LETTERS

PUBLISHED ONLINE: 21 JULY 2013 | DOI: 10.1038/NNANO.2013.134

Observation of Dirac plasmons in a topological

insulator
P. Di Pietro1,2, M. Ortolani2,3, O. Limaj2,4, A. Di Gaspare3, V. Giliberti2,3, F. Giorgianni2,4, M. Brahlek5,
N. Bansal5, N. Koirala5, S. Oh5, P. Calvani1,2 and S. Lupi2,4,6 *

Plasmons are quantized collective oscillations of electrons and
have been observed in metals and doped semiconductors. The
plasmons of ordinary, massive electrons have been the basic
ingredients of research in plasmonics and in optical metamater-
ials for a long time1. However, plasmons of massless Dirac elec-
trons have only recently been observed in graphene, a purely
two-dimensional electron system2. Their properties are promis-
ing for novel tunable plasmonic metamaterials in the terahertz
and mid-infrared frequency range3. Dirac fermions also occur in
the two-dimensional electron gas that forms at the surface of
topological insulators as a result of the strong spin–orbit inter-
action existing in the insulating bulk phase4. One may therefore
look for their collective excitations using infrared spectroscopy.
Here we report the first experimental evidence of plasmonic
excitations in a topological insulator (Bi2Se3). The material
was prepared in thin micro-ribbon arrays of different widths
W and periods 2W to select suitable values of the plasmon
wavevector k. The linewidth of the plasmon was found to
remain nearly constant at temperatures between 6 K and
300 K, as expected when exciting topological carriers.
Moreover, by changing W and measuring the plasmon
frequency in the terahertz range versus k we show, without
using any fitting parameter, that the dispersion curve agrees
quantitatively with that predicted for Dirac plasmons.
A topological insulator is a quantum electronic material with an
insulating gap in the bulk that is of spin–orbit origin, and gapless
surface states at the interface with the vacuum or another dielectric.
The latter states are metallic and associated with massless Dirac
quasiparticles4–6. They exhibit chirality, are protected from backscat-
tering by time-reversal symmetry, and cannot be destroyed or
gapped by scattering processes that do not involve magnetic impu-
rities. Since their discovery, topological insulators have generated a
great deal of interest, not only because of their outstanding physical
properties, but also for their potential applications in quantum com-
puting7,8, terahertz detectors9 and spintronic devices10. As with
other compounds, some of these foreseen applications may
benefit from the exploitation of electron collective excitations, that
is, plasmonics. Of the two-dimensional electron systems, topological
insulator surface states have the advantage that they spontaneously
provide a two-dimensional Dirac system from the bulk material,
without the need to physically implement an atomic monolayer,
as in graphene. Moreover, thanks to momentum–spin lockage,
topological insulator plasmons may potentially preserve the coher-
ence of the electronic states up to room temperature. This would be
a major step forward in quantum mechanics applications. However,
topological insulator plasmons have yet to be observed.
In the two-dimensional electron gas at the topological insulator
surface, plasmons like those recently detected in graphene2 are
indeed expected to exist. However, they cannot be excited directly
by electromagnetic radiation because their dispersion law is such
as to prevent the conservation of momentum in the photon absorp-
tion process. In other two-dimensional systems, the necessary extra
momentum was obtained by patterning the surface with a subwave-
length grating2,11. In this way, confined plasmons are obtained, and
their interaction with radiation has been discussed for graphene in
ref. 12. Here, we have applied this methodology to thin films of
Bi2Se3 , which were patterned in the form of microribbon arrays of
different widths W and periods 2W (Fig. 1a,c), so that a series of dis-
crete values of k  p/W (refs 13,14) were obtained. Six thin films of
Bi2Se3 were grown by molecular beam epitaxy on 0.5-mm-thick sap-
phire (Al2O3) substrates15,16. Three had a thickness of d ¼ 120 quin-
tuple layers (that is, 120 nm) and the other three d  60 nm.
Transport characterization by means of resistivity and Hall
measurements showed that Dirac electrons generated by the top-
ology and bulk massive electrons due to the band-bending
effect15,16 participate in surface conduction. One film was kept as
grown for the purpose of comparison, and the other five were pat-
terned by electron-beam lithography and reactive-ion etching in
the form of parallel ribbons with different widths W and periods
L ¼ 2W, as shown in Fig. 1a,c.
The transmittance T(n) of the six films was measured in the
terahertz range by a Fourier-transform interferometer at tempera-
tures from 6 K to 300 K. The corresponding extinction coefficient
E(n) ¼ 1 2 T(n) is reported in Fig. 1b for the as-grown sample
and for the patterned films (Fig. 1d,e), normalized by the respective
peak values, both at 6 K (blue lines) and 300 K (red lines). The as-
grown film in Fig. 1b exhibits both of the phonon peaks observed in
Bi2Se3 single crystals, namely the a mode at 1.85 THz (61 cm21)
and the barely discernible b mode at 4.0 THz (132 cm21)17. They
broaden at high T and are superimposed to a Drude absorption,
which was mainly attributed to Dirac surface states18. A plot of
the E(n) of the patterned samples is presented in Fig. 1d for the
radiation electric field E parallel to the ribbons, where it is shown
that the phonon frequencies do not change appreciably with W or
d. Moreover, the spectra are very similar to those in Fig. 1b,
which confirms that the patterning procedure does not affect the
physical properties of the samples at all (Supplementary Table S1).
Figure 1e presents the extinction coefficient E(n) with E perpen-
dicular to the ribbons. As this direction is the same as that of the
reciprocal-lattice vectors needed for energy–momentum conserva-
tion in the transmittance measurement, the plasmon can now be
observed. As one can see, both at 6 K and at 300 K the a phonon

