Photogating Effect of Atomically Thin Graphene/MoS₂/MoTe₂ van der Waals Heterostructures

Abstract

The development of short-wave infrared photodetectors based on various two-dimensional (2D) materials has recently attracted attention because of the ability of these devices to operate at room temperature. Although van der Waals heterostructures of 2D materials with type-II band alignment have significant potential for use in short-wave infrared photodetectors, there is a need to develop photodetectors with high photoresponsivity. In this study, we investigated the photogating of graphene using a monolayer-MoS₂/monolayer-MoTe₂ van der Waals heterostructure. By stacking MoS₂/MoTe₂ on graphene, we fabricated a broadband photodetector that exhibited a high photoresponsivity (>100 mA/W) and a low dark current (60 nA) over a wide wavelength range (488–1550 nm).

1. Introduction

Since the first report of the exfoliation of monolayer graphene [1], many two-dimensional (2D) materials, such as hexagonal boron nitride (hBN) [2], black phosphorous [3], and transition metal chalcogenides, including molybdenum disulfide (MoS₂) [4] and molybdenum ditelluride (MoTe₂), have been discovered [5,6]. Various optoelectronic devices, such as optical modulators, photovoltaics, waveguides, light-emitting diodes, and photodetectors, have been fabricated using these 2D materials [7,8,9,10,11,12,13,14]. In particular, significant efforts have been made to fabricate short-wave infrared (SWIR) photodetectors using graphene and other 2D materials [15,16,17,18,19,20,21,22,23,24,25]. These devices are widely used for biological imaging, remote sensing, night vision, and telecommunications, and the use of 2D materials eliminates the need to cool these devices to cryogenic temperatures. To this end, several van der Waals (vdW) heterostructures of 2D materials have been proposed, wherein the 2D materials are held together by vdW forces. Among the most promising vdW heterostructures of 2D materials are those that exhibit type−II band alignment owing to interlayer optical excitation, which allows for infrared absorption even if the bandgaps of the components are too large to allow them to absorb infrared light [26]. Moreover, type-II band alignment promotes charge separation at the interface, which is essential for photodetection [27]. It has recently been demonstrated that a monolayer-MoS₂/monolayer-MoTe₂ vdW heterostructure showed a distinct photovoltaic current in the infrared region at room temperature (1550 nm) [28]. However, this photodiode suffered from low photoresponsivity (<0.02 mA/W).

In this study, we investigated the modulation of the charge-carrier density of graphene in a vertical graphene/MoS₂/MoTe₂ vdW heterostructure. We fabricated a broadband graphene/MoS₂ photodetector with a MoS₂/MoTe₂ vdW heterostructure as the gate stack. We found that the graphene was significantly doped by the photoexcited charge generated in the MoS₂/MoTe₂ heterostructure and that type−II band alignment between MoS₂ and MoTe₂ resulted in a photoresponsivity greater than 100 mA/W with a dark current of 60 nA over a wide wavelength range. Thus, we were able to realize a SWIR graphene photodetector with high photoresponsivity.

2. Results and Discussion

Figure 1a shows a schematic of the process for fabricating the photodetector. After monolayer graphene was exfoliated and placed on a wafer substrate, monolayer MoS₂ was transferred onto the graphene layer using the PMMA transfer method, such that it partially covered the graphene. The overlapping graphene/MoS₂ region formed the Schottky junction of the device. Next, monolayer MoTe₂ was transferred onto the graphene/MoS₂ junction. To passivate the Schottky junction region, we covered it with a thick hBN layer. The vdW heterostructures were then annealed in a vacuum. Finally, we deposited a source electrode on the graphene layer and a drain electrode on the MoS₂ layer using e−beam lithography. The source and drain electrodes were neither connected to the MoTe₂ layer nor to the hBN layer. Therefore, the device was a graphene/MoS₂ Schottky diode in which the MoTe₂/hBN layer was stacked on the Schottky junction.

