tunable bandgaps, cater to near-infrared and mid-infrared light detection, respectively. In conjunction with thermal radiometers for long-wavelength detection, these semiconductor families form the bedrock of commercial photodetectors.

With the advent of quantum computing [6], high-speed communication [10], autonomous intelligent driving [11], and augmented reality [12], the demand for photodetectors characterized by miniaturization, broadband capabilities, and functional diversity has consistently increased. These requirements pose formidable challenges to conventional photodetectors owing to their inherent trade-offs. For instance, miniaturization implies reduced light-receiving areas and decreased light-matter interaction lengths, limiting light absorption and causing low responsivity. Applications such as thermal imaging and medical diagnosis necessitate mid-infrared detection [10,13], but conventional long-wave photodetectors must operate at low temperatures to mitigate thermal noise, resulting in integration complexities and elevated costs. Traditional photodetectors are also limited to measuring only light intensity, neglecting critical parameters such as phase, polarization, wavelength, and orbital angular momentum (OAM) [11]. Incorporating these parameters could significantly enhance optical communication capabilities. Moreover, beyond only measuring the light field intensity, parameters such as polarization state hold substantial significance in diverse fields, including remote sensing, spintronics, and drug detection [14], [15], [16]. Traditionally, functional photodetection has been contingent upon precisely aligned optical paths and voluminous optical components, encompassing elements such as 1/4 wave plates, polarizers, and lenses. However, this conventional approach impedes miniaturization, necessitating the evolution of photodetectors.

Numerous design strategies have been proposed to realize miniaturized, monolithically integrated functional photodetectors. These encompass waveguides [17], two-dimensional material heterojunctions [18,19], and quantum dots [20]. Among them, metasurfaces, characterized as subwavelength-patterned surfaces with robust light interactions, emerge as promising candidates for the succeeding iteration of photodetectors. The metasurfaces, which consist predominantly of periodic subwavelength structures arranged on a two-dimensional plane, afford the pliable manipulation of light phase and mode distribution within an exceptionally compact configuration [11].

Metasurfaces can be broadly categorized into two main classes based on their underlying mechanisms and material compositions: plasmon-based [21,22] and Mie-scattering-based metasurfaces [23]. Plasmonic metasurfaces leverage plasmonic materials, predominantly metals, to manipulate light. In contrast, Mie-scattering-based metasurfaces primarily rely on dielectric materials, often directly machined from optoelectronic substrates. (Note that this article focuses mainly on metasurfaces combined with photodetectors. It's crucial to differentiate these metasurfaces from those used for wavefront shaping, which typically rely on geometric phases and can function in non-resonant bands. Refer to reference [24] for further details.) Plasmonic metasurfaces exploit metallic materials to support surface plasmon polaritons (SPPs), and collective excitations of electrons at the conductor-insulator interface. These SPPs can be categorized as propagating and localized, as illustrated in Fig. 1. Propagating SPPs exist on the metal-dielectric surface but require grating assistance to meet phase-matching conditions for direct excitation. Conversely, localized SPPs reside within metallic nanostructures and can be directly excited by external fields. Their inherent adjustability, minute mode volumes, and substantial near-field enhancement have propelled them to the forefront of plasmonic metasurface design. Recent advancements have also yielded plasmonic metasurfaces based on graphene [25] and transition metal dichalcogenides [26]. Meanwhile, high-refractive-index dielectric nanoparticles can support a plethora of multipolar modes predicted by Mie scattering. Fig. 1 exemplifies the first-order Mie resonance mode supported in a dielectric microsphere. Owing to their low loss characteristics, Mie resonant modes in dielectric structures exhibit a high-quality factor and narrow bandwidth, enabling significant near-field enhancement. The equivalent permittivity near the resonance peak can be adjusted over a wide range, providing dielectric metasurfaces with the capability for the flexible manipulation of light fields [27]. This renders them particularly well-suited for applications demanding low loss and a narrow operating band, such as color-resolved photodetection.

Fig. 1.

Metasurface-enhanced photodetectors for improved performance and multifunctional photodetection[27], [28], [29], [30], [31], [32], [33], [34], [35], [36].

Leveraging their potent light field manipulation capabilities, metasurface- assisted photodetectors offer significant advantages in terms of high performance and multifunctionality, holding immense promise for applications in optical communications, sensing, and computing. On the one hand, metasurfaces enhance various photodetector figure of merits (FOMs) through diverse mechanisms. Carefully designed nanostructures enable the achievement of local field enhancement and optimal absorption at desired wavelengths, leading to a substantial improvement in responsivity [28]. Additionally, metasurfaces can enhance response times owing to their reduced dimensions [29]. Notably, plasmonic metasurfaces introduce a novel photodetection mechanism based on hot-carrier injection, enabling the ultra-high-speed detection of low-energy photons [30]. On the other hand, metasurface-assisted photodetectors possess the capability to achieve on-chip multidimensional information detection by flexibly designing nanostructures, as exemplified in Fig. 1. By manipulating near-field modes, they can distinguish light fields with varying incident angles [31], polarization states [32], wavelengths [33] and OAMs based on parameters such as transmittance, enhancement factor and diffraction angle. Substantial research has been conducted on metasurface-assisted photodetectors to achieve high-performance, integrated, and multifunctional photodetection. This review presents a comprehensive overview of metasurfaces' diverse roles in photodetectors, including FOM optimization and functional photodetection for polarization, angle, wavelength and OAM detection. We elucidate the interaction mechanism between metasurfaces and photodetectors, while summarizing the latest research progress in this burgeoning field.

2. Key figure of merits

The performance of photodetectors is typically evaluated based on several key FOMs, including responsivity, response speed, and noise level, as outlined in Table 1. Responsivity R, response time τ, and detectivity D* are the most commonly employed metrics.

Table 1.

Figures of merit.

Parameter	Expression	Unit	Definition
Responsivity	$R=\frac{R_{ph}}{P}$	A/W or V/W	Ratio of the photoelectric signal $R_{ph}$ (photovoltage or photocurrent) to incident light power $P$
External quantum efficiency	$EQE=\frac{N_e}{N_{ph}}=\frac{hc{I}_{ph}}{e\lambda P}$	/	Number of photogenerated electrons $N_e$ divided by the number of incident photons $N_{ph}$ per unit time. Here, h, c, and e represent the Planck constant, speed of light in vacuum, and elementary charge, respectively. $I_{ph}$ and $\lambda$ signify the photocurrent and the wavelength of light.
Response time	${\tau }_r/{\tau }_f$	s	The rise/fall time is defined as the duration for the photocurrent to increase/decrease when the light is turned on/off, typically measured when the photocurrent reaches 90%/10% of its maximum value.
Cut-off frequency	$f_C$	Hz	Modulation frequency of the optical signal at which the responsivity decreases to −3 dB.
Noise equivalent power	NEP = $\frac{i_N}{R}$	WHz−1/2	The minimum detectable light power of a detector, that is, the incident power when the signal-to-noise ratio is equal to 1. $i_N$ is the noise current spectrum at 1 Hz bandwidth.
Specific detectivity	$D^{*}=\frac{\sqrt{AB}}{i_N}R$	Jones	Specific detectivity represents the sensitivity of a photodetector. Here, A stands for the effective area of the detector, and B corresponds to the bandwidth.
Polarization ratio	$PR=\frac{R_{max}}{R_{min}}$		The PR characterizes the polarization sensitivity of a device. $R_{max}$/$R_{min}$ represents the ratio of the maximum and minimum responsivities of the device at different polarization states.
Dissymmetry factor	$g=2\frac{R_{LCP}-R_{RCP}}{R_{LCP}+R_{RCP}}$		$g$ characterizes the discrimination of circularly polarized light by photodetectors. In some articles, half of the dissymmetry factor is referred to as the discrimination ratio.

