www.nature.

com/scientificreports

OPEN An anomalous interlayer exciton in

MoS2
Dilna Azhikodan1, Tashi Nautiyal1, Sam Shallcross2 & Sangeeta Sharma3

The few layer transition metal dichalcogenides are two dimensional materials that have an intrinsic
received: 02 August 2016
gap of the order of ≈2 eV. The reduced screening in two dimensions implies a rich excitonic physics and,
accepted: 24 October 2016 as a consequence, many potential applications in the field of opto-electronics. Here we report that a
Published: 14 November 2016 layer perpendicular electric field, by which the gap size in these materials can be efficiently controlled,
generates an anomalous inter-layer exciton whose binding energy is independent of the gap size. We
show this originates from the rich gap control and screening physics of TMDCs in a bilayer geometry:
gating the bilayer acts on one hand to increase intra-layer screening by reducing the gap and, on the
other hand, to decrease the inter-layer screening by field induced charge depletion. This constancy
of binding energy is both a striking exception to the universal reduction in binding energy with gap
size that all materials are believed to follow, as well as evidence of a degree of control over inter-layer
excitons not found in their well studied intra-layer counterparts.

The experimental fabrication of graphene in 2004 marked the onset of a new science of low dimensional materi-
als. Of the many materials that have followed graphene, one of the most remarkable has been the few layer transi-
tion metal dichalcogenides1–3 (TMDCs). Similar to graphene in some ways, for instance both have a valley degree
of freedom4, these materials differ in a number of key respects. While graphene is a semi-metal the TMDCs have
an intrinsic band gap, an inheritance from the bulk crystal which is a semi-conductor. Furthermore, a high degree
of control over this gap exists in bilayer TMDCs5, with theory6,7 reporting that layer perpendicular electric fields
of only 2–3 V/nm can close a gap of ≈​2 eV. As a band gap plays a crucial role in the on-off ratio of field-effect
transistors, the presence of an intrinsic and controllable band gap is a profound technological advantage8,9. Most
importantly, the TMDCs host a rich physics of excitons (a bound state formed of a valence band hole and a
conduction band electron) with exceptionally large binding energy5,10–16. This opens up a vast arena of potential
opto-electronic applications for these materials17. In this paper, through the example of MoS2, we will investigate
the theory of the interplay between these two key aspects of TMDCs: the electric field control of the band gap and
the excitonic physics.
Theoretical understanding of the TMDCs is non-trivial. Firstly, there are many indications that electron cor-
relation plays a decisive role in the properties of these materials6,18, a point of view we will richly confirm in this
paper. Secondly, excitonic physics by its very nature demands a many-body treatment. A standard density func-
tional theory approach will not suffice for the investigation of these materials and recourse must be made to the
full machinery of many-body physics: the GW method must be deployed for the ground state followed by the
solution of the Bethe-Salpeter equation19 (BSE), or a time dependent density functional theory20, to access the
excitonic physics. These many-body techniques are computationally very demanding and thus only a few such
calculations6,21,22 exist, all of which are for the case of zero electric field.
A recent experimental investigation reports an intriguing excitonic physics of MoS2 in the presence on a layer
perpendicular electric field, with intra- and inter-layer excitons appearing to behave quite differently5. It is the
purpose of the present paper to apply the full “tool kit” of many-body ab-initio science (GW in conjunction with
the BSE) to this system, and to fully reveal the remarkable nature of this excitonic physics of MoS2 in an electric
field.

Results and Discussion

Band gap engineering. The behaviour of bilayer MoS2 in a layer perpendicular electric field has been well
established by DFT studies: the band gap can be continuously reduced with increasing field strength5,23–25. This
tuning of the band gap is so dramatic that both (i) a transition from indirect to direct band gap and (ii) full

1
Department of Physics, Indian Institute of Technology Roorkee, Uttarakhand - 247 667, India. 2Lehrstuhl für
Theoretische Festkorperphysik, Staudtstr. 7-B2 91058 Erlangen, Germany. 3Max Planck Institute of Microstructure
Physics, Weinberg 2, D-06120 Halle, Germany. Correspondence and requests for materials should be addressed to
S.S. (email: sharma@mpi-halle.mpg.de)

Scientific Reports | 6:37075 | DOI: 10.1038/srep37075 1

www.nature.com/scientificreports/

Figure 1. Direct and indirect band gaps (in eV) as a function of electric field (in V/nm). Calculations are
performed using GW method and DFT.

