Light generation and harvesting in a van der Waals heterostructure.

Two-dimensional (2D) materials are a new type of materials under intense study because of their interesting physical properties and wide range of potential applications from nanoelectronics to sensing and photonics. Monolayers of semiconducting transition metal dichalcogenides MoS2 or WSe2 have been proposed as promising channel materials for field-effect transistors. Their high mechanical flexibility, stability, and quality coupled with potentially inexpensive production methods offer potential advantages compared to organic and crystalline bulk semiconductors. Due to quantum mechanical confinement, the band gap in monolayer MoS2 is direct in nature, leading to a strong interaction with light that can be exploited for building phototransistors and ultrasensitive photodetectors. Here, we report on the realization of light-emitting diodes based on vertical heterojunctions composed of n-type monolayer MoS2 and p-type silicon. Careful interface engineering allows us to realize diodes showing rectification and light emission from the entire surface of the heterojunction. Electroluminescence spectra show clear signs of direct excitons related to the optical transitions between the conduction and valence bands. Our p-n diodes can also operate as solar cells, with typical external quantum efficiency exceeding 4%. Our work opens up the way to more sophisticated optoelectronic devices such as lasers and heterostructure solar cells based on hybrids of 2D semiconductors and silicon.

Molybdenum disulfide (MoS2) is a typical representative of layered transition metal dichalcogenide (TMD) semiconductors[1] with electronic properties and a potential range of applications complementary to those of graphene. Bulk TMD crystals are stacks of layers held together via weak van der Waals interaction, allowing the extraction of single 2D atomic layers using the adhesive-type-based micromechanical cleavage technique[2] originally developed for the preparation of graphene. Because it has a band gap, monolayer MoS2 can be used as the basic building block of room-temperature field-effect transistors[3] with an on/off ratio exceeding 108 as well as logic circuits[3] and amplifiers[4] with high gain. Large-area MoS2 can also be grown using CVD-like growth techniques[5,6] or deposited using liquid phase exfoliation.[7−9]

The electronic and optical properties of monolayer MoS2 and other semiconducting dichalcogenides are fundamentally different from those of their bulk counterparts. Because of the lack of inversion symmetry, charge carriers in monolayer MoS2 behave as massive Dirac fermions,[10] while the conduction band of MoS2 shows strong spin–orbit-induced spin splitting[11] and strong coupling of spin and valley degrees of freedom that can be detected using circularly polarized light[12−14] and could be used in novel devices based on the valley Hall effect.[15]

A transition from an indirect band gap to a direct band gap occurs in the monolayer limit,[16−19] manifesting itself in strong photoluminescence.[17,18] The direct band gap in MoS2 can also be harnessed for the realization of vertical optoelectronic devices[20] as well as phototransistors[21] and photodetectors[22] with high responsivity and low noise-equivalent power.[22] Sundaram etal. recently demonstrated[23] that monolayer MoS2 can also be used as a light emitter in an electroluminescent device with light emission occurring due to hot carrier processes in a region near locally gated contacts. While this result showed that monolayer MoS2 could be used for the fabrication of light-emitting devices, the device geometry was limited by a relatively high power threshold for light emission and only a small portion of the device, restricted to the contact edge, was active in electroluminescence.

One way to overcome these factors limiting the exploitation of monolayer MoS2 for practical applications in optoelectronic devices is to build light-emitting diodes based on vertical p–n junctions, resulting in a natural increase of the junction area that can easily be scaled. Reports on vertical p–n junctions based on TMD materials have been published before,[24] but these devices were not capable of electroluminescence and included thicker, indirect band gap TMD materials, which are less suitable for optoelectronic applications than their monolayer counterparts. We demonstrate here a vertical p–n junction in the form of a vertical heterostructure composed of n-type MoS2 and p-type silicon serving as the hole injection layer.[25] We choose p-type silicon for this purpose because it is readily available and easy to pattern and handle. No reports on p-type monolayer MoS2 have been published so far.

Our device shows a decreased threshold power for light emission, while the entire heterojunction surface is active as a light emitter. The device is also capable of operating as a solar cell.

