Introduction

VO2 is a strongly correlated functional oxide exhibiting an IMT at 340 K1,2,3,4. In the so-called insulating state at room temperature, VO2 is a semiconductor with a 0.6 eV band gap and monoclinic crystal structure, while above the transition temperature it exhibits a rutile structure and a metallic behavior with a resistivity drop up to 5 orders of magnitude in the bulk material1,2. The IMT can be induced by thermal, electrical, magnetic or optical excitations3 and is typically hysteretic. The electrically induced phase transition results in an extremely steep drop of the resistivity that has been exploited to realize VO2 two-terminal switches3,4. Coupled VO2 oscillators have been recently demonstrated together with their potential for neuromorphic pattern recognition applications5. The IMT of VO2 resistors connected in series to the source or gate of field-effect devices has been exploited to demonstrate switching slopes well below the 60 mV/dec Boltzman limit constraining the MOSFET subthreshold slope6,7. VO2 resistance-to-capacitance reversible reconfigurability is also appealing for RF applications4. However, the small band gap of VO2 in its insulating state determines a small resistivity and high leakage currents in VO2 switches. Moreover, the electrical induced phase transition requires typically several hundreds of µW to few mW of power3,4. VO2 three terminal gated devices have been proposed and investigated with the purpose of decreasing the voltage and current required to trigger the IMT, but the large electron density necessary for the phase transition at room temperature is difficult to achieve by exploiting the gate control only8,9. Large insulating state leakage and reduced gate control on the carrier density are currently two fundamental challenges on the road for VO2 based devices and circuits.

Here, we demonstrate for the first time a two terminal device based on a vdW heterostructure junction formed by VO2 and multilayer MoS2 flakes. Bulk-like MoS2 is an indirect band-gap semiconductor with a 1.3 eV band gap10,11, which forms a type II n-n heterojunction when in contact with VO2. The proposed device shows an excellent tunable rectification behavior and can be turned into a Schottky rectifier at high applied voltage or increased temperature by triggering the VO2 IMT. With respect to pure VO2 switches, this heterostructure delivers a lower leakage current in the diode subthreshold region and it could lead to the realization of more energy efficient VO2 devices. Compared to monolayer MoS2, multilayer flakes reduce the risk of damages due to the large power flowing in the devices when electrically actuated and provide lower contact resistance12. The fabricated devices proved to be photosensitive in the visible spectral range as a result of light absorption in the MoS2 side of the junction. We report here a complete photoresponse characterization including the impact of the thermally induced IMT of VO2. The extracted responsivities exceed the results reported for other photodiodes based on multilayer MoS2 flakes11,13,14.

Results

Electrical characterization of the heterojunction

The schematic of the conceptual vdW MoS2/VO2 heterojunction together with the basic biasing scheme used for the experiments are depicted in Fig. 1a: the MoS2 contact is grounded and the voltage is applied and swept on the metal contact on VO2. An optical microscopy image of a fabricated MoS2/VO2 heterojunction with the electrical contacts to the two sides of the junction is reported in Fig. 1b. Two separate set of gold contacts were deposited on the MoS2 flake to verify their ohmic behavior (see Supplementary Information Fig. S1a). The thickness of the exfoliated flakes has been estimated by atomic force microscopy, and the devices described in the following have been realized with flakes thicknesses ranging from 80 to 100 nm. The VO2 and the SiO2 films are respectively 75 nm and 2 µm thick. Details on device fabrication are reported in the methods section. The X-ray diffraction study and resistivity curve of as deposited VO2 film, which demonstrates the absence of other vanadium oxide phases potentially detrimental for the device operation, are shown in Supplementary Information Figs S2a and S2b.

Figure 1
figure 1

VdW MoS2/VO2 heterojunction schematic. (a) Three-dimensional schematic view of the MoS2/VO2 heterojunction with the bias configuration used in the experiments. The MoS2 contact is grounded while the bias is applied and swept on the metal contact on VO2. (b), Optical image of the fabricated heterojunction composed of a multilayer MoS2 flake and VO2 pre-patterned structure. Gold contacts to the two sides of the junction have been deposited to characterize the device. Two distinct set of contacts were deposited on the MoS2 flakes to verify their electrical behavior.

