Majorana Fermions on Zigzag Edge of Monolayer Transition Metal Dichalcogenides
Abstract
Majorana fermions, quantum particles with nonAbelian exchange statistics, are not only of fundamental importance, but also building blocks for faulttolerant quantum computation. Although certain experimental breakthroughs for observing Majorana fermions have been made recently, their conclusive dection is still challenging due to the lack of proper material properties of the underlined experimental systems. Here we propose a new platform for Majorana fermions based on edge states of certain nontopological twodimensional semiconductors with strong spinorbit coupling, such as monolayer groupVI transition metal dichalcogenides (TMD). Using firstprinciples calculations and tightbinding modeling, we show that zigzag edges of monolayer TMD can host well isolated single edge band with strong spinorbit coupling energy. Combining with proximity induced swave superconductivity and inplane magnetic fields, the zigzag edge supports robust topological Majorana bound states at the edge ends, although the twodimensional bulk itself is nontopological. Our findings points to a controllable and integrable platform for searching and manipulating Majorana fermions.
Majorana fermionsWilczek () are quantum particles that are their own
antiparticles, and satisfy nonAbelian exchange statistics. The latter is
the key for their potential use in faulttolerant topological quantum
computationTQC (), which makes their experimental realization an
extremely important task from the longterm technological perspective. In
the past two decades, some exotic condensed matter/cold atom systemsTQC (); Fuliangmajo (); CW (); Sauprl (); Lutchynprl (); Oregprl (); Aliceaprb (); Mao () have been
proposed to support Majorana fermions. The experimental breakthrough occurs
only recentlyKouwenhoven (); Dengnano (); Dasnat (); Rokhinson (), using
heterostructures consisting of conventional swave superconductors and
semiconducting nanowires subjected to an external magnetic field, where
certain signatures of Majorana fermions were observed. However, there are a
few material complications inherent to semiconductor nanowires that may
prevent the experimental signature from being conclusiveLiujieprl (); Altlandprl (); Kellsprb (); JHLeeprl (); Pikulinnjp (); Sarmarapid (): (i)
The large diameters of the nanowires yield multiple occupied transversal sub
bands, resulting in complications for the superconductor proximity effect
and the chemical potential levelWimmer10prl (); Potter10prl (); Lutchyn11prl (); (ii) The spinorbit coupling in
these wires is rather weak, which renders the Majorana physics extremely
vulnerable to disorder, making it challenging to exclude alternative
interpretations of the experimental signature based on disorder effectStanescu11prb (); Potter11prb (); Liujieprl (); (iii) The random growth process of
nanowires also makes it hard to build a nanowire networkAliceanat () to
detect the statistics of Majorana fermions.
Amid the above difficulty, it is critically important to look for other 1D
conducting states to realize Majorana fermions. A natural and more
controllable way is to consider 1D edge states of a 2D material. In this
context, helical edge states of 2D quantum spin Hall insulators (QSHIs) have
been proposed to support Majorana fermionsFuliangqshimajo (); du1 (); xlq ().
However, so far the QSHIs have only been realized in semiconductor
heterostructures and are subjected to stringent growth conditions.
Furthermore, the bulk itself of a QSHI is generally not a good insulator
because of the relatively small band gap (about 10 meV). It is therefore
natural to ask whether the edge states of nontopological 2D materials with
a large bulk band gap can support Majorana fermions. While in pursuit of
such platforms a few key material properties are of particular interest: (i)
The compounds must have heavy elements that can generate strong spinorbit
coupling (SOC) necessary for robust 1D topological superconductorsPotter12prb (); sau2 (); and (ii) 2D atomically thin materials with
honeycomblike lattice structures (i.e., similar as graphene),
which are more likely to support single band edge states.
In this article we demonstrate this idea by showing that 1D zigzag edges of
a new class of 2D semiconductors, monolayer groupIV transition metal
dichalcogenides (TMDs), provide a promising new platform for studying 1D
topological superconductors with a single transversal band, strong SOC
energy, and controlled Majorana network generation. Using both
firstprinciples calculations and tightbinding modeling, we show that the
chalcogenterminated zigzag edges of these 2D semiconductors support edge
bands with strong Rashbatype SOC and are well separated from the large bulk
bands (1.5 to 1.8 eV). By utilizing a minimal realistic tightbinding
model, we numerically confirm the existence of zeroenergy Majorana states
at the two ends of the edge in the presence of proximity induced swave superconductivity, and their robustness against disorders. Our
findings point out a new pathway for searching for Majorana fermions using
edge states of widely existing 2D nontopological semiconductors.
