1. Introduction

Graphene, sp²-hybridized carbon atoms with a hexagonal lattice, has been demonstrated to be an appealing photoresponse material due to its unique properties, including high carrier mobility, ultrabroadband optical absorption, mechanical flexibility, and dynamic tunability in optical and electrical performances. [1–5] It was already used for various optical devices such as optical modulators [4,5], solar cells [6,7], ultrafast lasers [8,9], polarizers [10,11], and resonators [12,13]. Among them, much effort has been devoted to developing graphene phototransistors and photodetectors (PDs). [14–21] To date, the physical mechanisms of photodetection in graphene are mainly among the photovoltaic effect, photo-thermoelectric effect, photogating effect, bolometric effect, and plasmon-assisted mechanism [22–26]. Unfortunately, the responsivity of PDs using only graphene is notoriously low because of merely 2.3% optical absorption for single-layer graphene. [22] To solve this problem, the addition of other materials (e.g., quantum dots [20] or organic molecules [27]) onto the graphene may improve the photoresponse of the PDs. Moreover, different optical structures, such as microcavity [23,28], waveguide [29,30], metamaterial [31–33], and photonic crystal [34,35], have been widely used to enhance the light absorption of the graphene. For instance, Fang et al. [36]. fabricated a graphene-antenna sandwich photodetector that achieves an 800% enhancement of the photocurrent compared to the antennaless graphene device at a drain bias of 1 mV. In addition, the carrier density in a graphene sheet can be tuned by an applied gate voltage, thus resulting in a voltage-controlled Femi energy. [14,37] And the source-drain bias contributes to extracting photogenerated hot electrons. [15]

Plasmonic metamaterials are often ideal for enhancing the properties of graphene. [38] Metamaterials can enhance optical absorption through surface plasmon resonances (SPRs) and generate hot electrons, which may be efficiently collected by the metallic subwavelength structure. [39] Many studies demonstrated that the incorporation of metamaterials in graphene PD not only has the potential to trap light at a subwavelength scale but also could suppress photogenerated carrier recombination [40]. Moreover, the properties of the graphene PD, including bandwidth, polarization, etc., can also be tuned by the intrinsic properties of the SPRs. [41,42] There is no doubt that the SPRs-based graphene PDs have excellent potential to be applied in numerous domains. However, the previously developed plasmonic enhancement methods are mostly low-absorption or polarization-dependence.

In this work, we proposed and experimentally demonstrated a graphene PD assisted by a metamaterial with strong light absorption in the visible and near-infrared bands. The metamaterial absorber is composed of a gold square-ring array on a silicon wafer with gold and SiO₂ layers, where the gold layer serves as a mirror and enhances the absorption of square rings. All measurements were done with zero gate voltage (V_G=0) and source-drain bias (V_SD=0). The location-dependent photocurrent measurements were performed at normal incidence. Also, there is a linear relationship between the photocurrent and input power. The MGPD has achieved up to 3751% of photocurrent enhancement at λ=1020 nm with the input power of 625 μW. The symmetric square-ring enables polarization-insensitive absorption that results in the polarization-independent photoresponse. Besides, the spectral-dependent photocurrent enhancement was experimentally confirmed and understood by simulated electric field distribution based on the finite-difference time-domain (FDTD) method. The combination of graphene and plasmonic metamaterials has promise in developing photodetectors and optoelectronic devices.

