Semiconducting two-dimensional crystals are currently receiving significant attention because of their great potential to be an ultrathin body for efficient electrostatic modulation, which enables to overcome the limitations of silicon technology. Here we report that, as a key building block for two-dimensional semiconductor devices, vertical p–n junctions are fabricated in ultrathin MoS2 by introducing AuCl3 and benzyl viologen dopants. Unlike usual unipolar MoS2, the MoS2 p–n junctions show ambipolar carrier transport, current rectification via modulation of potential barrier in films thicker than 8 nm and reversed current rectification via tunnelling in films thinner than 8 nm. The ultimate thinness of the vertical p–n homogeneous junctions in MoS2 is experimentally found to be 3 nm, and the chemical doping depth is found to be 1.5 nm. The ultrathin MoS2 p–n junctions present a significant potential of the two-dimensional crystals for flexible, transparent, high-efficiency electronic and optoelectronic applications.
Since the rediscovery of stable monolayer graphite or graphene, two-dimensional (2D) layered materials or van der Waals materials have led to remarkable interest in the physics and applications of the materials1, 2, 3. Graphene provides a variety of fascinating properties, including an ultrahigh carrier mobility, large mechanical strength, a linear dispersion relation, long-range ballistic transport, quantum Hall effects at room temperature and tunable optical absorption properties4, 5, 6. Beyond graphene, other 2D materials provide a rich variety of more flexible electronic properties, including wide band gap insulators, such as hexagonal boron nitride7, semiconductors and even superconductors, as may be observed in black phosphorus8, 9 or transition metal dichalcogenides (TMDCs)10, 11, 12.
Unlike graphene, which cannot provide low current-off or saturated current-on states because of its zero band gap, the semiconducting TMDCs, such as n-type molybdenum disulfide (MoS2), possess sizable band gaps in the range of 1–2 eV with subnanometre thickness, and provide high on/off ratios as well as more efficient control over switching10, 11, 12. MoS2 has an indirect band gap of 1.3 eV in bulk structures but a direct band gap of 1.8 eV in the monolayer form. The tunable electronic properties of MoS2 enable electron tunnelling and negative differential resistance for use in low-power electronics13, 14. The material is not susceptible to short-channel effects, and this could be helpful in breaking through the scaling limits for transistor miniaturization15, 16, 17. Theoretical simulations indicate that a MoS2 field effect transistor (FET) could operate in the ballistic regime to yield excellent device performances, including an on/off ratio of 1010 and a subthreshold swing of ~60 mV dec−1 (ref. 18). MoS2 and its hybrid heterostructures formed with other 2D materials have demonstrated significant potential for use in flexible, transparent, low-power electronics and optoelectronics, such as tunnelling transistors13, memories19, 20, 21, photodetectors22, 23, 24, 25, electroluminescent devices26, light-emitting devices27 and integrated circuits28, 29. Although the carrier mobility of MoS2 is relatively low, it can be improved significantly by functionalizing the substrate30, passivating the surface31, applying high-k dielectric engineering32, 33 or forming inversion channels34.
Chemical doping has been shown to offer an effective approach to doping in electronic low-dimensional material applications including carbon nanotubes, mono- or few-layer graphene35. Chemically doped TMDC materials and the applications of these materials, however, have not been extensively studied36, 37. In this work, we successfully fabricated unipolar p-type doped MoS2(p-MoS2), n-type doped MoS2 (n-MoS2) and pristine MoS2 (pristine-MoS2) FETs using chemical doping of gold chloride (AuCl3) and benzyl viologen (BV).
In the following, we investigate the thickness-dependent electrical behaviour of a vertical p–n homogeneous junction composed of MoS2. The few-layer MoS2 p–n junctions show ambipolar carrier transport. The potential barrier in a MoS2 p–n junction can be effectively modulated in films with a thickness exceeding 8 nm as they are in conventional semiconductor p–n diodes, giving rise to current rectification in which carrier transport is permitted under a forward bias; however, films with a thickness of less than 8 nm, a ‘reversed’ current rectification is clearly observed in which a tunnelling-dominated current through the ultrathin potential barrier is favoured under a reverse bias. The ultimate thickness and scaling limits of the vertical MoS2 p–n junctions are experimentally determined to be 3 nm (four layers). The chemical doping depth in the direction perpendicular to the layers is found to be 1.5 nm (two layers). Reducing the film thickness below 3 nm, for example, in monolayer MoS2, compromises the p- and n-type doping, and one type of doping eventually overwhelms other types throughout the entire flake. The small film thickness, on the order of 1 nm, renders the ultrathin vertical p–n junction of MoS2 potentially useful in flexible, transparent, high-efficiency electronic and optoelectronic applications, such as phototransistors and solar cells.