Figure 1b,c show atomic force microscopy (AFM) images of the MoS₂ and MoTe₂ monolayers on the Si/SiO₂ wafer used to fabricate the photodetector. The height of the MoS₂ layer was 0.63 nm and that of the MoTe₂ layer was 0.66 nm; this confirmed that the MoS₂ and MoTe₂ structures were monolayers. To further assess the thicknesses of the graphene, MoS₂, and MoTe₂ layers, we determined their Raman spectra before they were transferred (Figure 1d, 532 nm). The ratio of the intensity of the 2D peak (Pos(2D) = 2682 cm⁻¹) of graphene to that of its G peak (Pos(G) = 1592 cm⁻¹) was slightly larger than 1. The height of the graphene layer was smaller than 1 nm, which indicated that the graphene layer was a monolayer (see Figure S1) [29]. In addition, the D peak of graphene conventionally observed at approximately 1350 cm⁻¹, whose intensity is proportional to the defect density of graphene, was not present [30]. The two peaks of MoS₂ seen at 387.4 and 405.7 cm⁻¹ were the $E_{2g}^1$ and $A_{1g}^{}$ peaks, respectively. The distance between these two peaks is indicative of the number of MoS₂ layers, and it increases as the number of MoS₂ layers increases [31]. In this study, this distance was ~18 cm⁻¹, which confirmed that the MoS₂ layer was also a monolayer (Figure 1e). Figure 1f shows the Raman spectrum of the MoTe₂ layer. Only one distinctive peak was present at approximately 240 cm⁻¹. This was the $E_{2g}^1$ peak of the MoTe₂. The absence of a peak at approximately 280 cm⁻¹ ($B_{2g}^{}$) indicated that this layer was also a monolayer [32].

Raman spectroscopy is a powerful tool, not only for the characterization of isolated 2D materials, but also for the vdW stacking of 2D materials, because the Raman active phonon modes of 2D materials are sensitive to changes in the degree of doping and strain of the materials, as well as their vdW interactions with other layers. To evaluate the quality of the vdW stacking at the Schottky junction, we performed a Raman spectroscopy analysis after fabricating the device, as is shown in Figure 1. First, we obtained the Raman intensity maps of the graphene G peak, MoS₂ A_1g peak, and MoTe₂ E¹_2g peak, which allowed for the delineation of the edges of the graphene, MoS₂, and MoTe₂ layers with precision (Figure 2a). Because monolayered MoS₂ and MoTe₂ can be degraded by exposure to air, we focused on the MoS₂, graphene, graphene/MoS₂, and graphene/MoS₂/MoTe₂ regions that were passivated by hBN (Figure 2b). The integration time for one spot in the Raman image was 130 ms, which was sufficiently long for the photoexcited charge carriers in one material to transfer to the other materials. Figure 2c,d show the positions of the MoS₂ E¹_2g (Pos(E¹_2g)) and A_1g (Pos(A_1g)) peaks, respectively. The Raman maps clearly show that the differences in Pos(E¹_2g) and Pos(A_1g) depended on the stacking. Pos(E¹_2g) and Pos(A_1g) were 387.2 and 405.5 cm⁻¹, respectively, in the case of the MoS₂ region; these values were similar to those for MoS₂ before the transfer process. On the other hand, in the region where MoS₂ was stacked on graphene, Pos(A_1g) was blue−shifted by 2 cm⁻¹, while Pos(E¹_2g) remained the same (Figure 2e). Thus, Pos(E¹_2g) and Pos(A_1g) could be changed by changing the degree of doping [33], strain [34], and vdW interactions with the neighboring materials [35]. However, a shift in the A_1g peak of the MoS₂ by 1 cm⁻¹ would require the removal of electrons in a density of 1 × 10¹³. In addition, the biaxial strain in MoS₂ changes both Pos(E¹_2g) and Pos(A_1g), which was not the case in this study. Therefore, we attributed the blue−shift of the Pos(A_1g) to the stiffening of the A_1g phonon by the graphene−MoS₂ vdW interaction [35]. This was indicative of interlayer coupling between graphene and MoS₂. For graphene/MoS₂/MoTe₂, the E¹_2g and A_1g peaks were red-shifted by ~3 and ~2 cm⁻¹, respectively, from those of graphene/MoS₂, which was consistent with previous reports [36]. We believe that the red-shifting of the peaks was attributable to the tensile strain or the relaxation of the compressive strain owing to the mismatch in the lattice constants of the MoS₂ (0.316 nm) and MoTe₂ (0.352 nm).

