LETTERS

PUBLISHED ONLINE: 9 MARCH 2014 | DOI: 10.1038/NNANO.2014.26

Electrically tunable excitonic light-emitting diodes

based on monolayer WSe2 p–n junctions
Jason S. Ross1, Philip Klement2,3, Aaron M. Jones3, Nirmal J. Ghimire4,5, Jiaqiang Yan5,6,
D. G. Mandrus4,5,6, Takashi Taniguchi7, Kenji Watanabe7, Kenji Kitamura7, Wang Yao8,
David H. Cobden2 and Xiaodong Xu1,2 *

The development of light-emitting diodes with improved There are two reasons for this. First, efficient electroluminescence
efficiency, spectral properties, compactness and integrability requires effective injection of both electrons and holes into the
is important for lighting, display, optical interconnect, logic active region, which should therefore be within a p–n junction.
and sensor applications1–8. Monolayer transition-metal dichal- Second, MoS2 is known to have poorer optical quality than other
cogenides have recently emerged as interesting candidates group-VIB TMDs, possibly due to impurities. In contrast, it has
for optoelectronic applications due to their unique optical prop- been shown previously that monolayer WSe2 has excellent optical
erties9–16. Electroluminescence has already been observed from properties16,24. Here, we demonstrate that combining monolayer
monolayer MoS2 devices17,18. However, the electroluminescence WSe2 with a p–n junction architecture using electrostatic doping
efficiency was low and the linewidth broad due both to the poor produces efficient and electrically tunable excitonic light-emitting
optical quality of the MoS2 and to ineffective contacts. Here, we diodes (LEDs).
report electroluminescence from lateral p–n junctions in mono- Figure 1a,b presents a schematic and optical image of a device
layer WSe2 induced electrostatically using a thin boron nitride made by a combination of electron-beam lithography and the trans-
support as a dielectric layer with multiple metal gates beneath. fer of exfoliated sheets. A monolayer WSe2 sheet sits on a sheet of
This structure allows effective injection of electrons and holes, hexagonal boron nitride (BN), typically 10 nm thick, which acts
and, combined with the high optical quality of WSe2 , yields both as a smooth, disorder-free substrate to minimize non-radiative
bright electroluminescence with 1,000 times smaller injection energy relaxation pathways and as a high-quality gate dielectric.
current and 10 times smaller linewidth than in MoS2 Applying voltages to the two 7 nm palladium gate electrodes
(refs 17,18). Furthermore, by increasing the injection bias we beneath the BN can create two separate electrostatically doped
can tune the electroluminescence between regimes of impur- regions in the WSe2 separated by a 300-nm-wide undoped strip.
ity-bound, charged and neutral excitons. This system has the Gold/vanadium (60/6 nm) source and drain contacts were evapor-
required ingredients for new types of optoelectronic device, ated on top. Importantly, they overlap the gates to reduce the
such as spin- and valley-polarized light-emitting diodes, Schottky barriers at the contacts. For electrical transport measure-
on-chip lasers and two-dimensional electro-optic modulators. ments, a d.c. bias V is applied to one contact (the source) and the
Few-layer group-VIB transition-metal dichalcogenides (TMDs) current I from the other (the drain) is measured by a virtual-earth
are a class of semiconductors in the two-dimensional limit9,10,19. current preamplifier. The silicon substrate is grounded.
Because of their large carrier effective mass and reduced screening We start by showing that a p–n junction can be created electro-
in two dimensions, electron–hole interactions are much stronger statically. First, we set the gate voltages Vg1 and Vg2 to the
than in conventional semiconductors. This leads to large binding same value. Figure 1c shows the current produced with a bias of
energies for both charged and neutral excitons, which, as a result, V ¼ 0.5 V as the gate voltage Vg1 ¼ Vg2 is swept at 60 K (room-
are spectrally sharp, robust and amenable to electrical manipu- temperature measurements are shown in Supplementary Fig. 1).
lation16,20,21. In addition, the large spin–orbit coupling22 and acen- The current increases rapidly for gate voltages .þ 6.5 V (electron
tric structure of TMDs provides a connection between spin and doping) and ,2 6.5 V (hole doping), demonstrating ambipolar
valley degrees of freedom14, light polarization11,13,15,16 and magnetic operation. In the inset, the red I–V curve, taken at Vg1 ¼ Vg2 ¼ þ8.0 V,
and electric fields23 that can be exploited for new kinds of is almost symmetric, as expected for both gated regions being
device operation. equally electron-doped. The nonlinearity near zero bias can be
Although in bulk TMDs the bandgap is indirect, in the limit of a associated with the undoped gap between the gates and residual
single monolayer it becomes direct9,10, fulfilling the most basic Schottky barriers at the contacts. In contrast, the blue I–V curve,
requirement for efficient light emission. Indeed, electrolumines- taken with Vg1 ¼ þ8.0 V and Vg2 ¼ 28.0 V, shows the strong
cence has already been reported from monolayer MoS2 field-effect rectification behaviour expected for a p–n junction.
transistors (FETs), occurring near the Schottky contact with a The p–n junction can be investigated in detail using scanning
metal17 or with highly doped silicon18. However, the efficiency optical measurements25,26. Figure 2a presents a microscope image
and spectral quality was much lower than has been demonstrated of a device and Fig. 2b a corresponding scanning photocurrent
in other nanoscale light emitters such as carbon nanotubes7. image, measured with zero bias at 100 K using a 10 mW

