Picosecond electrical response in graphene/MoTe2 heterojunction with high responsivity in the near infrared region

Understanding the fundamental charge carrier dynamics is of great significance for photodetectors with both high speed and high responsivity. Devices based on two-dimensional (2D) transition metal dichalcogenides can exhibit picosecond photoresponse speed. However, 2D materials naturally have low absorption, and when increasing thickness to gain higher responsivity, the response time usually slows to nanoseconds, limiting their photodetection performance. Here, by taking time-resolved photocurrent measurements, we demonstrated that graphene/MoTe2 van der Waals heterojunctions realize a fast 10 ps photoresponse time owing to the reduced average photocurrent drift time in the heterojunction, which is fundamentally distinct from traditional Dirac semimetal photodetectors such as graphene or Cd3As2 and implies a photodetection bandwidth as wide as 100 GHz. Furthermore, we found that an additional charge carrier transport channel provided by graphene can effectively decrease the photocurrent recombination loss to the entire device, preserving a high responsivity in the near-infrared region. Our study provides a deeper understanding of the ultrafast electrical response in van der Waals heterojunctions and offers a promising approach for the realization of photodetectors with both high responsivity and ultrafast electrical response.


Introduction
Understanding the ultrafast charge carrier dynamics in various electronic and optoelectronic devices, especially in photodetectors with high response speed and efficient detection responsivity [1][2][3] , is indispensable for enhancing their performance.Two-dimensional (2D) layered materials hold great promise for the next generation of high-efficiency photodetectors, owing to their ultrathin thickness, broadband photoresponse and high on/off ratio [4][5][6][7][8][9] .When these 2D materials are thin, their intrinsic response speed is only approximately several picoseconds because of their atomic thickness [ 10 , 11 ].However, to enhance the light-matter interaction for a larger photo-generated carrier concentration, an increase in thickness is required.Unfortunately, this could extend the intrinsic response time of the materials to the nanosecond time scale [ 12 , 13 ], and lead to a photodetector with high responsivity but slow response speed.Therefore, it is desirable to realize 2D layered photodetectors that possess high responsivity while maintaining an intrinsic response speed in the order of picoseconds.
One alternative approach is to construct 2D van der Waals heterojunctions that not only combine the merits of a single material but also lead to novel optical and electronic properties [ 14 , 15 ].Regarding the dynamics of photocurrent, several processes contribute to the response time, including photocarrier recombination (  r ), photocarrier drift (  d ), exciton dissociation (  s ) and charge transfer (  s ) [12] .Among these processes, charge transfer (CT), which serves as an interlayer interaction in heterojunctions, involves a series of transient carrier dynamic processes [16][17][18][19] .CT processes have been intensively studied using ultrafast alloptical pump-probe measurements [20][21][22] .Typically, a faster carrier decay in a strongly coupled graphene/WS 2 heterojunction compared to a pure WS 2 flake is observed, and direct CT absorption from graphene to a WS 2 flake is explained under excitation below the WS 2 bandgap [22] .These dynamic processes in heterojunctions indicate that interfacial CT https://doi.org/10.1016/j.fmre.2021.09.018 2667-3258/© 2021 The Authors.Publishing Services by Elsevier B.V. on behalf of KeAi Communications Co. Ltd.This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ) can effectively decrease charge carrier decay time and is significant for photovoltaic and photodetector device applications.Although pure optical measurements provide information on the photo-generated carrier dynamics, a complete picture of the photoresponse involving photocurrent generation and collection in a practical device is still elusive.Recently, ultrafast time-resolved photocurrent (TRPC) measurements have been demonstrated to enable the achievement of a picosecond intrinsic response time for different 2D materials and their heterojunction-based photodetectors [ 10 , 11 , 23 ].For instance, in graphene-covered WSe 2 heterojunction photodetectors, a 5.5 ps response speed was realized, but with the increase of the WSe 2 flake thickness from monolayer to 40 nm; an extended response time of 10 ns was also observed [12] .Meanwhile, using the same measurement technique, another study suggested that there were two different response speeds for the respective materials in graphene/1L MoS 2 heterojunctions [24] .Although the drift process (  d ) of charge carriers in photocurrent generation has been studied, the contribution of the CT (  s ) process to the response speed has not been clarified.
In this work, we demonstrated a comprehensive understanding of interlayer interactions and qualitatively explained the photogenerated charge carrier transfer behaviors in graphene/MoTe 2 van der Waals heterojunction photodetectors.Using a TRPC setup, the ultrafast response speed in pure MoTe 2 and graphene/MoTe 2 heterojunctions was investigated.Our results show a significantly reduced response time of MoTe 2 in graphene/MoTe 2 devices, which can be scaled down to 10 ps by an additional decay component introduced by the formation of the heterojunction, while the intrinsic response time of pure thick MoTe 2 is approximately 1 ns.The heterojunction devices take full advantage of the fast response speed in graphene as well as high responsivity in MoTe 2 , and present a possible method to enhance photodetection performance by decreasing photocurrent recombination loss.

