LETTERS PUBLISHED ONLINE: 10 AUGUST 2014 | DOI: 10.1038/NNANO.2014.150
Atomically thin p–n junctions with van der Waals heterointerfaces Chul-Ho Lee1,2,3, Gwan-Hyoung Lee4, Arend M. van der Zande5, Wenchao Chen6, Yilei Li1, Minyong Han7, Xu Cui8, Ghidewon Arefe8, Colin Nuckolls2, Tony F. Heinz1,9, Jing Guo6, James Hone8 and Philip Kim1 * Semiconductor p–n junctions are essential building blocks for electronic and optoelectronic devices1,2. In conventional p–n junctions, regions depleted of free charge carriers form on either side of the junction, generating built-in potentials associated with uncompensated dopant atoms. Carrier transport across the junction occurs by diffusion and drift processes influenced by the spatial extent of this depletion region. With the advent of atomically thin van der Waals materials and their heterostructures, it is now possible to realize a p–n junction at the ultimate thickness limit3–10. Van der Waals junctions composed of p- and n-type semiconductors—each just one unit cell thick—are predicted to exhibit completely different charge transport characteristics than bulk heterojunctions10–12. Here, we report the characterization of the electronic and optoelectronic properties of atomically thin p–n heterojunctions fabricated using van der Waals assembly of transition-metal dichalcogenides. We observe gate-tunable diode-like current rectification and a photovoltaic response across the p–n interface. We find that the tunnelling-assisted interlayer recombination of the majority carriers is responsible for the tunability of the electronic and optoelectronic processes. Sandwiching an atomic p–n junction between graphene layers enhances the collection of the photoexcited carriers. The atomically scaled van der Waals p–n heterostructures presented here constitute the ultimate functional unit for nanoscale electronic and optoelectronic devices. Heterostructures based on atomically thin van der Waals materials are fundamentally different and more flexible than those made from conventional covalently bonded materials, in that the lack of dangling bonds on the surfaces of van der Waals materials enables the creation of high-quality heterointerfaces without the constraint of atomically precise commensurability9,13,14. The ability to build artificial van der Waals heterostructures, combined with recent rediscoveries of transition-metal dichalcogenides (TMDCs) as atomically thin semiconductors3–5, provides a route to a wide variety of semiconductor heterojunctions and superlattices. The strong light–matter interaction and distinctive optical properties of TMDCs render them promising candidates for optoelectronic devices such as photodiodes, photovoltaic cells and light-emitting devices3,8,15–20. The availability of TMDCs with different bandgaps and workfunctions allows for bandgap engineering of heterostructures21. Furthermore, the ability to electrostatically tune the carrier
densities and band alignments of two-dimensional semiconductors offers an alternative way to design functional heterostructures22–26 and to understand the mechanisms underlying their operation. Here, we have realized the ultimate limit of p–n junction scaling in a heterostructure consisting of semiconducting TMDC monolayers. We use individually contacted layers of p-type tungsten diselenide (WSe2) and n-type molybdenum disulphide (MoS2) to create an atomically thin p–n junction. The difference in workfunction and bandgap between the two monolayers creates an atomically sharp heterointerface, predicted to be of type II band alignment21. Figure 1a presents a schematic of the van der Waals heterostructure of MoS2/WSe2. Each TMDC monolayer consists of a hexagonally packed plane of transition-metal atoms sandwiched between two planes of chalcogen atoms6. This heterojunction is created by van der Waals assembly of individual monolayers on a SiO2/Si substrate (Supplementary Section 1). To measure the electrical properties of the junction, we fabricated metal contacts on each layer (optical image in Fig. 1a). Al and Pd were chosen to inject electrons and holes into the n-MoS2 and p-WSe2 layers, respectively27. A gate voltage Vg applied to the Si substrate adjusts the carrier densities in each semiconductor layer. As shown in the inset of Fig. 1b, the individual MoS2 and