We also monitored Pos(G) (Figure 2f) and Pos(2D) (Figure 2g) for graphene. Because Pos(G) and Pos(2D) are highly sensitive to the strain and doping level of graphene, these factors can be estimated from the peak positions (Figure 2h) [37]. Pos(G) and Pos(2D) for the graphene−only region were 1585.7 and 2671.5 cm⁻¹, respectively; these values correspond to a tensile strain of ~0.05% and hole doping level of ~6 × 10¹² cm⁻². However, in the graphene/MoS₂ region, the G and 2D peaks were both blue-shifted compared with those in the graphene-only region, which indicated that interlayer coupling with MoS₂ induced a compressive strain of 0.1% and the electron doping of graphene at the ~3 × 10¹² cm⁻² level. The compressive strain induced in graphene by MoS₂ also suggests that the graphene/MoS₂ heterostructure was well formed in the analyzed area [38]. The electron doping of graphene implies that electrons were transferred from MoS₂ to graphene. After MoTe₂ was stacked on graphene, the compressive strain in graphene was relaxed slightly, probably because of the tensile strain generated or the relaxation of the MoS₂−induced compressive strain by MoTe₂. In addition, the graphene became less hole-doped compared with the graphene/MoS₂ region, which meant that the stacking of MoTe₂ enhanced the transfer of electrons to graphene. The underlying mechanism of graphene doping under the different heterostructures is discussed in more detail later in this paper.

To investigate the charge transfer at the graphene/MoS₂ and MoS₂/MoTe₂ interfaces, we performed photoluminescence (PL) spectroscopy (Figure 3). The PL spectrum of monolayered MoS₂ originates from the radiative recombination of three types of quasiparticles: A excitons (~1.85 eV), B excitons (~2.03 eV), and A− trions (~1.80 eV) [39]. Figure 3a shows the band diagrams of the three quasiparticles. The relative contributions of the A excitons and A− trions to the PL spectrum of MoS₂ depend on the Fermi level of MoS₂ [40]. When MoS₂ is hole-doped or less electron-doped, the PL peak from the A excitons is very strong compared with that of the A− trions. On the other hand, the A exciton peak decreases as MoS₂ becomes n-doped, because an excessive number of electrons in MoS₂ bind to the photoexcited electron-hole pairs to form trions.

Figure 3b,c show the total PL intensity and position of the A peak (A exciton + A− trion, respectively). The efficient PL quenching of graphene differentiates the graphene/MoS₂/MoTe₂ region from the MoS₂/MoTe₂ region. The MoTe₂ PL peak at approximately 1 eV was not observed, thus confirming that efficient charge separation had occurred between MoS₂ and MoTe₂ (Figure 3d).

To further investigate charge transfer at the interfaces, we analyzed the PL spectra of the MoS2-only (Figure 3e), graphene/MoS₂ (Figure 3f), and graphene/MoS₂/MoTe₂ (Figure 3g) regions in the 1.70–2.05 eV range. The PL spectrum of the MoS2-only region exhibited A exciton, B exciton, and A− trion PL peaks (Figure 3e). Specifically, the A− trion peak was stronger than the A exciton peak, which was in keeping with the fact that MoS₂ is an n-type semiconductor. The PL spectrum of MoS₂ changed when it was placed over graphene. In this case, the exciton PL peak was stronger than the A− trion peak. This means that the photoexcited electrons in MoS₂ were transferred to graphene, while the photoexcited holes were accumulated in MoS₂. This charge transfer can inhibit the radiative recombination of A− trions in MoS₂ by spatially separating the photogenerated electrons and holes. In addition, it simultaneously induced the electron doping of graphene during the optical measurements. This was consistent with the Raman spectroscopy analysis, which showed that the formation of the graphene/MoS₂ structure resulted in the n-type doping of graphene (Figure 2h). The work function of the MoS₂ was larger than that of graphene [41]. Thus, the electrons in the graphene were transferred to the MoS₂ layer after contact, resulting in an electric field whose direction was toward MoS₂ at the graphene/MoS₂ interface, as is shown in Figure 3h. Therefore, the photogenerated electrons in MoS₂ could be easily transferred to graphene.