Results and Discussion

Figure shows the structure of our device. Fabrication starts with exfoliation of MoS2 (ref (2)) onto an SiO2/Si substrate. MoS2 is then transferred[26] onto a prepatterned target highly doped p-type Si substrate[27] covered with SiO2 with 1 μm × 1 to 100 μm × 100 μm windows through which the underlying Si is exposed. The native oxide on the substrate is removed and the Si surface passivated with hydrogen using a second wet etch step.[28]

Figure 1

Geometry of the MoS/Si heterojunction light-emitting diode. (a) Optical image of the device in an intermediate state of fabrication. Monolayer MoS2 is placed across the sidewall of a square window etched into a SiO2 layer exposing the underlying p-doped silicon. Scale bar is 10 μm long. (b) Cross-sectional view of the structure of the device together with electrical connections used to induce light emission from the heterojunction. Electrons are injected from n-type MoS2, while holes are injected from the p-Si substrate.

In order to avoid degradation of the passivation layer, monolayer MoS2 is immediately transferred across the edge of a window, exposing the Si surface (Figure a), and contacted on one side with a gold electrode. On the basis of AFM imaging we can see that MoS2 is transferred on top of H–Si in a conformal fashion, with no visible voids or wrinkles. Both the 2D MoS2 film and the H–Si substrate are terminated and have no dangling bonds at their surfaces, allowing the formation of a van der Waals heterostructure.[29] Because the nature of the interface is similar to that in graphene/BN heterostructures, we expect that most of the interface in our device is clean and free of contaminants,[30] allowing direct charge injection between Si and MoS2. For diode characterization and electroluminescence measurements, we use the measurement scheme shown in Figure b. On some of the devices we also deposit a 30 nm thick HfO2 or Al2O3 layer in order to encapsulate the device and increase the mobility in monolayer MoS2.[31] This includes both devices presented here. More than 10 functioning devices were produced showing similar characteristics. We have observed that unencapsulated devices show a significant reduction of device currents and emitted light intensity when exposed to the ambient over the course of a week. The initial performance level can be restored by performing a vacuum anneal, indicating that the observed performance degradation could be due to adsorbed water and oxygen rather than an irreversible degradation of the interface. Encapsulated devices presented here show no significant change of performance over a period of at least one month.

Figure a shows the current vs bias voltage (I–V) characteristic of our MoS2/Si heterojunction diode with 30 nm HfO2 on top, exhibiting rectifying behavior with a current of 346 nA for a forward bias of 10 V and a junction area of 19 μm2. This shows that classical diodes and all related optoelectronic devices could be prepared using a combination of an atomically thin 2D semiconductor and a 3D semiconductor, which should allow for a rapid fabrication and development of this type of device on industrial scales. We find that in the reverse bias regime a breakdown does not occur below −10 V.

Figure 2

Electrical characteristic of the device. (a) Current vs bias voltage characteristic of the MoS2/Si heterojunction diode. (b) Band diagram of the MoS2/Si heterojunction in equilibrium conditions and under forward bias. Electrons injected from the n-MoS2 and holes from p-Si can radiatively recombine in the junction. (c) Intensity map showing the electroluminescent emission with superimposed outline of the most important device components. The entire surface of the heterojunction is emitting light. Scale bar is 5 μm long. (d) Integrated light intensity as a function of device current. The inset shows the emitted light intensity as a function of electrical power. The threshold current for light emission is ∼100 nA, corresponding to a threshold power of 3.2 W/cm2 for a device with an active area of 19 μm2.

In Figure b we outline the proposed band structure of our device, typical of type-II abrupt heterojunctions.[32] The junction is characterized by conduction (ΔEC = 200 meV) and valence (ΔEV = 900 meV) band offsets due to different electron affinities[33] and band gaps of Si and MoS2. Under the application of forward bias V to the heterojunction, electrons injected from the MoS2 side and holes injected from p-Si can radiatively recombine in the junction, resulting in light emission. Due to the direct band gap nature of MoS2,[16−19] we expect the emitted light to be characterized by radiative transitions in MoS2, as radiative transitions in Si are expected to be much less efficient due to its indirect band gap. Due to valence and conduction band offsets, discontinuities could occur in the bands with valence and conduction band cusps that can impair charge carrier injection efficiency and limit the device current.