Full size image

The current voltage characteristic of a first device (D1) at room temperature is shown in Fig. 2a. The MoS2/VO2 heterojunction exhibits a clear rectifying behavior with a rectification ratio IF/IR larger than 103 and an extracted IF/IS exceeding 105, where IF, IR and IS are respectively forward, reverse and saturation current. Under positive bias, the current shows an exponential increase that becomes linear under larger voltages because of the series resistance. The behavior of the junction can be explained in terms of the qualitative band diagram in Fig. 2b (see Supplementary Information Fig. S3 for more information). Both MoS2 and the insulating phase of VO2 are intrinsically n-type materials15,16. Since MoS2 workfunction is smaller than VO2 one, at the formation of the heterostructure electrons are transferred from the first to the second, resulting in the depicted band alignment, Fig. 2b. The large conduction band discontinuity (ΔEC), estimated by the affinity rule to be roughly 1 eV, is the main responsible for the asymmetric conduction of the junction. Applying a positive bias to VO2, electron injection from MoS2 conduction band to VO2 conduction band is favored, resulting in an exponential increase of the current. Conversely, under reverse bias electrons are injected from VO2 to MoS2, but they face the large ΔEC barrier. Overall, the realized junction behaves as an n-n heterostructure with type II band alignment and is able to provide a good rectification. The device I–V characteristic in the exponential region can be approximated with the Shockley diode equation providing an extracted ideality factor of 1.75, while the non-saturating reverse current is better described by the Fang and Howard model17 specifically developed for n-n junctions and capable of analytically capturing the increase of diode current with the absolute value of the reverse bias (see Supplementary Information Fig. S4b,c).

Figure 2
figure 2

Electrical characteristic of the MoS2/VO2 heterojunction. (a) Electrical I–V double-sweep characteristic of device D1 at room temperature in semi-logarithmic (black) and linear (red) scales. The device presents a rectification behavior with no significant hysteresis. (b) Qualitative band diagram of the MoS2/VO2 heterojunction for the insulating and metallic phase of the functional oxide. The VdW gap at the junction is not represented and the polycrystalline nature of VO2 is not considered. (c) Evolution of the electrical characteristic of D1 with increasing temperature. Both forward and reverse current are boosted by the temperature increase, with a more pronounced enhancement in the window around the IMT temperature (60–80 °C). (d) Extracted conductance slope from the exponential regime of the forward current. The conductance slope increases across the VO2 IMT. Inset: conductance of device D1.

Full size image

The characterization of the heterojunction device in a wide range of increasing temperature is of particular interest because VO2 transition from the semiconductor to the metallic phase around 68 °C is expected to increase the conductance of the junction by collapsing the VO2 bandgap. The I–V curves measured with the temperature as a parameter are shown in Fig. 2c. Both forward and reverse current increase with temperature and a significant current enhancement is observed close to the IMT temperature. Above it, the temperature dependence of the characteristic is less pronounced and the device behaves as a Schottky rectifier. The band diagram of the junction with metallic VO2 is drawn in Fig. 2b. The Schottky barrier (ΦM) is estimated to be 0.3 eV larger than ΔEC, and the built-in voltage should increase slightly since VO2 work function is reported to increase across the IMT2. The reverse current increase with raising temperature can be explained with the growth of electron density in VO2 conduction band, changing from 1018 cm−3 in the insulating phase to 1023 cm−3 in the metallic one18. The boost of forward current can be attributed to the stronger thermionic injection in VO2 conduction band from MoS2 conduction band, but it is mitigated by the slight increase of the built-in voltage. The rectification ratio with metallic VO2 is slightly degraded with respect to the semiconductor junction but it remains larger than 103. In order to characterize the junction conduction we extracted the conductance slope19 \({S}_{G}={({\rm{\partial }}logG/{\rm{\partial }}V)}^{-1}\) for the forward current at different temperatures (Fig. 2d). At room temperature, the device exhibits an excellent minimum S G value of 120 mV/dec, which suggests that the heterojunction is abrupt and has a reduced density of interface defects. This observation is further supported by the absence of any hysteresis of the I–V characteristics (see Fig. 2a) in double-sweep measurements and by a low extracted quality factor of 1.75 in forward bias. The conductance slope increases across the MIT and tend to decrease after the phase transition.