The effects of the chemical doping on carrier transport and device performance were investigated by fabricating and comparing the performances of p-MoS2, n-MoS2 and pristine-MoS2 FETs (see the Supplementary Figs 1 and 2). The excellent doping results made the fabrication of an ultrathin vertical MoS2 p–n homogeneous junction possible. Figure 1a–d showed the fabrication details of a vertical MoS2 p–n junction. A few-layer MoS2 flake with a thickness of 11 nm was obtained by mechanical exfoliation and was used as the channel in a back-gate FET. The bottom surface was doped to form an n-type semiconductor by introducing BV, and the top surface was doped to form a p-type semiconductor by introducing AuCl3. A Cr/Pd (5 nm/50 nm) top electrode and a Cr/Pd/Cr(5 nm/50 nm/5 nm) bottom electrode were contacted with the top and bottom surfaces of the MoS2flake, respectively, to provide symmetric metal contacts. Both optical microscopy and atomic force microscopy (AFM) images clearly revealed that the stacking structure was bottom electrode/MoS2/top electrode, as shown in Fig. 1e–g. The drain-to-source current (ID) was characterized as a function of the drain and gate voltages (VD and VG) using a semiconductor parameter analyser. A monochromator (655 nm, 15 mW) and a standard solar simulator (AM1.5 spectrum) were combined with electrical measurements to test the photoresponse.
Rectification of MoS2 p–n junction devices
Compared with the unipolar MoS2 films, such as p-MoS2, n-MoS2 and pristine-MoS2, the MoS2 p–n homogeneous junction provided several advantages. First, it provided a clear rectifying effect on carrier transport. The output characteristics revealed a current rectification ratio, defined as the ratio of the forward current to the reverse current, of ~100, and the theoretical fits suggested an ideality factor (n) of 1.6, as shown in Fig. 2a. The p–n junction properties varied depending on the applied VD, as illustrated by the energy band diagrams shown in Fig. 3a–d. At equilibrium, a potential barrier was established within the channel that prevented electron and hole transport from the source to drain. As with conventional semiconductor p–n diodes, the barrier height could be increased by applying a reverse bias (VD<0 V), or it could be reduced by applying a forward bias (VD>0 V), giving rise to a rectifying effect on carrier transport.
Ambipolar characteristics of vertical MoS2 p–n junction devices
Second, unlike the p-MoS2, n-MoS2 and pristine-MoS2 FETs, which showed unipolar carrier transport, ambipolar carrier transport with a hysteresis window of 60 V was observed in the p–n MoS2 FET, as shown in Fig. 2b. Electron and hole transport were attributed to the presence of n-MoS2 at the bottom surface and p-MoS2 at the top surface, respectively, as shown in Fig. 3e,f. Under a positive VG, the majority carriers were generated via accumulation, which were electrons generated at the bottom (n-MoS2) and holes generated at the top (p-MoS2) of the MoS2 p–n junction. By contrast, the minority carriers were generated via inversion under a negative VG, which were holes generated at the bottom (n-MoS2) and electrons generated at the top (p-MoS2). In both cases, the current flow at a positive VD was contributed by electrons and holes with lateral in-plane transport along the n-MoS2 or p-MoS2 layers, and vertical interlayer tunnelling. It should be noted that electron transport, with a maximum current of the order of 10 nA, was dominant over the hole transport, with a maximum current of the order of 0.1 nA. Because the charge carrier density was controlled by capacitive coupling to the back gate, the modulation of electron transport in n-MoS2 by the gate, which was close to the dielectric layer, was more effective. By contrast, hole transport was not effectively modulated by the gate in the p-MoS2 because of the additional capacitance of the pristine-MoS2 (Ci). Assuming that the capacitance for electron transport in n-MoS2 (Cn) was equal to the oxide capacitance (Cox), as expressed by Cn=Cox=εox/t, the capacitance for hole transport in p-MoS2 (Cp) could be approximated as Cp=(Cox−1+Ci−1)−1. Therefore, Cp was smaller than Cn, which reduced the coupling between the hole carriers and the gate. Here εox is the oxide permittivity and t is the oxide thickness. In addition to the electron