Methods
Device fabrication: Pure MoTe 2 devices and graphene/MoTe 2 devices were fabricated from mechanically exfoliated materials.In the pure MoTe 2 device, a MoTe 2 flake was mechanically exfoliated onto transparent polydimethylsiloxane (PDMS), which was then transferred onto a silicon substrate with a 300 nm-thick silicon dioxide layer.In the graphene/MoTe 2 heterojunction devices, graphene flakes were first mechanically exfoliated onto a silicon/silicon dioxide substrate.The MoTe 2 flake was placed on the PDMS using a similar method, and it was then aligned and transferred onto the graphene using a microscope.Cr/Au (10 nm/50 nm) conducting electrodes on top of 2D MoTe 2 or graphene/MoTe 2 with a channel length of 5 m were fabricated using standard electron beam lithography (EBL), metal thermal evaporation and lift-off processes.
Basic characterization: Atomic force microscopy (AFM) (Bruker Dimension Icon) in the taping mode was used to identify the thickness of the samples.Raman measurements of the samples were taken using a confocal microscope (WITec, alpha-300) equipped with a 100 × objective lens.The excitation source of the Raman spectra was a 532 nm continuous wave laser, and the laser beam was focused to 1 m on the devices.The electrical properties were measured with an Agilent-B1500 semiconductor analyzer in a LakeShore vacuum chamber of 10 − 4 Pa.
SPCM and TRPC measurements: Scanning photocurrent microscopy (SPCM) and time-resolved photocurrent measurements (TRPC) were performed on our home-built setup.In SPCM measurements, a 780 nm fiber laser (NPI Rainbow 780 OEM) with a pulse width of 80 fs and a 488 nm continuous wave laser were chopped by a mechanical chopper at 1050 Hz, and then focused onto the sample by a long working distance objective (Olympus LMPLFLN 50 ×) near the diffraction limit.The generated photocurrent was collected by a lock-in amplifier (Stanford SR830) at the chopped frequency with a background noise of around 0.2 pA.The SPCM measurements were performed by raster scanning the entire device mounted on a piezoelectric translation stage (Piezoconcept LT3) according to the fixed laser spot.In TRPC studies, a 780 nm pulse laser was split into two independent beams to form a pump-probe measurement configuration.The pump beam was delayed by different path lengths, with the delay time precisely controlled by a mechanical delay stage (Thorlabs DDSM100/M).The pump and probe beams were recombined by a beam splitter after the delay line stage, and focused onto the sample using the same long working distance objective.

