The non-volatile electrostatic doping effect in MoTe2 field-effect transistors controlled by hexagonal boron nitride and a metal gate.

The electrical and optical properties of transition metal dichalcogenides (TMDs) can be effectively modulated by tuning their Fermi levels. To develop a carrier-selectable optoelectronic device, we investigated intrinsically p-type MoTe2, which can be changed to n-type by charging a hexagonal boron nitride (h-BN) substrate through the application of a writing voltage using a metal gate under deep ultraviolet light. The n-type part of MoTe2 can be obtained locally using the metal gate pattern, whereas the other parts remain p-type. Furthermore, we can control the transition rate to n-type by applying a different writing voltage (i.e., - 2 to - 10 V), where the n-type characteristics become saturated beyond a certain writing voltage. Thus, MoTe2 was electrostatically doped by a charged h-BN substrate, and it was found that a thicker h-BN substrate was more efficiently photocharged than a thinner one. We also fabricated a p-n diode using a 0.8 nm-thick MoTe2 flake on a 167 nm-thick h-BN substrate, which showed a high rectification ratio of ~ 10-4. Our observations pave the way for expanding the application of TMD-based FETs to diode rectification devices, along with optoelectronic applications.

Introduction

Graphene, which is one of the most intriguing two-dimensional (2D) materials for electronic applications due to its high electron mobility, flexibility, thermal conductivity, large surface area, and impermeability to gases, has been studied extensively over last two decades[1-8]. Despite its many merits for use in electronic materials, the application of graphene for switching devices is restricted due to its gapless nature in the pristine state[9]. However, transition metal dichalcogenides (TMDs), which are 2D semiconductor materials, exhibit a wide range of doping and band structure dynamics, allowing them to be used in a broad range of optoelectronics and nanoelectronics[10-13]. TMDs are composed of atomic layers bound together by van der Waals forces[14] and have good electronic transport channels with minimal scattering centers because they do not possess any interlayer covalent bonds[15,16]. Therefore, the bandgap and atomically thin layered structure of 2D TMDs render them a viable material for the active channel of field-effect transistor applications, such as ultra-fast photodetectors[17], electro- and photo-catalysis, supercapacitors[18], biosensors, energy storage devices, and memory devices, among others [19-23].

In the context of TMDs for use in electronic material applications, MoTe2 has gained significant interest owing to its fascinating semiconducting, metallic, and superconducting characteristics[24-27]. The direct bandgap of MoTe2 varies between 0.88 and 1.1 eV depending on the lattice configuration and number of layers[28-30]. In addition, since the bandgap of MoTe2 is significantly smaller than those of MoS2[31,32] and WSe2[33,34], MoTe2 is a good candidate for optoelectronic devices that provide a response covering the near-infrared wavelength region[35]. Moreover, in comparison to sulfur-terminated TMDs, Fermi-level pinning at the MoTe2-metal interface is significantly weaker[36]. Despite the narrow energy bandgap of this material, numerous methods have been reported for band modulation and control of the charge carrier polarity[37,38]. On the other hand, hexagonal boron nitride (h-BN), an insulating 2D material, has recently attracted growing interest because of its mechanical robustness, its exceptional thermal conductivity due to its strong BN covalent bonds, and its donor/acceptor-like defect states that control the doping mechanism[39,40].

It is therefore important to develop an efficient doping method for 2D TMDs to promote their application in semiconducting electronic applications. In this context, the doping of MoTe2 can be categorized into two types. The first method employs local electrostatic gating, which has been used successfully to create a p–n junction by polarizing a local area where the charge carrier type is opposite to that in other parts of the MoTe2 flake[41,42]. Although this technique is extremely adaptable, it is particularly volatile when gate voltage is turned off. The second method consists of atomic doping and surface modification using physical and chemical processes[43,44]. These processes permanently transform the material; however, but p-type and n-type doping are difficult to combine in the local areas of a single device. There is another way to to manupolate the carrier’s type in MoTe2. This method involves metal contacts engineering, which makes use of low and high work function metal electrodes[45,46]. For example, platinum, which is a high work function metal, has been used for as a source and drain contact and ambipolar MoTe2 was converted into a unipolar p-type field-effect transistor (FET)[47]. However, the unipolar n-type transport of MoTe2 is exceedingly difficult to accomplish owing to Fermi-level pinning and a limited variety of low work function metals. Thus, to modulate the carrier type and concentration in MoTe2, the development of a stable, nonvolatile, and controlled technique is necessary to adjust the properties of MoTe2 from the broad perspective of electronic devices.

