Letter pubs.acs.org/NanoLett
Electrically Tunable Bandgaps in Bilayer MoS2 Tao Chu,†,‡ Hesameddin Ilatikhameneh,† Gerhard Klimeck,† Rajib Rahman,† and Zhihong Chen*,† †
School of Electrical and Computer Engineering and Birck Nanotechnology Center, Purdue University, West Lafayette, Indiana 47907, United States ‡ Globalfoundries Advanced Technology, Hopewell Junction, New York 12533, United States S Supporting Information *
ABSTRACT: Artificial semiconductors with manufactured band structures have opened up many new applications in the field of optoelectronics. The emerging two-dimensional (2D) semiconductor materials, transition metal dichalcogenides (TMDs), cover a large range of bandgaps and have shown potential in high performance device applications. Interestingly, the ultrathin body and anisotropic material properties of the layered TMDs allow a wide range modification of their band structures by electric field, which is obviously desirable for many nanoelectronic and nanophotonic applications. Here, we demonstrate a continuous bandgap tuning in bilayer MoS2 using a dual-gated field-effect transistor (FET) and photoluminescence (PL) spectroscopy. Density functional theory (DFT) is employed to calculate the field dependent band structures, attributing the widely tunable bandgap to an interlayer direct bandgap transition. This unique electric field controlled spontaneous bandgap modulation approaching the limit of semiconductor-to-metal transition can open up a new field of not yet existing applications. KEYWORDS: transition metal dichalcogenides (TMD), bilayer MoS2, dual gate FET, tunable bandgap, photoluminescence, interlayer transition
F
interaction in these 2D layered structures are the key to enable new functionalities and novel device operation mechanisms. It was experimentally8−14 and theoretically15−17 demonstrated that the band structure of bilayer graphene could be continuously altered by electric fields, transforming the zero bandgap material into a semiconductor with a bandgap up to ∼300 meV. However, the small bandgap size of bilayer graphene limits its use to the mid-infrared frequency range. On the other hand, 2D TMDs have inherent bandgaps.18,19 More importantly, the bandgaps of bilayer TMDs can be continuously reduced to zero by applying a vertical electric field across the two layers, as predicted by DFT calculations. This bandgap tuning is reversible and significantly larger than that of bilayer graphene.20 In this paper, we report the first experimental realization of the above-mentioned bandgap tuning in bilayer MoS2. The transfer characteristics of dual-gated bilayer MoS2 FETs show continuously increasing off-state currents with the increasing gate field, suggesting an electric field induced semiconducting to metallic transition. PL spectra of the same are also investigated. The prominent excitonic transition peak associated with the direct bandgap evolves into two resonances
or traditional bulk semiconductors, the bandgap is determined by the chemical composition and specific arrangement of the crystal lattices. Although it is desirable for many optoelectronic and electronic applications to have materials with continuously tunable bandgap available, sophisticated material growth systems are normally needed to design and engineer band structures with the preferred properties. For example, tremendous effort has been made on forming semiconductor alloys to adjust the bandgaps within the range of those of the associated compounds.1 However, lattice mismatch and variations in mixing ratios usually cause defects in the grown materials, making the material design very challenging. Another well-known artificial material synthesis effort is to create superlattices consisting of various thin semiconducting layers.2−4 Although energy gaps can be artificially defined by introducing new Brillouin zone boundary conditions, material deposition with atomic scale thickness control is required to fabricate these structures. Moreover, the bandgaps realized by the above methods are fixed for a specific device/material sample and are not dynamically tunable during device operation. In the past decade, 2D layered materials including graphene and TMDs have sparked tremendous interest in the scientific community, owing to their unique material properties and potential in various applications.5−7 Among many newly discovered properties that do not exist in bulk materials, the strong in-plane bonding and weak van der Waals interplanar © XXXX American Chemical Society
Received: August 12, 2015 Revised: November 8, 2015
A
DOI: 10.1021/acs.nanolett.5b03218 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 1. (a) Optical and AFM image of a bilayer MoS2 flake. Scale bars are 1 μm. (b) Raman spectra of a single and bilayer MoS2. (c) Photoluminescence spectra of pristine single and bilayer MoS2 without electric field. (d) Optical image of two MoS2 devices (inset is the illustration of the device structure).