TRPC measurement technique and CT hypothesis
Ultrafast carrier dynamics in materials can be probed by the alloptical pump-probe technique, as shown in the transient absorption measurements ( Fig. 1 a).In these measurements, a pulsed pump laser beam with a specific energy is typically used to selectively excite the sample, and another pulsed white laser serving as a probe beam can be applied to trace the desired signals, such as the interfacial CT process.The time-resolved differential transmission spectrum can be expressed as △T / T 0 = ( T -T 0 )/ T 0 , where T and T 0 are the white lightinduced transmission signals with and without the pump beam, respectively.Hence, positive signals represent absorption saturation and photo-induced bleaching in the ground state.With different time delays between the pump and probe pulse, the differential transmission collected by the spectrograph indicates the carrier dynamic changes.In pure 2D materials, the differential transmission generally exhibits a slow exponential decay with an increase in the delay time, corresponding to the exciton formation process and its recombination lifetime [25] ( Fig. 1 a lower plane, gray curve).In the construction of heterojunctions, an enhanced signal intensity and an exponential decay with a much faster process than pure materials can be observed, demonstrating a CT process to the adjacent material ( Fig. 1 a lower plane, light blue curve).In addition, with specific probe energy (usually below the bandgap of the charge carrier extraction material), the rise time in the differential transmission spectrum of the carrier injection material can further reflect the CT time scale [26] .Although all these optical measurements provide important information on the intrinsic photogenerated carrier dynamics, which can be considered the ultimate limit of the photoresponse in the materials, these processes may not contribute to the photocurrent or other electric responses in a practical device.
Compared to the all-optical pump-probe experiments, the TRPC measurements follow the conventional pump-probe scheme, where a pulsed laser beam (80 fs in this work) is split into two independent beams to form a pump-probe configuration, and an ultrafast photo-generated current can be detected with a sub-picosecond resolution (Fig. S1).Crucially, the pump in the TRPC measurements must be sufficiently strong to saturate the sample.When the pump beam and the probe beam coincide spatially and temporally, the generation of photocurrent by the probe beam is suppressed because of the saturation in the ground state, causing a prominent dip to appear at the zero-time delay ( Fig. 1 b lower plane).With an increase in the delay time between the pump and probe beam, parts of the pump-induced charge carriers relax, and can be excited again by the probe beam.Accordingly, the time-resolved photocurrent displays an exponential decay (recovery) with  corresponding to the device response time.In all the processes contributing to the response time, photocarrier recombination (  r ) leads to photocurrent loss, whereas photocarrier drift (  d ) in combination with exciton dissociation and CT (  s ) leads to the generation of photocurrent.Meanwhile, the response time can be expressed using the equation [12] .Hence, the TRPC generation in pure 2D materials follows a single-exponential decay with only CT to the electrodes ( Fig. 1

b left).
In contrast, according to the studies on all-optical pump probes, when forming a heterojunction, such as graphene and MoTe 2 flakes ( Fig. 1 b right), the graphene layer provides an additional CT channel, and the band alignment of the straddling gap (type I) can lead to photo-excited charge carriers in MoTe 2 transferring to the graphene (  s ).With less exciton recombination in the light-absorption material, the photocurrent recombination loss (  r ) of the entire device is reduced, and a new

