Enhanced photoresponse in MoTe2 photodetectors with asymmetric graphene contacts

Atomically thin two dimensional (2D) materials are promising candidates for miniaturized high-performance optoelectronic devices. Here, we report on multilayer MoTe2 photodetectors contacted with asymmetric electrodes based on n- and p-type graphene layers. The asymmetry in the graphene contacts creates a large (Ebi ~100 kV cm-1) built-in electric field across the short (l = 15 nm) MoTe2 channel, causing a high and broad (400 to 1400 nm) photoresponse even without any externally applied voltage. Spatially resolved photovoltage maps reveal an enhanced photoresponse and larger built-in electric field in regions of the MoTe2 layer between the two graphene contacts. Furthermore, a fast (~0.01 ms) photoresponse is achieved in both the photovoltaic and photoconductive operation modes of the junction. Our findings could be extended to other 2D materials and offer prospects for the implementation of asymmetric graphene contacts in future low-power optoelectronic applications.


Introduction
2D van der Waals crystals have received great attention due to their excellent properties and versatility for a wide range of potential applications in optoelectronics. [1,2] In particular, transition metal dichalcogenides (TMDs) with their finite and channels and/or charge traps at the metal/2D material interface. [14,15,[26][27][28][29] Thus, both the length of the channel and the quality of the contacts should be carefully chosen to optimize the photoresponse.
In this study, we report on photodetectors based on MoTe 2 with vertical asymmetric graphene contacts. We use p-type graphene grown by chemical vapor deposition (CVD) as the top contact and n-type exfoliated graphene as the bottom contact. This asymmetry in the graphene contacts is adopted to break the mirror symmetry of the internal electric field profile, thus creating a large built-in electric field E bi . This feature combined with the short length of the MoTe 2 optically active channel enables an efficient and fast photoresponse. The heterostructure exhibits a high and broad spectral photo response from the VIS to the NIR range of the electromagnetic spectrum (λ = 400-1400 nm) without any external applied voltage: the photoresponsivity is R = 12.38 mA W −1 at λ = 1064 nm and R = 27.64 mA W −1 at λ = 550 nm. Through scanning photovoltaic mapping, an enhanced light absorption is clearly observed in the overlapping region of the graphene and MoTe 2 layers. Furthermore, because of the short (l = 15 nm) MoTe 2 channel between the top and bottom graphene contacts in the vertical heterostructure, the response time of the device can be as short as ≈6. 15 µs, which is one to three orders of magnitude faster than that reported before for MoTe 2 -based photodetectors. [1,9,15,30,31] Figure 1a shows the schematic layout of our MoTe 2 -graphene based photodetectors. To fabricate the MoTe 2 -graphene heterostructure, a multilayer graphene flake was first mechanically exfoliated from a bulk crystal using adhesive tape and then transferred onto a Si/SiO 2 substrate (300 nm thick SiO 2 ). [32,33] Using the same mechanical exfoliation and transfer method, a flake of MoTe 2 was then transferred onto the bottom graphene layer. Finally, a microstamp of CVD-graphene was transferred on top of the MoTe 2 flake to create the top electrode. [33] Metallic contacts (Ta/Au) were fabricated on the substrate using standard photoetching, magnetron sputtering, and lift off. Single crystals of 2H-MoTe 2 were purchased from HQ graphene (Netherlands); CVD-grown graphene was provided by G-CVD (Xiamen, China) and the bulk graphite was purchased from 2D Semiconductors (US). All mechanical exfoliation and transfer processes were conducted inside a glove box. Figure 1b shows the atomic force microscopy (AFM) image of one of our devices. The large and uniform exfoliated multilayer graphene serves as a bottom electrode and has a thickness of 3.6 nm; the MoTe 2 flake has a thickness of about 15 nm; the top CVD-graphene layer