under the application of an electric field, showing a wide red shift of the lower energy resonance. The impact of the electric field on the band structure of bilayer MoS2 and corresponding PL spectrum is studied using ab initio simulations. As discussed in the last section of the paper, the simulation results are in good agreement with the experimentally extracted bandgaps, further verifying our analysis. In our experiments, MoS2 flakes are mechanically exfoliated from bulk MoS2 crystals onto 90 nm SiO2 substrates with heavily doped Si underneath serving as the back gate electrode. The selected oxide thickness enhances the optical contrast of MoS2 flakes on the substrate, as shown in the top left image of Figure 1a. The layer number is confirmed by atomic force microscopy (AFM); a step height of 1.4 nm is expected for bilayer MoS2. Single and bilayer samples are also verified by Raman spectroscopy,21,22 as shown in Figure 1b. A Raman shift of 22 cm−1 between the E12g and A1g modes in the bilayer is apparently different from the 19 cm−1 shift in the single layer. PL spectrum shows an indirect bandgap peak in the bilayer case, in contrast to the direct bandgap nature of the single layer (Figure 1c). An optical image of two dual-gated FETs fabricated on a single and a bilayer MoS2 flake (see the Method Section) is presented in Figure 1d. Both the top and the bottom gates are used to apply vertical fields to the MoS2 channel, simultaneously tuning the field strength and charge carrier density. Transfer characteristics as a function of the top gate voltage (VTG) at various back gate biases (VBG) for a single and a bilayer MoS2 device are compared in Figure 2. Figure 2a illustrates the expected ambipolar characteristics for an Schottky barrier (SB) FET, exhibiting a high current electron (n) branch and a low current hole (p) branch for the case of the contact metal work function having a closer line-up to the
conduction band (CB). The device off-state is defined at the point where the two branches intersect. Because single-layer MoS2 has a larger bandgap than bilayer,7,23 lower off-state currents are expected if the noise floor of the instrument or the leakage current through the gate oxide does not limit the resolution of the measurement. In practice, the off-state of a single layer MoS2 device is too low compared to the noise floor and the gate leakage current (∼fA, shown in Figure 2b). Hence, we cannot read the off-state currents directly but are able to identify the off-state intersection through an artificial extension of the p/n branches (dotted lines in Figure 2b). On the other hand, the bilayer device shows well-defined off-states above the noise floor because of the smaller bandgap, as shown in Figure 2c. Given the identical device structure, the transfer characteristics of both devices follow the same electrostatics dictated by the top and bottom gates and shift to the left with increasing VBG, as expected. The slope of the constant current line from the VTG vs VBG plot (Figure 2d) is consistent with the ratio of εBGdTG/ εTGdBG, where dBG = 90 nm, εBG = 3.9 for the SiO2 back gate dielectric, and dTG = 12 nm, εTG = 12 for the HfO2 top gate dielectric. Further evaluating the minimum current levels for the two devices, we notice that the off-state of the bilayer quickly increases with VBG, whereas it remains almost a constant for the single layer. Because of the top gate dielectric’s break down limit (VTG ≥ −6 V), we cannot obtain the p branch characteristics for the large positive back gate voltage curves (VBG ≥ 30 V) to identify the off-states for the bilayer device anymore, but the trend described above is clear and continuous, that is, with the back gate voltage changing from VBG = −40 V to VBG = +20 V, the device off-state increases from 9 × 10−14 A to 3 × 10−11 A. The average vertical displacement field, Dav, of the top and bottom gates, can be expressed as Dav = (1/ 2)[εSiO2(−(VBG − VB0))/tSiO2 + εHfO2(VTG − VT0)/tHfO2]. VB0 B
DOI: 10.1021/acs.nanolett.5b03218 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 2. (a) Expected ambipolar transfer characteristics for an SB FET with its contact metal work function lined up closer to CB. The threshold voltages for both the electron (n) and the hole (p) branches are labeled. The device off-state is defined at the intersection of the n branch and p branch. (b) Transfer characteristics for a dual-gated single layer MoS2 device, measured at VBG = −50 V to 50 V with a step of 10 V. The device off states are identified by the dotted lines. (c) Transfer characteristics for a dual-gated bilayer MoS2 device, measured at VBG = −50 V to 60 V with a step of 10 V. The off-state current increases with increasing VBG. (d) A set of (VTG, VBG) to obtain a constant current of IDS = 10 × 10−4 μA/μm for the device from (b).