Results
Isolated single edge band. We consider four different TMDs: MoS, MoSe, WS and WSe, but will present our results in
the following mainly using WSe as a representative because of its much
larger SOC energy. Monolayer TMDs are atomically thin 2D directbandgap
semiconductors with exotic coupled spin and valley physics xiaodimo ()
and excellent optical properties, as demonstrated in recent experimentsMak10prl (); AKis (); caonatcom (); Mak2 (); xiaodongcui (); xu1 (); xu2 (). Structurally
monolayer MX is a trilayer XMX sandwich. Within each layer, M and X
atoms form 2D hexagonal lattices. When viewed from top it shows a honeycomb
structure. The 2D bulk of monolayer MX has a direct band gap of 1.5
1.8 eV located at the corners of its 2D hexagonal Brillouin zone called
valleysxiaodimo (). The bulk’s edges can be classified as zigzag and
armchair types like in graphene. Due to the lacking of inversion symmetry in
the monolayer, the zigzag edges can be further classified as Xterminated
and Mterminated, which correspond to the lower and upper edges of the
ribbon shown in Fig. 1a. We refer to them as Xedge and Medge respectively.
It is already known from STM measurements that the zigzag edges of
triangular shaped monolayer MoS nanoflakes support multiple pairs of
1D metallic edge statesHelveg (). The edges of these nanoflakes are
later identified as Moedge with passivated S atomsBollinger (). For a
zigzag MX ribbon shown in Fig. 1a, the edge states exist on both the
Medge and Xedge.
In addition to the density functional theory (DFT) calculations, insight
into the underlying physics can be obtained from a minimal tightbinding
model that is constructed by considering the lattice symmetry and the
corresponding crystal field splitting. It is known that the valence band
maximum and conduction band minimum of monolayer MX consist mainly of
M atom’s d orbitals. Thus to describe the low energy band structure
of the monolayer’s bulk it is sufficient to consider the d orbitals
from the M atomsxiaodimo (). The trigonal prismatic coordination of the
M atom splits its d orbitals into three groups: , , ) and , . The monolayer’s mirror symmetry in the
direction permits hybridization only between the and groups. This allows us to consider three orbitals of , , and for a minimal tightbinding model. We
refer the readers to RefLiuguibin () for detailed descriptions of this
tightbinding model including the symmetry analysis and material specific
parameters fitted from firstprinciple calculations. The tightbinding model
is able to capture the essential physics of the monolayer, including the
direct band gaps at the and valleys, the degeneracy of the
band edges and the valley contrast spin splitting of the valence band due to
SOC, etcxiaodimo (); Liuguibin ().
In simple languages this tightbinding model only considers M atoms’ onsite energies and electron hopping along the six vectors connecting the nearest M atoms (marked as in Fig. 1a). Without considering SOC and the spin degree of freedom, the tightbinding Hamiltonian can be written as a matrix
in the space. Here the basis is taken as and represents
the matrix element between and , and is obtained from the Fourier transformation of
the real space tunneling matrix between neighboring sites.
Rashbatype SOC in edge band. We proceed to include SOC and demonstrate how the Rashbatype SOC in the edge band is generated. The type SOC in MX originates from the d orbitals of the heavy M atoms (Mo or W)zhuzhiyong (); xiaodimo (). In monolayer;s Bulk, the spioribt term can be described as
(1) 
where and represent the spin and orbital angular momentum operator respectively. It turns out in the basis of the and are both matrices, which enables us to write the total Hamiltonian in a spindecoupled form
Here the upper and lower subblocks represent the spin up and down
respectively. It is noted that is a
timereversalinvariant Bloch Hamiltonian meaning , where is the
timereversal operator. The total Hamiltonian also respects the timereversal symmetry(TRS) in this sense. However,
when view each spin block individually, the TRS is explicitly broken by the
SOC term .