corresponds to single layer graphene. Raman spectroscopy studies were conducted to assess the quality of the individual flakes and heterostructure over an extended frequency range (from 200 to 2800 cm −1 ) with an  exfoliated graphene layers. c) Raman spectra for different regions of the heterostructure including CVD-graphene, exfoliated graphene, MoTe 2 , and the vertical overlapping region. The diameter of the laser spot for the Raman studies is approximately 1.5 µm. d) Current-voltage I ds -V ds curves in the dark and under illumination with a laser of wavelength λ = 1064 nm and power P = 3.13 W cm −2 . The diameter of the laser beam is about 45 µm, which is larger than the device size. The inset shows the I ds -V ds curves in the dark of MoTe 2 photodetectors with symmetric graphene contacts. www.advopticalmat.de excitation laser wavelength λ = 532 nm and power P = 100 µW. A large magnification (100 × ) objective lens was used to focus the laser to a spot diameter of 1.5 µm and probe different regions, including the MoTe 2 , exfoliated graphene, CVD-graphene and the region where all layers overlap. Figure 1c shows typical Raman spectra of our samples. The Raman spectrum for the isolated MoTe 2 flake shows two Raman-active modes: the E 2g 1 mode at 232.14 cm −1 and the B 2g 1 mode at 288.40 cm −1 , as observed in the literature. [26] For bulk MoTe 2 , only the E 2g 1 mode is observed ( Figure S1a, Supporting Information), consistent with previous reports. [5,7] The difference between the bulk and 2D layers is assigned to the breakup of the translation crystal symmetry in few layer MoTe 2 . [9] In the Raman spectra of CVD and exfoliated graphene, the most intense features are the G and 2D peaks. The 2D peak from CVD graphene is approximately twice more intense than the G peak, confirming that CVD graphene is single layer. [34] In contrast, the G peak for exfoliated graphene is higher than the 2D peak, indicating that the flake is multilayer, in agreement with the AFM data. Furthermore, the Raman spectrum of exfoliated graphene reveals an additional stronger peak at 2716.4 cm −1 , as also reported in the literature for multilayer graphene. [34][35][36] The peak is assigned to a q 11 double-resonance Raman process. [35] In the Raman spectra of the vertical overlapping region, all the above peaks can be clearly observed, demonstrating the good quality of the heterostructure.

Results and Discussion
To investigate the electrical properties of our devices, the current, I ds , was measured for different voltages, V ds , applied between the drain (top CVD graphene) and source (bottom exfoliated graphene) contacts. All the electrical measurements were carried out in vacuum (≈5 mbar) at room temperature. Figure 1d shows the I ds -V ds curves in the dark and under illumination: in the dark, the I ds -V ds curve shows nonlinear rectifying characteristics, consistent with asymmetric contact barriers between the MoTe 2 flakes and the two graphene electrodes. Under continuous-wave laser illumination (λ = 1064 nm), the current increases and the I ds -V ds shows a clear photovoltaic effect. For comparison, we fabricated photodetectors based on MoTe 2 with symmetric graphene contacts. In this case, the rectification behavior is not observed or is much smaller (inset of Figure 1d).
We now examine in detail the photoresponse properties of our photodetectors. Figure 2a shows the I ds -V ds characteristics with 1064 nm laser illumination at powers ranging from P = 0 to 6.25 W cm −2 . The photocurrent, I ph , is defined as I ph = |I light − I dark |, where I light and I dark are the currents Adv. Optical Mater. 2019, 7, 1900190 Figure 2. a) I ds -V ds curves with illumination at various laser powers P at room temperature (λ = 1064 nm). b) Photocurrent I ph as a function of P at different reverse biases V ds . c) Dependence of the exponent β on the reverse bias (λ = 1064 nm). d) Photoresponsivity R as a function of P at different V ds . The diameter of the laser spot is about 45 µm, which is larger than the device size.

