The two-dimensional (2D) layered materials provide an opportunity to engineer high photodetection capabilities due to various intriguing physics [1–3] and facile construction of van der Waals heterostructures (vdWHs) without the lattice-mismatch problem [2,4,5]. Until now, many efforts have been performed to improve the photodetection performance [6–8]. In particular, the high light/dark ratio is required with the high photocurrent and low dark current, which guarantees the high detectivity and responsivity [9–11].

Among the vdWH family, the transition metal dichalcogenides (TMDCs) have gained extensive attention in optoelectronic devices due to their excellent electron mobility, high absorption coefficient, and mechanical flexibility [12]. In particular, the rational stack of different TMDCs materials can engineer novel photodetectors [13,14], such as the unipolar barrier photodetectors [15], the out-of-plane charge transport phototransistors [16,17], the pn junction field effect transistor [18], and the photo-gate transistor [17,19].

In general, the operation mode of 2D photodetectors is fixed in either photoconductive or photovoltaic mode [9]. The photoconductive structures possess high responsivity due to increasing optical gain but suffer from a low detectivity with a large dark current and a slow response speed [20,21]. Compared with photoconductive structure, the photovoltaic photodetectors based on pn junctions have high detectivity and fast response speed [22,23]. However, it is greatly difficult to obtain the high responsivity due to the presence of the depletion region in the junction [24–26]. In addition, there are also limited materials due to the difficulties in tuning and controlling doping of materials. Therefore, in the fixed operation mode structures, it is greatly difficult to realize simultaneously high detectivity and responsivity due to the intrinsic tradeoff between detectivity and responsivity [27].

In this work, we report a photodetector with a local pn junction floating-gate (PNLFG-PD) based on the WSe2/MoS2 vdWHs, where the n-type MoS2 is partly floated on the WSe2 channel to form the local pn junction gate, exhibiting the switchable operation mode between the dark and light conditions. The high responsivity of 2.12×105A/W synchronizes the high detectivity of 1.25×1014 Jones, and is accompanied with the high Ilight/Idark ratio up to 1.9×106. This work provides a unique platform to realize the high-performance photodetection.

First, a few layers WSe2 exfoliated from bulk flakes were first stacked over the Si substrate. Then the drain and source electrodes (Au) were pre-deposited on the WSe2, and the MoS2 was partly covered on the WSe2. Note that the MoS2 was not contacted with the drain electrode. Subsequently, the WSe2 part was covered with the photoresist and only the MoS2 part was treated with ultraviolet ozone (UVO). The power of the UVO instrument (BZS250GF-TC) is 300 W, and the treatment is operated under atmospheric pressure at room temperature. The devices were globally back-gated through the highly p-doped Si substrate.

The optoelectronic measurements were performed at the air atmosphere using an Agilent B1500A semiconductor parameter analyzer with a probe station system, which could perform current region in the range of 0.1 fA–1 A. The source and drain terminals were provided by tungsten nanomanipulator tips. A 405 nm laser was used to measure the photocurrent mapping (a 405 nm laser with 500 μW power and 1 μm spot size). Raman spectra were tested by the Horiba micro-Raman system (Labraxm HR Evolution) with 532 nm laser excitation and the laser power of 1.28 mW. The topography device was characterized by atomic force microscopy (Bruker Dimension Icon) at tapping mode. XPS experiments were performed using the PHI5000 Versa Probe III (Scanning ESCA Microprobe) equipped with a monochromatic Al Kα mono source (hν=1486.6eV) and a concentric hemispherical analyzer. Charge neutralization was performed using both low-energy electrons (<5eV) and argon ions. Measurements were made at a takeoff angle of 45° and the laser spot has a diameter of about 50 μm. The sample diameter must be greater than 50 μm due to the limitation of spot area. It is difficult to obtain larger area 2D materials by the exfoliation method. To obtain the real information of MoS2 samples, we just tested MoS2 bulk materials treated by UVO.

