LETTER

pubs.acs.org/NanoLett

Raman Spectroscopy of Few-Quintuple Layer Topological Insulator

Bi2Se3 Nanoplatelets
Jun Zhang,†,r Zeping Peng,‡,r Ajay Soni,† Yanyuan Zhao,† Yi Xiong,§ Bo Peng,† Jianbo Wang,§
Mildred S. Dresselhaus,||,^ and Qihua Xiong*,†,#
†
Division of Physics and Applied Physics, School of Physical and Mathematical Sciences, Nanyang Technological University,
Singapore 637371
‡
School of Physical Science and Technology, MOE Key Laboratory on Luminescence and Real-Time Analysis, Southwest University,
Chongqing 400715, China
§
School of Physics and Technology, Center for Electron Microscopy and MOE Key Laboratory of Artificial Micro- and Nano-Structures,
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

Wuhan University, Wuhan 430072, China

Department of Physics and ^Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology,
)
Downloaded via LEIBNIZ IFW DRESDEN on May 31, 2023 at 13:58:33 (UTC).

Cambridge, Massachusetts 02139, United States

#
Division of Microelectronics, School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798

bS Supporting Information
ABSTRACT: We report on Raman spectroscopy of few quintuple
layer topological insulator bismuth selenide (Bi2Se3) nanoplatelets
(NPs), synthesized by a polyol method. The as-grown NPs exhibit
excellent crystalline quality, hexagonal or truncated trigonal mor-
phology, and uniformly flat surfaces down to a few quintuple
layers. Both Stokes and anti-Stokes Raman spectroscopy for the
first time resolve all four optical phonon modes from individual
NPs down to 4 nm, where the out-of-plane vibrational A11g mode
shows a few wavenumbers red shift as the thickness decreases
below ∼15 nm. This thickness-dependent red shift is tentatively
explained by a phonon softening due to the decreasing of the
effective restoring force arising from a decrease of the van der
Waals forces between adjacent layers. Quantitatively, we found
that the 2D phonon confinement model proposed by Faucet and Campbell cannot explain the red shift values and the line shape of
the A11g mode, which can be described better by a BreitWignerFano resonance line shape. Considerable broadening (∼17 cm1
for six quintuple layers) especially for the in-plane vibrational mode E2g is identified, suggesting that the layer-to-layer stacking affects
the intralayer bonding. Therefore, a significant reduction in the phonon lifetime of the in-plane vibrational modes is probably due to
an enhanced electronphonon coupling in the few quintuple layer regime.
KEYWORDS: Topological Insulators, bismuth selenide, Raman spectroscopy, electron-phonon coupling, phonon softening, polyol
synthesis

materials with diverse applications.8 Bi2Se3 has been shown to

T opological insulators (TIs) are a new class of quantum
matter for which their bulk phases are ordinary insulators
but possess robust, nontrivial, and conducting surface or bound-
ary states with nondegenerate spins.1,2 These surface states,
arising due to strong spinorbit coupling,1,2 are topologically
protected against back scattering and are immune to defect
driven localization as long as the disorder potential does not
violate time reversal symmetry.1,2 According to theoretical pre-
dictions or experimental observations, the TIs family includes the
2D HgTe/CdTe quantum wells,3 half-Heusler compounds4,5 as
well as some bismuth and antimony chalcogenides (such as
Bi2Se3, Bi2Te3, and Sb2Te3) 3D layered materials,6,7 the latter of
which were known historically as excellent thermoelectric
be an ideal candidate for studying room temperature topological
insulating behavior as it has a topologically nontrivial band gap of
0.3 eV, much larger than the room temperature energy scale.1,2,7
Therefore, Bi2Se3 is considered to be a promising topological
system toward unique applications in next generation electronics
such as in quantum computation. However, those topologically
protected surface states are often masked by the bulk states due to
high carrier density in 3D TIs,1,2,6 which impedes the experimental
Received: March 8, 2011
Revised: April 27, 2011
Published: May 23, 2011