current and hole current, a small current between those two was observed near zero gate bias (see Fig. 2a,b). This current was introduced by the pristine MoS2. Since the chemical doping depth was only 1.5 nm (as discussed below) for both n-type and p-type doping, the MoS2 moieties in the middle of a few-layer structure can be remained as pristine (n-type), and can form a p+–n–n+multijunction along the vertical direction. The current contribution from the middle pristine MoS2moieties was relatively small because of its low carrier density compared with those of the chemically doped MoS2 moieties. The current map collected under dark conditions as a function ofVD and VG indicated that electron transport proceeded at positive VG and hole transport at negative VG, as shown in Fig. 2c. Carrier multiplication and avalanche effects were clearly observed under a reverse bias (VD<0 V). The mapping of the corresponding photocurrent (PC), defined as the difference between the values of ID under dark or illuminated conditions, revealed two peaks under a positive VD, as shown in Fig. 2d. The positions and magnitudes of the PC peaks indicated electron and hole transport and reflected the presence of gate-controlled metal–semiconductor barrier modulation38, 39.
Optoelectronic characteristics of vertical MoS2 p–n junction devices
Third, the MoS2 p–n homogeneous junction had the potential to be made ultrathin, transparent and flexible, and its vertical junction structure gave rise to a relatively large junction area that was beneficial for optoelectronic applications. For example, the strong PC generation at positive VGand VD suggested that the p–n MoS2 FET could be used as a phototransistor for light detection, as shown in Fig. 4a,b. Under a forward bias applied at VG=60 V and VD=1 V, the magnitude of IDunder illumination (ID,light) in the p–n junction was about two orders of magnitude larger than the magnitude of ID under dark conditions (ID,dark). The time-resolved characteristics revealed a reliable photoresponse with a stabilized PC ON/OFF ratio of ~100. Moreover, the vertical MoS2 p–n homogeneous junction was demonstrated to be useful in photovoltaic applications, as shown inFig. 4c,d. Under illumination with a standard solar simulator, the MoS2 p–n junction functioned as a solar cell when the gate was grounded, and its energy-conversion performance, including its efficiency (η), fill factor (FF) and photoresponsivity (R) were estimated to be 0.4%, 0.22 and 30 mA W−1, respectively. Considering that the junction was only 11-nm thick, the chemically doped MoS2p–n junction could potentially be quite useful in future flexible, transparent and high-efficiency optoelectronics if the device parameters, including the layer thickness, electrode layout, doping agent and concentration, were optimized.
The vertical MoS2 p–n homogeneous junction in this work showed its own natural advantages, compared with other solar energy-harvesting devices on the basis of MoS2 p–n junction and MoS2hybrid systems, including the lateral MoS2 p–n junction37, MoS2-Au (ref. 40), MoS2-graphene41, MoS2-WS2 (ref. 41), MoS2-WSe2 (ref. 42) and MoS2-Si (ref. 43) systems (see the Supplementary Table 1). For example, in contrast to the lateral MoS2 p–n junction, the vertical p–n junction can provide a much larger planar junction area (or active area). This was very important for optoelectronic applications since the larger active area would absorb more photons, generate more photo-excited charge carriers and increase the conversion efficiency. Compared with the heterogeneous systems, the homogeneous junction can provide the maximized carrier transport efficiency. The photo-excited charge carriers could be very easily lost at the heterogeneous interface because of a variety of factors, including the mismatch of the geometric morphology and lattice structure, the presence of the dangling bonds, surface defects, chemical residuals, absorbed H2O and O2 molecules and so on. Those factors could result in a high contact resistance at the interface and a low carrier transport efficiency through the interface in the heterogeneous systems. In contrast, the homogeneous junction can naturally exclude all those deleterious factors, minimize the carrier lost through the junction and maximize the carrier transport efficiency.