The configuration of the PNLFG-PD is schematically illustrated in Fig. 1(a). Note that the MoS2 flake is partly floated on the WSe2 channel source terminal, which guarantees the combined structure of the pn junction with the channel. The Si with 300 nm SiO2 is used as the substrate.

The optical microscope image is shown in Fig. 1(b), and an atomic force microscopy (AFM) image is shown in Fig. 1(c). The thicknesses of MoS2 and WSe2 along the white dotted line are 45 nm and 60 nm, respectively. The Raman spectral [Fig. 1(d)] shows the good quality of WSe2/MoS2 heterojunctions stacking. The comparative current-voltage (I-V) curves of the WSe2 PD and WSe2 PD with floated MoS2 are tested and shown in Fig. 1(e). The electrodes 1, 2 are accorded to the WSe2 PD with floated MoS2, exhibiting the significantly reduced current by 2 orders of magnitude compared to that of the WSe2 transistor accorded to the electrodes 4, 5. These results identify the role of floating n-type MoS2, which can reduce effectively the current of the WSe2 channel. Note that in the WSe2 PD with 4–5 electrodes, the asymmetrical current about the zero bias point is attributed to the unequivalent contact area for the source and drain terminals [28].

To further investigate the modulating ability of local-floated MoS2 on the electrical characteristics of WSe2, the typical transfer curves of drain current (Ids)-gate voltage (Vg) were tested, as shown in Fig. 1(f), where the drain voltage (Vds) was fixed at 3 V, and the Vg ranged from −80 to 80 V, and the device was set in dark condition. The Λ-shaped curves are exhibited with the sweeping of Vg from negative to positive, showing the anti-ambipolar transport characteristics. The Λ-shaped curves identified into the pn junctions have been reported in previous research [29,30]. These further suggest the partly floated MoS2 in PNLFG-PDs changes the p-type WSe2 into pn junction. It also can be seen that the multilayer MoS2 possesses larger carrier density than WSe2, so that the region of WSe2 covered with MoS2 will be completely depleted and inversed into the weak n-type.

In addition, to further modulate the out-of-plane doping effect of local-floated n-MoS2, the UVO treatment is adopted by the different treatment time, which has been proved to affect p-type doping [31]. In Fig. 2(a), the threshold voltage (Vth) is changed from −70V to −57V (electron, n) as the oxidation time increases from 0 min to 5 min, reflecting the reduction of electron concentration. The UVO treatment induces the formation of MoOx about 2 nm, which plays the p-doping role of UVO due to a high work function of MoOx [32]. The consistent phenomenon has been proved in the previous report [31]. The carrier concentration and carrier field-effect mobility μFE in PNLFG-PDs can be effectively regulated from 8 to 32cm2/(V·s) by the top oxidizing treatment.

The optoelectronic characteristics of the device are then investigated under the irradiation of a focused laser beam (λ=405nm). Shown in Fig. 2(b), the Λ-shaped transfer curves disappeared when the device was illuminated with light intensity of 1.92 mW and showed the obvious p-type dominated characteristics. These results exhibit the light-switchable polarity from the anti-ambipolar characteristics in the dark condition to p-type dominated characteristics in light, which is distinctive from any conditional photovoltaic or photoconductive device. In addition, in Fig. 2(c), the Λ-shaped transfer curves gradually disappeared as the light intensity increased from 0 to 1.92 mW, further suggesting the light switches the polarity of the PNLFG-PD from anti-ambipolar to p-type conduction.

These results further confirm the n-type doping role of the local-floated MoS2 on the p-type WSe2 channel. Moreover, the reduced dark current is beneficial to reduce shot noise under the operating condition, which is badly required for high-performance photodetection.

In order to further investigate the photoelectric characteristics, shown in Fig. 2(d), the current output characteristics are investigated in the dark, showing that the current at the forward voltage is gradually reduced as the oxidation time increases, while the backward voltage lightly increases. The photoelectric characteristics are also investigated under 5 mW laser light illumination at 405 nm, shown in Fig. 2(e). It can be seen that the photocurrent is enhanced by 3 orders at Vds=3V (from 3×10−8 to 1.11×10−5A).