The integration of metasurfaces with photodetectors primarily influences the device responsivity by manipulating absorption. This, in turn, optimizes FOMs related to R, such as external quantum efficiency, noise equivalent power, and specific detectivity. Moreover, the varying metasurface response to light fields of different wavelengths, polarizations, incident angles, and other parameters imparts a dependence on this information to the photodetector, affecting its operating band and polarization resolution capabilities (including polarization ratio PR and dissymmetry factor g). In addition, some studies have reported the impact of metasurfaces on parameters related to response time [29], and their working mechanisms require further investigation.

3. Photodetection mechanisms

The detection mechanism of photodetectors is usually divided into two types, photonic photodetection and thermal photodetection, depending on whether the photon-generated carriers thermalize their energy in the photoelectric conversion process. We delve deeper into the specific detection mechanisms of each type below.

3.1. Photonic photodetection

In photonic photodetection, the optical response relies on the collection of photoexcited carriers. Traditional photocurrent mechanisms achieve carrier separation via a built-in electric field or bias voltage, including the photovoltaic (PV) effect, photoconductivity (PC) effect, and photogating (PG) effect. Recently, photocurrent generation mechanisms such as the bulk photovoltaic effect (BPVE), which utilizes the symmetry breaking of materials, have garnered significant attention and are being extensively investigated due to their unique advantages and functionalities compared to conventional approaches.

3.1.1. Photovoltaic effect

In the PV effect, electron-hole (e–h) pairs excited by the light field in the material are separated by the built-in electric field of a p–n or Schottky junction, forming a photocurrent. This mechanism can operate at zero bias, making it ideal for commercial photodetectors requiring low dark current and high switching ratios. A noteworthy subcategory of photodetectors based on the photovoltaic effect is the avalanche photodiode (APD). APDs achieve ultra-high sensitivity detection through carrier multiplication, enabling them to perform ultra-low power detection. They typically operate under large reverse bias voltages. The photogenerated carriers, under the combined action of the built-in electric field and the external bias voltage, impact the crystal lattice, leading to the generation of additional electron-hole (e-h) pairs, resulting in a significant gain in photocurrent.

3.1.2. Photoconductivity effect

The PC effect is characterized by photogenerated excess carriers augmenting the concentration of free carriers within semiconductors, thereby increasing conductivity. The applied bias voltage facilitates the separation of these photoinduced carriers, ultimately generating a photocurrent. The magnitude of the photocurrent is obtained by subtracting the dark current from the total current. This underscores the dependence of PC-induced photocurrent on a bias voltage, representing a significant distinction from the PV effect.

3.1.3. Photogating effect

The PG effect can be considered a particular case of the PC effect. In this scenario, the channel conductance is further modulated by the photoinduced trapped carriers. When e-h pairs are excited by the light field, one type of carrier becomes trapped, forming a localized field that gates the channel materials. The other carrier type generates the photocurrent. This effect can be achieved through defect states generated by impurities or defects or based on heterogeneous integrated structures. In hybrid systems such as semiconductor–quantum dots or 2D material heterojunctions, e-h pairs are excited by the optical field and separated by the built-in electric field. One carrier type becomes entrapped in the trapping layer, while the other enters the channel to generate the photocurrent. This spatial separation substantially reduces carrier recombination rates and improves carrier lifetimes. Therefore, photodetectors based on the PG effect can achieve high responsivity at the expense of response speed, necessitating a balance between these parameters in practical applications.

3.1.4. Plasmon-induced hot carrier injection

The energy of photons absorbed by semiconductors is limited by their bandgap, a phenomenon known as bandgap limitation, which presents challenges for detecting long-wavelength light. Hot carriers, generated by the non-radiative decay of surface plasmons, offer an alternative means of photodetection that circumvents this limitation. Nonradiative plasmon relaxation and subsequent e-e scattering can generate numerous non-equilibrium hot electrons; this occurs within one ps [37,38], much faster than the thermalization process. These hot electrons transport to the metal-semiconductor interface through e-e and electron-phonon scattering. However, this process is accompanied by hot electron relaxation, leading to significant transport losses. Only electrons that reach the interface with sufficient energy before thermalization can overcome the Schottky barrier and contribute to the photocurrent. The hot-electron injection photocurrent is limited by the Schottky barrier height rather than the semiconductor bandgap. It is particularly well-suited for detecting low-energy photons, such as in the near-infrared bands. Additionally, the thermal electron injection mechanism offers an exceptionally fast response speed. However, owing to transport loss and carrier injection efficiency limitations, the quantum efficiency of thermionic photodetectors is typically low, on the order of 10⁻⁴ [38].

3.1.5. Bulk photovoltaic effect

In non-centrosymmetric materials with a single component, photocurrents are generated without external fields and spatial inhomogeneities through exposure to uniform light. This phenomenon is referred to as the BPVE [39]. Recent research has reexamined the BPVE in various ferroelectric materials owing to its potential to surpass the Shockley–Queisser limit and enhance the efficiency of optoelectronic devices. BPVE was also reported to play an important role in infrared detection based on Weyl semimetal [40], [41], [42], [43].

The photocurrent generated by the BPVE consists of two components: shift current $J_{sh}$ and ballistic current $J_b$. Shift current $J_{sh}$ arises from the real-space shift R following carrier band-band transitions, while ballistic current $J_b$ results from the asymmetric distribution of the momenta of excited nonthermalized carriers [39].

3.1.6. Circular photogalvanic effect (CPGE)

The CPGE involves the generation of photocurrent where its direction depends on the helicity of the incident light field. The CPGE has multiple mechanisms of origin. It was initially discovered and studied in gyrotropic media. Owing to spin-orbit interactions in an asymmetric potential field, a perturbation term proportional to the electron wave vector and the Pauli matrix arises in the Hamiltonian of the system, leading to the degeneration of energy bands with different spins. Different spin optical fields excite non-equilibrium electron distributions in k-space, resulting in helicity-dependent photocurrent [44,45].

The CPGE can also occur in semiconductor materials when a similar Hamiltonian is generated. For instance, in silicon nanowires (NWs) with specially designed orientations, the CPGE can be realized under the influence of the built-in electric field of a Schottky contact [46].

3.1.7. Orbital photogalvanic effect (OPGE)

In addition to carrying spin angular momentum (SAM), light also possesses OAM. The interaction between the OAM of light and atomic media can create new selection rules and optical responses. Research has demonstrated that the OAM of light can induce a robust, nonlocal interaction between electromagnetic waves and matter, resulting in a photocurrent that carries the phase information of the OAM of light. This phenomenon is referred to as the OPGE [47,48].

3.2. Thermal photodetection

Thermal photodetection encompasses various effects, including the photothermoelectric effect (PTE), the pyroelectric effect, and the bolometric effect [49]. Compared to photon-type detectors, thermal detectors boast a significantly broader detection bandwidth. However, their response time is relatively long due to phonon-dominated transmission. The primary noise source in thermal detectors is Johnson–Nyquist noise, with 1/f noise additionally introduced depending on whether external bias is required.

3.2.1. Photothermoelectric effect (PTE)

PTE detectors harness the Seebeck effect. In the presence of a temperature differential across two points in a conducting material, an electric potential (emf) arises between those points. This emergent potential, the Seebeck potential, is characterized by the ratio between the emf and temperature difference, denoted as the Seebeck coefficient. In PTE detectors, photons are absorbed to create a temperature difference, which can be significantly enhanced through metasurface. This temperature gradient drives charge carriers to diffuse in a direction opposite the temperature gradient, thereby establishing a potential difference [50]. Since PTE detectors operate without external bias, the 1/f noise is negligible, and the dominant noise is Johnson–Nyquist noise [51].

3.2.2. Pyroelectric effect

The inherent electric dipole moment, arising from the asymmetric charge distribution, is sensitive to temperature fluctuations within pyroelectric crystals. Consequently, heating or cooling these crystals induces a temporary voltage on their surface, a phenomenon known as the pyroelectric effect. This unique property renders pyroelectric materials particularly suitable for radiation detection, especially in the infrared and millimeter-wavelength regimes.