r 2011 American Chemical Society 2407 dx.doi.org/10.1021/nl200773n | Nano Lett. 2011, 11, 2407–2414
Nano Letters LETTER

realizations of many predicted novel phenomena and appli-

cations.1,2 Several approaches have been explored to tune the
carrier density, e.g., compensate doping in the bulk crystals,9
using a gate voltage to tune the Fermi level of nanostructures,10
or grow/fabricate ultrathin (few quintuple layer (QL)) TIs
nanobelts to enhance the surface to volume ratio.11,12
On the other hand, despite many excellent works concerning
electronic properties,13,6,7,917 phonons in TIs, especially down
to a few nanometer thickness regime, are not well understood.
Investigations on phonons and electronphonon interactions
are critical toward applications based upon TIs. In this regard,
Raman scattering spectroscopy is a powerful and sensitive
technique to study phonons in such TIs. Another strong
motivation to investigate phonons in TIs by Raman spectroscopy
is based on the significant success of Raman spectroscopy in the
characterization of graphene.18,19 The band structure of TI
surface states is similar to that of graphene,20 which both exhibit
Dirac-like linear band dispersion. Therefore we propose that
Raman spectroscopy can also be a prospective tool to identify
the number of quintuple layers in TIs accurately and rapidly, just
like counting graphene layers.21 It is also intriguing to probe the
Dirac-like linear band dispersion surface states of TIs by inves-
tigating double resonant Raman scattering, reminiscent of the
G and 2D modes in graphene18,19 and to determine the Fermi
level position by probing electronphonon coupling.22,23
Recent work on Bi2Te3 nanoflakes indeed showed nontrivial
consequences to their vibrational modes by exploring them from
3D bulk to a 2D atomically thin sample.24 However, a few recent
studies on Bi2Se3 unfortunately could not present the whole
picture on these phenomena, since several low frequency vibra-
tional modes (below 100 cm1) are absent in their Raman
spectra due to the limitation of their spectrometers.25,26 Our
Figure 1. Morphology of the as-grown thin Bi2Se3 NPs dispersed on Si
work presented in this paper is motivated to investigate the substrates: (a) SEM image; (b) AFM image, the top frame is the
phonon properties of layered Bi2Se3 systematically and to study topography image, while the bottom frame is the section profile along
how critical parameters such as the frequency and lifetime of the the red lines shown in the image. The height (thickness) profile indicates
vibrational modes vary as the thickness decreases from bulk to approximately 7 nm for this batch of synthesis.
the atomically thin quintuple layer (QL) regime of a single unit
cell. Further investigations are anticipated toward gaining a better lacey carbon grids similarly for transmission electron microscopy
understanding of electronphonon and phononphonon inter- analysis (JEOL 2010F).
actions in such TI systems. Figure 1a shows a representative SEM image of chemically
Various methods such as chemical vapor transport,27 mechanical grown NPs with a reaction time of 100 min. The product evolved
exfoliations by Scotch tape24 or peeling by an atomic force from a powder-like morphology at the beginning of the reaction
microscope tip,28 and molecular beam epitaxial growth have and then developed into a flower-like intermediate state and
been used to synthesize or prepare few layer Bi2Se3.12,17,29 Here, finally reached a nanoplatelet morphology with a very uniform
we present a wet chemical synthesis of Bi2Se3 nanoplatelets contrast (Supporting Information, Figure S-1ac). Most good
(NPs) by a polyol method with controllable thickness down to a thin NPs exhibit hexagonal morphologies with planar dimen-
few QLs. It enables a systematic investigation of their Raman sions extending up to several micrometers, while some of them
spectra as a function of sample thickness and excitation laser show a truncated trigonal morphology. The XRD pattern
power density. Surprisingly, in addition to facile and scaling- (Supporting Information, Figure S-1d) indicates that the product