1
CNR–SPIN, Corso F. Perrone, 16152 Genoa, Italy, 2 Dipartimento di Fisica, Università di Roma ‘La Sapienza’, Piazzale A. Moro 2, I-00185 Rome, Italy,
3
CNR–IFN, Via Cineto Romano, 42 00156 Rome, Italy, 4 INFN, Piazza dei Caprettari 70, 00186 Rome, Italy, 5 Department of Physics and Astronomy Rutgers,
The State University of New Jersey, 136 Frelinghuysen Road, Piscataway, New Jersey 08854-8019, USA, 6 CNR–IOM, Area Science Park, Basovizza, Ed. MM,
Strada Statale 14 Km 163,5, I-34149 Trieste, Italy. * e-mail: stefano.lupi@roma1.infn.it

NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology 1

© 2013 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2013.134

a b d = 60 QL unpatterned
1.0
T = 300 K
T=6K

E(ν)
0.5

2 μm
0.0
0 2 4 6
ν (THz)
c d W = 2 μm e
1.0 E || 1.0 W = 2 μm E

E(ν)

E(ν)
0.5 0.5

d = 120 QL; W = 2 μm
0.0 0.0
1.0 W = 2.5 μm 1.0 W = 2.5 μm
E(ν)

E(ν)
0.5 0.5

d = 60 QL; W = 2.5 μm 0.0

0.0
1.0 W = 4 μm 1.0 W = 4 μm
E(ν)

E(ν)
0.5 0.5

d = 60 QL; W = 4 μm 0.0 0.0

1.0 W = 8 μm 1.0 W = 8 μm
E ||
E(ν)

0.5 E(ν) 0.5

E
d = 120 QL; W = 8 μm 0.0 0.0
1.0 W = 20 μm 1.0 W = 20 μm
T = 300 K T = 300 K
T=6K T=6K
E(ν)

E(ν)

0.5 0.5

d = 120 QL; W = 20 μm 0.0 0.0

0 2 4 6 0 2 4 6
ν (THz) ν (THz)