Demonstration of electrically induced reversible IMT

The capability of electrically inducing the IMT of the VO2 layer in the heterostructure in a room temperature experiment is shown in Fig. 3. This measurement has been performed on another device (D2) with the same structure of D1. The biasing configuration used for the measurements is the same shown in Fig. 1b, with the addition of a series resistance RS of 1 kΩ to limit the maximum current. The I–V curve of the heterojunction (Fig. 3a) exhibits two steep jumps in correspondence of the IMT and MIT of VO2, separated by a hysteresis window. The voltage required to trigger the IMT is significantly larger with respect to the values that can be achieved in pure VO2 switches4 (see Supplementary Information Fig. S5). The high actuation voltage is due to the resistance of the heterojunction and to the relatively large distance between the MoS2 edge and VO2 metallic contact (more than 1 μm for D2). Remarkably, the proposed device delivers a lower leakage current in the diode subthreshold region with respect to pure VO2 switches. This feature combined with the possibility of modulating dynamically the IMT threshold by gating the MoS2 side of the junction could pave the way to more energy efficient VO2 switches. For the device of Fig. 3, the power thresholds required to electrically induce the IMT and MIT transitions from an n-n heterojunction to a Schottky diode and vice versa are respectively 1.83 mW and 2.89 mW. We verified that the actuation power density required to trigger the IMT did not alter the electrical behavior of the heterojunction by measuring the electrical characteristic of D2 before and after the electrical induced IMT. Figure 3b shows a direct comparison of the two I–V curves: no major variations are observed, demonstrating a full reversibility of the VO2 phase change and the stability of the heterojunction conduction.

Figure 3
figure 3

Electrically induced IMT of the VO2 side of the heterojunction. (a) Electrical characteristic of device D2 under large applied bias, sufficient to trigger the IMT of VO2 close to 17.5 V. Inset: biasing scheme used for the experiment. A discrete resistor RS of 1 kΩ is connected in series to the device to limit the current in the low resistance state. (b) I–V curve of D2 measured before and after having electrically induced the IMT of VO2. The comparison between the two curves shows that the phase change of VO2 and the actuation power has no permanent effect on the electrical behavior of the heterojunction.

Full size image

Photoresponse characterization

The photoresponse of device D1, whose dark I–V curve is reported in Fig. 2a, was characterized by illuminating the device with a focused light beam with controllable wavelength and power. The incident power density has been maintained well below the values required3 to optically trigger or affect VO2 IMT throughout all the measurements performed. The photosensitive area of the device corresponds to the overlap between the MoS2 flake and VO2 (see Fig. 1a), and it is estimated to be 104 µm2. Contributions to the photoresponse at zero applied bias from the MoS2 region far from the heterojunction have been found to be negligible coherently with the ohmic nature of the realized contacts (see Supplementary Information Fig. S1a,b). Moreover, the MoS2 region far from the junction shows a small contrast between the dark and illuminated IV curves measured across the set of contacts deposited on MoS2 (see Supplementary Information Fig. S1c). Figure 4a shows the impact of different illumination wavelengths with similar incident power density on the I–V curve of D1, resulting in a non-zero short circuit current ISC and open circuit voltage VOC. A schematic representation of the origin of the photovoltaic effect in the heterojunction is reported in Supplementary Information Fig. S6. The measured ISC shows a linear dependency with respect to the incident power density, as shown in Fig. 4b for illumination at 600 nm wavelength. We characterized the time domain photoresponse of the device by measuring the evolution of ISC in response to ON/OFF and OFF/ON transitions of the light source. The device shows a symmetrical rise and decay time and the extracted response time at room temperature is 3.5 ms (see Supplementary Information Fig. S7). This value is considerably smaller than the ones reported in literature for several other implementations of MoS2 based photodetectors20,21 but still significantly larger than the response time in conventional Si photodiodes21.