Moreover, the photocurrent at 0–1 V is smaller than the reverse current in the long oxidation time shown in Fig. 2(e). In order to further investigate the impact of MoOx, in Fig. 2(g), the forward conduction voltage (Vf) gradually shifts to large voltage as the oxidation time increases in the dark. In Fig. 2(h), the light induces Vf gradually reducing as the oxidation time increases. The values of dark Vf and illuminated Vf are extracted from Figs. 2(e) and 2(g) and are plotted in Fig. 2(i). The opposite modulation of oxidation for dark Vf and illuminated Vf is exhibited, i.e., the photocurrent at 0–1 V is smaller than the reverse current in the long oxidation time, but the larger photocurrent is obtained as long as the bias is larger than Vf. These results induce a high Ilight/Idark ratio, shown in Fig. 2(f), which is up to 1.9×105, and higher than that of WSe2-based photodetectors.

The photodetection performances are then evaluated under the irradiation of the focused laser beam (λ=405nm). Figures 3(a)–3(c) demonstrate the linear and semi-log I-V curves of the PNLFG-PD at varied light intensity from 0 to 5920 μW. The forward conduction voltage at the dark station is about 1.81 V, which is gradually reduced as the light intensity increases, as shown in Fig. 3(b), suggesting that the channel electric field regulation effect by the MoS2 is weakened. The large difference of dark Vf and illuminated Vf induces a sudden increase of photocurrent at the bias of 1.81 V, as shown in Fig. 3(c).

Responsivity (R) is an important parameter for photodetection, which is defined as the ratio of the output photocurrent to the input light power and is calculated by Eq. (1): R=Ilight−IdarkPinA,(1)where R is the photoresponsivity, and Idark is the dark drain current with 1.7×10−11A. A is the sensing area of 132.2μm2. Pin is the incident light intensity ranged from 370 μW to 5920 μW.

This high value of R results from the increase in photocurrent and the decrease in dark current. Another key argument for evaluating the performance of photodetectors, the specific detectivity (D), is calculated by Eq. (2): D=RA2qIdark,(2)where q is the unit of charge. The light intensity is 370 μW.

In Fig. 3(e), the dependence of R and D on the drain voltage, where the bias is at 2 V, Idark is about 7×10−11A, and the light intensity is 370 μW, demonstrates simultaneously increase as the drain voltage increases, which is completely different from trend of the pn junction detector [Fig. 3(d)]. In conditional configurations of pn junctions and photoconduction, R and D have been obtained and reported in previous reports [24,25], and simultaneous high values of R and D have been observed in few reports [26].

The dependence of both R and D on the light intensity is shown in Fig. 3(f). The R and D are 2.12×105A/W and 1.25×1014 Jones, respectively. The simultaneous high R and D are attributed to the noteworthy reduction of the dark current and increase of the photocurrent.

The on-off characteristics of the PNLFG-PD are shown in Fig. 3(g), where the device exhibits reliable and fast on-off switching performance. Figure 3(h) is the magnification of Fig. 3(g), and the response (τrise) and recovery (τdecay) times are 94 and 117 μs, respectively. The response time is defined as the time for the photocurrent to rise from 10% to 90% or fall from 90% to 10% of the peak. The good response speed immediately benefits from the fast separation of photogenerated electron–hole pairs by the built-in field in the floated pn junction. It is worth mentioning that a faster response speed may be obtained considering the optimized contact of the device and large parasitic impedance of the experimental setup.

Finally, Fig. 3(i) benchmarks the values of R and D in this work, other single 2D WSe2-based phototransistors, and 2D pn junction devices. The obvious tradeoff between R and D is the high R for the photoconductive device, but the D value tends to be low, while the values of R and D are simultaneously high in the PNLFG-PD, which exhibits the advancement of the local pn junction gate.

To gain insight into the mechanism for the high R and D values in the PNLFG-PD, we investigated the spectra response characteristics shown in Fig. 4(a), showing the device wavelength operation range from 375 nm to 810 nm, which corresponds to the bandgap of WSe2 and MoS2.