Here, we present a promising strategy to address the aforementioned difficulties. More specifically, we employ a localized metal gate on a specific region of MoTe2, wherein h-BN is used as a dielectric material in the metal gate, and its thickness plays a vital role in the electrostatic doping of MoTe2. One region of the MoTe2 is placed on an h-BN substrate with a localized metal gate underneath, while the other region is placed on h-BN without a gate to allow control of the gate effect on a specific region of the MoTe2. Subsequently, illumination with deep ultraviolet (DUV) light is carried out to induce charge transfer to the defect states of h-BN with the localized metal gate underneath. Then, h-BN with charged defect states functions as a gate electrode to cause electrostatic doping of the localized MoTe2 region. We also investigate the characteristics of p–n diodes consisting of p-MoTe2 and n-MoTe2, which are fabricated using h-BN and a metal gate.

Results and discussion

Photo-induced doping effect of h-BN/MoTe2 FET

h-BN and MoTe2 nanoflakes were fabricated using adhesive tape and a conventional mechanical exfoliation process, and the dry transfer technique was used to prepare stacks of the h-BN/MoTe2 heterostructures. Figure 1a,b show a schematic diagram and an optical microscope image of the h-BN/MoTe2 heterostructure-based FET, respectively. We also examined the 2D flakes using Raman spectroscopy, which is a non-destructive and precise technique for determining the strain effect, thermal conductivity, band structure, and adsorption of chemicals on material surfaces[48-50]. To prevent the heating effect, the Raman spectra were recorded at room temperature using a laser with a wavelength of 514 nm and a low laser power of 1.0 mW. Figure 1c shows the Raman spectra of MoTe2 and three peaks assigned to A1g (174.63/cm), E12g (237.87/cm), and B12g (291.97/cm). The Raman spectra of h-BN are provided in supplementary information Fig. S1, where we observed a dominant E2g peak (1364.47/cm). Figure 1d shows the topographical atomic force microscopy (AFM) image and height profile of the h-BN/MoTe2 heterostructure, indicating that the thicknesses of the h-BN and MoTe2 components were 2 and 0.8 nm, respectively.

Figure 1

(a) Schematic diagram of an h-BN/MoTe2 FET. (b) Optical image of an h-BN/MoTe2 FET. (c) Raman spectrum of MoTe2. (d) AFM image and height profile of an h-BN/MoTe2 FET.

The charge carrier type of a TMD plays an important role in the interface resistance between the contact metal and semiconductor. Pristine MoTe2 can be either ambipolar or unipolar, being n-type or p-type, depending on its natural doping state[36,51-56]. We found that our thin MoTe2 flakes were p-type in the pristine state. Thus, we initially fabricated a thin layer of MoTe2 (0.8 nm) on a thick h-BN layer (167 nm). A Si/SiO2 substrate was employed, where Si was degenerately doped for use as the back gate. The AFM images and height profiles of both h-BN and MoTe2 are shown in supplementary information Fig. S2. Pristine MoTe2 (0.8 nm) was found to exhibit p-type behavior, as shown in the transfer curves (Ids − Vg−m) and (Ids − Vg−Si) given in Fig. 2a and supplementary information Fig. S3a, respectively. During the transfer curve measurements, which were performed in a vacuum, the drain-source voltage (Vds) was fixed at 0.5 V. In addition, we have investigated the output characteristics of pristine thin p-type MoTe2 and found that I–V curves are nonlinear as shown in Fig. S3b, which indicates the existence of a Schottky barrier between thin MoTe2 and metal contact (Cr/Au). Subsequently, the photo-induced doping effect was investigated when h-BN/MoTe2 was illuminated by DUV for various time intervals with the application of a writing voltage (Vw.v) ranging from − 2 to − 10 V, as shown in Fig. 2a. The writing voltages are applied through a localized metal gate (Cr/Au, 3/13 nm) to fill or deplete electrons in the defect sites of the h-BN layer with the help of a DUV in a vacuum. To achieve this photo-induced doping effect, the use of both a DUV and a writing voltage are essential[57]. Figure 2a shows a pristine MoTe2 FET on h-BN that was initially p-type, but that had been converted into n-type by DUV illumination and the application of a writing voltage. Initially, the application of − 2 V writing voltage under DUV light illumination resulted in a change in the polarity of the pristine MoTe2 from p-type to n-type, as shown in Fig. 2a. Upon further increasing the writing voltage, the MoTe2 region above the localized metal gate became completely n-type at a − 10 V writing voltage[58-60]. In addition, higher writing voltages resulted in more positive charges on the h-BN flake, which eventually provided an additional positive gate voltage. This photo-induced doping effect of MoTe2 can be attributed to a mechanism involving the electron depletion of donor-like defects in the h-BN flakes, which are generated by the negative gate voltage upon DUV optical excitement[61,62]. The depleted electrons enter the conduction band of the h-BN and then transfer to the MoTe2, leaving positively ionized defects inside the h-BN layer, which can be observed under an external electric field (V). Consequently, these positively charged donor-like defects in the h-BN resulted in the electrostatic doping effect of MoTe2.