Comparison of photodetection performance in pure MoTe 2 and Gr/MoTe 2 devices
To elucidate the distinct charge carrier dynamic processes, we first investigated the ultrafast photocurrent behaviors in photodetectors made of pure MoTe 2 layers (see Methods).Fig. 2 a shows an optical image of a pure MoTe 2 photodetector with a thickness of 4 nm, and the corresponding Raman spectrum ( Fig. 2 b) exhibiting three characteristic peaks located at ∼171, ∼234, and ∼289 cm − 1 were assigned to the A 1g , E 1 2g and B 1 2g vibration modes, respectively [27] .In the electrical transport characteristics, the I ds -V ds output characteristic curve at room temperature in this device shows a linear current change at the back-gate varying from -50 V to 50 V ( Fig. 2 c), indicating a small contact barrier between MoTe 2 and the metal electrodes.With the increase in V g , the conductance of the MoTe 2 device is monotonically enhanced, demonstrating an n-type transport behavior.Meanwhile, in the I ds -V g transfer characteristic curve, this device exhibits an on/off ratio of approximately 10 4 with a threshold voltage of -40 V, manifesting an electron-doped intrinsic property ( Fig. 2 d).Based on the transconductance, we calculate the electron mobility of this two-terminal device using the equation [28] , where L / W is the ratio between the channel length and width, and C i is the capacitance between the back gate per unit area (in our case C i = 1.15 × 10 − 8 F cm − 2 for 300 nm thick silicon dioxide).The result reveals a value around 29.2 cm 2 V − 1 s − 1 , which is among the high mobilities in various transition metal dichalcogenides (TMDCs) [29] .
The spatially resolved photocurrent response (see Method) of the MoTe 2 device was characterized by a home-built scanning photocurrent microscopy (SPCM) at zero bias under the excitation of a continuouswave 488 nm laser ( Fig. 2 e) and a pulsed 780 nm laser ( Fig. 2 f).To better determine the position of the generated photocurrent, the optical reflection images of the devices were superimposed on the photocurrent images in Fig. 2 e, f with the white dashed rectangles indicating the locations of the electrodes.Apparently, these two SPCM images recorded at zero bias display the photocurrent appearing at the electrode edges, which is attributed to the contribution of both PV and photo-thermoelectric effect (PTE) [30][31][32] .While the broadened photocurrent in Fig. 2 f can be ascribed to the enhanced photothermoelectric effect and reduced optical resolution by the excitation of the 780 nm pulse laser.The excitation laser power-dependent photocurrent response was also investigated for the MoTe 2 photodetector, and similar SPCM images were obtained with an increase in laser power (Supplementary Section 2).Based on the photocurrents measured by the SPCM, we have calculated the responsivity of the 4 nm thick MoTe 2 photodetector at zero bias and relatively low excitation powers, yielding 0.6 mA W − 1 at 488 nm and 0.23 mA W − 1 at 780 nm.These obtained responsivities are comparable to those of previously reported thin-layered TMDCs under a similar experimental configuration [ 24 , 33 , 34 ].
To gain more insight into the photocurrent detection performance of these infrared materials, we fabricated other MoTe 2 photodetectors with the thickness L varying from 2 nm to 35 nm.Fig. 3 a presents the photoresponse of a representative 25 nm thick photodetector using a single probe laser beam at 780 nm.We observe a typical sublinear power dependence (black spheres), which can be fitted by the equation PC ∝ ln(1 + N 0 ) [ 13 , 23 ] (blue line), where  is the Auger recombination rate,  is the response time, and N 0 is the initial exciton population for each pulse.We further calculated the responsivity of this 25 nm MoTe 2 photodetector at different excitation powers, according to the experimentally obtained photocurrents.At zero source-drain bias, the highest value of 12.5 mA W − 1 is obtained at the excitation power of 0.36 W, almost two orders of magnitude larger than the 4 nm MoTe 2 device and several times higher than those of other high-speed photodetectors operating in the near-infrared region [ 14 , 35 , 36 ].
The response speed of the two-terminal pure MoTe 2 photodetectors with different thicknesses L was then examined by pump-probe ultrafast photocurrent measurement (see Method).Here, the probe beam at 780 nm was chopped so that the lock-in amplifier could only measure its photocurrent.When increasing the thickness L from 2 to 35 nm, the response time exhibits a distinct extended trend ( Fig. 3 b).We extract the decay  with respect to thickness L using the equation , and the result shows a power law relationship of approximate 1.3 (blue line in Fig. 3 c).These results indicate that the extended response speed derived from the transient time  tran is unavoidable, even in the two-terminal lateral device configuration.We attribute this phenomenon to the fact that the photogenerated charge carriers in thicker MoTe 2 photodetectors are initially in the middle of the samples rather than at the surface [37] , and therefore, they still need to experience a longer out-of-plane distance to drift to the electrodes on the top.Thus far, the thickness-dependent photoresponse properties of pure MoTe 2 detectors have been demonstrated, which indicate that a device with one 2D material can not possess both high responsivity and fast response speed.
Considering the ultrahigh electron mobility of graphene and the efficient CT at the heterojunction interface, we further fabricated vertical graphene/MoTe 2 heterojunction photodetectors.Fig. 4 a presents the optical image of a typical heterojunction device with a MoTe 2 thickness of 25 nm.The corresponding SPCMs of this device obtained using different excitation lasers are shown in Fig. 4 b and c.Compared with the pure 25 nm MoTe 2 photodetector (Supplementary Section 5), the SPCMs of the graphene/MoTe 2 heterojunction exhibited a similar photocurrent intensity and the photocurrent profile appeared only at the electrodes on top of MoTe 2 , indicating that the integration of graphene/MoTe 2 does not form a strong band-bending at the junction (Supplementary Section 6) and does not influence the photoresponsivity of intrinsic MoTe 2 .It should be noted that the photoresponse of graphene in our heterojunction is one to two orders of magnitude weaker than that of MoTe 2 .Meanwhile, because the energy of the 780 nm pulse laser (1.59 eV) is closer to the multilayered MoTe 2 exciton absorption (0.9 eV) [38] , the heterojunction has a larger photoresponse to the red laser than the blue laser.In the TRPC experiments, we observed a distinct photocurrent decay in the MoTe 2 region of the heterojunction.This phenomenon is also observed in the junction region where graphene and MoTe 2 fully overlap.Compared with the pure 25 nm MoTe 2 photodetector (green curve in Fig. 4 d), a prominent fast new decay component appears at the graphene/MoTe 2 photodetector, while the slow component is preserved (black curve in Fig. 4 d  | ∆PC| data, each decay curve was normalized.The decay curve in the graphene/MoTe 2 heterojunction yields a new decay component of approximately 10 ps with a relative weight percentage of 15.5% ( Fig. 4 f), which is one order of magnitude faster than the intrinsic response time of ∼400 ps in 25 nm MoTe 2 and implies a photodetection bandwidth as wide as 100 GHz.We further measured other series of heterojunction photodetectors with different MoTe 2 thicknesses, with almost every device having this fast decay component.In the thicker (35 nm) samples, the fast response time slightly increased to 16 ps, and the relative weight percentage decreased to 10% (Supplementary Section 8).Meanwhile, in these vertical heterojunction devices, we observed a longer intrinsic MoTe 2 response time (the slow component) than the two-terminal devices with the same thickness, which shows a power law relationship of 1.8 between  and L (red line in Fig. 3 c), corresponding to the transient time expression of  tran = L 2 / μV bias [ 12 , 24 ].To further study this fast decay component, we performed probe power and gate-dependent TRPC measurements in a graphene/25 nm MoTe 2 heterojunction photodetector.Fig. 4 g shows the normalized TRPC spectrum for different probe powers at gate bias of 0 V.With the increase in probe power from 108 W (19 J cm − 2 ) to 252 W (45 J cm − 2 ), two timescales appear at all TRPC curves and there is only a slightly enhanced dip at zero time delay.We also plotted the time-resolved | ∆PC| for different probe powers and put them in one graph.Most prominently, all the normalized TRPC curves seem to overlap and show an almost unchanged fast decay component (Supplementary Fig. S7).The fitted fast response time and | ∆PC| for all the probe powers are shown in Fig. 3 i.The results reveal that the response time in graphene/25 nm MoTe 2 remains at 10-15 ps independent of the probe power (black dots in Fig. 3 i).In the gate-dependent TRPC measurement, we changed the back gate voltage from -5 V to 5 V.With the different gate biases, a similar result was obtained (red triangles in Fig. 4 i), where the fast response time did not show any significant changes.The preserved fast decay component under different power and gate conditions indicates that the high-speed graphene/MoTe 2 heterojunction photodetector can perform under various circumstances.
To prove the origin of the fast photocurrent decay component, we further conducted TRPC measurements in the graphene region of the heterojunction.In contrast to the pure graphene (blue curve in Fig. 5 a) or the MoTe 2 region in the heterojunction (red curve in Fig. 5 a), a photocurrent decay comparable to that of intrinsic graphene ( ∼5 ps, 63.5% of weight) with a slower component ( ∼70 ps, 36.5% of weight) is observed in the graphene region of the heterojunction.According to previous reports, the photogenerated electron-hole pairs in some unique Dirac semimetals such as graphene [ 11 , 39 , 40 ] or Cd 3 As 2 [35] can relax their absorbed energy through rapid electron-electron interactions without any change in the lattice temperature.Hence, intrinsic graphene possesses an ultrafast photocurrent response of several picoseconds [41] , independent of the pump laser with an increase in time delay (blue square in Fig. 5 b).Accordingly, the observed slow decay component is attributed to the interaction between graphene and MoTe 2 ( Fig. 5 c), which also gives rise to the fast photocurrent decay in MoTe 2 of the graphene/MoTe 2 heterojunction photodetector.