Figure 1 | Extinction coefficients of the microribbon arrays of Bi2Se3 topological insulators in the terahertz range. a, Scanning electron microscope (SEM)
image of the W ¼ 2.5 mm patterned film. b, Extinction coefficient of the as-grown, unpatterned film at 6 K (blue lines) and 300 K (red lines). c, Optical
microscope images of the five patterned films with different widths W and periods 2W. Red arrows indicate the direction of the radiation electric field E,
either perpendicular or parallel to the ribbons. Film thickness is indicated under the images, where 1 QL  1 nm (QL, quintuple layers). d, Extinction coefficient
at 6 K and 300 K for the five patterned films, with the radiation electric field parallel to the ribbons. e, Extinction coefficient of the five patterned films, with
the radiation electric field applied perpendicularly to the ribbons, at 6 K and 300 K. All data are normalized by their respective peak values.

is replaced by a double absorption, where both peak frequencies
strongly depend on W. We assign these features to the a phonon
and to the plasmon of Bi2Se3 , mutually interacting via a Fano inter-
ference. This produces a renormalization of both the phonon and
the plasmon frequency, with a hardening of the mode at higher fre-
quencies and a softening of the one at lower frequencies, indepen-
dent of their nature. A similar effect is reported in the literature
for doped graphene on SiO2 (ref. 19). A Fano shape of the weak b
phonon line is also observed in the top panel of Fig. 1e, as it
becomes closer to the plasmon resonance through an inflection
point in E(v) at the bare phonon frequency (line).
To extract the bare plasmon (np) and phonon (nph) frequency
from the data in Fig. 1e, we fit the experimental data to the following
equation from Giannini and colleagues20:
(n′ + q(n′ ))2 g2
E(n′ ) = ·  2
n′2 + 1 n − np
1+
Gp /2
The expressions for the renormalized frequency n′ and for the
(Fano) parameter q—the ratio between the probability amplitude
of exciting a discrete state (phonon) and of exciting a continuum
or quasi-continuum state (plasmon)—are reported in the
Methods. Gp and np are the plasmon bare linewidth and frequency,
respectively, and g is its coupling factor with the radiation.
The above model satisfactorily reproduces the experimental data,
as shown by the black lines in Fig. 2. This figure also shows the bare
phonon and plasmon contributions, reconstructed in the form of
Lorentzians (Supplementary Tables S1, S2 and S3). The phonon
frequency does not change with W, being the same as in Fig. 1d
(parallel polarization), but for the plasmon, both np and Gp decrease
as W increases (see also the inset in the bottom panel of Fig. 2). Gp
can be assumed to be the sum of the following independent con-
tributions: (i) the Drude linewidth obtained by fitting the spectra
(1) of Fig. 1d (Supplementary Fig. S2, Table S1); (ii) the Landau
damping21 rate, through both the creation of hole–electron pairs
and a phonon-assisted process22; (iii) radiative decay into
photons; and (iv) finite size effects22. Contribution (i) is

2 NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology

© 2013 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2013.134 LETTERS
5
1.0 E W = 2 μm
4
T=6K

νp(THz)
E(ν)

0.5
2
4
0.0
0
1.0 W = 2.5 μm 0.0 0.5 1.0
W−1/2 (μm−1/2)
E(ν)

3
0.5

νp(k) (THz)
0.0
1.0 W = 4 μm
2
E(ν)

0.5

0.0 1
1.0 W = 8 μm
E(ν)

0.5
0
0 5 10 15 20
0.0
W = 20 μm k (×103 cm−1)
2
Γp (THz)

1.0
1 Figure 3 | Experimental and theoretical dispersion of plasmons in Bi2Se3.
E(ν)

0.5 0 Main panel: experimental values of np versus k at 6 K (blue circles)