Figure 2

(a) Transfer characteristics of MoTe2 (0.8 nm) FET on a 167 nm-thick h-BN substrate before and after photo-induced doping under DUV illumination (5 min) with writing voltages ranging from − 2 to − 10 V. (b) Transfer characteristics of the MoTe2 (2.4 nm) FET on a 42 nm-thick h-BN substrate. (c) Transfer characteristics of the thin MoTe2 (1.6 nm) FET on a 2 nm-thick h-BN substrate. (d) Electron mobility and carrier concentration of the MoTe2 (0.8 nm) FET on a 167 nm-thick h-BN substrate after photo-induced doping with different metal gate voltages.

Effect of the h-BN thickness

To investigate whether donor-like defects exist at the h-BN/MoTe2 interface or inside the h-BN itself, we measured the photo-induced doping characteristics of MoTe2 films with various h-BN thicknesses. If the photo-induced doping rate is proportional to the h-BN thickness, then it can be assumed that the defects exist inside the h-BN body; however, if the photo-induced doping effect originates from defects at the interface, it must be independent of the h-BN thickness. Thus, we fabricated thin (0.8–2.4 nm) MoTe2 FETs with different h-BN thicknesses to reveal the role of the h-BN thickness in the photo-induced doping effect. The transfer characteristics of the MoTe2 (2.4 nm)/h-BN(42 nm) heterostructure were measured with a drain-source voltage of 0.5 V and a metal gate voltage from − 4 to + 4 V as shown in Fig. 2b. AFM confirmed the thicknesses of the MoTe2 and h-BN layers, as shown in supplementary information Fig. S4. As the writing voltage was increased from − 2 to − 10 V, the polarity of MoTe2 changed from p-type towards n-type, but it did not convert completely to n-type, remaining ambipolar. Similarly, the transfer characteristics of another MoTe2 (1.6 nm thickness) FET on a thin (2 nm) h-BN layer were evaluated and are shown in Fig. 2c. In this case, we also observed that the p-type pristine MoTe2 did not completely change its polarity to n-type and again remained ambipolar. The photo-induced doping effect rates in Fig. 2b,c are in contrast to those in Fig. 2a, where the underlying h-BN flakes are particularly thick. Our findings therefore imply that photo-induced doping in h-BN/MoTe2 heterostructures is attributed to the optical stimulation of electronic states within the h-BN layer, and the thickness of this h-BN layer plays an important role in determining the extent of photo-induced doping. It is also possible that donor-like defect states may exist at different depths inside the h-BN flakes; Fig. S5 in supplementary information shows a schematic representation of the remaining positive defects in thin and thick h-BN layers after DUV illumination with the application of a writing voltage. Since DUV is illuminated from the top side of the h-BN flake, the positive defects are found more in the upper part of the h-BN flake. To compare the photo-induced doping effect at different writing voltages, we estimated the carrier density of the MoTe2 FET. The charge-carrier concentration (ne) can be calculated as follows[63]:

where Vth is the electron transport threshold voltage, Vg−m is the metal gate voltage, and e is the charge of an electron (1.602 × 10−19 C). The capacitance value (Cg) of h-BN per unit area can be calculated as Cg = ε0 εr/d, where d is the thickness of the h-BN layer, ε0 is the vacuum permittivity, and εr is the dielectric constant of h-BN. Figure S6a in supplementary information shows the gate capacitance vs frequency graphs for different thicknesses of h-BN, which demonstrates that the capacitance decreases with an increasing h-BN layer thickness as shown in Fig. S6b. Figure 2d shows the electron carrier concentration ne after photo-induced doping under the application of a writing voltage (Vw.v) in combination with DUV for a MoTe2 (0.8 nm) FET on a thick (167 nm) h-BN layer. The carrier concentration (ne) was estimated at Vg−m =  + 4 V after photo-induced doping. Similarly, we estimated ne at Vg−m = 0 V as shown in Fig. S6c, which shows a similar behaviour but the number of charge carriers is less as compared to ne at Vg−m = + 4 V. In addition, we calculated the field-effect mobility of the MoTe2 FET using the following equation.