1
Department of Materials Science and Engineering, University of Washington, Seattle, Washington 98195, USA, 2 Department of Physics, Justus Liebig
University, 35392 Giessen, Germany, 3 Department of Physics, University of Washington, Seattle, Washington 98195, USA, 4 Department of Physics and
Astronomy, University of Tennessee, Knoxville, Tennessee 37996, USA, 5 Materials Science and Technology Division, Oak Ridge National Laboratory, Oak
Ridge, Tennessee, 37831, USA, 6 Department of Materials Science and Engineering, University of Tennessee, Knoxville, Tennessee, 37996, USA, 7 Advanced
Materials Laboratory, National Institute for Materials Science, Tsukuba, Ibaraki 305-0044, Japan, 8 Department of Physics and Center of Theoretical and
Computational Physics, University of Hong Kong, Pokfulam Road, Hong Kong, China. * e-mail: [email protected]

268 NATURE NANOTECHNOLOGY | VOL 9 | APRIL 2014 | www.nature.com/naturenanotechnology

© 2014 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2014.26 LETTERS
a Vsd
A c 5

− + WSe2
−
− +
4 5
+

Vg1

Current (nA)
BN Vg2 3
SiO2 0

Current (nA)
Si Pd

b 2
BN SiO2
−5
Vg1
D
Vg2 S 1 −2 0 2
Vsd (V)
Vg1 WSe2 S
Vg2 D
0

−8 −4 0 4 8
4 μm
Vg1 = Vg1 (V)

Figure 1 | Monolayer WSe2 p–n junctions. a,b, Cartoon (a) and optical micrograph (b) of multiple monolayer WSe2 p–n junction devices with palladium
backgates (Vg1 and Vg2) and source (S) and drain (D) contacts. The source–drain voltage (Vsd) is applied to one contact and the current (A) is read out of
the other. During electroluminescence in the WSe2 , electrons (blue) and holes (yellow) move towards each other (arrows) and recombine. The backgates are
separated from the WSe2 by hexagonal BN. The device sits on a layer of silicon dioxide on a silicon substrate. c, Main panel: Variation of current with both
bottom gate voltages set equal (Vg1 ¼ Vg2 ¼ Vg) for one junction at bias Vsd ¼ 500 mV, showing ambipolar behaviour. Inset: I–Vsd characteristics when the
two gate voltages are set to be equal (Vg1 ¼ Vg2 ¼ 8 V, red) and opposite (Vg1 ¼ 28 V, Vg2 ¼ 8 V, blue) values above the injection threshold.

a b e
X−

4
Photocurrent (nA)

WSe2
Photoluminescence intensity (a.u.)