Figure 2. The imaginary part of the in-plane dielectric function of bilayer MoS2. Results are presented for
four different values of the layer perpendicular field as indicated. The inset presents the behaviour of ε2 over a
wide energy range. Two distinct types of excitonic behaviour may be observed: an exciton (A) whose resonance
energy does not change as a function of the electric field, and a second exciton (II) whose resonance energy
changes strongly with applied field. All results are calculated by solving BSE in presence of an electric field and
spin-orbit coupling.

metalization can be observed (at 1–2 V/nm and 2–3 V/nm respectively, depending on the stacking type of the
bilayer). As the ground state spectrum is known to be strongly - and non-uniformly - renormalized by many-
body effects, we will first revisit the band gap behaviour using the GW method. In Fig. 1 we present both DFT (we
deploy the PBE functional) and GW calculations of the band gap as a function of field; we show results only for
the AA′​stacking, known to be the lowest energy stacking configuration26 of the MoS2 bilayer. Clearly, many-body
effects play a crucial role in the behaviour of the band gap: while the gap reduces at 604 meV per unit field with
the DFT calculation (in good agreement with earlier work23,25, this value is almost half −​384 meV - if many-body
effects are included via the GW method. These GW results are much closer to the experimental value5 of 260 meV
per unit field. We also note that the band gap error from the DFT calculation appears to be increasing with field:
an error of 30% at zero field becomes an error of almost 70% at 2 V/nm. This strongly suggests that many-body
effects are both crucial to the treatment of bilayer MoS2 in an electric field, and indeed become more important
with increasing field. As the difference between the two methods lies in the much better treatment of screening in
the GW method, and as excitonic physics depends crucially on the screening of the excitons, we conclude that a
GW treatment of the ground state is essential.

Field induced excitonic physics. The imaginary part of the dielectric function ε2, calculated by solving the
BSE, is exhibited in Fig. 2 for a range of electric fields. Below the fundamental gap and in absence of the electric
field ε2 is dominated by two main excitonic peaks, marked as A and B at energies 1.99 eV and 2.19 eV respectively
(these peak positions are in excellent agreement with previous theoretical works18,21). When the field is applied
two things occur: (a) the A and B peaks split, but otherwise remain largely unchanged in energy as the field
changes and (b) above a critical field of 0.3 V/nm a new spin - orbit split excitonic peak (II-1 and II-2) emerges
whose resonance energy changes strongly as a function of field. The second of these features is the novel inter-
layer excitonic peak described in the experimental work of Chu et al. (the spin - orbit splitting is likely too small
to have been observed, see the broad peaks in Fig. 4c and error bars in Fig. 4e of ref. 5). While the experiment
clearly observed a dramatically different behaviour of peak positions A and II, measurement accuracy precluded

Scientific Reports | 6:37075 | DOI: 10.1038/srep37075 2

www.nature.com/scientificreports/

Figure 3. (a) Excitonic resonance energy and (b) binding energy (difference between the position of the
excitonic peak and the GW direct band gap) as a function of the electric field applied to a MoS2 bilayer.
Experimental data for the peak position from ref. 5 are also presented for comparison. Note the opposite
trend of peak position and binding energies for the excitons: while exciton II has a peak position that depends
strongly on applied field its binding energy is constant, and vice versa for exciton A.