WSe2 layers exhibit, respectively, n- and p-type channel characteristics due to the unintentional doping present in each crystal. From the measured threshold voltages (Vth), we estimate (using n (electron) or p (hole) = q −1Cg|Vth – Vg|, where q = 1.6 × 1019 C and Cg = 1.23 × 10−8 F cm−2) carrier densities of 1 × 1012 to 3 × 1012 cm−2 at Vg = 0 V for both materials. Accordingly, a p–n junction is formed between the two atomically thin semiconductors. Figure 1b presents the I–V curves of the junction at various gate voltages, as measured between metal electrodes on the WSe2 (D1) and MoS2 (S1). We observe gate-voltage-tuned rectification of current as electrostatic doping modulates the density of free electrons and holes in the junction. Although the observed I–V characteristics are similar to those of a conventional p–n junction diode, the underlying physical mechanism of rectification is expected to differ in view of the lack of a depletion region in the atomically thin junction. Figure 1c,d presents band profiles in the lateral and vertical directions, respectively, at a forward bias of 0.6 V, based on simulations for the electrostatic configuration of the device (Supplementary Section 3). We find that most of the voltage drop occurs across the vertical p–n junction, leaving no appreciable potential barriers in the lateral transport direction
1
Department of Physics, Columbia University, New York, New York 10027, USA, 2 Department of Chemistry, Columbia University, New York, New York 10027, USA, 3 KU-KIST Graduate School of Converging Science and Technology, Korea University, Seoul 136-701, Korea, 4 Department of Materials Science and Engineering, Yonsei University, Seoul 120-749, Korea, 5 Energy Frontier Research Center (EFRC), 1001 Schapiro Center (CEPSR), Columbia University, New York, New York 10027, USA, 6 Department of Electrical and Computer Engineering, University of Florida, Gainesville, Florida 32611-6200, USA, 7 Department of Applied Physics and Applied Mathematics, Columbia University, New York, New York 10027, USA, 8 Department of Mechanical Engineering, Columbia University, New York, New York 10027, USA, 9 Department of Electrical Engineering, Columbia University, New York, New York 10027, USA. * e-mail:
[email protected] 676
NATURE NANOTECHNOLOGY | VOL 9 | SEPTEMBER 2014 | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved
NATURE NANOTECHNOLOGY
LETTERS
DOI: 10.1038/NNANO.2014.150
a
b 0.6 10−4
0.4
Ids (A)
10−6 MoS2
10−10
Ids (nA)
WSe2
10−8
10−12 −60
0.2
−30
0
30
60
Vg (V) MoS2 S2
0.0 −30
D1 Pd
Al
ate
str
ub Si s
/
iO 2
WSe2
S
D2
) (V Vg
WSe2
S1 MoS2
0 1.0 0.5 0.0
30 −0.5
c
d CB
MoS2
Vds (V)
WSe2 2
1.0
E (eV)
CB
e− e−
e− e−
e−
ΔEc
τ
e−
0.0 B −1.0
−2.0
ΔEv
h+ h+ h+ h+ h+
VB
h+
VB 3 μm
n-MoS2
p−n junction
Two unit cells
p-WSe2
Figure 1 | Charge transport in an atomically thin p–n heterojunction. a, Bottom left: Schematic diagram of a van der Waals-stacked MoS2/WSe2 heterojunction device with lateral metal contacts. Top: enlarged crystal structure, with purple, red, yellow and green spheres representing Mo, S, W and Se atoms, respectively. Bottom right: Optical image of the fabricated device, where D1 and D2 (S1 and S2) indicate the metal contacts for WSe2 (MoS2). Scale bar, 3 µm. b, Current–voltage curves at various gate voltages measured across the junction (between contacts D1 (+) and S1 (–) in the optical image in a). Inset: Gate-dependent transport characteristics at Vds = 0.5 V for individual monolayers of MoS2 (blue curve, measured between S1 and S2) and WSe2 (red curve, measured between D1 and D2). c,d, Band profiles in the lateral (c) and vertical (d) directions, obtained from electrostatic simulations. Under forward bias (Vds = 0.6 V), electrons (blue circles with ‘e−’) in the conduction band (CB) of MoS2 and holes (red circles with ‘h+’) in the valence band (VB) of WSe2 undergo interlayer recombination (black arrows in c) via Shockley–Read–Hall (red dashed arrows in d) or Langevin (blue arrow in d) mechanisms, contributing the dark current. τ and B are, respectively, the tunnelling-assisted recombination lifetime and the Langevin recombination constant. Note that there is no significant band bending in the lateral transport direction. Band offsets for electrons (ΔEc) and holes (ΔEv) across the junction are indicated in c.