In Figs. 1b and 1c and Fig. 2 we present both the DFT and tightbinding band
structure of the WSe zigzag ribbon, where the edge states localized on
different edges are marked correspondingly. The electrons from these edge
bands dominantly resides on the M atoms of the zigzag edges. Comparing with
the DFT band structure, we see the tightbinding model can successfully
captures the parabolic edge bands on both the Xedge and Medge. It’s worth
to mention that the effective SOC we find in the parabolic Medge band is
generally larger than in the Xedge band. Nevertheless the Xedge is
preferable for two main reasons. First, Xedge hosts a single edge band
while the Medge hosts multiple edge bands; Second, the Xedge is
structurally very stable while the Medge can be dramatically affected by
edge passivationsSeifert (); PanHui (); Wangjacs (); Chenjacs (). Here after, we
will focus on the Xedge band.
In Fig.2b we compare the Xedge band before and after turning on the SOC. Apparently the can be viewed as a TRS breaking perturbation term, whose effect is slightly shifting the spin up branch to the left and spin down branch to the right. The whole band structure nevertheless remains symmetric about because of the TRS. Accordingly the low energy effective 1D Hamiltonian for the Xedge band can be written as
(2) 
where is the component of the Pauli matrix, , and is a constant. Up to a unitary transformation, this
Hamiltonian is equivalent to that for the semiconductor nanowires with
Rashbatype SOCLutchynprl (); Aliceanat (); Wimmer10prl (); Potter10prl (); Lutchyn11prl (); Kouwenhoven (); Dengnano (); Dasnat (). Here is the effective Rashba velocity. In Table 1
the effective mass and fitted from our firstprinciple
calculations for MoS, MoSe, WS and WSe are listed.
These parameters are orders of magnitude larger than their semiconductor
nanowire counterparts, especially for WS and WSe.
MoS  MoSe  WS  WSe  InAsAliceaprb ()  InSbKouwenhoven ()  
0.28  0.24  0.33  0.31  0.04  0.015  
0.12  0.11  0.33  0.46  0.06  0.2  
0.26  0.2  2.3  4.3  0.01  0.05 
Existence of Majorana end states. To create a 1D topological
superconductor, we introduce superconducting pairing through proximity
effects by depositing the MX monolayer on top of a conventional
swave superconductor (Nb, NbSe, etc.), as illustrated in
Fig. 3a. The MX monolayers have XMX layer thickness 3 Å, which is well within the superconducting coherence
length of a typical swave superconductor. The selected monolayer
can either be (1) a zigzag nanoribbon or (2) a large monolayer sample with
an identified Xedge. A top gate can then be applied locally to tune the
chemical potential. We note that in the first case there are coexisting edge
states on the Medge as shown in the ribbon’s band structures (Figs. 1b and
1c). These Medge states are well localized on the Medge’s M atoms. To
realize a topological superconducting state, the chemical potential
need be tuned to be around the Xedge band bottom near (see Fig.
2b). The corresponding Medge states at this chemical potential occur at
momenta far away from . In that region, even number of Moedge
bands are cut at the Fermi surface, which do not affect the topological
properties of the systemLutchynprl (). Therefore the Medge states do
not interfere with the topological superconducting state on the Xedge,
which is also confirmed in our following numerical simulations. Such
coexistence of edge bands is completely eliminated for the second case,
where the Xedge can be well isolated. For this case the gate is only
required to cover the selected segment of the edge since the bulk maintains
a large band gap.
To demonstrate the functionality of the proposed setup, we carry out a
numerical simulation in 2D with the tightbinding model. We adopt the ribbon
structure for this purpose as illustrated in Fig. 3a. Because of their
excellent structural stability, TMD zigzag nanoribbons can be synthesized
with uniform width and smooth edges without defects Wangjacs (); WangJMC (); ZhengLiu (). To drive the system into a topological
superconducting state, an inplane magnetic field is applied to create a
Zeeman splitting gap at the band crossing point.
The Zeeman term induced by the magnetic field reads
(3) 
where is the creation operator for electron on site with orbital index and spin index and . We have assumed the magnetic field is in the direction, but nevertheless any inplane magnetic field would work equivalently. The proximity effect induced superconducting paring term writes
(4) 
where for simplicity we assume a uniform intraorbit pairing strength. Denote the lattice version of the ribbon’s Hamiltonian in Eq. LABEL:totH as , we then solve the corresponding BdG equation for the full Hamiltonian (see Methods)
(5) 
to get the low energy spectrum. The emergence of the zeroenergy mode with
increasing Zeeman splitting is shown in Fig. 3b. In Fig. 3c we show the
evolution of the 6 lowest energy modes with the ribbon’s length. As a
signature of Majorana fermionsTudor13prb (), the zeroenergy modes show
an oscillating energy splitting with an exponentially decaying envelope.