Figure 4. (a) Exciton binding energy and (b) static screening as a function of the band gap. In panel (a) we
see that the binding energy of the inter-layer exciton (black circles) is almost independent of the gap size,
dramatically violating the universal law that all other materials follow. The two lines are best fits to the inter-
layer exciton and all other materials. In panel (b) we display static screening a diverse set of two and three
dimensional materials have been studied. 3D materials in the order of increasing band gaps: Ge, Si, GaAs, AlAs,
ZnTe, SiC, ZnSe, SrTiO3, BaTiO3, ZnS, Diamond, BN, AlN, NaH, Kr, LiF, Ar, Ne. 2D materials: mono- and bi-
layers of MoS2, MoSe2, WS2 and WSe2 in zero field. Evidently, there is a very well followed universal linear law
for the dependence of static screening on gap size.

a statement being made about the binding energies (difference between the gap and the position of the excitonic
peak) of these excitons. Theoretically we have no such restriction, and can now proceed to a careful analysis of
this novel excitonic behaviour.
We first confirm the intra- and inter-layer character of the exciton types A and II. This may be easily checked
as it is well established that all bands near the gap approach pure layer character as the field increases. Thus by
restricting the sub-space in which the Bethe-Salpeter equation is solved to conduction and valence manifolds
from the same, or different, layers we may immediately deduce which excitons result from intra- or inter-layer
transitions. We are thus, in this way, able to confirm that exciton peak A (and B) are purely intra-layer excitons,
while exciton peak II is an inter-layer exciton.
Having established this, in Fig. 3(a) we first present the peak position as a function of applied field, finding
agreement with experiment in the fact that the A-1 and A-2 peaks remain almost unchanged with field while, on
the other hand, the position of the II-1 and II-2 peaks changes strongly. The absolute positions of all resonances
are blue shifted with respect to experiments5,10 by ~0.14 eV, a fact that has already been noticed at zero field21. This
deviation from experiment is almost independent of the applied field, and we therefore find a rate of change of
peak II with applied field to be very close to the experimental value of 275 meV per unit field (the best find line to
the theoretical data yields 384 meV per unit field). We now turn to the binding energy of excitons A and II, pre-
sented in Fig. 3(b). Remarkably, the evolution in the position of peak II with field is almost perfectly compensated
by a corresponding change in the band gap. On the other hand the constant peak position of exciton A-2, due to
the changing band gap, corresponds to a continuously reducing binding energy with field (exciton A is not bound

Scientific Reports | 6:37075 | DOI: 10.1038/srep37075 3

www.nature.com/scientificreports/

Figure 5. Oscillator strength, calculated by solving the BSE, as a function of electric field for peak II. At
each point the oscillator strength for peak A-1 is written in brackets.

for field values ≥​0.6 V/nm). We therefore conclude that the trends in binding energy - the physically relevant
property of the exciton - are the opposite of the peak position trends: the binding energy of exciton II is weakly
dependent on the field, while that of A is strongly dependent on the field.
An exciton binding energy that is almost independent of the gap is a strikingly anomalous behaviour, a point
that may be vividly brought out by comparison to the gap dependence of binding energies found generally in
two dimensional materials. In Fig. 4(a) we present the binding energy as a function of gap size for a wide range
of two dimensional materials. In the absence of the anomalous exciton II, all binding energies fall close to a best
fit line, a universal behaviour first noticed by Choi et al.27. In Fig. 4(a) we include several additional materials to
those considered in ref. 27 which both changes the slope of the universal line and worsens the fit, however to a
good approximation the linear law still holds. What lies behind this universal relation discovered by Choi et al. To
answer this in panel (b) of Fig. 4 we present macroscopic ε−1 (ω →​  0, q =​  0) - a direct measure of the screening -
plotted against the gap size for a diverse set of materials (including both two and three dimensional cases). An
almost perfect universal linear behaviour can be observed, and thus the linear law for the exciton binding energy27
represents a manifestation of a deeper linear screening law. The constancy of the binding energy of exciton II as
a function of the band gap in bilayer MoS2, in conjunction with the “regular” behaviour of exciton A, is therefore
indicative of an unusual screening physics.
To further elucidate this physics we note that in an increasing electric field the charge accumulates in-plane
while being depleted from the interlayer region23. As a consequence of this field induced change in charge density
intra-layer screening will become stronger with increasing field (more charge to screen with) while, at the same
time, having a tenancy to reduce inter-layer screening (less charge to screen with). This explains the binding
energy trend of both excitons: the in-plane exciton (A) is “screened away” due to this increased intra-layer screen-
ing while, in contradistinction, once the interlayer screening is sufficiently weak the binding energy of exciton II is
insensitive to further reduction in screening. This critical value occurs at a field strength of ≤​0.3 V/nm and below
this value distinct II and A excitons cannot be determined. By performing a BSE calculation in the subspace of
only inter-layer bands we confirm that the exciton II is created at a field of 0.18 V/nm.
It is important to mention that the new inter-layer exciton II is bright but, as is already clear from Fig. 2,
comes with a very low intensity. To probe this further we plot in Fig. 5 the oscillator strength (OS)19 for peak II as
a function of field. Evidently, the exciton II OS is not only very low, for instance at 0.1 V/nm the OS for peak A is
about 20 times greater, but furthermore continuously reduces with increasing field strength; at 1.5 V/nm this ratio
of oscillator strengths is 140. These results are in perfect consonance with the experimental finding of a very low
photoluminescence yield for the inter-layer exciton as compared to the intra-layer exciton.