In addition, we investigated systematically the ln (I/V2) versus 1/V plots under the different light intensity. It is well known that the quantum tunneling effect refers to the behavior of a particle penetrating or crossing the potential barrier beyond its own kinetic energy. Figure 4(b) investigates three typical transport mechanisms: direct tunneling (DT) at low voltage, Fowler-Nordheim tunneling (FNT), and space-charge-limited current at the different bias voltage (V).

The direct tunneling current is proportional to the voltage and is expressed as Eq. (3) [33], where d, m*, and φ are the tunneling thickness, effective electron mass, and tunneling barrier (φ); ℏ is the simplified Planck constant: I∝Vexp(−2d2m*φℏ).(3)

The FN tunneling currents are widely modeled by the FN law following Eq. (4) [33]: I∝V2exp(−4d2m*φ33ℏeV).(4)

From the above equation, FN tunneling has a linear region with a negative slope from the curve of ln (I/V2) versus 1/V. Thus, the φ is defined as φ=Vtranse [34], which is estimated and extracted from Fig. 4(b). The Vtrans of the threshold voltage for FN tunneling corresponds to φ/e, shown in Fig. 4(c), which is the triggered FNT barrier, showing that φ is decreased from 1.38 eV to 0.65 eV as the light intensity increases, indicating that the light illumination can reduce the φ barrier. The observed photoinduced barrier-lowering (PIBL) in the PNLFG-PD is consistent with the structure of the molybdenum-based double heterojunction phototransistor [25].

To elaborate on the PIBL mechanism of the PNLFG-PD, the band alignment of the PNLFG-PD is schematically plotted according to the previously reported conduction band energy (Ec) and valence band energy (Ev) of MoS2 and WSe2. In the dark, shown in Fig. 4(a), the difference of Fermi level leads to charge transfer between WSe2 and MoS2, which naturally results in the full depletion at the p-type WSe2 part covered with the high electron concentration of n-type MoS2. Consequently, shown in the upper panel of Fig. 4(c), the full depletion regions on the vertical and lateral sides induce a large barrier φ with the two built-in fields, which prevents the injection of electrons from the source terminal (S) [schematically shown in the bottom panel of Fig. 4(c)]. Therefore, the large barrier φ in the PNLFG-PD channel guarantees the low dark current [about 10 pA shown in Fig. 2(f)], which is urgent to realize the high value of D.

Under illuminating, shown schematically in Fig. 4(d), the electron–hole pairs are dynamically generated in the WSe2 and MoS2 layers, and subsequently separated by the lateral and vertical built-in field. In other words, there is an equivalent charge transfer between the WSe2 channel and floated MoS2. A large quantity of photogenerated holes and electrons originated are accumulated into the WSe2 channel in the side of the source and floating MoS2. The energy band diagram at the light-on state to illustrate the photoinduced barrier-lowering is shown in Fig. 4(f). The green symbol (1) represents the photogenerated electron–hole pairs, the orange symbol (2) represents the electrons and holes separating and accumulating process to reduce the barrier, which rapidly lowers the Fermi level (EFp) at the source and alleviates the strength of the built-in field in the lateral direction. The lowering of EFp in the side of the source also causes a decrease in the barrier of φ in the vertical direction. Therefore, the EFp is thus variant and floating with the change of light intensity, which induces feedback of lowered barrier resulting in the electrons being much easier to inject from the source. Therefore, the light-induced lowering of the barrier in the PNLFG-PD is completely different from that of the conditional PN junction, which can gain the high photocurrent and guarantees the high value of R.

In conclusion, we present the 2D operation mode-switchable photodetector with the local-floated gate of the pn junction, which exhibits the anti-ambipolar conduction in dark and unipolar conductive mode in light, and enables the simultaneous realization of high responsivity (2.12×105A/W) and detectivity (1.25×1014 Jones), as well as high Ilight/Idark ratio of 106. This helps to study various other types of heterostructures and can be further applied to electronic and optoelectronic devices.