where W is the channel width, L is the channel length, and represents the slope of the linear part of the transfer characteristics of the MoTe2 FET at an applied Vds of 0.5 V. Figure 2d shows the mobility of the MoTe2 (0.8 nm) FET on a thick (167 nm) h-BN layer after the application of a writing voltage Vw.v in combination with DUV. Furthermore, the photo-induced doping effect was found to be stable for several days. The MoTe2 FET demonstrated a stable n-type doping effect as shown in supplementary information Fig. S7a.

Effect of the MoTe2 thickness

We also investigated the dependence of the MoTe2 flake on the photo-induced doping effect. For this purpose, two different thicknesses of MoTe2 flakes were placed on an h-BN layer, and the transfer curves were measured after photo-induced doping with various writing voltages. Figure 3a shows the transfer curves of the MoTe2 (6.4 nm) FET on h-BN (160 nm), which exhibits ambipolar behavior in the pristine state. Further, we have examined the output characteristics of pristine thick n-type MoTe2 and found that I–V curves are nonlinear as shown in Fig. S8. However, the writing voltage was increased from − 2 to − 10 V, the n-type characteristics of the MoTe2 FET were enhanced after photo-induced doping. For comparison, we examined the photo-induced doping effect in a thicker MoTe2 (46 nm) FET on h-BN (165 nm), as shown in Fig. 3b. The transfer curve indicated the n-type characteristics of the pristine MoTe2 FET, and it was observed that the photo-induced doping treatment enhanced the n-type properties. More specifically, the pristine MoTe2 flake exhibited p-type characteristics when its thickness was < 2.4 nm, as shown in Fig. 2a–c, whereas the thick (46 nm) MoTe2 flake exhibited n-type characteristics in the pristine state. These results indicate that a p–n junction can be formed in thin MoTe2 flakes using a combination of photo-induced doping treatment and a local metal gate.

Figure 3

(a) Transfer characteristics of the MoTe2 (6.4 nm) FET on a 160 nm-thick h-BN substrate. (b) Transfer characteristics of the MoTe2 (46 nm) FET on a 165 nm-thick h-BN substrate.

MoTe2 p–n diodes with different h-BN thicknesses

Subsequently, we employed the photo-induced doping technique to prepare n-type regions at local regions of thin MoTe2 flakes (0.8–2.4 nm thick) on h-BN mounted on metal gates, while the other regions remained p-type, similar to the pristine state of MoTe2. Although the DUV illuminated the entire area of the h-BN layer, only the donor-like defects at the local regions over the metal gate could be charged. Consequently, a p–n diode was obtained in the MoTe2 flake between terminals S and D2, as shown in Fig. 1a. In addition, Fig. 4a shows the output characteristics of the MoTe2 p–n diodes with different h-BN thicknesses after photo-induced doping; the inset of Fig. 4a shows the log scale Ids − Vds curves, indicating the rectification characteristics. Since the photo-induced doping rate of MoTe2 depends on the thickness of the h-BN layer (see Fig. 2), the function of the p–n diode is expected to also be dependent on this thickness. Figure 4b shows the rectification ratio (RR) of the MoTe2 p–n diode for different h-BN thicknesses, where the RR is defined by Ion at Vds = + 5 V divided by Ioff at Vds = − 5 V. The highest RR value (~ 1.5 × 103) was found for the MoTe2 flake mounted on the thickest h-BN layer (167 nm). We also investigated the MoTe2 p–n diode characteristics for different thicknesses of MoTe2 flakes. Thus, Fig. 4c shows the output characteristics of the MoTe2 p–n diodes with various thicknesses of MoTe2, and the inset shows the Ids−Vds curves on a logarithmic scale. As expected, diode characteristics were generally not observed in MoTe2 flakes with thicknesses > 16 nm due to the fact that the majority of the flakes will be in the n-type state (i.e., that of the pristine state). As shown in Fig. 4d, a higher RR was achieved for thinner MoTe2 flakes.