0 10 20
W (μm)
0.0
0 2 4 6
ν (THz)
Figure 2 | Extraction of the bare plasmon and phonon contributions from
the extinction data through a Fano fit. Normalized extinction coefficient
E(n) versus frequency n for the five patterned films, for the radiation electric
field E perpendicular to the ribbons, at 6 K (circles), as well as fits to
equation (1) (black lines). Bare plasmon and a(b) phonon contributions,
extracted through the fits, are shown by the red and green (magenta) lines,
respectively. In the top and bottom panels, the bump at 2.6 THz is due to
insufficient compensation of the Mylar beamsplitter absorption. Inset
(bottom panel): plasmon linewidth Gp versus ribbon width W at 6 K. The
dotted line is the Drude contribution to Gp as extracted from data with
polarization parallel to the ribbons. The dashed line is a guide to the eye.
independent of W and provides the background marked by the
black dotted line in the inset. The other effects are expected to
increase with increasing plasmon frequency (decreasing W ) and
therefore should be responsible for the behaviour of Gp shown in
the inset.
The bare plasmon frequencies obtained from the Fano fits are
plotted versus the wavevector k in Fig. 3 and versus W 21/2 in its
inset. As shown in the inset, np / W 21/2, as expected for a two-
dimensional plasmonic excitation. In the main panel of Fig. 3, the
same points are plotted versus k, based on the relation k  p/W.
Here, the approximation accounts both for a possible depletion in
the electron density at the ribbon edges, as observed in graphene22,
and for scattering from the edges22. As both phenomena act in a
range smaller than our W by two orders of magnitude, the corre-
sponding correction can be neglected. The effective value of k
may also be influenced by the excitation of edge modes13,14, as in
graphene. However, those calculations cannot be easily extended
to the topological insulators where, for example, the electron mobi-
lity is lower by three orders of magnitude. In any case, by assuming
k ¼ p/W we find the good agreement between data and the dis-
persion law np / k 1/2 shown in Fig. 3.
compared with the plasmon dispersion for Dirac (dashed black line) and
massive (dotted blue line) electrons calculated with no fitting parameters
using equations (2) and (3), respectively. The additional point (green
diamond) refers to a seventh sample with W ¼ 1.8 mm and period L ¼ 4 mm
(L ¼ 2.2W), for which the raw data are reported in Supplementary Fig. S1.
Inset: linear dependence of np on W 21/2, where W ¼ p/k is the
ribbon width.
One has still to show that the two-dimensional plasmon can be
ascribed to the Dirac (massless) fermions, rather than to the
massive electrons also present at the surface. As discussed
below, both qualitative and quantitative arguments support
the former assignment. The qualitative argument is based
on the behaviour of Gp in Fig. 1e, where it is nearly the same
at 6 K and 300 K (see also the Fano parameters in Supplementary
Tables S2, S3).
The plasmon linewidth appears remarkably robust to a tempera-
ture variation by a factor of 50. This suggests a weak effect of
phonon scattering in the topological insulator plasmonic absorp-
tion, at variance with conventional two-dimensional electron
gases. The infrared absorption of non-topologically protected
plasmons in two-dimensional systems has been reported only at
low temperatures11 and for plasmon energies below the main
optical phonon energy, consistent with the assumption that
electron–phonon scattering is their main decay channel23,24.
Indeed, because contradictory results have appeared in the literature
on the strength of the electron–phonon scattering in topological
insulators25,26, this issue remains open to further experiments
and calculations.
The quantitative argument proceeds as follows. In topological
insulator thin films, the free-electron layers at the topological insu-
lator–substrate and topological insulator–vacuum interfaces interact
via an effective Coulomb potential27,28. This interaction, if one
neglects interlayer tunnelling, produces two longitudinal collective
excitations, that is, an optical plasmon (with np / k 1/2), and an
acoustic plasmon (with np / k)27,28. The latter mode is degenerate
with the continuum27, that is, strongly Landau-damped21, and then
unobservable. In principle, both Dirac (D) and massive (M) fermions
may contribute to the remaining optical mode. In the long

NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology 3

© 2013 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE NANOTECHNOLOGY
wavelength limit k  0 their dispersion laws can be written27,28 as
 
1 √ e2  1/2
nD (k) = k v
− F 2pn g g
D s v (2)
2p 4p10 1h
and
 1/2
1 √  e2
nM (k) = k nM /m∗ (3)
2p 210 1
where the spin ( gs) and valley ( gv) degeneracies are both equal to 1
(ref. 27). Moreover, in Bi2Se3 thin films grown in the same
conditions as the present ones, nM ¼ 7.5+3.5 × 1012 cm22,
nD ¼ 3+1 × 1013 cm22, vF ¼ 6+1 × 107 cm s21 (ref. 29), and the
effective mass of the parabolic bulk band is m* ¼ 0.15+0.01me
(refs 15,16). Neither equation (2) nor equation (3) contain the film
thickness or the bulk dielectric function. They instead depend on
the average 1 ¼ (11 þ 12)/2 between the vacuum dielectric function
(11 ¼ 1) and that of the substrate (12 ≈ 10). The theoretical disper-
sions were calculated using equations (2) and (3) at wavevectors
k ¼ p/W by applying the central experimental values for nD , vF
(Dirac), nM and m* (massive) reported above. The resulting dis-
persion lines are reported in Fig. 3. The line for the Dirac plasmons
is found to be in very good agreement with the experimental data,
without using free parameters.
In conclusion, we have reported the first observation of Dirac
plasmon excitations in thin films of Bi2Se3 topological insulator
and have reasoned that those plasmons should be assigned to
Dirac quasiparticles of the conducting two-dimensional edge state.
Our assignment is based on qualitative arguments and comparison
with the theoretical dispersion relation. We have also reported a
strong Fano-like interference between the Dirac plasmons and the
phonons of Bi2Se3. A similar phenomenon has been observed in
conventional plasmonic systems between bright (dipole) modes
and dark (quadrupole) modes20 and used here, for instance, to
increase the quality factor of resonances. It is expected that novel
opportunities for sensing and on-chip signal transmission in the ter-
ahertz range will emerge using topological insulators. Finally, we
note that plasmon wavevector tunability, which we achieve here
by varying the widths and periods of the microribbon arrays, can
be generalized to produce more complex plasmonic designs.
Methods
The high-quality Bi2Se3 thin films were prepared by molecular beam epitaxy
using the standard two-step growth method developed at Rutgers University15,16.
The 10 × 10 mm2 Al2O3 substrates were first cleaned by heating to 750 8C in an
oxygen environment to remove organic surface contamination. The substrates were
then cooled to 110 8C, and an initial three quintuple layers of Bi2Se3 were deposited.
This was followed by heating to 220 8C, at which the remainder of the film was
deposited to achieve the target thickness. The Se:Bi flux ratio was kept to 10:1 to
minimize Se vacancies. Once the films were cooled, they were removed from the
vacuum chamber, and vacuum-sealed in plastic bags within 2 min, then shipped to
the University of Rome.
Bi2Se3 ribbons were fabricated by electron-beam lithography and subsequent
reactive ion etching. The Bi2Se3 film was spin-coated with a double layer of electron-
sensitive resist polymer poly(methyl methacrylate) (PMMA) to a total thickness of
1.4 mm. Ribbon patterns with different W were then written in the resist by electron-
beam lithography. To obtain a lithographic pattern with a realignment precision
below 10 nm over a sample area suitable for terahertz spectroscopy of 10 × 10 mm2,
we used an electron beam writer equipped with an x–y interferometric stage (Vistec