When the ribbon is sufficiently long the zeroenergy mode as well as the
excitation gap become well defined. We notice that the fast alternating
modes in the excitation gap of a short ribbon are contributions from the
coexisting Medge bands, which confirms our prediction that they do not
affect the topological superconductor on the Xedge. It is well established
that in a 1D topological superconductor the Majorana fermions appear as end
states in real spaceSauprl (); Lutchynprl (). We confirm this by plotting
the particle component of the zeroenergy mode wave function in Fig.3d,
where the ribbon’s size is in length and in width.
Effect of disorder. It is important that Majorana fermions can sustain certain amount of disorders since in realistic experimental conditions disorders are unavoidable. To explore the disorder effect in this system, we add random onsite potential
(6) 
to the tightbinding model, where are normally
distributed in the range [,]. We have simulated two kinds of
disorders: (1) disorder covers both bulk and edge; (2) disorder only covers
the bulk but not the edge. As shown in Fig.4ac, the zeroenergy modes as
well as the excitation gap are robust against edge disorders up to ( can be much larger for a smaller ). For
even stronger edge disorders the excitation gap starts to diminish and the
zeroenergy modes gain splitting. An important advantage of this proposed
system is that the topological superconductor resides only on the edge, as a
result it gains strong immunity from bulk disorders. This is demonstrated in
Fig.4d where the zeroenergy modes and excitation gap remain totally intact
despite of the strong disorder in the bulk.
Material considerations. Although both the zigzag and armchair
edges of monolayer MX support edge states, the zigzag edges are
generally more stable than the armchair edges. Consequently during
nanoribbon synthesis the zigzag nanoribbons dominatesWangjacs (); WangJMC (); ZhengLiu (). In particular, the Xedge shows maximal
stability among all the edge configurations of monolayer MXWangjacs (); PanHui (); Chenjacs (). In the experimentally grown WS nanoribbon, the Sedge was found to be perfect without defects ZhengLiu (). Furthermore, edge states have been observed in MoS
nanoflakes using STMHelveg (); Bollinger ().
In the DFT band structures of MX zigzag ribbon (Fig.1b and
supplementary note), three pairs of ingap edge states exist connecting the
two inequivalent valleys. All of them consists dominantly of the d
orbitals from the M atoms of the ribbon’s outermost zigzag chain. Notably
two of these edge states have parabolic dispersions at their band
minimum/maximum near , which suits our purpose to find an effective
Hamiltonian like Eq.2. The two parabolic bands are also fully
captured in the tightbinding model. The edge state phenomenon is similar in
all MX zigzag ribbons we have calculated. After including SOC, the
Rashbatype SOC in the edge band is evident when zooming in the band bottom
at (Fig.2c and supplementary note)). We have also calculated
ribbons with different width. We find these well localized edge states start
to exist in very narrow ribbons ().
The Fermi surface in DFT calculations of suspended MX is typically in
the band gap and close to the valence band top. However, both ntypeAKis (); Xuacsnano (); Zandenatm () and ptypeptype1 (); ptype2 () conductivities
have been reported in transport measurements, suggesting a widerange
tunability for the chemical potentialsDolui (). The material synthesis
and device fabrication of monolayer MX are rapidly developing because
of their potential applications in the next generation of electronics. The
nanoribbon samples for a possible experimental realization for our proposed
setup have become readily availableWangjacs (); WangJMC (); ZhengLiu (). The
monolayer’s 2D nature and similarity with graphene also makes many well
developed 2D device engineering and fabrication techniques directly
applicable.
Discussion
The advantage of the proposed platform can be summarized in such a few
aspects: (1) The single edge band is well isolated in the middle of a
relatively large bulk band gap, which would lead to a minimal background
signal in the zerobias peak measurements for detecting Majorana fermions.
The well known multiband problem in the semiconductor nanowires is
successfully avoidedPotter10prl (). More importantly, the platform is a
true 1D system localizing on a single atomic chain resulting in strong
immunity from bulk’s disorders as we have demonstrated; (2) The system is
atomically thick, which would lead to robust and uniform superconducting
proximity effect when placed on top of the swave superconductor. It would
also result in efficient gate tunability, which has already been
demonstrated in transport measurements with the monolayerAKis (); YZhang ();
(3) The large effective mass and large effective RashbaSOC in this platform
is unparalleled to the conventional semiconductor nanowires. As a result,
the proximity induced superconducting pairing and the associated Majorana
fermions are robust against disordersStanescu11prb (); Potter11prb (); Potter12prb (); sau2 (); (4) The 1D edge states can
be controllably generated from 2D materials, making it possible to construct
a Majorana networkAliceanat () for studying the nonAbelian statistics
of Majorana fermions and implementing topological quantum computation.