Figure 4

(a) Output characteristics of the MoTe2 p–n diodes on h-BN substrates of different thicknesses, where the thicknesses of the MoTe2 flakes ranged from 0.8 to 2.4 nm. (b) Rectification ratio of the MoTe2 p–n diodes on h-BN substrates of different thicknesses, where the thicknesses of the MoTe2 flakes ranged from 0.8 to 2.4 nm. (c) Output characteristics of the MoTe2 p–n diodes for MoTe2 flakes of different thicknesses. (d) Rectification ratio of the MoTe2 p–n diodes for MoTe2 flakes of different thicknesses, where the thicknesses of the h-BN flakes ranged from 160 to 167 nm.

Following our examination of the photo-induced doping effect with a negative writing voltage of the metal gate, which mainly relies on the presence of donor-like defects in the h-BN layer, we moved on to address the possibility of reverse photo-induced doping. For this purpose, a MoTe2 (0.8 nm) FET on an h-BN layer (167 nm) was subjected to DUV illumination for 5 min with a positive writing voltage for the metal gate, as shown in supplementary information Fig. S7b. The same system was used in combination with a writing voltage of − 10 V before starting the experiment, and reverse photo-induced doping was investigated with positive writing voltages ranging from + 2 to + 10 V. It was found that the transfer curve changed toward p-type as the writing voltage increased, but it remained more like n-type even with the highest writing voltage of + 10 V. It should also be noted here that the density of acceptor-like defects was lower than that of the donor-like defect states in the h-BN layer.

Materials and methods

Fabrication of MoTe2 field-effect transistors on h-BN

The natural bulk crystals of h-BN and MoTe2 were provided by HQ graphene. Using adhesive tape in a cleanroom environment, the mechanical exfoliation method was used to obtain ultrathin nanoflakes of h-BN and MoTe2 from their bulk forms. A photoresist (SPR) and ethyl lactate (EL) were spin-coated onto Si/SiO2 (SiO2: 300 nm) substrates in the initial stage of the photolithography process. Subsequently, the obtained patterns were exposed to oxygen plasma for 5 min to eliminate the SPR and EL residues. A thermal evaporator was then used to evaporate Cr/Au (3/30 nm) for the large patterns, while the bottom electrode composed of Cr/Au (3/13 nm) was fabricated using conventional e-beam lithography and thermal evaporation techniques. Subsequently, a large h-BN flake was dry-transferred onto the top of the bottom electrode, while the other remainder was present on the Si/SiO2 substrate. The MoTe2 flake was then transferred onto the h-BN layer using a micromanipulator, as shown in Fig. S9 in supplementary information. At the end of the transfer procedure, the substrate was placed on a hot plate at 90 °C to eliminate vapor from the external surfaces and interfaces. After each transfer process, the samples were cleaned with acetone and methanol, and finally dried under a flow of N2 gas. The source/drain electrodes were fabricated using conventional e-beam lithography. Finally, Cr/Au (10/80 nm) metal contacts were deposited using a thermal evaporation technique.

Photo-induced doping and measurements

For the photo-induced doping treatment, the MoTe2 FETs on h-BN were illuminated by DUV light (λ = 220 nm, 11 mW cm−2). Optical microscopy and Raman spectroscopy were used to examine the MoTe2 flakes, and their thicknesses were measured by AFM. The electrical transport properties were measured in a vacuum using a source meter (Keithley 2400) and a picoammeter (Keithley 6485).

Conclusion

We herein reported the fabrication of MoTe2 field-effect transistors (FETs) on hexagonal boron nitride (h-BN) with a localized metal gate and found that the photo-induced doping treatment was most effective for thinner MoTe2 flakes mounted on a thicker h-BN layer. The use of a negative writing voltage under deep-ultraviolet (DUV) illumination induced n-doping of the MoTe2 FET, while the use of a positive writing voltage under DUV illumination induced p-doping; this difference was attributed to the donor- and accepter-like defects present in the h-BN. In addition, it was found that the photo-induced doping effect became stronger as the writing voltage was increased. Furthermore, a negative writing voltage resulted in a stronger doping effect than a positive writing voltage, which indicates that donor-like defects are more dominant than acceptor-like defects in the h-BN. These observations clearly demonstrate the success of this selectable local doping technique, which is applicable as a post-fabrication treatment method.

Supplementary Figures.