EBPG 5000). The patterned resist served as mask for the removal of Bi2Se3 by
reactive ion etching at a low microwave power of 45 W to prevent heating of the
resist mask. Sulphur hexafluoride (SF6) was used as the active reagent. The Bi2Se3
film was etched at a rate of 20 nm min21, which was verified by atomic force
microscopy after soaking the sample in acetone to remove the PMMA. The in-plane
edge quality after reactive ion etching, inspected by atomic force microscopy, closely
follows that of the resist polymer mask, that is, with an edge roughness of ,20 nm.
The vertical profile of the edge forms an angle of 458 with the substrate plane,
because our reactive ion etching process has no preferred etching direction.
4 NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2013 Macmillan Publishers Limited. All rights reserved.
DOI: 10.1038/NNANO.2013.134
The absorption spectra in the terahertz range were obtained using a Bruker
IFS-66V Michelson interferometer and a liquid-helium-cooled bolometer. The
Bi2Se3 film and a co-planar Al2O3 bare substrate were mounted on the cold
finger of a helium-flow cryostat and kept at a pressure of 1 × 1026 mbar.
The radiation was polarized either along, or perpendicular to, the ribbons by a
terahertz polarizer with a degree of polarization of .99.5%. The extinction
coefficient reported in Fig. 2 was obtained from the film transmittance T, defined
as the ratio between the intensity transmitted by the thin film and that transmitted
by the bare substrate.
The Fano fits were obtained by replacing v′ in equation (1) (ref. 20) by
n − nph n − np
n′ = − (4)
Gph (n)/2 Gp /2
the plasmon-coupled phonon linewidth by
Gph (n) = 
2pv2
2 (5)
n − np
1+
Gp /2
and the Fano factor by
vw/g n − np
q(n) = + (6)
Gph (n)/2 Gp /2
Here, w and g are the coupling factors of the radiation with the phonon and the
plasmon, respectively, v measures the phonon–plasmon interaction, and Gp is the
width of the plasmon line, assumed to be Lorentzian.
Received 25 January 2013; accepted 12 June 2013;
published online 21 July 2013
References
1. Maier, S. A. Plasmonics: Fundamentals and Applications (Springer, 2007).
2. Ju, L. et al. Graphene plasmonics for tunable terahertz metamaterials. Nature
Nanotech. 6, 630–634 (2011).
3. Grigorenko, A. N., Polini, M. & Novoselov, K. S. Graphene plasmonics. Nature
Photon. 6, 749–758 (2012).
4. Hasan, M. Z. & Kane, C. L. Colloquium: topological insulators. Rev. Mod. Phys.
82, 3045–3067 (2010).
5. Kane, C. L. & Mele, E. J. Quantum spin Hall effect in graphene. Phys. Rev. Lett.
95, 226801 (2005).
6. Moore, J. E. The birth of topological insulators. Nature 464, 194–198 (2010).
7. Collins, G. P. Computing with quantum knots. Sci. Am. 294, 57–63 (2006).
8. Kitaev, A. & Preskill, J. Topological entanglement entropy. Phys. Rev. Lett. 96,
110404 (2006).
9. Zhang, X., Wang, J. & Zhang, S.-C. Topological insulators for high-performance
terahertz to infrared applications. Phys. Rev. B 82, 245107 (2010).
10. Chen, Y. et al. Experimental realization of a three-dimensional topological
insulator, Bi2Te3. Science 325, 178–181 (2009).
11. Allen, S. J., Tsui, D. C. & Logan R. A. Observation of the two-dimensional
plasmon in silicon inversion layers. Phys. Rev. Lett. 38, 980–983 (1977).
12. Koppens, F. H. L., Chang, D. E. & de Abajo, J. C. Graphene plasmonics:
a platform for strong light–matter interactions. Nano Lett. 11,
3370–3377 (2011).
13. Nikitin, A. Yu., Guinea, F., Garcia Vidal, F. J. & Martin Moreno, L. Edge and
waveguide terahertz surface plasmon modes in graphene microribbons. Phys.