Methods
Realspace BdG equation. Here we elaborate how the BdG calculation for Fig. 3 is implemented on the zigzag ribbon’s lattice. Denote the spinindependent hopping matrix from the nearest M atoms at site to as and the onsite potential matrices as diagLiuguibin (). Then the realspace BdG equation can be written as
where , when and , . and each being a vector
are the components of the th quasiparticle wave function at site , is the corresponding energy eigenvalue. The low energy
spectrum and wave function is then obtained by using the sparse matrix
eigensolver in MATLAB.
DFT calculation. Our firstprinciple DFT calculations are performed
using the allelectron fullpotential linearized augmentedplanewave
(FPLAPW) methodELK (). The SOC is included in terms of the
secondvariational method with scalarrelativistic orbitals as a basis.
Construction of our tightbinding model is described in details in RefLiuguibin ().
References
 (1) Wilczek, F. Majorana returns. Nature Phys. 5, 614618 (2009).
 (2) Nayak, C., Simon, S. H., Stern, A., Freedman, M. & Das Sarma, S. Nonabelian anyons and topological quantum computation. Rev. Mod. Phys. 80, 1083 (2008).
 (3) Fu, L. & Kane, C. L., Superconducting proximity effect and Majorana fermions at the surface of a topological insulator. Phys. Rev. Lett. 100, 096407 (2008).
 (4) Zhang, C., Tewari, S., Lutchyn, R. M. & Das Sarma, S. superfluid from save interactions of fermionic cold atoms. Phys. Rev. Lett. 101, 160401 (2008).
 (5) Sau, J. D., Lutchyn, R. M., Tewari, S. & Das Sarma, S. Generic new platform for topological quantum computation using semiconductor heterostructures. Phys. Rev. Lett. 104, 040502 (2010).
 (6) Lutchyn, R. M., Sau, J. D. & Das Sarma, S. Majorana fermions and a topological phase transition in semiconductorsuperconductor heterostructures. Phys. Rev. Lett. 105, 077001 (2010).
 (7) Oreg, Y., Refael, G. & Oppen, F. V. Helical liquids and Majorana bound states in quantum wires. Phys. Rev. Lett. 105, 177002 (2010).
 (8) Alicea, J. Majorana fermions in a tunable semiconductor device. Phys. Rev. B 81, 125318 (2010).
 (9) Mao, L., Gong, M., Dumitrescu, E., Tewari, S. & Zhang, C. Holedoped semiconductor nanowire on top of an wave superconductor: a new and experimentally accessible system for Majorana fermions. Phys. Rev. Lett. 108, 177001 (2012).
 (10) Mourik, V. et al. Signatures of Majorana fermions in hybrid superconductorsemiconductor nanowire devices. Science 336, 10031007 (2012).
 (11) Deng, M. T. et al. Anomalous zerobias conductance peak in a NbInSb nanowireNb hybrid device. Nano Lett. 12, 64146419 (2012).
 (12) Das, A. et al. Zerobias peaks and splitting in an AlInAs nanowire topological superconductor as a signature of Majorana fermions. Nature Phys. 8, 887895 (2012).
 (13) Rokhinson, L. P., Liu, X. & Furdyna, J. K. The fractional a.c. Josephson effect in a semiconductorsuperconductor nanowire as a signature of Majorana particles. Nature Phys. 8, 795799 (2012).
 (14) Liu, J., Potter, A. C., Law, K. T., and Lee, P. A., ZeroBias Peaks in the Tunneling Conductance of SpinOrbitCoupled Superconducting Wires with and without Majorana EndStates, Phys. Rev. Lett. 109, 267002 (2012).
 (15) Bagrets, D., & Altland, A. Class D Spectral Peak in Majorana Quantum Wires, Phys. Rev. Lett. 109, 227005 (2012).
 (16) Pientka, F., Kells, G., Romito, A., Brouwer, P. W., and von Oppen, F., Enhanced ZeroBias Majorana Peak in the Differential Tunneling Conductance of Disordered Multisubband QuantumWire/Superconductor Junctions, Phys. Rev. Lett. 109, 227006 (2012).