are below the noise floor, as shown in the Supporting Information. As mentioned above, in this case we can only identify the bilayer device off-states by extending the p/n branches similarly to the single layer case. However, regardless of which gate dielectric is being used, the same trend of increasing off-state current with increasing electric field is always observed in bilayer MoS2 devices−suggesting a robust phenomena. To extract the size of the bandgap, we utilize the threshold voltages of the ambipolar characteristics. As shown in Figure 2a, the top band diagram shows a line-up between the source Fermi level and the CB of the MoS2 channel, which is defined as the threshold voltage for the n branch, Vn‑th (labeled as n-Vth in the figure). Similarly, at the threshold voltage for the p branch, Vp‑th (labeled as p-Vth), the Fermi level at the drain is aligned with the valence band (VB) of the channel. Simply, the bandgap can be estimated using the following formula:
and VT0 are the offset voltages due to unintentional doping. VB0 = −40 V and VT0 = −2.3 V are roughly identified at the lowest off-state from Figure 2c. This implies that the dual-gate bilayer MoS2 device under a zero external field is unintentionally doped, which results in a built-in field between the two MoS2 layers. To compensate for this doping, VB0 and VT0 need to be applied to reach charge neutrality, under which the net electric field is close to zero and the extracted bandgap approaches the value expected for pristine bilayer MoS2. We speculate that doping originates from the trap charges in the substrate or is introduced during the ALD process. Another interesting observation is that, when VBG = −50 V is applied, the offstate increases as in the VBG = −30 V case (VBG = −40 V is the zero field case due to unintentional doping and its off-state is the lowest), suggesting that the bandgap reduction in bilayer MoS2 is independent of the vertical field polarity. It is apparent from our experiment that the off-state current of the bilayer MoS2 device continuously increases with increasing |Dav|, indicating an electric field controlled bandgap engineering and ultimately a semiconductor-to-metal transition. Please note, compared to most MoS2 device characteristics published in literature showing only n branches,6 our devices from Figure 2b and c show true ambipolar characteristics with noticeable p branches. The scaled top gate dielectric (12 nm HfO2) employed in our devices ensures thin enough Schottky barriers to allow sufficient hole current injection.24 When thicker dielectric (20 nm HfO2) is used, the p-branch currents become so small that even the off-states of the bilayer devices
⎛ Vn‐th − Vp‐th ⎞ Eg = e⎜Vds + ⎟ β ⎝ ⎠
(1)
where Vds is the drain-to-source voltage, β is the band movement factor: β = 1 + (CT/Cox) . CT and COX are the interface trap capacitance and oxide capacitance, respectively. Figure 3a and b are a set of selected transfer characteristics from Figure 2 for the single and bilayer MoS2 devices, respectively. Close to the device off-states, the subthreshold swing (SS = dVGS/log(IDS)) of the n branch equals β × 60 mV/ decade. Hence, β can be extracted for individual devices from C
DOI: 10.1021/acs.nanolett.5b03218 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 3. Extraction of bandgap change by threshold voltages. (a) Transfer characteristics of the single layer MoS2 device at VBG = −50, −40, and −30 V. (b) Transfer characteristics of the bilayer MoS2 device at VBG = −10, 0, and 10 V. (c) Extracted bandgap change as a function of Dav, for both the single and bilayer MoS2 devices. An approximately 260 meV bandgap reduction per 1 V/nm displacement field is extracted for bilayer MoS2, whereas a constant bandgap is obtained for the single layer. (d) Output characteristics of a bilayer MoS2 device at VBG = 60 and 10 V.