www.advopticalmat.de
measured with and without illumination, respectively. As shown in Figure 2b, the photocurrent I ph increases with increasing P, revealing a sublinear behavior, i.e., I ph ∝ P β , where β increases from 0.855 to 0.911 when V ds decreases from 0 to −0.20 V. Thus even without any applied source-drain voltage, the photogenerated electron-hole pairs can be effectively separated by the built-in electric field of the heterostructure to generate a photocurrent. Also, although β increases with decreasing V ds , its value remains always smaller than 1. This is the fingerprint of a photoconductivity gain that is influenced by charge traps in the layers and/or their interfaces. [9,37,38] A similar phenomenon was reported in heterostructures based on different vdW crystals, such as WS 2 /MoS 2 and WSe 2 /GaSe heterojunctions. [39,40] Figure 2d plots the photoresponsivity (R = I ph /PS) of the heterostructure at different applied voltages as a function of the incident laser power P and λ = 1064 nm. Here S is the in-plane area (≈400 µm 2 ) of the heterostructure device. The photoresponsivity increases with increasing reverse bias, reaching a value of R = 65.56 mA W −1 at V ds = −0.20 V and P = 12.50 mW cm −2 . The photoresponsivity remains high at V ds = 0 V: R = 12.38 mA W −1 , which is larger than that reported for MoTe 2 layers. [41] The corresponding photodetectivity D* and photoconductive gain g are plotted for different applied biases and powers in Figure S2 in the Supporting Information. The photodetectivity has a value of 2.24 × 10 10 Jones at P = 12.50 mW cm −2 and V ds = 0 V. This corresponds to a photoconductive gain of g ≈ 0.2. The photoresponse retains similar characteristics at different wavelengths ( Figure S3, Supporting Information) with a photoresponsivity of up to R = 27.64 mA W −1 at λ = 550 nm and P = 12.50 mW cm −2 . Thus rectification and photovoltaic effects can be clearly observed in MoTe 2 layers with vertical asymmetric graphene contacts. The nonzero open-circuit voltage (V oc ) and short-circuit current (I sc ) increase with increasing laser power ( Figure S4a,b, Supporting Information). Heterostructure devices based on MoTe 2 layers of different thickness (from ≈8 to 21 nm) show a similar behavior ( Figure S5, Supporting Information) with the larger photoresponsivity being observed in the thicker layers (Table S1, Supporting Information). For photodetectors based on MoTe 2 with symmetric exfoliated graphene contacts, the photoresponse is instead significantly weaker ( Figure S6, Supporting Information).  www.advopticalmat.de λ ≈ 1150 nm corresponds to the excitonic absorption associated with the lowest direct optical transition at the K-point of the Brillouin zone. [5,42] The peak at λ ≈ 600 nm arises from high energy excitonic transitions influenced by interlayer interactions. [5,43] A similar peak was also observed in the reflectance spectra of thin MoTe 2 layers. [5] The weak photoresponse at long wavelengths (λ ≈ 1400 nm) is assigned to the optical absorption from band tail states due to charge traps. [8,9] To investigate the photoresponse in further detail, we carried out a series of measurements of the spatially resolved photovoltage maps. These were obtained by scanning a focused laser beam across the plane of the device without any externally applied voltage. Figure 3b shows the optical image of the heterostructure. A larger optical image of the device that includes the Ta/Au metal contacts is in Figure S7 in the Supporting Information. The corresponding normalized photovoltage maps at λ = 550 and 1064 nm are shown in Figure 3c,d, respectively. In each figure, the different parts of the heterostructure are marked in different colors: the bottom exfoliated graphene is marked with a white solid line; the MoTe 2 flake is marked with a yellow solid line, and the top CVD graphene is highlighted by a blue solid line. As shown in Figure 3c,d, the photovoltage is nonuniform and is enhanced in the region of the heterostructures where all layers overlap. This enhancement was not observed in previous studies of MoTe 2 with symmetric CVD graphene electrodes in the literature. [41] Also, in our samples, regions with exfoliated graphene/MoTe 2 and MoTe 2 /CVD graphene, the photovoltage is much weaker. The measured photovoltage is positive in all regions, implying the existence of a built-in electric field pointing in the same direction, e.g., from the bottom exfoliated graphene to MoTe 2 in exfoliated graphene/MoTe 2 , from the bottom exfoliated graphene to CVD graphene in exfoliated graphene/MoTe 2 /CVD graphene, and from MoTe 2 to CVD graphene in MoTe 2 /CVD graphene.