The electroluminescent emission intensity map for a forward bias of 10 V is shown in Figure c, superposed on the outline of the device. Most of the heterojunction surface is active, in contrast to MoS2 electroluminescent devices based on hot carrier processes in a region near locally gated contacts[23] or previously reported observations in a similar device geometry,[34] where the light emission was localized only at the heterojunction edge. The presence of a large active area in our device can be attributed to hydrogen passivation of the Si substrate, resulting in the formation of a true heterojunction with an efficient charge transfer. Large-area emitters such as the one presented here are also more attractive from a practical point of view because the total emitted light intensity could be more easily scaled up by simply increasing the device area.

Figure d shows the integrated electroluminescence intensity as a function of device current and electrical power for the active area centered on the heterojunction surface. The results show light emission from the device for bias voltages exceeding 5.5 V, corresponding to an electroluminescence threshold current of ∼109 nA. The equivalent threshold power density is 3.2 W/cm2, significantly lower than the previously reported threshold power of 15 kW/cm2 for MoS2 electroluminescent devices based on hot carrier injection.[23] This shows that the defect-free vertical heterojunction geometry and Si/MoS2 band alignment are favorable for reducing the emission threshold and increasing the surface area of the emitter. The presence of a threshold is probably due to the existence of cusps in the heterojunction band diagram under forward bias conditions (Figure b) and could probably be further decreased with careful band engineering of the interface.

The photoluminescence (PL) spectrum of monolayer MoS2 is acquired in the region where the flake is supported by SiO2 and is shown in Figure a. The spectrum shows two peaks at 685 nm (1.81 eV) and 624 nm (1.98 eV). They are associated with excitonic transitions between the bottom of the conduction band and the top of the valence band, split due to spin–orbit coupling.[17,18,11]

Figure 3

Light emission characteristics of the device. (a) Photoluminescence spectrum of the region of the monolayer MoS2 flake supported by SiO2. (b) Electroluminescence spectrum acquired under a forward bias V = 15 V and a current of 1.8 μA. The spectrum is fitted with three Lorentzian lines, which correspond to A and B excitons at 694 and 644 nm and the A– trion resonance at 721 nm.

In Figure b we show the electroluminescence spectrum, together with a fit to a multiple peak Lorentzian model. The main feature of the spectrum is a peak with a position of 694 nm (1.78 eV), which has a full width at half-maximum of 56 nm. The position of this peak matches well with the observed PL peak at 685 nm and is associated with the A exciton[12,17,18] in monolayer MoS2. This shows that the relevant energy for the radiative recombination process in the MoS2/Si heterojunction is the direct band gap in monolayer MoS2. We observe an additional feature at 721 nm (1.72 eV), which can be related to the trion (negatively charged exciton) resonance in monolayer MoS2.[35]

In addition to the A exciton, because of the low emission threshold, the electrical power density at which our device operates provides enough energy through impact ionization to excite the higher energy B exciton, which we can distinguish as an additional feature in the electroluminescence spectrum, located at 644 nm (1.92 eV). The small red shift with respect to the related PL peak at 624 nm could be attributed to differences in the dielectric environments: the EL peak is acquired in the heterojunction region where MoS2 is in direct contact with silicon, while the PL peak is acquired in the region where MoS2 is supported by SiO2. The difference in these two dielectric environments could affect the exciton binding energy through screening of the Coulomb interaction between electrons and holes.

The photoluminescence in the heterojunction area is strongly reduced in comparison to portions of the MoS2 layer that are supported by SiO2. This indicates the presence of an internal, open circuit voltage that separates the electrons and holes in the junction during the short period between light absorption and emission. This is favorable for operating our van der Waals heterostructures as solar cells. In Figure a we show current as a function of voltage in a second device with an area of 8 μm2 for different illumination powers, showing the increase of short-circuit current under illumination, indicating power generation. The electrical power, P, generated in the device defined as P = I × V is shown in Figure b, with a peak power of ∼1 nW for a bias voltage V = ∼0.1 V and illumination power Pinc = 861 nW. We characterize the spectral response of the solar cell by measuring its short-circuit current, Isc (obtained for a bias voltage of 0 V), as a function of illumination wavelength, λ, using a supercontinuum light source. The external quantum efficiency, EQE, of the device, defined as EQE = (Isc/Pinc) × (hc/eλ), where h is Planck’s constant, c the speed of light, and e the elementary charge, is shown in Figure c. It is characterized by a sharp drop above 1000 nm, coinciding with the absorption edge of silicon, and a broadband response in the 500–1000 nm region, indicating that MoS2 and Si form a true p–n heterojunction instead of a Schottky contact and operate in tandem. The complementary absorption profiles of these materials result in a device with spectral response that is extended with respect to the response of monolayer MoS2.[36] The maximum recorded EQE is 4.4%, which is a promising value for a device based on a two-dimensional monolayer and is more than an order of magnitude higher than in lateral p–n junctions based on the dichalcogenide WSe2, which also shows a much narrower spectral response limited by its band gap.[37,38] The EQE could be further enhanced by careful control over doping levels of MoS2 and Si, which would reduce the series resistance of the device, by use of large-area grown or deposited materials,[7,6,39] and by incorporating additional 2D semiconducting layers such as WSe2 with complementary absorption spectra. Because of the direct band gap nature of monolayer MoS2 and other dichalcogenide materials, we expect such future device cells to surpass the efficiencies of previously demonstrated solar cells based on bulk TMD semiconductors.[40]