X−
0
X+

Xo
c d
Photoluminescence (×105 counts)

Photoluminescence energy (eV)

X+
1.68
7
X−
6
1.65
5 X+

4
1.62
1.55 1.65 1.75
Photon energy (eV)

Figure 2 | Photoresponse of monolayer p–n junction at 100 K. a, Microscope image of a monolayer p–n junction device. The source and drain contacts are
white, the two bottom gates are red, and the BN is blue. b, Corresponding scanning photocurrent image showing pronounced photocurrent generation
localized to the junction. The orange dashed lines outline the source and drain contacts whereas the blue dashed lines outline the backgates. c, Integrated
photoluminescence map. d, Peak photoluminescence energy map showing p and n regions as a result of the different energies of oppositely charged
excitons. X2 (Xþ) represents the negatively (positively) charged exciton found in the n (p) region e, Top to bottom: Selected spectra from n-doped region
(blue), junction (black) and p-doped region (red) with negative, neutral (Xo), and positive exciton features indicated. All maps taken with Vg1 ¼ þ8 V,
Vg2 ¼ 28 V. Scale bars, 4 mm.

diffraction-limited 660 nm laser spot scanned over the sample. We of WSe2 monolayers at 660 nm (ref. 27), the internal quantum effi-
see a large photocurrent signal localized between the gates, with a ciency reaches a maximum of 5%. Such a photocurrent is the
peak magnitude of 5 nA. Taking into account the 1% absorption natural result of the junction functioning as a photodiode, with

NATURE NANOTECHNOLOGY | VOL 9 | APRIL 2014 | www.nature.com/naturenanotechnology 269

© 2014 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2014.26

a b dominates, implying an excess of electrons. Above the other

intensity (a.u.)
gate, held at Vg2 ¼ 28.0 V (red trace), the higher-energy positively
charged Xþ trion (two holes and one electron) dominates, implying
Normalized

an excess of holes. In the gap between the gates (black trace) the
neutral exciton Xo peak can also be seen, consistent with no
doping in that region. Here, the superimposed Xþ and X2 peaks
WSe2 may come from the gated regions, because the laser spot is larger
monolayer than the gap.
1.55 1.60 1.65 1.70 1.75 When the device is configured as a p–n junction (Vg1 ¼ 2Vg2 ¼ 8 V),
Photon energy (eV) but not otherwise, we observe bright electroluminescence. Good
spectra can be obtained even at room temperature with a current
Figure 3 | Photoluminescence and electroluminescence. a, At 300 K, the of 5 nA, as illustrated (blue) in Fig. 3a. This current is three
electroluminescence spectrum (blue) generated by a current of 5 nA closely orders of magnitude smaller than in MoS2 FETs17,18. In fact, in
resembles the photoluminescence spectrum (red). b, Electroluminescence our best device we observed electroluminescence at an injection
image (red) superimposed on device image (greyscale). The dashed lines current as low as 200 pA (Supplementary Fig. 3). To understand
outline the WSe2 monolayer. Scale bar, 2 mm. the nature of the electroluminescence, we superimpose a normalized
photoluminescence spectrum of undoped monolayer WSe2 (red).
photogenerated carriers separated by a strong depletion field con- It is known that the photoluminescence of WSe2 arises from the
centrated in the undoped gap. The sensitivity of the photodiode recombination of direct-gap excitons; thus, the similarity between
can be tuned over a wide range by varying the gate voltages and the electroluminescence and photoluminescence spectra implies
bias (Supplementary Fig. 2). that the injected electrons and holes form excitons before recombin-
Figure 2c presents a corresponding map of the integrated photo- ing radiatively. This is a natural consequence of the large exciton
luminescence intensity, which indicates the high optical quality of binding energy due to the strong Coulomb interaction in monolayer
the whole WSe2 sheet and shows that the luminescence is not sub- TMDs. Figure 3b presents an image of the total electroluminescence
stantially quenched by the underlying gates. More revealingly, intensity (coloured) superimposed on a simple white-light reflection
Fig. 2d shows a colour map of the peak photoluminescence image (greyscale). It is clear that the electroluminescence emanates
photon energy, exhibiting two distinct regions clearly correlated from the entire length of the monolayer junction between the
with the expected n-doped (blue) and p-doped (red) parts of the two gates.
WSe2 above the gates. The reason for this is made clear in Fig. 2e, At low temperatures, the electroluminescence spectrum develops
which shows photoluminescence spectra taken at different pos- an interesting structure. Figure 4a shows a plot of electroluminescence
itions. The detailed origin of these spectral features has been estab- intensity as a function of current and photon energy. There are
lished previously16. Above the gate held at Vg1 ¼ þ8.0 V (blue trace) three main spectral features: a narrow higher-energy peak (green
the negatively charged X2 trion (two electrons and one hole) arrow), a broad central peak (brown arrow) and a lower-energy