To estimate the built-in electric field, E bi , in the heterostructure, we fabricated and measured the transfer characteristics of FETs based on individual MoTe 2 , exfoliated graphene, and CVD-graphene layers on a SiO 2 /Si substrate. According to the transfer curves in Figure S8 in the Supporting Information, CVD graphene is p-type, exfoliated graphene is n-type and multilayer MoTe 2 is p-type.
The Fermi energy, E F , for monolayer graphene, is derived from Equation (1) [39,44,45]  πεε ( ) Here E F is measured relative to the neutrality point of the graphene band structure, V g is the back gate voltage on the Si-substrate, V D is the charge neutrality point voltage from the transfer curves of graphene, v F ≈ 10 6 m s −1 is the Fermi velocity, ε 0 is the permittivity of free space, ε = 3.9 is the dielectric constant of SiO 2 , e is the electron charge, and t = 300 nm is the thickness of SiO 2 . [39,44,45] We set the Dirac point of graphene at −4.55 eV relative to the vacuum level. From the measured transfer curves in Figure S6 in the Supporting Information, we derive that the Fermi level of CVD graphene is at an energy of 0.19 eV below the Dirac point, i.e., E Fp = −4.74 eV, at V g = 0 V. We obtain a similar estimate of doping level from the analysis of the Raman spectra (Section S9, Supporting Information). For n-type exfoliated multilayer graphene, we obtain an upper estimate of the Fermi level by using Equation (1), E Fn = −4.44 eV. Thus from E Fp = −4.74 eV and E Fn < −4.44 eV, we infer a built-in electric field E bi = (E Fn − E Fp )/el < 150 kV cm −1 , where l = 15 nm is the thickness of the MoTe 2 layer. On the other hand, since exfoliated graphene is n-type, its Fermi level should be above the Dirac point, i.e., E Fn > −4.55 eV. Thus we conclude that the built-in electric field should be within the range 127 kV cm −1 < E bi < 150 kV cm −1 . The corresponding energy band diagram of the heterostructure at equilibrium is shown in Figure 4a. The junction can work under different conditions including the photovoltaic mode without any applied voltage (Figure 4b) and the photoconductive mode under reverse bias (Figure 4c). At zero bias, the built-in electric field points in same direction at the CVD graphene/MoTe 2 and exfoliated graphene/MoTe 2 interfaces. Thus the photoresponse of the regions where all layers overlap is stronger than in other regions. This is consistent with our scanning photovoltage maps (Figure 3c,d) and studies of additional devices including MoTe 2 with symmetric exfoliated graphene contacts and MoTe 2 with symmetric CVD graphene contacts (Figure 1d). The curves show nonlinear characteristics, which confirms that contact barriers exist at the interfaces of CVD graphene/MoTe 2 and exfoliated graphene/MoTe 2 . Finally, we note that under a reverse bias voltage, the externally applied electric-field points in the same direction as the built-in electric field. At V ds = −0.2 V, the electric field almost doubles, leading to a larger photocurrent, as measured experimentally (Figure 2).
The stability and speed of the photoresponse are crucial figures of merit of a photodetector. Figure 5a,  www.advopticalmat.de temporal response of the photocurrent. This is obtained with a square-wave modulation of the light intensity for different powers (P = 0.31, 1.25, and 3.13 W cm −2 ) at λ = 1064 nm. Under zero or reverse biases, the photocurrent can be switched on and off repeatedly and reproducibly. This switching behavior was also observed for photoexcitation under different laser wavelengths (λ = 400, 550, 635, 800, and 1064 nm) at zero bias ( Figure S9, Supporting Information). Also, the heterostructure exhibits a fast dynamic response (Figure 5c,d). To study the temporal response of the current, the heterostructure was illuminated with pulsed light generated by a light-emitting diode driven by a square-wave signal generator. The dynamic response for the rise and decay of the photocurrent is well described by the equations I(t) = I 0 [1 − exp(−t/τ r )] and I(t) = I 0 exp(−t/τ d ), where τ r and τ d are the time constants for the rise and decay of the current. By fitting the rising and falling edges of the current versus time in Figure 5c, we derive τ r = 16.58 µs and τ d = 14.96 µs at V ds = 0 V. A faster photoresponse with τ r = 6.15 µs and τ d = 4.35 µs is obtained at V ds = −0.20 V. This faster dynamics is assigned to the enhanced electric field of the heterostructure by the applied reverse bias. [33] As summarized in Table 1, the measured photoresponse times are shorter than those reported for MoTe 2 based FETs and several other heterostructure photodetectors in the literature. [1,9,17,[46][47][48][49][50][51][52] We assign the improved photoresponse to the short transport channel and enhanced built-in electric field of the heterostructure.