Rev. B 84, 161407 (2011).
14. Nikitin, A. Yu., Garcia Vidal, F. J. & Martin Moreno, L. Surface plasmon
enhanced absorption and suppressed transmission in periodic arrays of
graphene ribbons. Phys. Rev. B 85, 081405(R) (2012).
15. Bansal, N., Kim, Y. S., Brahlek, M., Edrey, E. & Oh, S. Thickness-independent
transport channels in topological insulator Bi2Se3 thin films. Phys. Rev. Lett. 109,
116804 (2012).
16. Bansal, N., et al. Epitaxial growth of topological insulator Bi2Se3 thin film
on Si(111) with atomically sharp interface. Thin Solid Film 520,
224–229 (2011).
17. Di Pietro, P. et al. Optical conductivity of bismuth-based topological insulators.
Phys. Rev. B 86, 4701 (2012).
18. Valdes Aguilar, R. et al. THz response and colossal Kerr rotation from the
surface states of the topological insulator Bi2Se3. Phys. Rev. Lett. 86,
045439 (2012).
19. Fei, Z. et al. Infrared nanoscopy of Dirac plasmons at the graphene–SiO2
interface. Nano Lett. 11, 4701–4705 (2011).
20. Giannini, V., Francescato, Y., Amrania, H., Phillips, C. C. & Maier, S. A. Fano
resonances in nanoscale plasmonic systems: a parameter-free modeling
approach. Nano Lett. 11, 2835–2840 (2011).
NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2013.134
21. Landau, L. On the vibration of the electronic plasma. J. Phys. USSR 10, 25 (1946).
22. Yan, H. et al. Damping pathways of mid-infrared plasmons in graphene
nanostructures. Nature Photon. 7, 394–399 (2013).
23. Baumberg, J. J. & Williams, D. A. Coherent phonon–plasmon modes in
GaAs:Alx Ga1–xAs heterostructures. Phys. Rev. B 53, R16140–R16143 (1996).
24. Cho, G. C., Dekorsy, T., Bakker, H. J., Hovel, R. & Kurz, H. Generation
and relaxation of coherent majority plasmons. Phys. Rev. Lett. 77,
4062–4065 (1996).
25. Pan, Z.-H et al., Measurement of an exceptionally weak electron–phonon
coupling on the surface of the topological insulator Bi2Se3 using angle-resolved
photoemission spectroscopy. Phys. Rev. Lett. 108, 187001 (2012).
26. Zhu, X. et al. Electron–phonon coupling on the surface of the topological
insulator Bi2Se3 determined from surface-phonon dispersion measurements.
Phys. Rev. Lett. 108, 185501 (2012).
27. Profumo, R. E. V. et al. Double-layer graphene and topological insulator thin-
film plasmons. Phys. Rev. B 85, 085443 (2012).
28. Das Sarma, S. & Hwang, E. H. Collective modes of the massless Dirac plasma.
Phys. Rev. Lett. 102, 206412 (2009).
29. Cao, Y., et al. In-plane helical orbital texture switch near the Dirac point in the
topological insulator Bi2Se3. Preprint at http://lanl.arxiv.org/abs/1209.1016 (2012).
NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2013 Macmillan Publishers Limited. All rights reserved.
LETTERS
Acknowledgements
The authors thank M. Polini for fruitful discussions about Dirac and massive
plasmonic dispersions. The work at Rutgers was supported by IAMDN of Rutgers
University, NSF DMR-0845464 and ONR N000141210456. The work at ‘La Sapienza’, was
supported by the INFN, TERASPARC experiment.
Author contributions
M.B., N.B., N.K. and S.O. fabricated and characterized the Bi2Se3 films. M.O., A.D.G.
and V.G. performed electron-beam lithography and etching. P.D.P., F.G., O.L. and M.O.
carried out the terahertz experiments and data analysis. P.C., M.O. and S.L. planned and
managed the project, with inputs from all authors. All authors discussed the results.
P.C., M.O. and S.L. wrote the manuscript.
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 S.L.
Competing financial interests
The authors declare no competing financial interests.
5