a c
35 X−
intensity (log counts)
Electroluminescence

XI d
8 p region Junction n region
30 I = 24 nA
I (nA)

7 −K +K −K +K −K +K
Electroluminescence intensity (a.u.)

×15
6 σ− σ+
25 X+

I = 27 nA
×5
1.58 1.62 1.66 1.70
Photon energy (eV)

b 8 Xo + + +
+ + + + − − − −
intensity (log counts)

+
Photoluminescence

9 ×1.5 + + + − − −
4 I = 31 nA −
X− − − −
8
Vg (V)

Xo −V
0
7 +V
XI
−4 6 I = 33 nA

X+
−8
1.58 1.62 1.66 1.70 1.5 1.6 1.7 1.8
Photon energy (eV) Photon energy (eV)

Figure 4 | Tuning valley-exciton electroluminescence at 60 K. a, Electroluminescence intensity plot as a function of bias current and photon energy. From
left to right, the arrows indicate the impurity-bound exciton (XI), the charged excitons (X2 then Xþ) and the neutral exciton (Xo). b, Plot of
photoluminescence intensity as a function of photon energy and gate voltage Vg ¼ V g1 ¼ Vg2. c, Selected electroluminescence spectra at different bias
currents. As the current increases, we observe electroluminescence tuning from XI through X2 and Xþ and finally Xo. The bottom spectrum is fit by four
Gaussian lineshapes, one for each exciton species. d, Band diagram and device schematic showing electroluminescence generation from valley excitons.
Wavy red arrows indicate electroluminescence. Dashed red (blue) arrow indicates the direction of hole (electron) flow. Filled and empty circles indicate
carriers in the þK and 2K valleys. Both valleys are shown to be populated, leading to electroluminescence that has both right (sþ) and left (s2)
circular polarization.