Summary and Discussion

Gating the bilayer creates a new inter-layer charge transfer exciton, recently reported in experiments5. We find
excellent agreement with experimental measurements of the exciton peak positions, and in addition our theoret-
ical analysis affords considerable additional insight into the nature of the excitonic physics. In particular, we find
that the interlayer exciton (II) has a binding energy that is, quite remarkably, independent of the gap size. We trace
this anomalous behaviour to the fundamental screening physics of the bilayer: an electric field acts to increase
intra-layer screening but decrease inter-layer screening. This new excitonic physics is therefore tied to the basic
physics of a gated bilayer geometry, and goes beyond the example we have presented. We conclude that the elec-
tric field induced excitonic physics of the TMDC bilayers will exhibit two distinct types of excitonic behaviour:
(i) in-plane excitons will become, following the universal trend, less strongly bound as the field closes the gap but,
(ii), inter-layer excitons will not follow this universal trend, and the binding energy will be unchanged once the
inter-layer screening falls below some critical value. The electric field induced excitonic physics of bilayer TMDCs
therefore offers an unexpectedly rich degree of control over the exciton binding energy, a feature that may be of
considerable importance in opto-electronic applications.

Computational Details
Geometry. The calculations are performed using a super-cell geometry with a large vacuum21 of 30 Å
(56.7 a.u.) and an in-plane lattice parameter fixed to the bulk experimental value of 3.160 Å28. The details of the

Scientific Reports | 6:37075 | DOI: 10.1038/srep37075 4

www.nature.com/scientificreports/

Figure 6. The imaginary part of the in-plane dielectric function of bilayer MoS2. Results are calculated with
different number of k-points.

AA’ stacked configuration that we calculate may be found in ref. 26. We fix the interlayer separation to the same
value at all fields; the gap reduction of the GW calculations (but not the DFT calculations) is then found to agree
closely with experiment. We note that relaxing this constraint has, in one work25, been reported to dramatically
impact the gap reduction with field, however in such a way that brings the results into sharp disagreement with
experiment5.

Electronic structure calculations. The ground-state density functional theory (DFT) calculations are
performed using the Perdew-Burke-Ernzerhof (PBE)29 exchange-correlation functional using the PAW method
as implemented in the VASP30 code. We have employed the non-self-consistent GW method31 to determine the
many-body gaps, with excitonic properties calculated by solving the Bethe Salpeter equation (BSE)19 in the q →​  0
limit. Optical spectra was calculated using an energy cutoff of 400 eV, a Gaussian smearing of 0.02 eV, 200 unoc-
cupied bands, and a force tolerance of 0.01 eV/Å. We have included 50 G-vectors to account for local field effects.
Spin-orbit interaction is included in all calculations. The ELK code32 has been deployed to verify the veracity of
the BSE results.

k-point convergence. In order to obtain reliable binding energies it is of vital importance to fully converge
the optical spectra. As in a previous work21, we find that excitonic binding energies are very sensitive to the num-
ber of k-points. As is clear from Fig. 6 a minimum grid of 18 ×​  18 ×​ 1 is required to obtain reliable binding ener-
gies (to an accuracy of 0.01 eV), which are then fully converged at a grid size of 24 ×​  24 ×​ 1. Hence all the results
presented in previous sections are calculated using a k-point grid of 24 ×​  24 ×​  1.