3.2.3. Bolometric effect

A bolometer is a device for measuring radiant heat by utilizing a material whose electrical resistance exhibits high sensitivity to temperature changes. It typically consists of an absorptive element with highly temperature-dependent resistance and a thermal reservoir. Measurement of the impedance change of the element, for example, through a bridge circuit, provides information on the temperature and the incident optical power. The necessity of an external bias inevitably introduces additional 1/f noise [52].

4. Metasurface-enhanced performance of photodetectors

Recent years have witnessed remarkable progress in the field of metasurface-assisted photodetectors. Leveraging the unparalleled design flexibility afforded by metasurfaces, researchers have implemented diverse strategies to enhance their responsivity significantly. Notably, the design of metasurfaces supporting bound states in the continuum (BIC) modes or multiple resonant modes enables the achievement of ultra-narrowband and ultrawideband optoelectronic responses. Importantly, unlike the PG effect, introducing metasurfaces for responsivity enhancement does not sacrifice response speed; as elaborated in Section 3.2, it can even lead to improved response times. Beyond the substantial responsivity gains, metasurface integration facilitates the tunability of response bands/regions and enables multidimensional detection, which will be further discussed in Section 4.

4.1. Metasurface-enhanced photoresponsivity

The development of novel 2D materials [13,[53], [54], [55] and semiconductors [56], [57], [58] has enabled the realization of ultracompact photodetectors boasting numerous advantages, including high speed [59], broadband response [60], and cost-effectiveness [61]. However, a fundamental trade-off exists between device miniaturization and achieving high responsivity. Ultrathin materials tend to exhibit reduced light absorption, resulting in diminished efficiency. Metasurfaces, with their ability to concentrate and manipulate light efficiently, present a promising solution for achieving high photo-responsiveness. By exploiting plasmonic or Mie resonances, metasurfaces generate localized field enhancements that significantly boost the absorption of optoelectronic materials positioned on or within their structure. Consequently, integrating metasurfaces into novel photodetector designs can dramatically increase their photogenerated carrier density and responsivity, paving the way for transformative commercial applications [1].

4.1.1. Plasmonic-metasurface-enhanced photoresponsivity

Plasmonic metasurfaces primarily enhance photodetector responsivity through absorption enhancement, temperature increase, and hot-carrier injection. Intuitively, SPPs can confine the optical field on a sub-wavelength scale, resulting in significant near-field enhancement and improved optical absorption, as depicted in Fig. 2a. In photodetectors relying on the PC [62] or PV effect [29], this absorption enhancement can play a decisive role in boosting responsivity. Chen et al., for instance, demonstrated that triangular gold disks fabricated using microsphere lithography enhance the absorption ability of graphene by approximately tenfold [63]. Furthermore, gold nanotriangles and silver nanodisks were reported to lead to 12- and 7.2-fold enhancements in photocurrent for InSe-based [29] and monolayer MoS₂-based [36] photodetectors, respectively. Beyond near-field enhancement, metal metasurfaces can also augment absorption by extending the light-matter interaction distance. Xu et al. utilized chirped metal gratings to deflect incident light into the optically active layer by a large angle, enhancing the device's light absorption. This approach achieved significant gains in responsivity ranging from 1.5x to 2x between 560 and 690 nm [64]. Furthermore, Chen et al. constructed a waveguide in the absorption layer by the doping-induced refractive index change. Through the diffraction of the Au pillar array, vertically incident light was coupled into the waveguide, achieving an EQE of 22% in the 2 µm mid-infrared band.

Fig. 2.

Metasurface-enhanced photodetection. (a) Schematic of SPP-enhanced absorption. Local field intensity is maximized at the tip, enhancing optical absorption. (b) Illustration of SPP-induced temperature rising for photodetection. (c) Nonradiative relaxation of plasmons produces hot electrons, which migrate to the Schottky junction and undergo carrier separation under the built-in electric field to create a photocurrent. (d) Responsivity enhancement induced by absorption enhancement in the PCE-based MoS₂ device [36]. The left figure depicts the device schematic and resonant mode simulation, while the right shows detector responsivity at different wavelengths before and after metasurface processing. (e) Responsivity enhancement arising from metal structure absorption and Seebeck coefficient tuning [28]. The left and right figures display the device structure and photocurrent mapping. (f) Photocurrent enhancement through plasmonic hot-electron injection. The left panel illustrates the device structure and photocurrent under different biases, while the right shows a schematic of the plasmon hot-electron injection mechanism [76]. (g) Field enhancement across a broad spectral range from 476 nm to 647 nm using snowflake-type fractal metasurfaces [68]. (h) Utilization of gold nanorod arrays with small gaps to achieve a substantial increase in mid-infrared responsivity. The device structure, mode simulation, and photovoltage enhancement are demonstrated [62]. (i) Enhancement of photocurrents with perovskite metasurfaces [80]. (j) Graphene plasmonic metasurfaces for mid-infrared high-responsivity photodetection. The device structure and mechanism are depicted [25].

Despite plasmonic metasurfaces significantly enhancing localized fields, the intrinsic loss associated with metal structures limits further absorption enhancement. However, in devices based on the PTE, where a light-induced temperature gradient drives the photocurrent, the metal absorption-induced temperature rise can promote device responsivity [65]. In these devices, the photocurrent arises from the Seebeck effect rather than photogenerated carriers, allowing the absorption of the metal structure to amplify this photocurrent (Fig. 2b). Echtermeyer et al., for example, achieved a 20-fold enhancement in photoresponse at 514 nm using plasmonic strips located on one side of the electrode. Under a 0 V bias voltage, the photoelectric responsivity reached approximately 5 V/W [66]. Gold nanoparticles were also reported to enhance the photocurrent by 1500% [67]. Furthermore, the Fermi level and corresponding Seebeck coefficient of 2D materials can be tuned through gate voltage. This principle allows for further enhancement of the PTE current by applying different gate voltages to various regions of the device, resulting in significant variations in Seebeck coefficients. Utilizing Ti/Au metasurfaces for temperature enhancement and adjusting the Seebeck coefficient via back gates, a 25-fold enhancement was achieved, with a responsivity reaching 52 µA/W without a bias voltage [28], as illustrated in Fig. 2e.

Intricately designed complex plasmonic metasurfaces can exhibit superior performance, expanding the bandwidth of absorption enhancement and amplifying its impact. Typically, a significant improvement in the responsivity of photodetectors can only be achieved at the plasmonic resonance frequency position. This presents a double-edged sword: it allows for precise response band tuning by adjusting the geometric dimensions of the metasurface, bypassing the need for bandgap material selection or doping operations. However, these resonance peaks limit broadband response capabilities. To overcome this limitation, fractal structures, characterized by self-similar branches with different geometric dimensions, can support multiple resonant modes and achieve enhanced absorption over a wide frequency range. This concept was exemplified by Fang et al., who fabricated a gold metasurface with a snowflake fractal structure on chemical vapor deposition-grown graphene. The meticulously chosen geometric parameters yielded 8–13-fold improvements in responsivity across the entire 476–647 nm band [68]. The device structure, underlying mechanism, and corresponding enhancements are presented in Fig. 2g. Subsequently, Francesco et al. employed a 5th-order gold Sierpinski carpet fractal metasurface in graphene photodetectors and achieved a maximum 100-fold enhancement at a 1.46 eV photon energy [69]. The device exhibited good responsivity enhancement between 522 nm (2.38 eV) and 1550 nm (0.8 eV), with a peak specific detectivity of 10¹¹ Jones. This broadband, polarization-insensitive response enhancement scheme based on fractal structures can be readily adapted to other photodetectors utilizing the PTE or PV effect. Additionally, chirped arrays integrated with structures of different geometric dimensions and resonant frequencies can engage in photodetection, thereby realizing broadband enhancement [70].