In the region where MoTe₂ was stacked on the graphene/MoS₂ structure, resulting in graphene/MoS₂/MoTe₂, the A exciton peak was not observed. This implies that holes did not accumulate in the MoS₂ layer in the resulting structure. This can be explained by the fact that MoS₂ and MoTe₂ exhibited type-II band alignment (Figure 3i) and that the photogenerated holes (electrons) in MoS₂ (MoTe₂) were transferred to MoTe₂ (MoS₂). The transfer of holes from MoS₂ to MoTe₂ aided the separation of the photoexcited electrons and holes in graphene, resulting in efficient PL quenching (Figure 3b). The type−II band alignment between MoS₂ and MoTe₂ also explains the increased electron doping of graphene in graphene/MoS₂/MoTe₂ after the stacking of MoTe₂ on graphene/MoS₂, as per the Raman spectroscopy analysis (Figure 2h). The photoexcited holes trapped in MoTe₂ could enter graphene through MoS₂, resulting in the additional electron doping of graphene. This is because the MoS₂ layer was too thin to screen for holes in MoTe₂.

Figure 4a shows the current-voltage (I–V) characteristics of the photodetector under continuous illumination with a 50 nW light over the graphene/MoS₂/MoTe₂ region. The laser spot was placed away from the metal electrodes to exclude the photocurrent from the graphene/metal and MoS₂/metal junctions. In the dark, the I–V characteristics were similar to those of a typical Schottky diode with series resistances (Figure S2). The series resistances can be attributed to the resistances of the graphene and MoS₂ layers and the contact resistances between graphene, MoS₂ layers, and the metal electrodes. The current increased under illumination, whose wavelength range was 488–1550 nm. A finite photocurrent was observed at zero voltage, showing the photovoltaic effect of the device [42]. However, the photocurrent was almost absent at zero voltage and increased as V was increased. It implies that the photocurrent mainly originated from the photogating effect rather than the photovoltaic effect.

For photodetectors based on 2D materials, the photoresponsivity, R, and specific detectivity, D*, are generally used as the figures of merit [42]. R is the photocurrent per incident unit optical power, and D* is a measure of the smallest detectable signal from the photodetector and is given by $D^{*}=R{A}^{1/2}/{\overline{I_n{}^2}}^{1/2}$, where A is the illumination area (1200 μm²) and ${\overline{I_n{}^2}}^{1/2}$ is the root-mean-square noise current. The area of the whole photodetector was larger than the laser spot size. Figure 4b shows the estimated R value as a function of V for various wavelengths. The device exhibited the maximum R value at V = 3 V; however, R increased when we applied a higher V. For instance, R exceeded 10⁴ mA/W at 488 nm and 10² mA/W at all the other wavelengths. Notably, R was 3 × 10³ and 1.8 × 10² mA/W at 980 and 1550 nm, respectively; these are the wavelengths at which MoS₂ is optically inactive. When the noise current is dominated by shot noise, D* can be estimated using the following equation [43]:

$$D^{*}=R{A}^{1/2}/{\left(2e{I}_{dark}\right)}^{1/2}$$

(1)

D* was estimated to be 3 × 10¹¹, 9 × 10¹⁰, and 5 × 10⁹ Jones at 488, 980, and 1550 nm, respectively. Figure 4c shows the time-resolved photocurrent of the device under illumination with a 980 nm light at V = 3 V. The rise and decay times were 0.37 and 1.32 s, respectively (see also Table S1 [15,22,24,44,45,46,47,48,49]).