within each semiconductor layer. In contrast, under reverse bias, substantial potential barriers result from band bending in the lateral direction (Supplementary Fig. 4). Consequently, under forward bias, the current is governed by tunnelling-mediated interlayer recombination between majority carriers at the bottom (top) of the conduction (valence) band of MoS2 (WSe2). This unusual interlayer recombination can be described by two possible physical mechanisms or a combination of both: (1) Shockley–Read–Hall (SRH) recombination (R ≈ nM pW/τ (nM + pW)) mediated by inelastic tunnelling of majority carriers into trap states in the gap; and (2) Langevin recombination (R ≈ BnMp W s) by Coulomb interaction. Here R is the recombination rate, nM and pW are, respectively, the electron and hole densities in MoS2 and WSe2 , and τ and B are, respectively, the tunnelling-assisted recombination lifetime and the Langevin recombination constant2. An exponent (1.2 < s < 1.5) is used in the modelling of the two-dimensional Langevin mechanism28, in contrast to the linear
dependence on p for the three-dimensional case (Supplementary Section 3). The rectifying I–V characteristics are understood with an increasing interlayer recombination rate at higher forward biases (Supplementary Fig. 5). In addition, the current under forward bias can be tuned by varying the carrier densities through electrostatic gating; this is maximized with nearly balanced nM and pW at Vg = 0 (Fig. 1b), as predicted for both SRH and Langevin recombination. We estimate a lower bound for lifetime τ > 1 µs from the measured current and the simulated carrier densities (Supplementary Section 4). This lifetime is relatively long because inelastic tunnelling processes occur between randomly stacked TMDC layers with lateral momentum mismatch13. This is beneficial for reducing the interlayer recombination in photocurrent generation, as discussed below. Based on this understanding of the charge transport mechanism, we next explore the optoelectronic response of the atomically thin p–n junction. Figure 2a presents representative I–V curves
NATURE NANOTECHNOLOGY | VOL 9 | SEPTEMBER 2014 | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved
677
LETTERS 0.3
b
Current density (mA cm−2)
(mA W−1)
0.0
0.8
20
0.0 −20
−0.3
d 30
(a.u.)
10
0
0
Jsc
−10 Vg (V)
−0.6 Vg = 0V
1
(mA cm−2) −1 1
DOI: 10.1038/NNANO.2014.150
c 105 Photoluminescence intensity (a.u.)
a
NATURE NANOTECHNOLOGY
104
103
102
−30
−0.9
0.0
−0.1
0.0
0.1
0.2
0.3
Vds (V)
0.5
0.4
101 1.4
0.5
1.6
Vds (V)
g 1017 (m−2)
e− e−
0.8
1013
1 × 1032 h+
f
0 τ
e−
2.2
1015
108
h+
2.0
h 1.0
|Photocurrent| (mA cm−2)
e
1.8
Photon energy (eV)
3 × 1013
0.6
0.4
0.2
B h+ 0 −20
0.0 −10
0
10
20
−20
Vg (V)
−10
0
10
20
Vg (V)
Figure 2 | Gate-tunable photovoltaic response. a, Photoresponse characteristics at various gate voltages under white-light illumination. Inset: Colour plot of photocurrent as a function of voltages Vds (x axis) and Vg (y axis). The dashed line represents the profile of short-circuit current density Jsc at Vds = 0 V. b, Photocurrent map of the device presented in Fig. 1a for Vds = 0 V and 532 nm laser excitation. The junction area and metal electrodes are indicated by dashed and solid lines, respectively. Scale bar, 3 µm. c, Photoluminescence spectra measured from the isolated monolayers (blue curve for MoS2; red curve for WSe2) and the stacked junction region (brown curve). d, Photoluminescence spatial maps for emission at 1.66 eV (top) and 1.88 eV (bottom), corresponding to direct gap transitions of monolayer WSe2 and MoS2 , respectively. The junction area is indicated by dashed lines. Scale bars, 3 µm. e,f, Schematic illustrations of exciton dissociation (e) and interlayer recombination (f) processes. In e, horizontal and vertical arrows represent charge transfer and intralayer recombination processes, respectively. In f, red and blue arrows indicate Shockley–Read–Hall (SRH) and Langevin recombination processes, respectively. g, Simulations of the gate-voltage-dependent majority (red curve for holes in WSe2; blue curve for electrons in MoS2) and minority (red dashed curve for holes in MoS2; blue dashed curve for electrons in WSe2) carrier densities in each layer (top), spatially averaged 〈nM pW 1.2〉 (middle) and 〈nM pW/(nM + pW)〉 (bottom). h, Measured (circles and dashed curve) and simulated (green curve for two-dimensional Langevin process and purple curve for SRH mechanism) photocurrent at Vds = 0 V as a function of gate voltages. For the fit, B = 4.0 × 10−13 m2 s−1 and τ = 1 µs are used for the twodimensional Langevin (s = 1.2) and SRH mechanisms, respectively.