Local field enhancement significantly contributes to high responsivity. Gap modes in plasmonic structures, also known as hot spots, enable tiny mode volumes and ultrahigh local field enhancement, widely used in applications such as tip-enhanced Raman scattering. Incorporating gap modes into photodetector design further improves responsivity. For example, Wu et al. separated arrays of gold disks and SiO₂-coated Ag nanoparticles with 5-nm-thick hBN [71]. This configuration facilitated a gap mode between the silver nanospheres and gold nanodisks, resulting in a field enhancement of 1.6 × 10⁵. Placing a single layer of MoS₂ at the gap to generate photocurrent based on the photoconductive effect achieved an 8.8-fold photocurrent enhancement and a high responsivity of 287.5 A W ^{− 1}. Another classical structure for generating plasmonic hotspots, the bowtie, has also been employed in photodetection, resulting in a 70% improvement in responsivity on a black phosphorus device [72]. Gold nanoribbons fabricated between electrodes can shorten the transit distance and time of carriers, allowing more photogenerated carriers to be collected before recombination. This strategy is particularly suitable for materials with short carrier recombination times. Meanwhile, the gap modes between nanoribbons can greatly enhance the light absorption. Yu et al. employed this structure to demonstrate graphene detectors with more than 200-fold responsivity enhancement at a 4.45 µm wavelength, as illustrated in Fig. 2h [62]. Subsequently, a similar strategy was applied to gold comb array–assisted graphene detectors [73]. Over a broad range of wavelengths from 800 to 20 µm, a considerable improvement in responsivity was achieved (0.6 A/W at 800 nm and 11.5 A/W at 20 µm). BICs, which coexist with a continuous spectrum radiating energy, have gained considerable attention from researchers owing to their counterintuitive physical mechanisms and extremely high-quality factors. A recent theoretical study entailed efforts to combine quasi-BICs with photodetection. Wang et al. realized a symmetry-protected BIC mode using a 1D Ag ridge array covered by TiO₂. By breaking the symmetry through skewed incident angles, quasi-BIC modes with perfect absorption properties were constructed, and simulations revealed improvements of at least an order of magnitude in responsivity [74].

Finally, hot electrons induced by the nonradiative relaxation of SPPs are also considered one of the enhancement mechanisms [75] (refer to Fig. 2c for the schematic). Hot-carrier-based photodetectors offer the advantage of broadband detection, as the bandgap does not restrict the photocurrent generation. Additionally, they exhibit rapid response speeds. Li [76] et al. achieved an eight-fold enhancement through SPP-induced ultra-hot electronics by fabricating a crescent-shaped plasmonic gold metasurface on a graphene–boron nitride heterojunction (Fig. 2f). The resulting photodetector exhibited a responsivity of 70 pA/mW and an EQE of 5 × 10⁻⁸ at 1550 nm, which, although relatively low compared to typical graphene devices, provides satisfactory detectivity considering the extremely low noise from the van der Waals heterojunction. In plasmonic-metasurface-based photodetectors, multiple mechanisms often work synchronously to achieve enhanced responsivity, as depicted in Fig. 2d, wherein plasmon-enhanced absorption and plasmon-induced hot carriers coexist to increase responsivity. Occasionally, plasmonic metasurfaces are combined with other design strategies, such as the PG effect [63], optical waveguides [77], heterojunction structures [78], and Fabry–Perot (F–P) cavities to enhance the performance of photodetectors further.

It is also worth noting that plasmonic modes in other materials can be harnessed for photodetectors. For instance, Ni et al. achieved photodetection in the ultra-long wavelength range, from ultraviolet to mid-infrared, by decorating graphene photodetectors with B-doped silicon nanospheres [20]. These silicon microspheres induce SPP resonance, enhancing optical gain. Graphene, being a Dirac semimetal, can also support the transmission of SPP modes, which has been applied in mid-infrared photodetector research. Unlike metals commonly employed in plasmonic metasurfaces, graphene presents a versatile platform: a material that simultaneously supports propagating plasmons and exhibits photoelectric responsiveness. Freitag et al. patterned a 1D array of stripes on graphene to excite plasmons at a wavelength of 10.6 µm [79]. In this case, graphene plasmons coupled with surface polar phonons of the SiO₂ substrate to form plasmon–phonon quasiparticles, resulting in a six-fold increase in responsivity under a − 2 V bias. This enhanced response was attributed to improved absorption due to plasmon resonance and increased phonon temperature from plasmon–phonon quasiparticle relaxation. Furthermore, Guo et al. connected a graphene nanodisk-resonator array with nanoribbons, creating a high-responsivity mid-infrared photodetector, as depicted in Fig. 1j [25]. In this setup, graphene plasmon resonances were generated within the nanodisk cavity and subsequently relaxed into hot carriers. Additionally, defects were introduced in the nanoribbons during fabrication, generating disorder potential. This disorder potentially contributed to the temperature-dependent carrier transport modes, with the conductivity being sensitive to carrier temperature. Combining these mechanisms, the researchers achieved a responsivity of 16 mA/W at 12.2 µm [25].

In summary, plasmonic metasurfaces offer versatile ways to enhance the responsivity of photodetectors, including absorption enhancement and hot-carrier generation. These capabilities hold immense promise for advancing optoelectronic applications. By implementing innovative designs, including gap and fractal structures, near-field intensity and response bandwidth can be further enhanced, respectively. Plasmonic metasurfaces have thus become an indispensable tool for improving photodetector performance.

4.1.2. Dielectric-metasurface-enhanced photoresponsivity

While plasmonic metasurfaces demonstrably excel in both mode volume compression and localized field enhancement, all-dielectric metasurfaces offer distinct advantages. Notably, they exhibit significantly lower intrinsic losses, reducing light competition with optoelectronic materials. This characteristic renders them particularly attractive for photodetectors based on the PV effect, where metal absorption significantly hinders plasmonic structures. Furthermore, the fabrication processes involved in depositing metal nanostructures onto materials can introduce substantial technological challenges, including the potential for damaging materials during lithography and the risk of detachment between the metal nanostructures and the photoelectric material. In such cases, dielectric metasurfaces offer a compelling strategy to enhance photodetector responsivity.

Due to the Mie resonances of high-index materials, dielectric structures can achieve substantial local field enhancements, enhancing detector absorption and improving responsivity. For example, He et al. fabricated a metasurface on perovskite films using a focused ion beam (FIB). This metasurface exhibited electric dipole resonance at 560 nm and magnetic dipole resonance frequency at 700 nm [80]. This incorporation of gold mirrors resulted in approximately tenfold responsivity enhancement across the ultraviolet to visible bands, as illustrated in Fig. 2i. Similarly, a 40% photocurrent enhancement was achieved by etching rectangular metasurfaces on hybrid organic-inorganic perovskite film [81]. Another notable study involved the heterogeneous integration of In₂S₃ on the surface of Si nanopillars [82]. The multiple Mie resonances within the silicon micropillars enhanced the local field, enhancing absorption. Simultaneously, a vertical built-in electric field was introduced in the In₂S₃ channel to capture photogenerated holes, leading to a high photogain through the PG effect. Combining these two strategies achieved an exceptionally high responsivity of 4812 AW⁻¹ at 405 nm, representing an approximately 71-fold increase compared to the pristine device (Fig. 2j). Additionally, numerous perfectly absorbing dielectric metasurfaces have been theoretically demonstrated for photodetection applications [83].

4.2. Metasurface-shortened response time

Beyond the improvements in responsivity, plasmonic-metasurface-assisted photodetectors have exhibited notable reductions in response time. For instance, the incorporation of gold nanotriangles into an InSe photodetector resulted in a more than 50% reduction in both the rise and fall times compared to devices lacking this structure [29]. This reduction can be attributed to the interplay between plasmons and excitons. Nevertheless, further research is required to comprehensively elucidate the specific mechanisms underpinning this acceleration in response speed.