2. Design and fabrication

A square-ring/SiO₂/gold sandwiched system has been recently demonstrated as a metamaterial absorber by plasmonic effect [43]. Consider that polarization property is related to the symmetry of the designed structure. Here, we choose square-ring antennas to explore the enhancement effect on the photoelectric detection of single-layer graphene. The designed graphene photodetector based on a metamaterial absorber is schematically illustrated in Fig. 1(a). The metamaterial absorber consists of periodic gold square-ring nanoantennas on top of a single-layer graphene sheet, which is transferred onto a silicon wafer with 120 nm gold and 75 nm SiO₂ films. A unit cell of the metamaterial absorber is also sketched in Fig. 1(b). The topmost layer is composed of 30 nm thick gold square-ring nanoantennas with outer and inner side lengths of 635 and 390 nm. The periods of the unit cell in both x and y-directions are 700 nm. Then there is a single-layer graphene sheet between the gold square-ring nanoantenna arrays and SiO₂ film interface. It acts not only as a light absorber but also as a charge-transport channel. The back layer marked by yellow is a gold film, which is a reflector on the silicon wafer. The SPRs, originating from optically exciting antennas, give rise to the generation of hot electrons and inject into the graphene sheet (0.3 nm), resulting in n-type doping, as displayed schematically in Fig. 1(b).

 

Fig. 1. Metamaterial-based graphene photodetector. (a) Structural diagram of the device, the gold square-ring nanoantennas on graphene is periodically distributed in both x and y-directions. (b) Schematic of a unit cell, where photogenerated hot electrons resulting from SPRs inject into the graphene sheet.

Download Full Size | PDF

The fabrication procedure of an MGPD is illustrated in Fig. 2. Firstly, a gold film of 120 nm was deposited on a silicon wafer by the electron-beam evaporation (EBE), and then a 75 nm SiO₂ insulating layer was grown by the chemical vapor deposition (CVD). There is a 5 nm titanium adhesive layer between the gold and SiO₂ films. Next, a large-area single-layer graphene sheet synthesized by the CVD was transferred onto the SiO₂ film. A layer of poly(methyl methacrylate) (PMMA) was then spin-coated to define the marker and large electrodes by the electron beam lithography (EBL). After developing, a 100 nm gold film was evaporated on the patterns in which etched the graphene by the oxygen plasma in advance. Followed by a standard lift-off procedure, the gold marker and large electrodes supported by SiO₂ film were created. To define a graphene ribbon, a layer of negative photoresist (AZnloF 2035) was thereafter spin-coated and subjected to standard maskless lithography using alignment technology. After developing, the graphene outside the photoresist mask was etched by the oxygen plasma again; and then the photoresist mask was removed by the acetone to obtain a graphene ribbon. Next, a standard EBL process was employed to define square-ring antennas and conductive wires on the PMMA layer. After developing, a 30 nm gold film was evaporated on the patterns. Finally, an MGPD was achieved by a standard lift-off technology.

 

Fig. 2. Schematic of the fabrication procedures for an MGPD.

Download Full Size | PDF

According to the above procedure, we experimentally fabricated an MGPD, as shown in Fig. 3(a). The gold square-ring antennas had been fabricated perfectly on a graphene ribbon by the alignment technology. The entire square-ring arrays have an area of 10×10 μm², as shown in Fig. 3(b). Also, we observed that the graphene ribbon on the SiO₂ film is in good condition. Figure 3(c) shows a scanning electron microscopy image (SEM) of the periodic gold square-ring antennas. Many studies demonstrated that the electronic properties of graphene could be affected due to in contact with the gold nanoantennas directly [37]. To evaluate the influence of the metal nanostructure, we performed Raman spectroscopy on graphene with and without the square-ring antennas. The low intensity of the D peak indicates the high quality of the graphene (Fig. 3(d)). And the coupling of square-ring antennas with graphene remained the structural integrity of the graphene, as displayed by the well-kept G and 2D peaks in the Raman spectrum. Meanwhile, we also observed changes in the position of the 2D peak, which were caused by additional electron doping from the gold square-ring antennas. [37]

 

Fig. 3. (a) Top-view photograph of an MGPD. (b) Zoom-in photograph of an MGPD at the blue dashed box of left figure. (c) SEM image of the gold square-ring nanoantennas. (d) Raman spectroscopy of the graphene with and without square-ring nanoantennas shows electron doping caused by the gold.