We characterized the electrical and optoelectronic performances of a vertical p–n homogeneous junction formed by chemically doping in few-layered MoS2 films. It was straightforward and interesting to investigate the thickness limits of a vertical p–n junction. A thickness-dependent study was carried out by fabricating a series of MoS2 p–n junctions from few-layered MoS2 films (18, 7 or 4 layers) or from the monolayer structure, as shown in Fig. 5. The potential barrier was varied as the MoS2 film thickness decreased (see Fig. 5a). The 18-layer MoS2 p–n junction behaved as a conventional semiconductor diode, with current rectification properties that allowed carrier transport to proceed under a forward bias because of a reduction in the potential barrier under a positive VD(see Fig. 5b). By contrast, conventional diode behaviour was not observed in the 7-, 4- and 1-layer MoS2 p–n junctions, in which the thickness of the p–n junction, that is, the width of the potential barrier, was reduced to several nanometres or even less than 1 nm, and a large tunnelling current was observed at a negative VD (see Fig. 5c–e). Under a low reverse bias, field-induced band bending was not severe, and direct tunnelling (DT) dominated carrier transport. The DT current (ID,DT) depended linearly on the bias according to44, 45
where Aeff is the effective contact area, B is the barrier height, m0 is the free electron mass, q is the electronic charge, h is Planck’s constant and d is the thickness of the MoS2 film (barrier width). By contrast, the tunnelling distance for electron transport from the drain to the source was further reduced by field-induced band bending under a high reverse bias, and Fowler–Nordheim tunnelling (FNT) became dominant. The FNT current (ID,FNT) followed a nonlinear relation to the bias according to44, 45
where m* (0.45m0) is the effective electron mass of MoS2 (ref. 18). Equation (2) could be further expressed in a linear relation as
According to equation (3), ln(ID/VD2) versus 1/VD could be plotted for each different MoS2 film thickness (see Fig. 5f–i). The strong linear dependence under a high bias suggested that FNT was dominant, and the logarithmic growth at a low bias indicated that DT was dominant. The effective value of B for FNT was estimated from the slope of the linear fits, which increased from 0.14 to 0.35 eV as the MoS2 film thickness decreased from 11 to 0.7 nm. The transition voltage from DT to FNT (VD,trans) also increased from −0.6 to −0.1 V (see Fig. 6a), suggesting that a smaller bias was needed to trigger FNT as the MoS2 p–n junction thickness decreased. Moreover, the current rectification ratio as a function of the MoS2 film thickness clearly indicated a threshold transition between conventional rectification (with a rectification ratio >1) and ‘reversed’ rectification (with a rectification ratio <1) at ~8 nm (12 layers), as shown in Fig. 6b. In other words, the tunnelling effects became dominant in vertical MoS2 p–n homogeneous junction as the film thickness dropped below 8 nm.
The strong in-plane bonding and weak van der Waals interplanar interactions yielded a chemical doping depth in MoS2 that differed from that seen in conventional semiconductors. As demonstrated previously, the MoS2 p–n junction showed ambipolar carrier transport as a result of enhanced hole transport by AuCl3 and enhanced electron transport by BV. Ambipolar carrier transport may be used as a fingerprint of a p–n junction. As the MoS2 film thickness was reduced from 18 to 4 layers, ambipolar carrier transport remained, indicating the appropriate formation of a p–n junction; however, in the monolayer MoS2, only unipolar electron transport was observed, as shown in Fig. 7. This result may reflect the overlap and recombination of both p- and n-type doping in the monolayer MoS2, which eventually results in a single dominant doping type (n-type doping in this work) throughout the entire monolayer film. In other words, a vertical p–n homogeneous junction could not be formed in the monolayer MoS2. We experimentally measured the thickness limit for a vertical MoS2 p–n junction to be 3 nm (four layers). The chemical doping depth along the direction perpendicular to the layers was estimated to be 1.5 nm (two layers) for both p- and n-type doping. In order to confirm the doping depth, a direct observation of the doping profile in MoS2flakes was made by using secondary ion mass spectroscopy (see the Supplementary Fig. 3). The doping depth was found to be 2 nm for p-type doping (see the depth profile of Au, which was originated from AuCl3) and to be 1.5 nm for n-type doping (see the depth profile of C and H, which were originated from BV). Those results were consistent with the value (1.5 nm) estimated from the electrical measurement.