up characteristics, this polyol wet chemical method also shown in Figure 1a has a rhombohedral phase (JCPDS card no.
produces very high crystalline quality Bi2Se3 NPs with a 33-0214) with good crystallinity, in good agreement with elec-
primarily regular hexagonal shape, and clean, flat surfaces tron diffraction analysis to be discussed in the following.
down to a few QLs. The thickness of NPs has been accurately measured by AFM,
The polyol synthesis was originally developed by Fi
evet and as shown in Figure 1b. The top panel is the AFM topography
co-workers as a simple and versatile route to synthesize colloidal image of three NPs with hexagonal shape, while the bottom panel
particles made of metals,30 alloys,31 and chalcogenides.32,33 The shows three traces of sectional analysis suggesting approximately
details can be found in Supporting Information. The as-grown 7 nm in thickness for this batch of synthesis. It is important to
product dispersed in an isopropyl alcohol solution was drop- note that the NPs have a very flat surface indicated by the uniform
casted onto silicon substrates for further scanning electron height in the AFM images (Figure.1b). The thickness of each QL
microscopy (SEM, JEOL 7001F), atomic force microscopy of Bi2Se3 is about 9.55 Å.34 Since the thickness determination is
(AFM, Veeco Dimension V), and XRD analysis (Bruker D8 crucial for interpretation of the Raman data, the AFM images
advanced diffractometer, Cu KR). NPs were dispersed onto have been carefully processed and analyzed by a linear fitting of
2408 dx.doi.org/10.1021/nl200773n |Nano Lett. 2011, 11, 2407–2414
Nano Letters
Figure 2. TEM analysis: (a) bright field TEM image of as-grown Bi2Se3
NPs: (b) SAD (selected area diffraction) pattern from the circled area in
(a), indexed as a 6-fold symmetry [0001] pattern: (c) HRTEM image of
a NP, lattice fringes were clearly resolved. The inset was a FFT (fast
Fourier transform) analysis showing a 2-fold symmetry [2311] pattern.
Figure 3. Optical and corresponding AFM images of Bi2Se3: (a) an
optical image of a few NPs dispersed on a Si marked substrate with
100 nm SiO2; (b) AFM height image of the regime highlighted as a
dashed box in (a), the thickness of each NP is labeled in the image.
the top surface of the NPs and the corresponding substrate
surfaces as schematically indicated by red lines in the bottom
panel.
We also conducted a detailed characterization of the TEM
studies in order to elucidate the crystalline orientation of the NPs
as shown in Figure 2. Figure 2a shows a low magnification bright
field TEM image of the Bi2Se3 NPs. A hexagonal morphology
was identified. The circle indicates the position where a selected-
area electron diffraction (SAD) pattern was taken as shown in
Figure 2b. This SAD pattern can be indexed as a 6-fold symmetry
[0001] zone axis pattern, in agreement with the layered structure
along the c axis. Figure 2c displays a high-resolution TEM
(HRTEM) image of the same NP but with a slightly different
zone axis, and single crystalline quality was identified. The inset is
a fast Fourier transform (FFT) analysis which shows a pattern
indexed as a 2-fold symmetry [2311] pattern.
It is an important step to make the ultrathin NPs visible under
an optical microscope for using Raman measurements to further
study their properties. In order to find good substrates that can
give the highest contrast under visible illumination, we have
calculated the contrast of Bi2Se3 on SiO2/Si substrates by
classical geometry optics method, which was successfully used
to study the contrast of graphene on silicon substrates.35 The
refractive indices of Bi2Se3 were obtained by a KramersKronig
analysis of the reflectance spectra measured by Greenaway and
2409
LETTER
Harbeke in the 1970s (Supporting Information, Figure S-2).36
We found that a Si wafer with 100 nm SiO2 provides excellent
contrast for most of the different thicknesses of Bi2Se3 NPs.
Figure 3a shows an optical image of hexagonal Bi2Se3 nano-
platelets on a Si substrate with 100 nm thick SiO2 and Au markers
(arrows), taken under an Olympus optical microscope (BX51,
100 objective with N.A. 0.9) with white light illumination. The
corresponding AFM height image (dashed box area highlighted