Figure 4
figure 4

Photoresponse of MoS2/VO2 heterojunction. (a) I–V characteristic of D1 in linear scale under different illumination wavelengths, from 500 nm to 800 nm. The measurements have been performed under an incident power density close to 330 nW/mm2 for all the wavelengths and at room temperature. (b) Short circuit current of the heterostructure as a function of the incident power density, showing a linear dependency. (c) Spectral evolution of the photoresponsivity measured at different temperatures. At room temperature, a responsivity larger than the one of conventional silicon p-n photodiode is measured in the spectral range 500/650 nm. The responsivity in the visible spectrum is boosted by the increase of temperature and it saturates for temperatures above the IMT of VO2. (d) Electrical power generated by the photodiode under 600 nm illumination with an incident power density of 1.48 μW/mm2. The electrical power that the heterojunction can effectively harvest decreases with the increase of the temperature.

Full size image

We characterized the impact of temperature on the wavelength resolved photoresponsivity R at zero applied bias, defined as the ratio between the photogenerated current and the incident power on the device (Fig. 4c). The cut-off wavelength is located between 750 and 800 nm and it does not change with temperature, suggesting that light absorption happens mostly in the MoS2 side of the heterojunction14. At room temperature, a maximum photoresponsivity of 1.25 A/W was measured at 550 nm, and values exceeding silicon photodiode performance have been obtained in the 500/650 nm range13. Responsivity values larger than one have already been reported in p-n junctions based on doped multilayer MoS2 flakes, and can be attributed to the efficient photocarrier separation operated by the built in voltage of the junction21,22. The photoresponsivity in the visible range is clearly boosted by the temperature rise and tends to saturate above the VO2 IMT temperature. This enhancement can be explained by different mechanisms. First, the increase of VO2 work function with temperature2,15 determines a larger built-in voltage and therefore a stronger electric field at the junction that provides a more efficient separation of photogenerated carriers. Secondly, the MoS2 depleted thickness could increase because of VO2 work function and carrier density boost18, resulting in an increase of the photosensitive volume. Moreover, another contribution could come from the variation of the optical properties of VO2 producing potential optical interference effects responsible for an enhanced light absorption. Indeed, recent studies reported that the photoluminescence monolayer MoS2 flakes on VO2 substrates increases upon heating VO2 above the IMT temperature, most likely because of constructive optical interference23,24.

We extracted the generated electrical power Pel as a function of the applied voltage at different temperatures under 600 nm illumination with an incident power density of 1.48 μW/mm2 (Fig. 4d). The increase of temperature produces a boost of ISC and a larger reverse current (Fig. 2c). This is related to a more favorable leakage path for photogenerated carriers, which causes a drop of VOC. Since VOC decreases at a faster rate than ISC grows (see Supplementary Information Fig. S8), the harvesting of the optical power becomes less efficient at high temperatures14. Therefore, the IMT of VO2 is beneficial for the heterostructure photosensitivity but appears detrimental for its photovoltaic performance.

Discussion

In conclusion, we investigated a novel two-terminal device based on a VdW n-n heterojunction between VO2 and multilayer MoS2 flakes. The demonstrated devices show a rectifying electrical behavior, with rectification ratio larger than 103, ideality factor of 1.75 and excellent forward conductance slope. The proposed heterojunction can be turned into a Schottky rectifier by electrically or thermally inducing the VO2 IMT. The electrically induced IMT is proved to be fully reversible and the MoS2/VO2 heterostructure grants a lower subthreshold leakage current with respect to pure VO2 switches. The envisioned possibility of modulating the IMT threshold voltage by gating the junction to electrostatically control the band alignment of the heterojunction could open a new category of VO2 based solid state devices with promising applications for steep slope electronic switches. Finally, we characterized the optical sensitivity of the heterojunction and obtained external photoresponsivities exceeding the results reported for several other implementations of photodiodes based on multilayer MoS2 flakes11,13,14. The fast transient response measured suggests that the device photoresponse is not dominated by trap-assisted optical absorption. The responsivity value is increasing with temperature and saturates above the VO2 IMT. The use of monolayer MoS2 for the realization of the heterostructure could result in a considerable enhancement of the photoresponsivity because of the direct band gap and the larger built-in voltage23,24,25. Large-area MoS2 synthesis techniques could allow a large scale, cheap fabrication of MoS2/VO2 heterojunction devices26,27. Moreover, recent reports of low-thermal budget deposition of VO2 thin films by pulsed laser deposition (PLD) are promising for the realization of CMOS compatible heterostructures28.