and a colour plot of the photocurrent (inset) in the gate range of –30 to 30 V under white-light illumination. We observe a gatetunable photovoltaic response. In particular, the short-circuit current density reaches its maximum at Vg = 0 V and decreases on varying the gate voltage in either polarity (the line profile at source–drain voltage (Vds) = 0 V in the colour plot). The maximum photoresponsivity is ∼2 mA W−1, as measured using a focused 532 nm laser with a power of 0.8 µW (corresponding to 100 W cm−2). Note that we observed a similar gate dependence for all light sources used, although the gate voltage was shifted by extrinsic substrate- or photo-induced doping effects29 (Supplementary Fig. 7). To further elucidate the underlying physical processes, we performed spatial mapping of the photoresponse by scanning the light source in a confocal optical microscope. Figure 2b displays the resulting photocurrent map taken at Vds = 0 V. The strongest 678
photoresponse is observed in the MoS2/WSe2 junction area, indicating spontaneous charge separation occurring at the junction. This charge separation process is also compatible with the photoluminescence characteristics of the MoS2/WSe2 junction. Figure 2c shows photoluminescence spectra obtained from the separated monolayers and from the overlapping layers forming the p–n junction. We find that emission from the direct gap transitions of both monolayer MoS2 (1.88 eV) and monolayer WSe2 (1.66 eV) is strongly quenched only at the MoS2/WSe2 junction area. The integrated intensities decrease, respectively, by 81% and 98% compared to isolated MoS2 and WSe2 monolayers. The spatially resolved photoluminescence maps at the fixed bandgap energies also reveal luminescence quenching in the p–n junction area (Fig. 2d). We attribute the observed strong decrease in photoluminescence to the rapid separation of charge carriers in the junction region. Although an energy transfer process could explain the decrease in photoluminescence by the
NATURE NANOTECHNOLOGY | VOL 9 | SEPTEMBER 2014 | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved
NATURE NANOTECHNOLOGY a
LETTERS
DOI: 10.1038/NNANO.2014.150
c
b
0
MoS2
0
−15 (mA W−1)
Ids (nA)
Graphene
−40
e−
WSe2
e− Pd
Cr/
SiO
h+
2 /S
−80
i
Graphene
h+
GR −0.4
0
GR
0.4
0.8
Vds (mV) Photon wavelength (nm) 800
e d
700
600
500
60
50
10
0
2L–2L EQE (%)
Photoresponsivity (mA W−1)
ML–ML
−50
1L–1L
1
−100 1L–1L
0.1
2L–2L ML–ML (10/9 nm) −150 −10
0
10
20
30
1.5
1.7
Voltage per thickness (mV nm−1)
1.9
2.1
2.3
2.5
Photon energy (eV)
Figure 3 | Graphene-sandwiched van der Waals p–n heterojunctions. a, Schematics of MoS2/WSe2 junction sandwiched between top and bottom graphene electrodes. b, Optical image and corresponding photocurrent map (for Vds = 0 V) of the graphene-sandwiched monolayer p–n junction device (1L–1L). Photocurrent is uniformly observed only in the junction area indicated by the dashed lines, separated from the metal electrodes (indicated by solid lines). Scale bars, 3 µm. c, Current–voltage curves of the device in b, measured in the dark (black) and under 532 nm laser excitation (red). Inset: Schematic of the bandstructure, with exciton dissociation and charge-collection processes indicated by horizontal arrows. The dashed line represents the Fermi level. Note that the bottom graphene (GR) is slightly p-doped from the SiO2 substrate. d, Photoresponse characteristics of graphene-sandwiched p–n junctions with different thicknesses. For comparison, the applied voltages (horizontal axis) are normalized with respect to the junction thicknesses. e, EQE plots as a function of excitation energy (wavelength) for the devices in d. The laser powers were in the range 3–7 µW, corresponding to 380–890 W cm−2. Results in d and e are shown for devices composed of monolayer–monolayer (1L–1L), bilayer–bilayer (2L–2L) and multilayer–multilayer (ML–ML (10/9 nm)) junctions.