5. Metasurface-assisted functional photodetection

The rich information content of light fields, including intensity, wavelength, polarization, phase, and OAM, renders them an ideal medium for communication and sensing applications. However, commercially available detectors are typically limited to the detection of intensity information solely, necessitating the use of numerous bulky optical components for extracting other modes of information. This presents significant challenges for device miniaturization and integration. Fig. 3a illustrates various approaches to realizing functional photodetection on a single chip. With the advancement of metasurfaces, it has become possible to distinguish light with different characteristics at sub-wavelength scales. This breakthrough enables the realization of ultra-compact, high-performance, multifunctional photodetectors based on metasurfaces. These detectors can discern polarization state, incident angle, and OAM information on a single chip. Note that the main focus of this paper is on metasurface photodetectors. Metasurfaces can also serve as beam splitters based on polarization [84], wavelength [85], or OAM [86]. For further details, please refer to the previous review [87].

Fig. 3.

Basic strategies for functional photodetection and a typical linearly polarized photodetector. (a) Several approaches to realize functional photoelectric detection. (b) Ultra-broadband, high-responsivity polarization detection using nanostrip array electrodes. [73] (c) Combining the polarization-sensitive material black phosphorus with asymmetric bowtie metasurface for high-polarization-ratio detection. [72] (d) Polarization-sensitive photodetection enabled by dielectric asymmetric metasurfaces. [91] (e) Polarization response in graphene ribbon arrays supporting graphene plasmons. Differently polarized light excites e–h pairs and plasmon-phonon quasiparticles, affecting current negatively or positively (right panel). [79] (f) Mechanism of the artificial BPVE. From top to bottom, the figures illustrate the T-shaped structure geometry, Seebeck coefficient gradient, electric field intensity, magnitude and direction of electron driving force, and local photocurrent. [95] (g) Schematic of a linearly polarized light detector based on T-shaped structures. False-color SEM images of two-port and three-port devices. [34] (h) Polarized light detector that can realize arbitrary PRs. A single superlattice consists of two tilted tapered nanoantennas. By altering the tilt angle, the PR can be adjusted within the range of 1 → ∞/−∞ → −1. [92].

5.1. Polarization state detection

Extensive research efforts have explored the captivating potential of metasurfaces for polarization-sensitive detection. Researchers have harnessed the varying absorption, transmission, and near-field responses of micro-nano structures to different polarized light fields, leading to the development of photodetectors sensitive to both linear and circular polarization, exhibiting high polarization ratios. Additionally, on-chip measurements of the complete polarization state have been achieved through metasurfaces. This section will delve into these advancements in greater detail.

5.1.1. Linearly polarized light detection

Introducing asymmetry into a metasurface along orthogonal axes imparts linear polarization sensitivity to the photodetector. However, certain applications necessitate polarization-insensitive detection, requiring specific design strategies to eliminate this inherent dependence [88]. Nevertheless, the demand for high-performance, practical linear polarized photodetectors continues to grow.

The difference in the light absorption for different polarizations can be effectively converted into a difference in photocurrent [89], as indicated by path ② in Fig. 3a. A typical structure for distinguishing the polarization of transmitted light involves one-dimensional metal gratings, wherein light polarized perpendicular to the gratings passes through while parallel polarizations are reflected, as demonstrated in Fig. 3b. A PR of approximately three was achieved near the plasmonic resonance band by processing a metal strip array at the electrode position on one side [66,90]. Subsequently, Cakmakyapan et al. developed gold electrodes in nanostrip arrays, resulting in high responsivity across a broad spectrum from 800 nm to 20 µm. Owing to the structural asymmetry, a polarization ratio of ∼6 was observed in the 800 nm band [73]. Other asymmetric plasmonic metasurfaces, leveraging hot carriers induced by nano rectangular arrays [21] and plasmons induced by nanodisk dimers [75], have also exhibited linear-polarization-sensitive photocurrents. Combining asymmetric metasurfaces with polarization-sensitive optoelectronic materials can enhance linear polarization detection capabilities. Venuthurumilli et al. fabricated bowtie-structured gold metasurfaces on black phosphorus, orienting the long axis perpendicular to the armchair direction, as depicted in Fig. 3c. By harnessing the polarization-selective absorption of black phosphorus [13] and the polarization-selective transmission of metasurfaces, they achieved a polarization ratio of 8.7 [72]. Asymmetric all-dielectric metasurfaces have also been shown to induce polarization-selective absorption. Xue et al. [91] developed an elliptical hole structure supporting Mie resonance on the surface of Cd₃As_2, demonstrating a PR of ∼2.1 at 532 nm (Fig. 3d). In graphene metasurfaces, polarization-sensitive plasmon mode excitations have led to photocurrent polarization ratios of ∼15 [79]. Importantly, the novel physical mechanism has correlated the polarity of the photocurrent with the polarization of the excitation light at a specific gate voltage, facilitated by the polarization-dependent graphene plasmons, as illustrated in Fig. 3e.

In the above mentioned study, varying polarization orientations of incident fields excited distinct near-field modes, leading to differences in scalar responses such as absorptivity and transmittance. These variations manifest in the magnitude of the generated photocurrent. In 2020, Wei et al. introduced the concept of directly converting the near-field mode into a vector response, specifically current density, using metasurfaces (Path ③ in Fig. 3a), termed the artificial BPVE [34]. The underlying mechanism is depicted in Fig. 3f. T-shaped gold nanostructures were fabricated on graphene. The contact sites of the graphene were metal-doped to produce a Seebeck coefficient gradient near the boundaries of the nanostructure, as depicted in Fig. 3f. Additionally, plasmon resonance caused variations in light field intensity and temperature at different tips of the gold nanostructure, depending on the polarization characteristics of the light field.

Based on the PTE, the photocurrent resulted from the Seebeck coefficient gradient and the difference in the light field. Through meticulous symmetry analysis, polarization-sensitive net photocurrent could be generated near non-centrosymmetric nanostructures, exhibiting characteristics similar to the BPVE. Utilizing this mechanism, they reported a photodetector with a polarization ratio of −1 and a responsivity of 36.3 mA/W [34], where orthogonally polarized light excited photocurrents of equal magnitude but opposite direction, as shown in Fig. 3g. Unlike other methods where polarization direction is reflected in the magnitude of the photocurrent, the photocurrent arising from the artificial BPVE effect exhibited a vectorial response to polarization, enabling calibration-free detection of polarization angle. As demonstrated in the right panel of Fig. 3g, the readout from each port of the three-port device was proportional to light intensity. It represented a trigonometric function of relative polarization angle. Analyzing these readouts could determine the polarization angle without requiring intensity calibration. Leveraging the flexibility of metasurface design, the same research group recently achieved a photodetector whose polarization ratio could be adjusted arbitrarily from −∞ to −1 (and 1 to ∞) through a technique called “bipolar detection,” as illustrated in Fig. 3h [92]. By adjusting the angle between the polarization direction and the long axis of the antenna, the polarization dependence curve of the photocurrent of the tapered nanoantennas in the x-direction could be adjusted arbitrarily. When two tilted tapered antennas were placed within a superlattice, the PR could be adjusted between −∞ and −1 and from 1 to ∞ by varying the two tilt angles. Although this strategy could theoretically achieve any PR, practical realization entailed limitations in fabricating accuracy. By processing two tilted tapered antennas with different metals and achieving varying graphene doping, different gate-voltage-dependent Seebeck coefficient gradients could be achieved. This allowed for the regulation of PR through gate voltage adjustment. Using this approach, an ∞ responsivity was achieved at a gate voltage of −23 V, with a zero-bias photoresponsivity of 15.6 V/W [92]. Furthermore, the device exhibited favorable dark current and response speed. Table 2 summarizes the parameters of some typical linearly polarized light detectors.

Table 2.