270 NATURE NANOTECHNOLOGY | VOL 9 | APRIL 2014 | www.nature.com/naturenanotechnology

© 2014 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2014.26 LETTERS
peak (black arrow). The shapes and relative intensities of these fea- current for a given voltage by reducing the contact resistance, by
tures change with current. Their origins can be deduced from a improving the WSe2 crystal quality, and by employing improved
comparison with the photoluminescence, the intensity of which is membrane transfer techniques.
plotted in Fig. 4b as a function of photon energy and common Finally, we mention an important implication of these results. It
gate voltage Vg ¼ Vg1 ¼ Vg2. Here we see the tuning of the dominant has been conclusively demonstrated using polarization-resolved
exciton species as the carrier density is changed by gating16. The photoluminescence that the excitons in monolayer WSe2 are
photoluminescence feature that is strongest at Vg ¼ 0 is due to Xo formed in the +K valleys16. The excellent match of the electrolumi-
recombination. It has a similar width to, and is at the same position nescence with the photoluminescence thus proves that the electro-
(1.69 eV) as, the narrow electroluminescence peak, which we there- luminescence also comes from such valley excitons. In the
fore identify as the Xo emission. The dominant photoluminescence experiments described here, the injected electrons and holes popu-
feature at Vg . 0, which shifts from 1.663 eV to 1.625 eV as Vg late both valleys equally, as shown in Fig. 4d, forming excitons in
increases from 0 to þ8 V, is due to X2 (of which there are multiple both valleys and thus producing unpolarized light, as we have
species16). The broad electroluminescence feature occurs in the verified (Supplementary Fig. 4). As a result of the spin-valley
same range of energies, implying that it is dominated by X2. The locking in monolayer TMDs, where the þK and 2K band edges
dominant feature in the photoluminescence at Vg , 0 is due to have opposite spin14, we expect that using ferromagnetic contacts
Xþ. This aligns with the high-energy shoulder (grey arrow) on the to obtain spin-polarized injection will allow the creation of
broad electroluminescence feature at 1.670 eV. There is also a spin- and valley-LEDs with controllably polarized emission32.
band of emission from impurity-bound excitons (XI) in the photo-
luminescence, which matches the lower-energy electroluminescence Note added in proof: During the preparation of this Letter we
feature centred at 1.59 eV. became aware of two similar studies33,34.
The finding that X2 dominates the electroluminescence in
Fig. 4a is consistent with the observation that X2 has much stronger Received 3 October 2013; accepted 21 January 2014;
photoluminescence than the other exciton species (Fig. 4b). The published online 9 March 2014
shifts of the trion peaks in the photoluminescence with Vg imply
that the trion binding energy depends on the perpendicular electric References
1. Chen, J. et al. Bright infrared emission from electrically induced excitons
field. Hence, the fact that the width of the broad peak in the electro- in carbon nanotubes. Science 310, 1171–1174 (2005).
luminescence matches the full range of X2 energies in the photolu- 2. Marty, L. et al. Exciton formation and annihilation during 1D impact excitation
minescence is explained by the variation of the perpendicular of carbon nanotubes. Phys. Rev. Lett. 96, 136803 (2006).
electric field across the junction. On the other hand, the Xo peak 3. Mann, D. et al. Electrically driven thermal light emission from individual
single-walled carbon nanotubes. Nature Nanotech. 2, 33–38 (2007).
shifts very little with Vg in the photoluminescence and is therefore 4. Kioseoglou, G. et al. Intershell exchange and sequential electrically injected
insensitive to the field; thus, the Xo electroluminescence peak is spin populations of InAs quantum-dot shell states. Phys. Rev. Lett. 101,
also sharp. The Xo electroluminescence linewidth is found to be 227203 (2008).
as narrow as 5 meV, which is an order of magnitude smaller than 5. Xia, F., Steiner, M., Lin, Y. & Avouris, P. A microcavity-controlled, current-
for electroluminescence from MoS2 (refs 1,7). driven, on-chip nanotube emitter at infrared wavelengths. Nature Nanotech.
3, 609–613 (2008).
Figure 4c presents the electroluminescence spectra at selected 6. Essig, S. et al. Phonon-assisted electroluminescence from metallic carbon