This finding was further supported by fabricating another monolayer MoS2 p–n junction device using the same doping process, but with double top electrodes and double bottom electrodes. This device was designed to confirm the carrier transport type at the top and bottom surfaces, respectively (see the Supplementary Fig. 4). Output characteristics showed ‘reversed’ current rectification in which a tunnelling-dominated large current was observed at the reversed bias. Transfer characteristics showed unipolar electron transport over a wide VG range. Both the features were consistent with the electrical behaviour of another monolayer MoS2 p–n junction (see Figs 5e and 7d), suggesting the good reproducibility and reliability of the vertical MoS2 p–n junction in this work. The individual transfer characteristics on both the top and bottom surfaces showed electron-dominated carrier transport, suggesting the compromise of p-type doping and the overwhelming of n-type doping. This also agreed with our theory. To quantitatively analyse the metal–semiconductor contact condition, the metal–semiconductor barrier height (φMS) was obtained by applying a temperature-dependent test. The maximum value of φMS obtained from both the top and bottom metal–semiconductor interfaces were ~40 meV at the positive VG, which was in agreement with our previous discussion on the electrical behaviour of a Schottky-like junction (see the Supplementary Fig. 2). Our work experimentally revealed the thickness limit of a vertical MoS2 p–n homogeneous junction and established the scaling limit for use in further design and development.
In conclusion, both the unipolar MoS2, such as the p-MoS2 and n-MoS2, as well as the ambipolar vertical MoS2 p–n homogeneous junction, were successfully fabricated by chemically doping AuCl3and BV. The thickness-dependent properties of the vertical MoS2 p–n junction suggested that normal diode behaviour occurred for a MoS2 film thickness exceeding 8 nm, and tunnelling-dominated ‘reversed’ rectification occurred for a film thickness smaller than 8 nm. The ultimate thickness and scaling limits for the vertical MoS2 p–n homogeneous junction were experimentally found to be 3 nm, and the chemical doping depth was found to be 1.5 nm. Given the small thickness, of the order of 1 nm, the vertical MoS2 p–n homogeneous junctions potentially have significant utility in flexible, transparent, high-efficiency electronic and optoelectronic applications.
The fabrication of the p-MoS2, n-MoS2 and pristine-MoS2 FETs began with mechanical exfoliation from bulk crystals. After transfer to a p-type Si substrate (1.0–10.0 Ω cm) coated with a 90-nm-thick thermal oxide layer, the MoS2 flakes were carefully selected by optical microscopy and AFM to have an approximate thickness of 10 nm for use in comparative studies. The p-MoS2 or n-MoS2films were obtained by spin-coating a layer of AuCl3 (20 mM) or BV (20 mM), respectively, followed by annealing on a hot plate at 100 °C for 10 min. The pristine-MoS2 sample reserved untreated as a reference sample. Metal Cr/Pd (5 nm/50 nm) source and drain contact electrodes were patterned using standard electron beam lithography (EBL) and electron beam evaporation techniques.