Key parameters of linearly polarized light photodetectors.

reference	Structure	Material	Polarization Ratio	responsivity
[66]	Au nano strip	Graphene	∼3	5 V/W@514nm
[73]	Au nano strip	Graphene	∼6	11.5A/W@20 µm
[93]	Au nano slit	Si	∼25@830nm
[21]	Au nanobar	Si	∼10 @1500nm	∼8nA/mW@ 1300nm
[91]	Elliptical holes	Cd3As2	∼2.1@532nm	1 mA/W@532nm
[79]	Graphene nanostrip	Graphene	∼15@10.6µm	∼7µA/W@10.6µm
[34]	Au T-shaped Structure	Graphene	−1	36.3 mA/W@4µm
[92]	Tilted Tapered nanoantenna	Graphene	−∞→ −1	15.6 V/W@4 µm
[94]	Orthogonal zigzag structure	Graphene	−∞ → −1	3.6 V/W@5.3µm

5.1.2. Circularly polarized light detection

Detecting circularly polarized light (CPL) poses more significant challenges compared to linearly polarized light. Photodetectors with intrinsic CPL detection capabilities have been proposed using chiral materials such as organic semiconductors [96], organic–inorganic hybrid perovskites [97], and novel two-dimensional materials [44]. However, natural materials exhibit limited circular dichroism, resulting in unsatisfactory discrimination ratios of the photoresponse (see definition in Table 1). To address this limitation, artificial micro- and nanostructures have been specially designed in metasurfaces, demonstrating the ability to access circular dichroism far exceeding that of natural materials [23], making high-discrimination-ratio direct CPL detection achievable in integrated photodetectors.

In 2015, Li et al. pioneered the combination of zigzag chiral silver NWs with silicon to realize direct CPL detectors with a high PR in the communication band, as depicted in Fig. 4a [32]. This device capitalizes on the absorption difference of orthogonal CPLs based on the chiral silver metasurface. CPL excites plasmons on the silver NWs, leading to the generation of hot carriers. These hot carriers separate at the Ag–Si Schottky junction and produce a photocurrent. This mechanism underpins the device's impressive PR of 3.4, albeit with a relatively modest responsivity. Another report highlighted a circular-polarization-sensitive photocurrent based on hot-electron injection using a nano-crescent chiral gold structure [98]. Chiral plasmonic metasurfaces can selectively transmit or enhance specific CPL, enabling circular-polarization-sensitive absorption in non-chiral materials. For instance, circular dichroism was achieved by placing “n”-shaped Au structures on MoSe₂, leading to photoconductive-based CPL-sensitive photodetection with a PR of 1.47 (1.41) for the right-handed (left-handed) structure [99]. These “n”-shaped Au structures were also integrated with silicon to achieve full Stokes parameter detection with a CPL polarization ratio below 1.5 [35]. Notably, a metal film is usually coated beneath the chiral metal metasurface, separated from the metasurface by a semiconductor or an insulator. This configuration imparts intrinsic chirality to the device by breaking mirror symmetry in the z-direction and exciting plasmonic gap modes, enhancing the ability to distinguish CPL.

Fig. 4.

Metasurface-based circularly polarized light detector and Stokes parameter detector. (a) CPL photodetectors based on zigzag silver NWs. The device structure and hot-electron injection mechanism are shown [32]. (b) CPL photodetector based on dislocated Si NW dielectric metasurface. The device structure and the external quantum efficiency under different CPL illuminations are shown. The inset presents the resonance modes of the Si NWs under different CPL excitations [101]. (c) CPL detector with a ∞ near-field chiral response discrimination ratio through an achiral T-shaped structure. The device structure, localized photocurrent, and response under different polarized light excitations are shown. [95] (d) Achiral-responsive metasurfaces placed on topological insulators, enhancing the CPL detection capabilities of devices. The figures illustrate the device mechanism, SEM images, and polarization detection results [106]. (e) Full Stokes detector based on a silicon metasurface ¼-wave plate and gold nanograting polarizer. The device structure and measurement results are shown [107]. (f) Polarization state detectors are based on n-shaped gold structures with different orientations. [35]. (g) Linear and circular polarization detectors based on the bipolar detection method and corresponding polarization state detectors [94].

In all-dielectric metamaterials, where plasmon resonance is absent, achieving a satisfactory PR is challenging and requires innovative design strategies. Li et al. achieved a chirality-resolved photoelectric response by etching an l-shaped metasurface on a perovskite film [100]. Hong et al. designed a dislocated silicon NW metasurface consisting of NWs with localized displacements, as illustrated in Fig. 4b [101]. Periodic artificial dislocations were introduced in the NWs to serve as secondary localized sources, enabling the coupling of both incident TM and orthogonal TE polarizations to the same nonlocal guided mode of the NWs. By precisely designing the width, height, and dislocation width of the NWs, the TE- and TM-polarized light could be coupled into the waveguide mode with the same coupling strength and a phase difference of π/2. Consequently, the guide mode experienced constructive or destructive interference when illuminated by left- and right-handed light, resulting in a substantial absorption difference within the device. This innovative design achieved a PR of ∼2 at 780 nm, with an EQE of ∼2 × 10⁻⁴ [101].

While the far-field response of achiral structures, such as transmittance and reflectance, lacks CPL sensitivity, recent research has unveiled remarkable chiral optical responses within the near-field modes of achiral structures [102,103]. If this inherent chiral information within near-field modes can be converted into photocurrent, CPL detection based on achiral metasurfaces will become achievable. Lu et al. pioneered this approach using the pyroelectric effect [104]. They designed a V-like structure with two tilted rectangular bars separated by a gap. By adjusting the gap, they could tune the resonant frequencies of the two modes, thereby controlling the coupling strength and phase of these modes when exposed to an external field. With well-designed geometric parameters, the plasmonic modes on the two rectangular bars interfered constructively and destructively under orthogonal CPL illumination. Consequently, a temperature gradient and photocurrent were generated, with their direction determined by the helicity of the optical field. This setup achieved a discrimination ratio of ∞ at 7.9 µm, with a responsivity of 43 mV/W. Furthermore, by sequentially connecting four V-shaped structures in distinct orientations, it was possible to eliminate responses to linearly polarized light and non-polarized light, obtaining a photocurrent proportional to the Stokes S3 parameter. Near-field mode chirality can also be converted into a localized photocurrent for CPL detection through the artificial BPVE described in Section 4.1. As shown in Fig. 4c, a T-shaped Au structure was fabricated on graphene, with electrodes collecting current in the x-direction [95]. Leveraging the extremely high carrier mobility of graphene, net local currents could efficiently reach the electrodes. Mode analysis was performed, and the symmetric/antisymmetric modes in the T-shaped structure were carefully tuned to maximize the asymmetry of the local mode when excited by CPL. Further analysis demonstrated that any plasmonic structure with a single mirror symmetry plane could generate mirrored mode distributions under a pair of orthogonal CPL excitations, leading to opposite photocurrents perpendicular to the symmetry plane, as depicted in Fig. 4c. By shaping the device as a ring or l-shape, responses to linearly polarized and unpolarized light could be eliminated, resulting in an extinction ratio of 84 and 13.9 for linearly polarized and unpolarized light, respectively. The incorporation of graphene strips further reduced short-circuit current and improved responsiveness. Ultimately, CPL detection achieved a responsivity of 392 V/W and a ∞ discrimination ratio at a wavelength of 4 µm [95].

Furthermore, when combined with certain new materials and mechanisms, metasurfaces without inherent CPL responses can enhance the detection of CPL. For instance, the CPGE can be generated at the boundaries of topological insulator materials [105]. As shown in the upper panel of Fig. 4d, spin-momentum-locked surface-state electrons exist at the interface of topological insulators. These electrons can be promoted to the conduction band by circularly polarized photons with specific spins. Consequently, the k-space distribution of electrons in the boundary state becomes asymmetrical, generating a CPL-dependent photocurrent owing to the spin-momentum-locked electrons with opposite spins. However, this process competes with the PTE induced by asymmetric illumination and the photon drag effect within the bulk material, resulting in a relatively small and concealed CPL-dependent photocurrent, which poses challenges for detection. To address this, Sun et al. designed a curved square ring metasurface at the interface of topological insulators. The resonant structures effectively localized the incident field at the device surface, giving boundary currents a substantial advantage over bulk currents. This led to a nearly 11-fold enhancement of the CPL-dependent photocurrent and a discrimination ratio of 0.87 [106]. Table 3 summarizes the critical parameters of representative CPL photodetectors discussed in this review.