To further confirm the bandgap tuning in bilayer MoS2 and understand its mechanism, we perform PL measurements under electric field and compare the results with DFT calculations. As discussed earlier, two peaks are observed in the PL spectrum of bilayer MoS2 (Figure 1c). One is associated with the direct bandgap luminescence at ∼1.83 eV, whereas the other one at ∼1.52 eV corresponds to the indirect bandgap. Unlike the direct bandgap in single layer MoS2, the indirect bandgap is responsible for the transport properties of bilayer MoS2 presented in Figures 2 and 3. However, the associated PL peak easily gets suppressed after the top gate stack fabrication. A very weak indirect peak can barely be detected from the dualgated bilayer MoS2 FET structure, as the blue curve shown in the inset of Figure 4c. As an alternative, we focus on the direct bandgap peak in the optical study here. It is known that electron−electron interactions are very strong in MoS2 thin films leading to strong exciton binding energies, which is what the PL spectroscopy probes.25,26 The prominent peak at ∼1.83 eV is normally referred as “A exciton peak”.21,27 It is also found in our fabrication process that the top gate stack can artificially alter the PL spectra. Therefore, a special gate dielectric stack with a buffer layer accompanied by a graphene top gate is used to preserve the optical signal in the device for PL measurements (see Method Section), as shown in Figure 4a and b. Figure 4c compares the PL spectra of a dual gated bilayer MoS2 device at various displacement fields: Dav = 0, 0.3, 0.6, and 1.2 V/nm. Because the PL intensity of the A exciton peak can be switched off by doping and Stokes shift is also found to increase with the doping level, all four PL measurements are
the experimentally measured SS. For the bilayer device in Figure 3a, β = 2.7 is obtained from SS = 164 mV/decade, whereas for the single layer device in Figure 3b, SS = 247 mV/ decade and β = 4.1. Unfortunately, we cannot quantitatively determine the bandgap size using eq 1, because Vp‑th is beyond the measurement window and cannot be identified as Vn‑th, for both single and bilayer devices. Nevertheless, we can determine the bandgap change, Eg − Eg0, as a function of VBG, by measuring the change of the threshold voltages, ΔVn‑th and ΔVp‑th. As explained above, Eg0 is the bandgap of the pristine bilayer MoS2 defined at VBG = −40 V, at which the internal field is zero. Plugging all the values into eq 1, Eg − Eg0 ≈ 0 eV is extracted for the single layer device, independent of VBG. In contrast, Eg − Eg0 ≈ 100 meV is extracted at VBG = −30 V for the bilayer MoS2 device. Obviously, the bandgap is no longer a constant material parameter for bilayer MoS2. Instead, it reduces ∼700 meV as VBG changes from −40 V to +20 V. Figure 3c shows Eg − Eg0 as a function of Dav for the bilayer device with a slope of ∼260 meV per 1 V/nm, compared with the constant bandgap in the single layer device. In principle, a zero bandgap can be approached in bilayer MoS2 at a larger field before the dielectrics’ break down. The bandgap reduction in the bilayer is also evident from the output characteristics at different VBG, as shown in Figure 3d. Both the current modulation and saturation level show drastic difference between the characteristics measured at VBG = 60 V (small bandgap) and 10 V (large bandgap). D
DOI: 10.1021/acs.nanolett.5b03218 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 4. (a) Illustration and (b) SEM image of dual-gated bilayer MoS2 devices with graphene top gates. (c) Photoluminescence of bilayer MoS2 at displacement field of 0, 0.3, 0.6, and 1.2 V/nm. Dotted lines are the Lorentzian peak fitting. The fitting error is