material with the larger gap, the observed decrease of photoluminescence of both materials suggests a charge transfer mechanism. Note that we can exclude the photo-thermoelectric effect as the origin of the photocurrent; its contribution is minor compared to the measured photoresponse as a consequence of the small Seebeck coefficient and very slight temperature gradient across the vertical van der Waals interface (Supplementary Section 6)30,31. Spontaneous dissociation of a photogenerated exciton into free carriers can be driven by large band offsets for electrons (ΔEc) and holes (ΔEv) across the atomically sharp interface, as shown in Fig. 2e. This charge separation may also be understood in terms of highly asymmetric charge transfer rates for electrons and holes in a heterostructure with type II band alignment; one process involves allowed transitions into the band, while the other involves forbidden transitions into the gap. These charge transfer and separation processes are analogous to those considered in excitonic organic solar cells11. In atomically thin p–n junctions, the (allowed) charge transfer processes are expected to be fast and
efficient because exciton (or minority carrier) diffusion is not required. Indeed, our observation of significant photoluminescence quenching and photocurrent generation in the p–n junction indicates that exciton dissociation at the junction is significantly faster than other non-radiative (or radiative) decay channels existing within the layers. For typical TMDC monolayers, such intralayer relaxation processes are present on a timescale of ∼10 ps (refs 3,32). For an atomically thin p–n heterojunction, even after exciton dissociation, the majority carriers spatially confined in each layer can undergo recombination though inelastic tunnelling. Such interlayer tunnelling-mediated recombination plays an important role in determining the photocurrent. Figure 2f represents interlayer SRH and Langevin recombination processes. Because the gate voltage tunes the majority carrier densities, we can model the gate modulation of the photocurrent observed in our experiments. Figure 2g presents plots of the simulated densities of majority and minority carriers in each layer as a function of gate voltage (top panel), under illumination. Here, the interlayer recombination rate will be
NATURE NANOTECHNOLOGY | VOL 9 | SEPTEMBER 2014 | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved
679
LETTERS
NATURE NANOTECHNOLOGY
proportional to nMpsW (middle panel in Fig. 2g) or nMpW/(nM + pW) (bottom panel in Fig. 2g), depending on whether the dominant recombination mechanism follows Langevin or SRH behaviour. These quantities, spatially averaged over the junction area, are minimized near Vg = 0 V. They increase with |Vg| due to accumulation of one type of majority carrier. Consequently, as shown in Fig. 2h, the sharply peaked photoresponse observed experimentally is modelled with both Langevin and SRH recombination because the photocurrent is determined by the difference of the recombination rate and the gate-independent generation rate (Supplementary Section 5 provides more details of the simulation). Note that although Langevin recombination may be a dominant mechanism due to the enhanced Coulomb interaction between electrons and holes spatially confined in the two-dimensional layered system33, SRH recombination could still be induced by imperfections at the junction interface, as well as defects in the materials. Therefore, we could not achieve full quantitative agreement using only a specific mechanism, despite reasonable qualitative agreement. In our devices, the relatively low carrier mobility in the lateral transport channel results in a diffusion time for the majority carriers to leave the junction region as long as ∼1 µs, a timescale comparable to the interlayer recombination lifetime. Because photocurrent collection competes with interlayer recombination, one strategy to reduce interlayer recombination losses is to increase the rate of collection of the photogenerated carriers. To realize this improvement, we used graphene electrodes directly on the top and bottom of the vertical p–n junction (Fig. 3a). This allows carrier collection by direct vertical charge transfer, rather than through lateral diffusion. For comparison, we fabricated van der Waals stacks using various thicknesses of MoS2/WSe2 layers between two graphene electrodes (Supplementary Sections 1 and 2)34. In this device geometry, graphene serves as a transparent van der Waals contact that also minimizes recombination compared to typical metal contacts8. Figure 3b shows that photocurrent (at Vds = 0 V) is only observed in the region where the two active TMDC monolayers overlap with the top and bottom graphene electrodes (Supplementary Section 8). Figure 3c