The p–n MoS2 FET was fabricated as shown in Fig. 1. First, a MoS2 flake was exfoliated from the bulk crystal on a Si substrate, on the surface of which had been spin-coated a water-soluble polyvinyl alcohol (PVA) layer and a hydrophobic polymethyl methacrylate (PMMA) film7. Then, on the top surface of the MoS2 flake was spin-coated a BV (20 mM) layer, and the assembly was annealed on a hot plate at 100 °C for 10 min to form an n-MoS2 surface (see Fig. 1a). Next, the Sisubstrate supporting the n-MoS2 flake was floated on the surface of a deionized water bath. Once the PVA layer had completely dissolved, the PMMA film was left floating on top of the water and could be transferred to a glass slide, the surface of which was coated with a thick polydimethylsiloxane (PDMS) film. Then, the glass slide was clamped on the arm of a micromanipulator mounted on an optical microscope. The MoS2 flake was optically aligned with the n-MoS2 surface downwards and was precisely stacked on a bottom electrode that had been deposited in advance on a target p-type Si substrate (1.0–10.0 Ω cm) coated with a 285-nm-thick thermal oxide layer using standard EBL and electron beam evaporation techniques (see Fig. 1b). During the transfer process, the target substrate was heated to 135 °C to drive off any water absorbed on the flake surface, as well as to promote adhesion between PMMA and the target substrate. After transfer, the PMMA and MoS2 flake remained on the target substrate, and the PMMA layer was dissolved in acetone (see Fig. 1c). Next, on the top surface of the MoS2 flake was spin-coated an AuCl3 (20 mM) layer. The structure was then annealed on a hot plate at 100 °C for 10 min to form a p-MoS2 surface. The top electrode was patterned using standard EBL and electron beam evaporation techniques (see Fig. 1d). The bottom electrode was composed of Cr/Pd/Cr (5 nm/50 nm/5 nm), and the top electrode was composed of Cr/Pd (5 nm/50 nm) in order to provide symmetric metal contacts to p–n MoS2 that were identical to the metal contacts used in the p-MoS2 and n-MoS2 devices, for comparison.
The electrical properties were characterized using a semiconductor parameter analyser under vacuum conditions (10 mTorr) at room temperature. The source and drain contacts were equivalent in the unipolar p-MoS2, n-MoS2 and pristine-MoS2 FETs. The bottom electrode in contact with the n-doped MoS2 in the ambipolar MoS2 p–n junction was set as the source and were grounded during all measurements. The top electrode in contact with the p-doped MoS2 in MoS2 p–n junction was set as the drain, and a drain bias was applied.
The optoelectronic properties were characterized using a monochromator (655 nm, 15 mW) in the phototransistor applications, and using a standard solar simulator (AM1.5 solar spectrum) in the solar cell application.
In solar cell applications, the vertical MoS2 p–n junction showed a short-circuit current (ISC) of 5.1 nA and an open-circuit voltage (VOC) of 0.6 V. The current and voltage obtained at the maximum output power (Imax and Vmax) were 2.2 nA and 0.3 V, respectively. Given the vertical p–n junction area (A), which was estimated from the optical microscopy image to be ~170 μm2, the short-circuit current density (JSC) could be approximated according to JSC=ISC/A=3.0 mA cm−2, and the current density at the maximum output power (Jmax) could be approximated according toJmax=Imax/A=1.3 mA cm−2. Assuming that the input power was equivalent to the solar spectrum (Pin) at 0.1 W cm−2, the maximum output power (Pmax) was estimated to bePmax=JmaxVmax=0.4 mW cm−2, the energy-conversion efficiency (η) was estimated to beη=Pmax/Pin=0.4%, the FF was estimated to be FF=Pmax/(JSCVOC)=0.22, and the R was estimated to be R=Jmax/Pin=30 mA W−1.
Du, Y., Liu, H., Neal, A. T., Si, M. & Ye, P. D.Molecular doping of multilayer MoS2 field-effect transistors: reduction in sheet and contact resistances. IEEE Electron Device Lett.34,1328–1330 (2013) .
This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) (2013-015516) and by Global Frontier Program through the Global Frontier Hybrid Interface Materials (GFHIM) of NRF funded by the Ministry of Science, ICT & Future Planning (2013M3A6B1078873).
Department of Nano Science and Technology, Samsung-SKKU Graphene Center (SSGC), SKKU Advanced Institute of Nano Technology (SAINT), Sungkyunkwan University (SKKU), Suwon 440-746, Korea
Jungjin Ryu &
Won Jong Yoo
Department of Electrical Engineering, Center for Low Energy Systems Technology (LEAST), University of Notre Dame, Notre Dame, Indiana 46556, USA
Hua-Min Li &
H.-M.L., A.S. and W.J.Y. conceived of the research project, supervised the experiment and wrote the paper. H.-M.L., D.L. and J.R. performed the device fabrication. H.-M.L. and X.L. performed the electrical and optoelectronic characterization. H.-M.L. and D.Q. performed the doping process and AFM analysis.
Competing financial interests
The authors declare no competing financial interests.