In the present study, the photogating effect was a photoinduced change in the Fermi level of the material in question, namely, the graphene under MoS₂. Under the photogating effect, the value of ${\phi }_b$ at the graphene/MoS₂ junction was modulated by light, which, in turn, changed the device current. The photogating of the graphene-neighboring MoS₂ layer was well known (Figure 4d,e) [50]. The MoS₂ layer absorbed light, resulting in the photoexcitation of electrons and holes. Owing to the electric field at the graphene/MoS₂ interface and the difference in the energies of the conduction band edge of MoS₂ and the Fermi level of graphene, the photoexcited electrons were transferred from MoS₂ to graphene. Meanwhile, the photoexcited holes of MoS₂ were transferred to MoTe₂, where they were trapped. This altered the electric field near the MoS₂ layer, causing the n-doping of graphene. Consequently, ${\phi }_b$ decreased under illumination, resulting in a reduction in the Schottky barrier height between graphene and MoS₂. Although this explains the photoresponse of the device under visible light, the photocurrent under infrared light (980 and 1550 nm) requires an additional explanation, because monolayered MoS₂ cannot absorb light with wavelengths larger than 800 nm. Considering the MoTe₂ monolayer, we propose the following mechanism to explain the photoresponse of the device under infrared light. When illuminated with 980 nm light (Figure 4f), MoTe₂ absorbs the light, generating photoexcited electrons and holes. These photoexcited electrons tunnel toward graphene directly or through the MoS₂ layer, while the photoexcited holes remain in MoTe₂. This leads to the n-doping of graphene, resulting in a photocurrent in the device. In the case of 1550 nm light (Figure 4g), both MoS₂ and monolayered MoTe₂ are optically inactive when they are separated. However, because monolayered MoS₂ and monolayered MoTe₂ exhibit type−II band alignment, an interlayer transition occurs between them when they are stacked. The photoexcited electrons in MoS₂ and holes in MoTe₂ are separated by the transfer of electrons to graphene or the extraction of electrons owing to the drain bias. The remaining holes in MoTe₂ cause the n−doping of graphene, thus reducing ${\phi }_b$ at the graphene/MoS₂ junction.

To confirm the role of the MoTe₂ layer, we fabricated a graphene/MoS₂ Schottky diode without a MoTe₂ layer on the Schottky junction and measured its photoresponse. As shown in Figure S3a, the current of the device barely changed under infrared light (980 and 1550 nm). Time-resolved current measurements were performed under infrared light illumination (Figure S3b). However, no change in the current was observed during the measurements. The absence of a photoresponse excludes the possibility that the detrapping of charge carriers near the graphene/MoS₂ interface was responsible for the photoresponse under infrared light. Thus, these results clearly indicate that the absorption of infrared light by graphene did not contribute to the photocurrent of the device and that monolayered MoS₂ alone did not induce the photogating of graphene under infrared light. It is well known that photogenerated charge carriers recombine within a few picoseconds because of plasmon emission and carrier-phonon scattering [51]. Consequently, the photogenerated charge carriers in graphene were annihilated before the charges at the graphene/MoS₂ interface were separated. In addition, the bandgap of MoS₂ inhibits the absorption of infrared light, making the graphene/MoS₂ junction inactive under infrared light illumination. In summary, the observed photoresponse of the photodetector shown in Figure 4 was attributable to the MoTe₂ layer deposited on the MoS₂/graphene Schottky junction.

In conclusion, we realized a graphene/MoS₂ barristor-based photodetector that exploited the photogating of graphene based on the type−II band alignment in the monolayered MoS₂/monolayered MoTe₂ structure. The device showed a photoresponsivity as high as 10⁴ mA/W and a detectivity of 3 × 10¹¹ Jones under visible light. More importantly, we were able to simultaneously achieve a photoresponsivity of more than 10² mA/W and detectivity of more than 5 × 10⁹ Jones in the 980–1550 nm range.