Download Full Size | PDF

3. Results and discussions

To comparing and analyzing the effect of the antennas, we also fabricated a graphene PD without the gold square-ring nanoantennas, as shown in the photograph in Fig. 4(a). The inset shows an SEM image of the graphene PD near the graphene ribbon, indicating this graphene ribbon is also uniform. Also, the length of the graphene channel between the source and drain electrodes is 10 μm. To characterize the photoresponse of these devices, we constructed a setup to perform photocurrent measurement at zero external bias (V_SD=V_G=0). Here, a white-light supercontinuum laser (NKT, Photonics) chopped at 1.1 kHz by an optical chopper (SR540) was coupled to a 50× IR objective (Olympus, NA 0.55) and was then focused on the devices with a beam spot diameter of ∼2 μm. The generated photocurrent of our device was ultimately measured by a lock-in amplifier (SR830) synchronized with the chopper. Firstly, we conducted the location-dependent photocurrent measurements for the graphene PD with and without the gold square-ring antennas. The spot center serves as the coordinate point, and the zero point corresponds to the center between source and drain electrodes. As shown in Fig. 4(b), the location-dependent photocurrent exhibits anti-symmetric response property, which is attributed to the nearly symmetric band structure at both contact electrodes [44]. The minimum photocurrent is observed when light is focused at the zero point. Moreover, the maximal photocurrent occurs near the electrode-graphene junction. The reason is that there exists a built-in potential barrier due to the different work functions of gold and graphene. The built-in potential barrier is similar to a traditional PN junction, where photogenerated carriers can be separated and form the measurable photocurrent. The built-in potential is highest at the electrode-graphene junction area, resulting in a maximal photocurrent near this position. Under illumination, the photogenerated carriers, including photoexcited electrons of graphene and field-induced hot electrons from square-ring antennas, can be extracted as a photocurrent by the built-in potential. Figure 4(b) also indicates that the maximal photocurrent occurs when the beam spot is located at position= ±4 μm for the MGPD. But the position of maximal photocurrent is ±5 μm for a graphene PD without the gold antennas. In fact, the photocurrent is determined by the optical absorption and built-in potential. For the PD with only a graphene sheet, due to the homogeneous absorption, therefore the photocurrent is only determined by the built-in potential, which has a maximum at position=±5 μm. but for an MGPD, the absorption is strongly dependent on the light localization and is not homogeneous. Therefore, the maximum photocurrent is determined by the spot size of the incident beam, period of the square-ring, and the position to the edge. In our experiment, the maximum photocurrent occurring at the position=±4 μm is achieved by an optimal combination between the absorption and built-in potential for an MGPD.

 

Fig. 4. (a) Top-view photograph of an antennasless graphene PD. The blue dashed box shows an SEM image of this PD near the graphene ribbon. (b) Photocurrents generated as functions of the laser spot position for the graphene PD with and without the gold square-ring antennas. The zero point corresponds to the center between source and drain electrodes. (c) Photocurrents detected as functions of the input power for the graphene PD with and without the gold antennas. (d) Normalized photocurrent as the polarization angles varied from 0 to 360° at a wavelength of 650 nm for an MGPD.

Download Full Size | PDF

Here, we choose the maximal photocurrent negative coordinate point as an incident position to explore the relationship between photocurrent and incident power. Figure 4(c) indicates that the MGPD's photocurrent is linear increase with the input power at the wavelength of 710 nm (red dot line) and 1020 nm (blue dot line). Furthermore, it is observed that the MGPD exhibits a higher photocurrent enhancement correspond to the antennasless graphene PD. The enhancement factor g is defined as I_M/I_G, where I_M and I_G represent the photocurrent of graphene PD with and without the square-ring antennas at the same incident power, respectively. The enhancement g was calculated to be 3751% at λ=1020 nm with the input power of 625 μW, and to be 2521% at λ=710 nm with the input power of 546 μW. Additionally, we also investigated the normalized photocurrent with the polarization angle varied from 0 to 360° at normal incidence (λ=650 nm). As shown in Fig. 4(d), the result reveals that the photoresponse of the MGPD is insensitive to the polarization angle. The reason for the polarization-independent will be analyzed later.