Table 3.

Key Parameters of CPL photodetectors.

Reference	Structure	Material	Bias	Polarization Ratio	Highest Responsivity
[101]	Dislocated silicon NW	Silicon	500 mV	2@780 nm
[32]	Zigzag Ag NW	Silicon	0 V	3.4@1340 nm	2.2 mA·W−1
[99]	n-shaped Au	MoSe2	1.5 V	1.47@790 nm	2.46 mA/W
[93]	Au apertures and spiral grooves	Si	0 V	1.13@830 nm
[104]	V-shaped Au	Au–Ni	0 V	−1@7.9 µm	43 mV·W−1
[95]	T-shaped Au	Graphene	0 V	−1@4 µm	392 V/W@Vg = 180 V
[106]	Square ring	Bi1.5Sb0.5Te1.8Se1.2	0 V	14.38@532 nm
[94]	Opposite zigzag Au	Graphene	0 V	−1@5.3 µm	95 mV/W

5.1.3. Full polarization state detection

Metasurface-assisted photodetectors have exhibited superiority in both circular and linear polarization detection, offering the promise of miniaturized, high-accuracy full-Stokes photodetection. Stokes parameters are values that describe the complete polarization state of electromagnetic waves, encompassing four parameters: S0, S1, S2, and S3, which respectively represent the intensity of the light field, the difference between the intensities of x- and y-polarized light, the difference between the intensities of +45°- and −45°-polarized light, and the difference between left- and right-handed light intensities.

Traditional polarimeters rely on the cumbersome combination of polarizers and quarter-wave plates to detect linearly polarized light and CPL, ultimately determining the polarization state of incident light. This approach necessitates the utilization of numerous bulk optical components and demands precise alignment between each element. Consequently, researchers are actively exploring the potential of metasurfaces for on-chip detection of Stokes parameters, replacing their bulk counterparts. For example, Basirid et al. employed a Si metasurface as a quarter-wave plate and a metal nanogrid as a polarizer to achieve full Stokes parameter measurement, as illustrated in Fig. 4e [107]. Within the metasurface, resonance modes are primarily distributed in the Si cube or gap, depending on the polarization direction of the light field. The refractive index difference results in a polarization-dependent phase difference, effectively functioning as a quarter-wave plate metasurface. This composite structure achieved high-PR polarization detection within the 1.4–1.55 µm communication band. The measurement accuracy for the parameters S1, S2, and S3 reached approximately 1.9%, 2.7%, and 7.2%. Although this device performed well, its fabrication posed significant challenges. Stokes parameter detection can also be accomplished using the polarization-dependent absorption/transmission of metasurfaces.

Afshinmanesh et al. pioneeringly proposed a metasurface-based approach for directly detecting full polarization states. They used gold nanoslits for linear polarization detection, and Archimedes spiral grooves along with nanoholes for circular polarization detection [93]. Subsequently, Li et al. adopted chiral n-shaped nanostructures with different orientations to achieve full Stokes parameter detection within the light field [35]. As depicted in Fig. 4f, this photodetector featured four ports capable of producing photocurrents with distinct dependences on the elliptic angle and polarization angle of the incident light field. Well-designed algorithms were employed to deduce the polarization state from the readout current, enabling the retrieval of all polarization parameters of the light field within an acceptable range of error. With post-data processing, the polarization ratio of an individual device for linearly polarized (circularly polarized) light was no longer the primary determinant in the Stokes parameter measurements. Nevertheless, enhancing the single-device resolution could undeniably improve the accuracy of polarization state measurements. By leveraging the “bipolar detection” method described earlier, Dai et al. achieved linear and circular polarization detectors with a polarization ratio of −1. They conducted Stokes parameter detection accordingly, as illustrated in Fig. 4g [94]. Chiral zigzag gold supramolecules were fabricated on graphene, capable of being excited by specific linearly polarized or circularly polarized light, leading to plasmon resonance and enhanced absorption. Supramolecules with opposite chirality or orthogonal orientation were applied to both sides of the photodetector. When subjected to specific circularly polarized (linearly polarized) light, absorption enhancement, and temperature rise occurred near one electrode. Photocurrents were subsequently generated through the PTE. When the metasurfaces on both sides were equal in area and exhibited opposite chirality (orthogonal orientation), the device produced photocurrents with equal magnitudes and opposite directions under light fields with opposite spins (vertical polarization). Modifying the metasurface area ratio allowed for the tuning of the polarization ratio from −1 to −∞. This structure facilitated the realization of a four-port full polarization detector, as displayed in the right panel of Fig. 4g. The average measurement errors for S1, S2, and S3 were 14.2%, 15.2%, and 5.4%, respectively.

Integrating liquid crystals into metasurfaces enables dynamic control of their polarization properties via an external voltage, effectively transforming a single device into a versatile tool for measuring the polarization state of the incident light field. Ni et al. [108] proposed a computational spectropolarimetry based on silicon grating embedded within liquid crystals. By collecting the response of a single photodetector at different liquid crystal rotation angles, the stokes parameters and spectrum of the incident light are reconstructed by a trust-region method based algorithm. This approach boasts an impressive polarization reconstruction error of less than 5°, as defined by the spatial separation on the Poincaré sphere.

5.2. Wavelength-resolved photodetectors

Scaling the size of structures allows for tuning the resonant peak position of plasmon or Mie resonance, offering flexibility in adjusting the operating frequency of photodetectors. This property makes metasurfaces adaptable for detecting multi-color light. For instance, silicon nanopillar arrays supporting sharp Mie resonances have been employed to fabricate color-resolving metasurfaces [109]. By altering the nanopillar radius, the resonant peak position of the device can be continuously adjusted, facilitating color (wavelength) detection. Furthermore, greater flexibility in band tuning can be achieved by utilizing graphene plasmons, whose resonant frequency can be adjusted not only by the geometry parameters but also by the gate voltage, as explored in previous simulation studies [110].

Plasmonic metasurfaces, relying on hot carrier injection, can broaden the response bandwidth of photodetectors [38]. Hot carriers generated during plasmon relaxation allow photons with energies below the bandgap to be absorbed, generating photocurrent as long as the photon energy exceeds the Schottky barrier. This extends the working band of the photodetector. For example, Wang et al. achieved 5.2 A/W high-responsivity detection at a 1070 nm wavelength using resonant NWs covered by bilayer MoS₂, which possesses an indirect bandgap of 1.65 eV, significantly higher than the working band [30]. Additionally, as the metal structure serves the dual role of plasmon resonance and carrier generation, hot-carrier-based devices offer advantages in wavelength tuning and multi-color detection. In Fig. 5a, intense plasmonic modes are excited in the near-infrared region with periodically arranged gold nanoslits, generating photocurrents through hot-electron injection at the Ti–Si interface [33]. By varying the slit period, tuning of the operating wavelength from 1295 nm to 1635 nm was achieved.

Fig. 5.

Metasurfaces for wavelength, incident angle, and OAM detection. (a) Plasmonic thermionic photodetectors based on gold nanoslits. The operational band of the device can be tuned by varying the period of the nanoslits [33]. (b) Color-resolved photodetection using silicon nanopillars. Multi-color imaging is realized through silicon nanopillar array p–i–n photodetectors with different radii [3]. (c) Incident angle detection based on coupled silicon NWs. The figure illustrates the light field distribution in coupled silicon NWs under different incident angles and presents the angular sensing results [112]. (d) Angle-resolved photodetection based on plasmons. Each superlattice consists of a coupler, reflector, and slits, allowing light to pass through only in specific directions [31]. (e) Photon angular momentum resolved structures based on plasmonic nanorings and concentric grooves [114]. (f) SAM and OAM resolution metasurface based on a plasmon grating and fishbone structure [116]. (g) Metasurface OAM beamsplitter [113].