Mechanism discussion
The observed different photo-response speeds with an extra component in the respective materials at heterojunctions suggest a possible method for the realization of photodetectors with high response speed and efficient detection responsivity.We used the photocurrent generation model to further interpret and discuss the ultrafast response time and the dynamic process of charge carriers in graphene/MoTe 2 heterojunction photodetectors.As proposed at the beginning, the response time in the generation of photocurrent can be expressed using the equation To analyze the complicated dynamics in the heterojunction, we first consider pure graphene and pure 25 nm MoTe 2 .For pure graphene, the average response time  is approximately several picoseconds for the multilayer used in this study.In our case, we take the maximum response time as the photocarrier recombination time  r and ignore  s , because the exciton dissociation time  s is far smaller than the response time.It follows that the photocarrier drift time  d in pure graphene is equal to  r ⋅ ∕(  r − ) = 8 ps (here, we take 3.75 ps for  and 7 ps for  r ).Likewise, a drift time  d of approximatedly 2000 ps was obtained for 25 nm pure MoTe 2 .The significantly shorter drift time  d obtained in pure graphene compared to that found in MoTe 2 is consistent with the fact that graphene has a significantly larger in-plane and out-of-plane mobility than MoTe 2 .
Next, we discuss the situation in the graphene/MoTe 2 heterojunction, where we observed two different response times in the respective materials.We believe that there are two possible mechanisms for the recombination of charge carriers in this heterojunction.One possibility is that the recombination time in the heterojunction is the average time between graphene and MoTe 2 , and the excited charge carriers of different materials in heterojunction have the same recombination process.In this case, the slower intrinsic response time component in thick MoTe Here we take 10 ps for  MoTe2 and 20 ps for  r MoTe2 ).If we further use a quick charge transfer time of 5 ps between MoTe 2 and graphene [25] , a drift time of 15 ps is obtained in the heterojunction, which is two orders of magnitude smaller than the thick intrinsic MoTe 2 .Therefore, we can conclude that the fast response time of the CT process in the graphene/MoTe 2 heterojunction decreases the photocurrent recombination loss in thick MoTe 2 , and the graphene leads these parts of the transferred carriers to have a shortened drift time.In principle, this structure makes full use of the ultrahigh mobility in graphene and breaks the limitation of transient time.