 (17) Lee, E. J. H. et al. ZeroBias Anomaly in a Nanowire Quantum Dot Coupled to Superconductors. Phys. Rev. Lett. 109, 186802 (2012).
 (18) Pikulin, D. I., Dahlhaus, J. P., Wimmer, M., Schomerus, H., and Beenakker, C.W. J., A zerovoltage conductance peak from weak antilocalization in a Majorana nanowire, New J. Phys. 14 125011 (2012).
 (19) Das Sarma, S., Sau, J. D., & Stanescu, T. D., Splitting of the zerobias conductance peak as smoking gun evidence for the existence of the Majorana mode in a superconductorsemiconductor nanowire, Phys. Rev. B 86, 220506 (2012).
 (20) Wimmer, M. et al. Majorana Bound States without Vortices in Topological Superconductors with Electrostatic Defects, Phys. Rev. Lett. 105, 046803 (2010).
 (21) Potter, A. C. & Lee, P. A. Multichannel generalization of Kitaev’s Majorana end states and a practical route to realize them in thin films. Phys. Rev. Lett. 105, 227003 (2010).
 (22) Lutchyn, R. M., Stanescu, T.D., and Das Sarma, S. Search for Majorana Fermions in Multiband Semiconducting Nanowires. Phys. Rev. Lett. 106, 127001 (2011).
 (23) Stanescu, T. D., Lutchyn, R. M., and Das Sarma, S., Majorana fermions in semiconductor nanowires, Phys. Rev. B 84, 144522 (2011).
 (24) Potter A. C. & Lee, P. A. Engineering a p+ip superconductor: Comparison of topological insulator and Rashba spinorbitcoupled materials. Phys. Rev. B 83, 184520 (2011).
 (25) Alicea, J., Oreg, Y., Refael, G., Oppen, F. V. & Fisher, M. P. A. NonAbelian statistics and topological quantum information processing in 1D wire networks. Nature Phys. 7, 412417 (2011).
 (26) Fu F. & Kane, C. L., Josephson Current and Noise at a SuperconductorQuantum Spin Hall InsulatorSuperconductor Junction, Phys. Rev. B 79, 161408(R) (2009).
 (27) Knez, I., Du, R.R. & Sullivan, G., Andreev Reflection of Helical Edge Modes in InAs/GaSb Quantum Spin Hall Insulator, Phys. Rev. Lett. 109, 186603 (2012).
 (28) Qi, X.L. & Zhang, S.C., Topological insulators and superconductors. Rev. Mod. Phys. 83, 1057 (2011).
 (29) Potter A. C. & Lee, P. A. Topological superconductivity and Majorana fermions in metallic surface states. Phys. Rev. B 85, 094516 (2012).
 (30) Sau, J. D., Tewari, S. & Das Sarma, S. Experimental and materials considerations for the topological superconducting state in electron and holedoped semiconductors: Searching for nonAbelian Majorana modes in 1D nanowires and 2D heterostructures. Phys. Rev. B 85, 064512 (2012).
 (31) Xiao, D., Liu, G.B., Feng, W., Xu, X. & Yao, W. Coupled spin and valley physics in monolayers of MoS and other groupVI dichalcogenides. Phys. Rev. Lett. 108, 196802 (2012).
 (32) Mak, K. F., Lee, C., Hone, J., Shan, J. & Heinz, T. F. Atomically thin MoS: A new directgap semiconductor. Phys. Rev. Lett. 105, 136805 (2010).
 (33) Radisavljevic, B., Radenovic, A., Brivio, J., Giacometti, V. & Kis, A. Singlelayer MoS transistors. Nature Nanotech. 6, 147–150 (2011).
 (34) Cao, T., Feng, J., Shi, J., Niu, Q. & Wang, E. Valleyselective circular dichroism of monolayer molybdenum disulphide. Nature Commun. 3, 887 (2012).
 (35) Mak, K. F., He, K., Shan, J. & Heinz, T. F. Control of valley polarization in monolayer MoS by optical helicity. Nature Nanotech. 7, 494–498 (2012).
 (36) Zeng, H., Dai, J., Yao, W., Xiao, D. & Cui, X. Valley polarization in MoS monolayers by optical pumping. Nature Nanotech. 7, 490–493 (2012).
 (37) Wu, S. et al. Electrical Tuning of Valley Magnetic Moment via Symmetry Control in Bilayer MoS. Nature Phys. 9, 149153 (2013).