Another wavelength-resolved photodetection approach involves fabricating optoelectronic materials into metasurfaces that support specific resonant wavelengths. As shown in Fig. 5b, Si-based p–i–n photodetectors were fabricated with nanopillar structures, enabling the tuning of the Mie resonant frequency via modification of the pillar diameter [3]. Integration of photodetectors with varying pillar diameters on a single chip facilitated color imaging, as illustrated in the right panel. Similarly, Li et al. achieved color-resolved photodetection using NWs of various sizes [12]. By designing the coupling between NWs, they eliminated backscattering, realizing transparent photodetectors suitable for applications such as augmented reality. Furthermore, NW photodetectors of different sizes exhibit distinct wavelength selectivities. This allows for reconstructing the incident spectrum by analyzing the photocurrent collected by individual detectors, paving the way for developing ultra-small spectrometers [111]. Leveraging the unique capabilities of active metasurfaces, whose transmission spectrum can be modulated by an applied bias voltage, the incident spectrum can be reconstructed through multiple measurements with a single pixel. Furthermore, this advancement has been combined with a metasurface lens, enabling the realization of a high-precision angle-resolved spectrometer with a footprint of only 4 × 4 µm² [4].

5.3. Photodetection of incident angle

Precise electrical detection of the incident angle (in-plane momentum) of the incoming light field can be accomplished through intricately designed artificial nanostructures. Yi et al. discovered that the light intensity distribution in two mutually coupled NWs is influenced by the incident angle, as depicted in Fig. 5c [112]. These insights led to calibration-free, accurate angle detection by assessing the light intensity ratio in two NWs via two pairs of electrodes.

When phase-matching conditions are met, the selective excitation of SPP modes offers a pathway for SPP-based angular detection. As illustrated in Fig. 5d, each metasurface unit comprises three components. The central region features a gold nanograting for phase compensation and angle selection. Asymmetrically arranged on both sides of the superlattice are nanoslits that penetrate the underlying gold film and reflectors designed to couple light into free space, ensuring that only light from one side couples into the Ge detector underneath the gold film [31]. This arrangement enables angle-resolved photodetection.

5.4. Photodetection of oam

Metasurfaces find extensive applications in splitting optical beams with OAM. By introducing a geometric phase through dielectric blocks, light can be diffracted in various directions corresponding to the OAM, as depicted in Fig. 5g [113]. Plasmonic metasurfaces are also employed for OAM beam splitting. Ring-shaped or Archimedean spiral grooves can convert the incident light into SPPs carrying angular momentum determined by the structure, OAM, and SAM of the light. These distinct plasmons are sorted through annular holes that support specific angular momentum eigenmodes, resulting in maximum transmittance. Ren et al. achieved three-dimensional wave multiplexing involving SAM, OAM, and wavelength using this method, as illustrated in Fig. 5e [114]. Subsequently, Zhang et al. demonstrated that light beams with different OAMs can be decomposed into SPPs propagating in different directions using two aligned one-dimensional gratings [115]. Combining this approach with the fishbone structure, which can distinguish SAM from the gratings mentioned above, a metasurface that can distinguish both SAM and OAM was realized [116] (Fig. 5f). Despite extensive research into metasurface-assisted OAM beam splitters, the development of direct OAM photodetectors remains limited. Ji et al. discovered the OPGE photocurrent generation mechanism, which is directly determined by the OAM and can be used for OAM detection [47]. However, metasurface-based OAM photodetector research remains unexplored, necessitating further investigation.

Research on metasurfaces for functional photodetection is rapidly advancing. Nevertheless, numerous challenges remain to be addressed. Notably, while satisfactory performance has been achieved in detecting linear and circular polarization, these approaches often rely on the dominance of polarization-sensitive photocurrents over polarization-insensitive photocurrents, the latter of which can readily arise from the PTE. Therefore, continuously exploring methods that effectively eliminate polarization-insensitive photocurrent is crucial for further enhancing detection accuracy. Furthermore, many polarization photodetection strategies may be challenging to extend to other spectral bands by merely adjusting geometric parameters. Achieving polarization detection across an ultrabroad spectral range, from ultraviolet to mid-infrared, necessitates further exploration. Beyond polarization, the direct detection of other optical information is still in infancy. Angle-resolved photodetection based on metasurfaces has been realized, albeit with limited responsivity owing to the coupling efficiency of plasmonic gratings. While metasurface-based OAM beam splitters have been widely reported, developing photodetectors for direct OAM information detection remains a largely unexplored frontier with significant potential. As functional photodetectors continue to evolve and improve, we may eventually achieve comprehensive light field information collection on a single pixel. This capability far surpasses the capabilities of existing photodetectors and may open a new era of intelligent vision.

6. Conclusion and outlooks

This article comprehensively surveys the remarkable strides in metasurface-assisted photodetectors, encompassing both performance enhancements and the burgeoning field of multifunctional photodetection. By manipulating light-matter interactions at the nanoscale, metasurfaces offer a powerful toolkit for optimizing key photodetector FOMs such as responsivity, EQE, response time, and bandwidth. These advancements are achieved through diverse mechanisms, including enhanced light absorption, amplified thermal gradients, shortened carrier transit times, and efficient hot electron injection. Moreover, the inherent sensitivity of metasurfaces to diverse optical parameters, such as wavelength, polarization, incident angle, and OAM, opens up exciting avenues for enriching the functionalities of photodetectors. By illuminating the significant progress achieved in this burgeoning area, the review concludes by outlining several key challenges and future research directions.

Several compelling avenues emerge for the continued advancement of metasurface-enhanced photodetection. Firstly, all-dielectric metasurfaces, while still in their nascent stages compared to their plasmonic counterparts, offer promising applications in PC- or PV-based devices and circumvent challenges associated with metal integration. Delving deeper into the potential of dielectric metasurfaces for photodetection is an avenue for further exploration. Secondly, integrating metasurfaces with novel structures and materials, such as waveguides and 2D material heterojunctions, holds immense promise for realizing photodetectors with exceptional responsivity, ultrafast response times, and reduced noise levels. Such multi-pronged design approaches can potentially accelerate the commercialization of novel photodetector technologies significantly. Thirdly, current research predominantly focuses on passive metasurfaces in the context of metasurface-assisted photodetection. Using active metasurfaces based on liquid crystals paves the way for achieving more complex, multidimensional, and multifunctional photodetection capabilities within a single pixel. Finally, the landscape of potential applications for metasurface-assisted photodetectors remains vast and largely unexplored. This includes high-speed, on-chip optical communication applications, integrated spectrometers and polarimeters, room-temperature single-photon detection, and ultra-high-sensitivity night vision devices [117], [118], [119], [120]. By combining metasurfaces with photodetectors, a transformative revolution in photoelectric conversion applications awaits to be unleashed.

Declaration of competing interest

The authors declare that they have no conflicts of interest in this work.

Biographies

Guanyu Zhang received his bachelor's degree from Tianjin University in 2021. He is now pursuing a Ph.D. degree in School of Physics, Peking university under the supervision of Prof. Guowei Lu. His research interests focus on functional photodetectors based on novel mechanisms and nano-structures.

Qinsheng Wang (BRID: 06066.00.16807) received his bachelor's degree from University of Science and Technology of China in 2010, and then obtained Ph.D. degree from ICQM, Peking University in 2016. He is currently an assistant professor at School of Physics, Beijing Institute of Technology. His research interests focus on the carrier dynamics and photocurrent response of 2d materials and topological materials.

Guowei Lu is interested in the interaction between light and matter at the nanoscale, as well as its applications. His research focuses on exploring the optical properties of nano-materials and nanostructures, with a particular emphasis on investigating novel phenomena and applications at both the single-nanostructure and single-molecule levels. This includes utilizing scanning probe-based nanomanipulation techniques for in-situ observation of nanoscale light emission direction regulation, surface enhancement spectroscopy, and designing nanostructures for photodetectors.