Conclusion
In summary, using time-resolved photocurrent measurements, we investigated the response time in pure MoTe 2 and graphene/MoTe 2 heterojunction photodetectors with different thicknesses.We demonstrated that the extended response time in pure thick MoTe 2 can be scaled down to 10 ps with the integration of graphene.By doing so, we realized a high-performance photodetector with both high responsivity and quick response by making use of the CT process and large carrier mobility in graphene.The design of an ultrafast device and the understanding of carrier dynamics in graphene/MoTe 2 heterojunction photodetectors pave the way for the next generation of ultrathin van der Waals optoelectronic devices.

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

Fig. 1 .
Fig. 1.Comparison of the ultrafast all-optical pump probe and ultrafast time-resolved photocurrent measurement.(a) Schematic illustration of the ultrafast pure optical pump-probe measurement.The heterojunctions (light blue) show a faster carrier decay than pure materials (grey) in the transient dynamic spectrum; (b) Schematic illustration of the time-resolved photocurrent (TRPC) measurement.Left: Photocurrent generation in pure 2D TMDCs by the photovoltaic effect, where the recovery of saturated photocurrent corresponds to a single component photocurrent response in the TRPC spectrum.Right: Photocurrent generation in Gr/TMDCs heterojunction by the photovoltaic effect, where the new charge carrier's transfer channel provided by graphene effectively decreases the recombination loss and brings about a fast photocurrent response component in the TRPC spectrum.