presents I–V curves of the vertical p–n junction consisting of both monolayers in the dark and under 532 nm laser excitation. The device produces linear I–V characteristics, apparently dominated by direct tunnelling between the two graphene electrodes (Supplementary Section 7). Nevertheless, we still observe a short-circuit current of ∼70 nA at an excitation power of ∼7 µW (corresponding to 920 W cm−2), corresponding to a photoresponsivity of ∼10 mA W−2. This value for vertical charge collection is larger than that for a typical laterally contacted device by a factor of ∼5. The p–n photodiode characteristics are improved significantly as the number of atomic layers in the MoS2/WSe2 heterojunctions increases and direct tunnelling between the top and bottom graphene electrodes is suppressed. Figure 3d shows the photoresponse characteristics of three devices with different junction thicknesses. As the thickness increases, the overall transport curve changes from a linear to a rectifying diode characteristic because direct tunnelling current decreases exponentially (Supplementary Section 7). In particular, the vertical p–n junction consisting of two multilayers (10/9 nm) shows a photovoltaic response with an open-circuit voltage of ∼0.5 V (corresponding to ∼26 mV nm−1). Additionally, the photocurrent gradually increases with thickness up to ∼120 mA W−1 because of increased light absorption. These photoresponse characteristics could also be tuned by varying the gate voltage (Supplementary Section 9). We also investigated the external quantum efficiency (EQE) as a function of excitation energy for graphene-sandwiched p–n junctions of different thicknesses (Fig. 3e). The measured EQEs at 532 nm were 2.4%, 12% and 34% for monolayer, bilayer and multilayer p–n junctions, respectively. The quantity does not scale 680
DOI: 10.1038/NNANO.2014.150
linearly with junction thickness because of the thickness-dependent competition between generation, charge collection and recombination, permitting further optimization of the overall efficiency. Regardless of thickness, all three p–n junctions show similar spectral responses, with two prominent peaks at 1.64 eV and 1.87 eV, corresponding to the excitonic absorption edges of WSe2 and MoS2 , respectively. This spectral photoresponse matches well with the absorption spectrum measured from the monolayer MoS2/WSe2 stack (Supplementary Section 10), indicating that excitons generated in both semiconductor layers contribute to the photocurrent. In conclusion, we have fabricated atomically thin p–n junctions from van der Waals-bonded semiconductor layers. Although the junction exhibits both rectifying electrical characteristics and photovoltaic response, the underlying microscopic processes differ strongly from those of conventional devices in which an extended depletion region play a crucial role. In particular, interlayer tunnelling recombination of the majority carriers across the van der Waals interface, which can be tuned by gating, is found to influence significantly both the electrical and optoelectronic properties of the junction. Further optimization of the band alignment, number of atomic layers and interface lattice mismatching will lead to unique material platforms for novel, high-performance electronic and optoelectronic devices.
Methods The MoS2/WSe2 stacks were fabricated on 280-nm-thick SiO2-coated Si substrates using a co-lamination and mechanical transfer technique. The thickness of the TMDC layers was confirmed by atomic force microscopy, photoluminescence and/or Raman spectroscopy (Supplementary Section 2). The vertical stacks of graphene/MoS2/WSe2/graphene were prepared using a polymer-free van der Waals assembly technique with edge electrical contact to the graphene layers34. Further details of the fabrication processes are provided in Supplementary Section 1. For model simulation of the device, Poisson and drift-diffusion equations were solved self-consistently using a two-dimensional finite difference method (Supplementary Section 3). The photoresponse characteristics were investigated under white-light illumination and laser excitation. Quantitative analysis of the response was obtained with focused laser excitation of known power. Scanning photocurrent measurements were performed using a confocal optical microscopy equipped with a two-axis scanning mirror or scanning mechanical stage. In these experiments, a 532 nm continuous-wave laser (Crystalaser) and a supercontinum laser source (NKT photonics) were used as excitation light sources. The laser radiation was focused to a 0.7–1 µm full-width at half-maximum (FWHM) spot through a ×50 objective. The powers used for wavelength-dependent EQE measurements were in the range 3–7 µW, where the devices exhibit a linear response (Supplementary Section 11).