Figure 4

MoS2/Si heterojunction as a solar cell. (a) Current as a function of bias voltage under different illumination powers from a 541 nm laser. The heterojunction area is 8 μm2. (b) Electrical power generated by the device as a function of bias voltage, recorded for different illumination powers, extracted from data shown in a. (c) External quantum efficiency as a function of wavelength in the 450–1100 nm range for an illumination power of 500 nW. The curve shows a broadband response with MoS2 and Si working in tandem and effectively extending the spectral response of MoS2 into the infrared region. At both ends of the wavelength range, our measurements are limited by the sharp drop in emission intensity of our supercontinuum light source.

Conclusion

To summarize, we show electroluminescent devices and solar cells based on heterojunctions composed of monolayer MoS2 and p-type silicon. This choice of materials combines the advantages of the direct band gap and small thickness of 2D MoS2 with the established silicon-based fabrication processes and could show the way to implementing 2D semiconductors as enabling materials in standard semiconductor fabrication lines. Furthermore, all the semiconducting materials used in our devices can be considered earth abundant and nontoxic. The entire junction area in our device participates in light emission with a low emission threshold power, allowing future large-area light emitters and lasers based on MoS2. The low threshold power allows us to distinguish features in the emitted light spectra related to three different optical transitions, A and B excitons and the A– trion resonance,[35] which could find valuable applications in the field of valleytronics. The heterojunction diode can also operate as a photovoltaic device, converting incoming light into electrical power with an external quantum efficiency of 4.4% and a broad spectral response, indicating that MoS2 and silicon operate in tandem.

Methods

Single layers of MoS2 are exfoliated from commercially available crystals of molybdenite (SPI Supplies Brand Moly Disulfide) using the Scotch-tape micromechanical cleavage technique method pioneered for the production of graphene[41] on silicon substrates covered by a 270 nm layer of thermal oxide. Monolayer samples were identified by optical microscopy.[42] Once identified, monolayers were transferred[26] onto p-type silicon substrates with a resistivity of 0.1–0.5 Ωcm, corresponding to a boron doping level between 3 × 1016 and 3 × 1017 cm–3, covered by a 100 nm thick layer of thermal SiO2 with patterned holes from 1 μm × 1 μm up to 100 μm × 100 μm. Windows in SiO2 are opened using 7:1 buffered oxide etch, resulting in sloped sidewalls. The initial etching step was followed by a 1 min 1% HF etch in order to remove the native oxide and passivate the Si surface.[28] The sample thickness was confirmed by photoluminescence measurements.

Monolayer MoS2 diodes were characterized at room temperature. For electrical characterization, we use a gold electrode deposited on MoS2 and a large-area electrode in direct contact with the p-Si substrate. A second gold electrode is deposited on top of Si near the MoS2 flake but not in direct electrical contact with it. We use this electrode to verify that charge carriers can be injected from the passivated Si substrate. The emitted radiation was collected and analyzed using a grating spectrometer (HORIBA Jobin Yvon) equipped with a liquid nitrogen cooled CCD camera (Triax 550). An Andor iXon Ultra camera was used to perform photon counting and map the light emission. Photoluminescence measurements were performed using a laser centered at 488 nm and a spectrometer (Princeton Instruments SP-2500i) with a liquid nitrogen cooled camera (PiXIS/Pylon/Spec-10:256).

Heterojunction band structures were based on modeling performed using the Adept 2.0 tool available at NanoHUB.org.