current values, illustrating sequential population of the excitonic nanotubes and graphene. Nano Lett. 10, 1589–1594 (2010).
states, which could be due to current pumping or to changes in 7. Mueller, T. et al. Efficient narrow-band light emission from a single carbon
the electric field at the junction under different source–drain nanotube p–n diode. Nature Nanotech. 5, 27–31 (2010).
biases. At the lowest current (24 nA), only the excitons with 8. Shambat, G. et al. Ultrafast direct modulation of a single-mode photonic
crystal nanocavity light-emitting diode. Nature Commun. 2, 539 (2011).
lowest energy (XI and X2) are seen. At a higher current (27 nA) 9. Mak, K. F., Lee, C., Hone, J., Shan, J. & Heinz, T. F. Atomically thin MoS2:
the Xþ shoulder appears and at 31 nA an Xo peak is also present. a new direct-gap semiconductor. Phys. Rev. Lett. 105, 2–5 (2010).
At the highest current (33 nA) we illustrate how the spectrum can 10. Splendiani, A. et al. Emerging photoluminescence in monolayer MoS2.
be decomposed into four Gaussian peaks. It is also apparent that Nano Lett. 10, 1271–1275 (2010).
the relative strength of the XI peak decreases as the current increases. 11. Cao, T. et al. Valley-selective circular dichroism of monolayer molybdenum
disulphide. Nature Commun. 3, 887 (2012).
The above observations reflect complex exciton dynamics, which 12. Feng, J., Qian, X., Huang, C-W. & Li, J. Strain-engineered artificial atom as
are not yet fully understood. Time-resolved photoluminescence a broad-spectrum solar energy funnel. Nature Photon. 6, 866–872 (2012).
measurements have shown that the lifetimes (for radiative and 13. Mak, K. F., He, K., Shan, J. & Heinz, T. F. Control of valley polarization in
non-radiative processes combined) of the free excitons and the monolayer MoS2 by optical helicity. Nature Nanotech. 7, 494–498 (2012).
14. Xiao, D., Liu, G-B., Feng, W., Xu, X. & Yao, W. Coupled spin and valley
impurity-bound excitons are 5 ps and 100 ps respectively28–31 in
physics in monolayers of MoS2 and other group-VI dichalcogenides. Phys.
monolayer MoS2 , which we expect to be similar to WSe2. The elec- Rev. Lett. 108, 1–5 (2012).
tron–hole pair injection rate is 1e (e is electron charge), which is one 15. Zeng, H., Dai, J., Yao, W., Xiao, D. & Cui, X. Valley polarization in MoS2
pair per 5 ps at 32 nA, comparable to the lifetime of a free exciton monolayers by optical pumping. Nature Nanotech. 7, 490–493 (2012).
but much shorter than that of XI. Therefore, there could be only 16. Jones, A. M. et al. Optical generation of excitonic valley coherence in
monolayer WSe2. Nature Nanotech. 8, 634–638 (2013).
of order one free exciton but many impurity-bound excitons 17. Sundaram, R. S. et al. Electroluminescence in single layer MoS2. Nano Lett.
present in the junction. We speculate that that the presence of mul- 13, 1416–1421 (2013).
tiple XI combined with the strong Coulomb interactions enhances 18. Ye, Y. et al. Exciton-related electroluminescence from monolayer MoS2.
non-radiative recombination, which limits the XI emission at Preprint at http://lanl.arXiv.org/abs/1305.4235 (2013).
higher currents. Alternatively, the saturation of the XI peak could 19. Novoselov, K. S. et al. Two-dimensional atomic crystals. Proc. Natl Acad. Sci.
USA 102, 10451–10453 (2005).
also be due to the filling of impurity states, as observed in standard 20. Mak, K. F. et al. Tightly bound trions in monolayer MoS2. Nature Mater.
photoluminescence and photocurrent experiments. 12, 207–211 (2013).
We measured the total photon emission rate at the largest 21. Ross, J. S. et al. Electrical control of neutral and charged excitons in a
applied current of 35 nA to be 16 million s21 (Supplementary monolayer semiconductor. Nature Commun. 4, 1474 (2013).
Note 1). This is 10 times larger, for 1,000 times smaller current, 22. Zhu, Z. Y., Cheng, Y. C. & Schwingenschlögl, U. Giant spin–orbit-induced
spin splitting in two-dimensional transition-metal dichalcogenide
than reported for MoS2 devices17,18. It corresponds to one photon semiconductors. Phys. Rev. B 84, 153402 (2011).
per 104 injected electron–hole pairs. We expect that the overall 23. Yuan, H. et al. Zeeman-type spin splitting controlled by an electric field.
device efficiency could be improved by increasing the injection Nature Phys. 9, 563–569 (2013).