Received 14 March 2014; accepted 24 June 2014; published online 10 August 2014
References 1. Sze, S. M. & Ng, K. K. Physics of Semiconductor Devices 3rd edn (Wiley, 2006). 2. Green, M. A. Solar Cells: Operating Principles, Technology, and System Applications (Prentice-Hall, 1982). 3. Mak, K. F., Lee, C., Hone, J., Shan, J. & Heinz, T. F. Atomically thin MoS2: a new direct-gap semiconductor. Phys. Rev. Lett. 105, 136805 (2010). 4. Splendiani, A. et al. Emerging photoluminescence in monolayer MoS2. Nano Lett. 10, 1271–1275 (2010). 5. Radisavljevic, B., Radenovic, A., Brivio, J., Giacometti, V. & Kis, A. Single-layer MoS2 transistors. Nature Nanotech. 6, 147–150 (2011). 6. Wang, Q. H., Kalantar-Zadeh, K., Kis, A., Coleman, J. N. & Strano, M. S. Electronics and optoelectronics of two-dimensional transition metal dichalcogenides. Nature Nanotech. 7, 699–712 (2012). 7. Bonaccorso, F. et al. Production and processing of graphene and 2D crystals. Mater. Today 15, 564–589 (December, 2012). 8. Britnell, L. et al. Strong light–matter interactions in heterostructures of atomically thin films. Science 340, 1311–1314 (2013). 9. Geim, A. K. & Grigorieva, I. V. Van der Waals heterostructures. Nature 499, 419–425 (2013). 10. Bernardi, M., Palummo, M. & Grossman, J. C. Extraordinary sunlight absorption and one nanometer thick photovoltaics using two-dimensional monolayer materials. Nano Lett. 13, 3664–3670 (2013). 11. Gregg, B. A. & Hanna, M. C. Comparing organic to inorganic photovoltaic cells: theory, experiment, and simulation. J. Appl. Phys. 93, 3605–3614 (2003). 12. Dimitrijev, S. Principle of Semiconductor Devices (Oxford Univ. Press, 2012).
NATURE NANOTECHNOLOGY | VOL 9 | SEPTEMBER 2014 | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved
NATURE NANOTECHNOLOGY
DOI: 10.1038/NNANO.2014.150
13. Britnell, L. et al. Field-effect tunneling transistor based on vertical graphene heterostructures. Science 335, 947–950 (2012). 14. Haigh, S. J. et al. Cross-sectional imaging of individual layers and buried interfaces of graphene-based heterostructures and superlattices. Nature Mater. 11, 764–767 (2012). 15. Zeng, H. L., Dai, J. F., Yao, W., Xiao, D. & Cui, X. D. Valley polarization in MoS2 monolayers by optical pumping. Nature Nanotech. 7, 490–493 (2012). 16. Cao, T. et al. Valley-selective circular dichroism of monolayer molybdenum disulphide. Nature Commun. 3, 887 (2012). 17. Mak, K. F. et al. Tightly bound trions in monolayer MoS2. Nature Mater. 12, 207–211 (2013). 18. Jones, A. M. et al. Optical generation of excitonic valley coherence in monolayer WSe2. Nature Nanotech. 8, 634–638 (2013). 19. Lopez-Sanchez, O., Lembke, D., Kayci, M., Radenovic, A. & Kis, A. Ultrasensitive photodetectors based on monolayer MoS2. Nature Nanotech. 8, 497–501 (2013). 20. Sundaram, R. S. et al. Electroluminescence in single layer MoS2. Nano Lett. 13, 1416–1421 (2013). 21. Kang, J., Tongay, S., Zhou, J., Li, J. B. & Wu, J. Q. Band offsets and heterostructures of two-dimensional semiconductors. Appl. Phys. Lett. 102, 012111 (2013). 22. Fontana, M. et al. Electron–hole transport and photovoltaic effect in gated MoS2 Schottky junctions. Sci. Rep. 3, 1634 (2013). 23. Yu, W. J. et al. Highly efficient gate-tunable photocurrent generation in vertical heterostructures of layered materials. Nature Nanotech. 8, 952–958 (2013). 