Fig. 2 .
Fig. 2. Electrical transport characterizations and photoresponse in the 4 nm pure MoTe 2 device.(a) Optical image of the fabricated MoTe 2 device.Inset: AFM line scan performed along the white dashed line shown in the optical image; (b) corresponding Raman spectrum, where the three prominent peaks A 1g , E 1 2g and B 1 2g are identified to be the fingerprint of the thin layered MoTe 2 ; (c) I ds -V ds output characteristic curve at gate voltages varying from -50 V to 50 V; (d) I ds -V g transfer characteristic curve at source-drain voltage from 0.2 to 1 V on the semi-logarithmic scale.Inset shows the mobility of this device around to be 29.2 cm 2 V − 1 s − 1 ; (e) and (f) Scanning photocurrent maps (SPCM) of the MoTe 2 device at 0 V source-drain voltage excited by a 488 nm CW laser and 780 nm pulse laser, respectively.

Fig. 3 .
Fig. 3. Responsivity and photocurrent response speed in pure MoTe 2 devices with different thicknesses revealed by TRPC technique.(a) Photocurrent (left) and responsivity (right) as the function of probe power in a 25 nm pure MoTe 2 device.The solid blue line in probe-induced photocurrent fits the formula I PC ∝ In(1 + N ); (b) TRPC measurements show extended response time with the increase of the MoTe 2 thicknesses; (c) Response time ( ) as the function of their thicknesses ( L ) on the log-log scale.The open squares are the  of the pure two-terminal MoTe 2 devices with the parallel contacts in Fig. (b), while the solid squares are derived from the longer  component in Gr/MoTe 2 heterojunction with vertical contacts.The red and blue solid fitting curves display the  ∝ L 1.83 and  ∝ L 1.3 in the vertical and parallel electrode configurations, respectively.The data points are the average values of the response time and the error bars indicate the minimum and the maximum values.

Fig. 4 .
Fig. 4. Enhancement of photocurrent response speed by integrating of graphene with a 25 nm MoTe 2 .(a) Optical image of the fabricated Gr/25 nm MoTe 2 heterojunction device; (b) and (c) SPCM of the device in (a) at 0 V source-drain voltage with the excitation of a 488 nm CW laser and 780 nm pulse laser respectively.The white and red dashed curves outline the position of the electrodes and the MoTe 2 , respectively; (d) comparing TRPC in pure 25 nm MoTe 2 (green) with the Gr/25 nm MoTe 2 (black) heterojunction, where the heterojunction shows two prominent response time scales; (e) and (f) normalized photocurrent (| ∆PC|) decay curves of pure 25 nm MoTe 2 and the Gr/25 nm MoTe 2 heterojunction, where the solid red lines fit the exponential decay.The | ∆PC| is the difference between the photocurrent (t = 0) and photocurrent (t → ∞), with each decay curve normalized for a better comparison of different | ∆PC| data; (g) Normalized TRPC in the Gr/25 nm MoTe 2 heterojunction at probe powers from 252 W to 108 W; (h) normalized TRPC in Gr/25 nm MoTe 2 heterojunction at gate voltages from -5 V to 5 V; (i) response time as a function of probe power and gate voltage in Gr/25 nm MoTe 2 heterojunction.

Fig. 5 .
Fig. 5. Demonstration of charge carrier transfer in the Gr/25 nm MoTe 2 heterojunction device.(a) TRPC of graphene (black) in Gr/MoTe 2 heterojunction, 25 nm MoTe 2 (red) in Gr/MoTe 2 heterojunction and pure graphene (blue); (b) normalized photocurrent (| ∆PC|) decay in pure graphene (blue squares) and graphene in heterojunction (black squares), where the two time constants in graphene of heterojunction correspond to  1 = 5.24 ps and  2 = 70.92ps; (c) schematic illustration of the photo-induced charge carrier transfer and two different response components in Gr/MoTe 2 heterojunction when the graphene region is excited.The dashed red arrow indicates the slower response component due to a relatively low mobility in thick MoTe 2 .