We simulated the absorption spectrum of an MGPD under x-polarized excitation by the FDTD method (FDTD solutions, Lumerical). Note that the 5 nm-titanium layer needs to be added to the simulated model. As shown in Fig. 5(a), it was observed that the absorption spectrum exhibits five peaks (red solid curve, marked by P1, P2, P3, P4, and P5). To further identify the origin of these modes, we plotted the cross-section electric field distribution along an x-y plane (interface of graphene and gold square-ring) and x-z plane (middle section). The electric field distributions at an absorption peak of P1 reveal that this mode results from the scatting effect of the gold square-ring antennas (Fig. 5(b)). Furthermore, as shown in Fig. 5(c-f), the electric fields are mainly localized near the square-ring antennas for the peaks of P2, P3, P4, and P5, indicating that they were primarily excited by the SPRs. It means that the enhanced light-trapping effect, including the SPRs and scatting effect, results in a greatly enhanced optical field near graphene to dramatically increase overall absorption. The scatting field was also used to enhance photoresponse, similar to the plasmonic enhancement effect observed in the 2D nanostructure arrays [44].

 

Fig. 5. (a) Normalized responsivity (blue dot curve, left axis), simulated (red solid curve) and measured absorptions (black solid curve) as functions of the laser wavelength for an MGPD. (b-f) Simulated cross-section electric field distributions along an x-y plane (interface of graphene and square-ring antenna) and x-z plane (middle section) for these optical modes (P1, P2, P3, P4, and P5). (g) Measured absorption spectra with polarization angle varied from 0 to 360° at a step of 15° for an MGPD.

Download Full Size | PDF

We also measured the normalized responsivity with various wavelengths when laser spot incidents on a position of -4 μm (see blue dot curve in Fig. 5(a)). It was found that the responsivity of the MGPD exhibits strong spectral-dependent. This is because the generated photocurrent originated from field-enhanced excitation of graphene electrons and plasmon-induced hot electrons from gold square-ring antennas [36]. The responsivity is proportional to the photocurrent, namely the maximal photocurrent of an MGPD occurs at a wavelength of 480 nm. Moreover, we measured the absorption spectrum of the MGPD, as shown in black solid curve of Fig. 5(a). It was observed that the maximum absorption corresponds to the maximum photocurrent. Note that the measured peak has a broader full width at half-maximum (FWHM) than the simulated one. This can be ascribed to the surface roughness of the square-ring antennas and material absorption. Since graphene has no bandgap, hot electrons created in the square-ring could transfer directly into the conduction band of graphene. Also, the responsivity spectrum peaks relatively agree with the simulated absorption ones (Fig. 5(a)), indicating that enhanced photocurrent originated from the improved light-trapping effect. Figure 5(a) also shows that the measured absorption spectrum relatively agrees with the simulated absorption one. To understand the polarization-independent photoresponse behavior of an MGPD, we also measured the absorption spectra with the polarization angle varied from 0 to 360° at a step of 15° (Fig. 5(g)). The result suggests that the polarization-independent of photoresponse has resulted from the polarization-insensitive absorption of symmetric square-ring antennas. In addition, the photoresponse of graphene may also be potentially improved through a perfect absorber (e.g., FP cavities [45] and metamaterials [46]), which achieve exceptionally strong optical absorption.

4. Conclusion

In summary, we developed polarization-insensitive graphene photodetectors enhanced by a broadband metamaterial absorber. Such an absorber comprises the periodic square-ring antennas on top of single-layer graphene, which is transferred onto a silicon wafer with the gold and SiO₂ films. Experimental results indicate that the photocurrent of an MGPD was location-dependent and achieved up to 3751% enhancement compared to the antennasless graphene PD at zero external bias. The symmetric square-ring antennas enable polarization-independent photoresponse at normal incidence. Furthermore, the enhancement of an MGPD originated from the enhanced light-trapping effect, including the SPRs and scatting effect. The combination of graphene and metamaterials has great promise in developing photodetectors and optoelectronic devices.