Electric field. The electric field is included as a scalar potential in the Kohn-Sham Hamiltonian. This potential
has a saw-tooth shape, a fact that follows from the imposition of periodic boundary conditions, and gradient of
this scalar potential, which is constant across the unit-cell, is then the electric field. Due to this extra term in the
Hamiltonian, the Kohn-Sham eigenvalues and eigenvectors are modified. These quantities then lead to an implicit
presence of the electric-field in the single particle Green’s function G0 and the RPA dielectric function (i.e. the
screening), ε−1, and hence the screened Coulomb potential W0 =​  ε−1v.

References
1. Butler, S. Z. et al. Progress, challenges, and opportunities in two-dimensional materials beyond graphene. ACS Nano 7, 2898–2926
(2013).
2. Chhowalla, M. et al. The chemistry of two-dimensional layered transition metal dichalcogenide nanosheets. Nature Chemistry 5,
263–275 (2013).
3. Wang, Q. H., Kalantar Zadeh, K., Kis, A., Coleman, J. N. & Strano, M. S. Electronics and optoelectronics of two-dimensional
transition metal dichalcogenides. Nature Nanotechnology 7, 699–712 (2012).
4. Xiao, D., Liu, G.-B., Feng, W., Xu, X. & Yao, W. Coupled spin and valley physics in monolayers of MoS 2 and other group-VI
dichalcogenides. Phys. Rev. Lett. 108, 196802 (2012).
5. Chu, T., Ilatikhameneh, H., Klimeck, G., Rahman, R. & Chen, Z. Electrically tunable bandgaps in bilayer MoS2. Nano Letters 15,
8000–8007 (2015).
6. Conley, H. J. et al. Bandgap engineering of strained monolayer and bilayer MoS2. Nano Letters 13, 3626–3630 (2013).
7. Splendiani, A. et al. Emerging photoluminescence in monolayer MoS2. Nano Letters 10, 1271 (2010).
8. Radisavljevic, B., Radenovic, A., Brivio, J., Giacometti, V. & Kis, A. Single-layer MoS2 transistors. Nature Nanotechnology 6, 147–150
(2011).
9. Yin, Z. et al. Single-layer MoS2 phototransistors. ACS Nano 6, 74–80 (2012).
10. Mak, K., Lee, C., Hone, J., Shan, J. & Heinz, T. Atomically thin MoS2: A new direct-gap semiconductor. Physical Review Letters 105,
136805–0 (2010).
11. Klots, A. R. et al. Probing excitonic states in suspended two-dimensional semiconductors by photocurrent spectroscopy. Scientific
Reports 4, 6608– (2014).
12. Li, Z. et al. Active light control of the MoS2 monolayer exciton binding energy. ACS Nano 10158 (2015).
13. Palummo, M., Bernardi, M. & Grossman, J. C. Exciton radiative lifetimes in two-dimensional transition metal dichalcogenides.
Nano Letters 15, 2794–2800 (2015).
14. Steinhoff, A. et al. Efficient excitonic photoluminescence in direct and indirect band gap monolayer MoS 2. Nano Letters
150831160549006– (2015).