24. Baugher, B. W. H., Churchill, H. O. H., Yang, Y. & Jarillo-Herrero, P. Optoelectronic devices based on electrically tunable p–n diodes in a monolayer dichalcogenide. Nature Nanotech. 9, 262–267 (2014). 25. Pospischil, A., Furchi, M. M. & Mueller, T. Solar-energy conversion and light emission in an atomic monolayer p–n diode. Nature Nanotech. 9, 257–261 (2014). 26. Ross, J. S. et al. Electrically tunable excitonic light-emitting diodes based on monolayer WSe2 p–n junctions. Nature Nanotech. 9, 268–272 (2014). 27. Fang, H. et al. High-performance single layered WSe2 p-FETs with chemically doped contacts. Nano Lett. 12, 3788–3792 (2012). 28. Greenham, N. C. & Bobbert, P. A. Two-dimensional electron–hole capture in a disordered hopping system. Phys. Rev. B 68, 245301 (2003). 29. Sercombe, D. et al. Optical investigation of the natural electron doping in thin MoS2 films deposited on dielectric substrates. Sci. Rep. 3, 3489 (2013).
LETTERS
30. Xu, X. D., Gabor, N. M., Alden, J. S., van der Zande, A. M. & McEuen, P. L. Photo-thermoelectric effect at a graphene interface junction. Nano Lett. 10, 562–566 (2010). 31. Buscema, M. et al. Large and tunable photothermoelectric effect in single-layer MoS2. Nano Lett. 13, 358–363 (2013). 32. Shi, H. Y. et al. Exciton dynamics in suspended mono layer and few-layer MoS2 2D crystals. ACS Nano 7, 1072–1080 (2013). 33. Cheiwchanchamnangij, T. & Lambrecht, W. R. L. Quasiparticle band structure calculation of monolayer, bilayer, and bulk MoS2. Phys. Rev. B 85, 205302 (2012). 34. Wang, L. et al. One-dimensional electrical contact to a two-dimensional material. Science 342, 614–617 (2013).
Acknowledgements This work was supported as part of the Center for Re-defining Photovoltaic Efficiency Through Molecule Scale Control, an Energy Frontier Research Center funded by the US Department of Energy, Office of Science, Office of Basic Energy Sciences (award no. DE-SC0001085), and in part by the National Science Foundation (DMR-1124894) and by the FAME Center, one of six centres of STARnet, a Semiconductor Research Corporation programme sponsored by MARCO and DARPA. G.H.L. and M.H. acknowledge support from the Basic Science Research Program (2014R1A1A1004632) (G.H.L.) and the Nano Material Technology Development Program (2012M3A7B4049966) (M.H.) through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT and Future Planning.
Author contributions C-H.L., G-H.L., M.H., X.C. and G.A. fabricated and characterized the van der Waals p–n heterostructures. C-H.L. and G-H.L. performed device fabrication and transport measurements. A.M.v.d.Z. and C-H.L. performed photocurrent measurements. W.C. and J.G. provided theoretical support. Y.L. performed optical characterization. P.K., J.H., T.F.H., J.G. and C.N. advised on experiments. C-H.L. and P.K. wrote the manuscript in consultation with G-H.L., A.M.v.d.Z., W.C., C.N., T.F.H., J.G. and J.H.
Additional information Supplementary information is available in the online version of the paper. Reprints and permissions information is available online at www.nature.com/reprints. Correspondence and requests for materials should be addressed to P.K.
Competing financial interests
The authors declare no competing financial interests.
NATURE NANOTECHNOLOGY | VOL 9 | SEPTEMBER 2014 | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved
681