Methods

VO2 deposition

The 75 nm thick VO2 film was deposited on SiO2 substrate by reactive DC magnetron sputtering in high-vacuum conditions. The chamber was pumped to a base pressure lower than 5·10-8 mbar. The power on the V metal target (2 inch diameter, 99.95% purity) was set to 150 W. Ar process gas (purity 99.999%) was introduced in the chamber and the flow was regulated by a mass flow controller. During deposition, the oxygen pressure was kept constant by a Proportional Integral Derivative (PID) feedback control. This regulated the oxygen flow based on the pressure readings of a Zirox XS22 lambda-probe oxygen sensor. The process pressure was 7.45 ± 0.01·10-3 mbar and the oxygen partial pressure 4.86 ± 0.03·10-4 mbar. The temperature was measured by a stationary thermocouple above the rotating substrate holder and was kept constant at 600 °C. During deposition the substrate was rotating at 15 rpm. An in situ annealing was adopted during the slow cooling of the sample (30 °C/min).

MoS2 synthesis

MoS2 powder was synthesized by heating a mixture containing stoichiometric amounts of molybdenum (99.9% pure, Alfa Aesar) and sulfur (99.999% pure, Alfa Aesar) at 1000 °C for 7 days in an evacuated and sealed quartz ampule. The mixture was slowly heated from room temperature to 1000 °C for 12 h in order to avoid any explosion due to the strong exothermic reaction and the high volatility of sulfur. From this powder, MoS2 crystals were grown using chemical vapor transport (CVT) with iodine as transport agent at ca. 5 mg/cm3. All quartz tubes used for vapor transport typically have an inner diameter of 16 mm and a length of 20 cm. The total powder charge is 5 g. A very slight excess of sulfur is always included (typically 0.5 wt % of the charge) to ensure the stoichiometry in the resulting crystals. The excess of sulfur is not incorporated into the dichalcogenide crystals but condenses as elemental sulfur onto the wall of the quartz tube at the end of the CVT process. The source and growth zones were kept at 1060 and 1010 °C, respectively, for 7 days in evacuated and sealed quartz ampules. After this time the furnace is turned off, a small fraction of the charge is transported toward the colder end of the tube, forming crystals with diameters of about 2−8 mm and thick tens of microns. The resulting crystals were washed with acetone and dried in vacuum. X-ray diffraction study has shown that MoS2 obtained in this way belongs to the 2 H polymorphism. A full characterization of the synthetized bulk samples is presented elsewhere29.

Device fabrication

Electron beam lithography (EBL) on negative AZ nLof 2020:PGMEA 1:1 resist and wet etching in diluted commercial Cr etch solution were used to pattern the VO2 film sputtered on a 2 μm thick SiO2 layer. MoS2 multilayer flakes have been exfoliated on a PDMS stamp by using the scotch-tape micromechanical cleavage method and then deterministically transferred30 on the edge between VO2 pre-patterned structures and the SiO2 layer. The two-terminal devices were completed with a further EBL step and metal evaporation to deposit by lift-off 100 nm thick gold contacts on VO2 and MoS2. Gold was selected for the electrical connections in order to provide ohmic contacts to both VO2 and MoS2 (see Supplementary Information Fig. S1). The thickness of the transferred flakes was measured by atomic force microscopy, and the devices discussed in the text were realized with flakes thickness ranging from 80 to 100 nm.

Electrical and optical characterization

All the measurements were carried on in ambient atmosphere and at temperatures ranging from ambient value to 100 °C. DC electrical measurements were performed using a HP4156A Semiconductor Parameter Analyzer and a Cascade Summit probe station. The light source used for optical characterization is a SuperK EXTREME supercontinuum white light lasers with a series SuperK SELECT multi-line tunable filter to select the wavelength within a minimum bandwidth of +/−5 nm. The optical power was measured with an 818-SL/DB Silicon Photodetector. The light beam was focused on the sample through optical microscopy lenses, resulting in a spot size of 0.25 mm2.