Conclusions
In summary, we have demonstrated high-performance heterostructure photodetectors based on multilayer MoTe 2 with Adv. Optical Mater. 2019, 7,1900190   vertical asymmetric graphene contacts. The heterostructure not only exhibits a broadband photoresponse from λ = 400 to 1400 nm, but also shows high responsivity of up to R = 27.64 mA W −1 under zero bias. Thus even without any externally applied voltage, the photogenerated carriers are efficiently separated by the built-in electric field of the heterostructure. The photoresponse is significantly enhanced in the vertical overlapping region (exfoliated graphene/MoTe 2 /CVD graphene). Furthermore, the heterostructure shows a fast temporal response with decay and rise times in the microsecond range. The improved photoresponse indicate that van der Waals heterostructures with asymmetric graphene contacts are promising candidates for high-speed and self-powered optoelectronic devices. Research on the epitaxial growth of graphene and the controlled doping of graphene are still in their infancy.
Progress in these important areas has the potential to further enhance the performance of these devices and accelerate the use of asymmetric graphene contacts for a wide range of optoelectronic devices.

Experimental Section
Device Fabrication: The multilayer graphene was mechanically exfoliated using adhesive tape from a bulk single crystal. The approximate thickness of the graphene flakes was found and identified by optical contrast using an optical microscope. The accurate thickness of the flakes was determined by using AFM. The optical microscope had a 20× objective lens and a 10× eyepiece, which enabled to identify clearly areas down to 2 µm × 2 µm. A three-axis water hydraulic micromanipulator (Narishige, Japan) was used to transfer a multilayer graphene flake (bottom contact) on a Si/SiO 2 substrate (300 nm thick SiO 2 ). The moving accuracy of the three-axis water hydraulic micromanipulator was 1 µm. Using the same mechanical exfoliation and transfer method, a MoTe 2 flake was then transferred onto the bottom graphene layer. Finally, a CVD graphene microstamp was transferred onto the MoTe 2 sheet to form the top contact. The CVD graphene was synthesized on a Cu substrate. By wet etching, electron beam lithography, and oxygen plasma etching, CVD-graphene was processed into mocrostamps prior to the mechanical transfer. Metallic contacts (Ta/Au) were fabricated on the substrate using standard photoetching, magnetron sputtering and lift off. All mechanical exfoliation and transfer processes were conducted inside a glove box. The methods for transferring and stacking the layers are described in the literature. [32,53,54] Electrical and Optoelectrical Studies: The I ds -V ds curves were measured by an Agilent Technology B1500A Semiconductor Device Analyzer. The measurement resolution of the Semiconductor Device Analyzer was down to 0.1 fA and 0.5 µV. The monochromatic illumination was provided by a Zolix Omni-λ300 monochrometer coupled to a Fianium WhiteLase Supercontinuum Laser Source. The Fianium WhiteLase Supercontinuum Laser Source generates laser wavelengths from 400 to 2400 nm whose power can be controlled accurately. An optical power meter (PM100D, Thorlabs) was used to monitor the output laser power on the sample. A objective lens (100×, Olympus), a light chopper, a micromechanical stage (MAX381, Thorlabs), and a lock-in amplifier (SR830, Stanford Research Systems) were used to carry out spatially resolved photovoltage mapping. A digital oscilloscope and pulsed light were used to investigate the response time. The pulsed light was generated by a light-emitting diode driven by a square-wave signal generator

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.