Scientific Reports | 6:37075 | DOI: 10.1038/srep37075 5

www.nature.com/scientificreports/

15. Zhu, B., Chen, X. & Cui, X. Exciton binding energy of monolayer WS2. Scientific Reports 5, 9218– (2015).
16. Hanbicki, A., Currie, G., Kioseoglou, M., Friedman, A. & Jonker, B. Measurement of high exciton binding energy in the monolayer
transition-metal dichalcogenides WS2 and WSe2. Solid State Communications 203, 16–20 (2015).
17. Bernardi, M., Palummo, M. & Grossman, J. C. Extraordinary sunlight absorption and one nanometer thick photovoltaics using two-
dimensional monolayer materials. Nano Letters 13, 3664–3670 (2013).
18. Komsa, H.-P. & Krasheninnikov, A. V. Effects of confinement and environment on the electronic structure and exciton binding
energy of MoS2 from first principles. Phys. Rev. B 86, 241201 (2012).
19. Salpeter, E. E. & Bethe, H. A. A relativistic equation for bound-state problems. Phys. Rev. B 84, 1232 (1956).
20. Sharma, S., Dewhurst, J. K., Sanna, A. & Gross, E. K. U. Bootstrap approximation for the exchange-correlation kernel of time-
dependent density-functional theory. Phys. Rev. Lett. 107, 186401 (2011).
21. Molina Sanchez, A., Sangalli, D., Hummer, K., Marini, A. & Wirtz, L. Effect of spin-orbit interaction on the optical spectra of single-
layer, double-layer, and bulk MoS2. Phys. Rev. B 88, 045412 (2013).
22. Singh, N., Jabbour, G. & Schwingenschlögl, U. Optical and photocatalytic properties of two-dimensional MoS2. Eur. Phys. J. B 85,
392 (2012).
23. Ramasubramaniam, A., Naveh, D. & Towe, E. Tunable band gaps in bilayer transition-metal dichalcogenides. Phys. Rev. B 84,
205325 (2011).
24. Zhang, Z. Y. et al. Indirect-direct band gap transition through electric tuning in bilayer MoS2. The Journal of Chemical Physics 140,
174707– (2014).
25. Liu, Q. et al. Tuning electronic structure of bilayer MoS2 by vertical electric field: A first-principles investigation. The Journal of
Physical Chemistry C 116, 21556–21562 (2012).
26. He, J., Hummer, K. & Franchini, C. Stacking effects on the electronic and optical properties of bilayer transition metal
dichalcogenides MoS2, MoSe2, WS2, and WSe2. Phys. Rev. B 89, 075409 (2014).
27. Choi, J.-H., Cui, P., Lan, H. & Zhang, Z. Linear scaling of the exciton binding energy versus the band gap of two-dimensional
materials. Phys. Rev. Lett. 115, 066403 (2015).
28. Boker, T. et al. Band structure of MoS2, MoSe2, and α–MoTe2 angle-resolved photoelectron spectroscopy and ab initio calculations.
Phys. Rev. B 64, 235305 (2001).
29. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple. Phys. Rev. Lett. 78, 1396 (1997).
30. Kresse, G. & Furthmuller, J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys. Rev.
B 54, 11169–11186 (1996).
31. Shishkin, M. & Kresse, G. Implementation and performance of the frequency-dependent GW method within the PAW framework.
Phys. Rev. B 74, 035101 (2006).
32. Dewhurst, J. K., Sharma, S. et al. URL http:/elk.sourceforge.net/.

Acknowledgments
S.S acknowledges Department of Physics, Indian Institute of Technology Roorkee for the visiting position.
D.A acknowledges Ministry of Human Resource Development, Government of India and the HPC facility at
Max Planck Institute of Microstructure Physics, Halle, Germany.

Author Contributions
S.S. framed the project and S.S. and T.N. supervised the calculations. D.A. and S.S. performed the calculations.
T.N., S.S. and S.Shal. participated in the analysis of the results. All authors wrote the manuscript together.

Additional Information
Competing financial interests: The authors declare no competing financial interests.
How to cite this article: Azhikodan, D. et al. An anomalous interlayer exciton in MoS2. Sci. Rep. 6, 37075; doi:
10.1038/srep37075 (2016).
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
This work is licensed under a Creative Commons Attribution 4.0 International License. The images
or other third party material in this article are included in the article’s Creative Commons license,
unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license,
users will need to obtain permission from the license holder to reproduce the material. To view a copy of this
license, visit http://creativecommons.org/licenses/by/4.0/

© The Author(s) 2016

Scientific Reports | 6:37075 | DOI: 10.1038/srep37075 6