 (38) Ross, J. S. et al. Electrical control of neutral and charged excitons in a monolayer semiconductor. Nature Commun. 4, 1474 (2013).
 (39) Helveg, S. et al. AtomicScale Structure of SingleLayer MoS Nanoclusters. Phys. Rev. Lett. 84, 951 (2000).
 (40) M. V. Bollinger et al. OneDimensional Metallic Edge States in MoS. Phys. Rev. Lett. 87, 196803 (2001).
 (41) Liu, G., Shan, W., Yao, Y., Yao, W., & Xiao, D. A threeband tightbinding model for monolayers of groupVIB transition metal dichalcogenides. arXiv:1305.6089.
 (42) Zhu, Z. Y., Cheng, Y. C. & Schwingenschlogl, U. Giant spinorbitinduced spin splitting in twodimensional transitionmetal dichalcogenide semiconductors. Phys. Rev. B 84, 153402 (2011).
 (43) Wang, Z. et al. Mixed LowDimensional Nanomaterial: 2D Ultranarrow MoS Inorganic Nanoribbons Encapsulated in Quasi1D Carbon Nanotubes. J. Am. Chem. Soc. 132, 13840 (2010).
 (44) Li, Y., Zhou, Z., Zhang, S., & Chen, Z. MoS Nanoribbons: High Stability and Unusual Electronic and Magnetic Properties. J. Am. Chem. Soc. 130, 16739 (2008).
 (45) Pan, H., and Zhang, Y. Edgedependent structural, electronic and magnetic properties of MoS2 nanoribbons. J. Mater. Chem. 22, 7280 (2012).
 (46) Erdogan, E., Popov, I.H., Enyashin, A.N., & Seifert, G. Transport properties of MoS nanoribbons: edge priority. Eur. Phys. J. B 85, 33 (2012).
 (47) Wang, Z. et al. Ultranarrow WS nanoribbons encapsulated in carbon nanotubes. J. Mater. Chem. 21, 171180 (2011).
 (48) Liu, Z. et al. Identification of active atomic defects in a monolayered tungsten disulphide nanoribbon. Nature Commun. 2, 213 (2011).
 (49) Stanescu, T., Lutchyn, R. M., & Das Sarma, S. Dimensional crossover in spinorbitcoupled semiconductor nanowires with induced superconducting pairing. Phys. Rev. B 87, 094518 (2013).
 (50) Wu, S. et al. VaporSolid Growth of High Quality MoS2 Monolayers With NearUnity Valley Polarization. ACS Nano. 7, 2768 (2013).
 (51) van der Zande, A. M. et al. Grains and grain boundaries in highly crystalline monolayer molybdenum disulphide. Nat. Mat. 12, 554(2013).
 (52) Zeng, Z. et al. SingleLayer Semiconducting Nanosheets: HighYield Preparation and Device Fabrication. Angew. Chem. Int. Ed. 50, 11093 (2011).
 (53) Zhan, Y., Liu, Z., Najmaei, S., Ajayan, P. M., & Lou, J. LargeArea VaporPhase Growth and Characterization of MoS Atomic Layers on a SiO Substrate. Small 8, 966 (2012).
 (54) Dolui, K., Rungger, I. & Sanvito, S. Origin of the ntype and ptype conductivity of MoS monolayers on a SiO substrate. arxiv:1301.2491
 (55) Zhang, Y. et al. Ambipolar MoS Thin Flake Transistors. Nano Lett. 12, 1136 (2012).

(56)
http://elk.sourceforge.net/
Acknowledgements
R.C., and C.Z. are supported by ARO (W911NF1210334), AFOSR
(FA95501310045), and NSFPHY (1104546). G.L acknowledges support by the
Basic Research Funds of Beijing Institute of Technology with Grant No.
20121842003 and the National Basic Research Program of China 973 Program
with Grant No. 2013CB934500. W.Y. acknowledges
support by HKSAR Research Grant Council (HKU705513P), and Croucher
Foundation under the Croucher Innovation Award. X.X is supported by US DoE, BES, Division of Materials Sciences and Engineering (DESC0008145). D.X. acknowledges support by the U.S.
Department of Energy, Office of Basic Energy Sciences, Materials Sciences
and Engineering Division, and AFOSR (FA95501210479).
Author contributions
All authors made critical contributions to the work.
Competing financial interests
The authors declare no competing financial interests.
Supplementary Materials