NATURE NANOTECHNOLOGY | VOL 9 | APRIL 2014 | www.nature.com/naturenanotechnology 271

© 2014 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2014.26

24. Zhao, W. et al. Evolution of electronic structure in atomically thin sheets of Acknowledgements
WS2 and WSe2. ACS Nano 7, 791–797 (2013). This work was mainly supported by the US Department of Energy, Basic Energy Sciences,
25. Yin, Z. et al. Single-layer MoS2 phototransistors. ACS Nano 6, 74–80 (2012). Materials Sciences and Engineering Division (DE-SC0008145). N.G., J.Y. and D.M. are
26. Lopez-Sanchez, O., Lembke, D., Kayci, M., Radenovic, A. & Kis, A. supported by the US DoE, BES, Materials Sciences and Engineering Division. W.Y. is
Ultrasensitive photodetectors based on monolayer MoS2. Nature Nanotech. supported by the Research Grant Council of Hong Kong (HKU705513P), the University
8, 1–5 (2013). Grant Committee (AoE/P-04/08) of the government of Hong Kong and the Croucher
27. Huang, J-K. et al. Large-area synthesis of highly crystalline WSe2 monolayers Foundation under the Croucher Innovation Award. D.C. is supported by the US DoE, BES,
and device applications. ACS Nano 8, 923–930 (2014). Materials Sciences and Engineering Division (DE-SC0002197). Device fabrication was
28. Korn, T., Heydrich, S., Hirmer, M., Schmutzler, J. & Schüller, C. Low- performed at the Washington Nanofabrication Facility and National Science Foundation-
temperature photocarrier dynamics in monolayer MoS2. Appl. Phys. Lett. 99, funded Nanotech User Facility.
102109 (2011).
29. Shi, H. et al. Exciton dynamics in suspended monolayer and few-layer MoS2 2D
crystals. ACS Nano 7, 1072–1080 (2013).
Author contributions
X.X. conceived the experiments. J.S.R. fabricated the devices and performed the
30. Lagarde, D. et al. Carrier and polarization dynamics in monolayer MoS2.
measurements, assisted by P.K. and A.M.J. J.S.R. and X.X. performed data analysis, with
Phy. Rev. Lett. 112, 047401 (2014).
input from D.C. and W.Y. N.G., J.Y. and D.G.M. synthesized and performed bulk
31. Sim, S. et al. Exciton dynamics in atomically thin MoS2: interexcitonic
characterization measurements on the WSe2 crystals. T.T., K.W. and K.K. provided boron
interaction and broadening kinetics. Phys. Rev. B 88, 075434 (2013).
nitride crystals. X.X, D.C., J.S.R and W.Y. wrote the paper. All authors discussed the results.
32. Yao, W., Xiao, D. & Niu, Q. Valley-dependent optoelectronics from
inversion symmetry breaking. Phys. Rev. B 77, 235406 (2008).
33. Pospischil, A., Furchi, M. M. & Mueller, T. Solar-energy conversion and Additional information
light emission in an atomic monolayer p–n diode. Nature Nanotech. Supplementary information is available in the online version of the paper. Reprints and
http://dx.doi.org/10.1038/nnano.2014.14 (2014). permissions information is available online at www.nature.com/reprints. Correspondence and
34. Baugher, B. W. H., Churchill, H. O. H., Yang, Y. & Jarillo-Herrero, P. requests for materials should be addressed to X.X.
Optoelectronic devices based on electrically tunable p–n diodes
in a monolayer dichalcogenide. Nature Nanotech. Competing financial interests
http://dx.doi.org/10.1038/nnano.2014.25 (2014). The authors declare no competing financial interests.

272 NATURE NANOTECHNOLOGY | VOL 9 | APRIL 2014 | www.nature.com/naturenanotechnology

© 2014 Macmillan Publishers Limited. All rights reserved.