in Figure 3a) was displayed in Figure 3b with noted thickness
values ranging from 6.5 to 30 nm. With white light illumination,
the contrast of the NPs exhibits three major categories: (a) bright
and shining contrast for very thick NPs (thicker than 16 nm); (b)
light blue contrast for thin NPs (below ∼9 nm); (c) dark blue
contrast for intermediate thickness (between 9 and 16 nm).
Layered Bi2Se3 has a rhombohedral crystal structure belong-
ing to the space group (R3m/D53d), composed of hexagonal
close-packed atomic layers periodically arranged along the
c axis.37,38 The atomic arrangement (Figure 4a) can be consid-
ered as repeating units with each consisting of five atomic
SeBiSeBiSe layers called quintuple layers, weakly bound
by van der Waals forces with a slightly covalent nature.37 There-
fore, crystalline Bi2Se3 can be exfoliated just like graphite. The
primitive unit cell of Bi2Se3 contains five atoms. Consequently,
there are 15 lattice dynamical modes at the center of the Brillouin
zone, three of which are acoustic modes and 12 are optical
modes. Group theory classifies these 12 optical modes into two
Raman-active modes of 2A1g, 2Eg symmetry and two infrared-
active modes of 2A1u and 2Eu symmetry, and the irreducible
representations for the zone-center phonons can be written as:
χ = 2Eg þ 2A1g þ 2Eu þ 2A1u. The corresponding atomic
displacements of the Raman-active modes are shown in
Figure 4b. Their corresponding Raman tensors are37
0 1
a 0 0
B C
A 1g : B C
@ 0 a 0 A,
0 0 b
0 10 1
c 0 0 0 c d
B C
CB C
Eg : B
@ 0 c d A, @ c 0
B 0 CA ð1Þ
0 d 0 d 0 0
One can observe that the Eg and A1g modes may be distinguished
from one to another as a result of the off-diagonal Raman tensor
components of the Eg modes. However, so far in the available
literature only one to three out of the four Raman modes have
been observed.25,26,37 The lowest frequency mode Eg1 has not
been observed yet since the early work on the lattice vibrations of
Bi2Se3 was published by Richter et al. in the 1970s.37
Raman scattering spectroscopy was carried out at room
temperature using a Micro-Raman spectrometer (Horiba-JY
T64000) in the backscattering configuration excited with a
HeNe laser (λ = 632.8 nm) and a solid state green laser (λ =
532 nm). The backscattered signal was collected through a 100
objective and dispersed by a 1800 g/mm grating under a triple
subtractive mode with a spectra resolution of ∼1 cm1 and the
lowest available frequency was 5 cm1. The laser power at the
sample surface was less than 70 μW as measured by a power meter
through a 10 objective. The power-dependent Raman scattering
(as shown in Figure S-3 in the Supporting Information) suggested
that such a low power excitation can effectively exclude any laser
heating effect in our further Raman measurements.
dx.doi.org/10.1021/nl200773n |Nano Lett. 2011, 11, 2407–2414
Nano Letters LETTER

Figure 4. Crystalline structure and Raman modes of Bi2Se3. (a) Crystalline structure (R3m) of Bi2Se3. The lattice vectors and the quintuple layer of the
rhombohedral unit cell are indicated. (b) The vibrational normal modes (only Raman active) of a quintuple layer Bi2Se3. (c) Raman spectra with parallel
(xx) and perpendicular (xy) polarizations of the configuration of the incident and scattered light, collected from a very thick NP. (d) Phonon dispersion
curves of Bi2Se3 along the ΓZ direction of the Brillion zone documented in ref 39.

A typical Raman spectrum consisting of both Stokes and anti-
Stokes contributions is shown in Figure 4c, which was collected
from a very thick NP (thickness larger than 20 nm). The upper
curve is the parallel polarization configuration, which corresponds
to the xx component of the Raman tensors, and the lower curve
is the perpendicular polarization configuration, which corresponds
to the xy component of the Raman tensors. According to the
selection rule and the Raman tensors (eq 1), we can assign the
two peaks in the xy polarization configuration to the Eg modes
because only Raman tensors of Eg modes contain off-diagonal
elements, and consequently the other two in the xx polariza-
tion configuration correspond to A1g modes. The anti-Stokes
spectrum also supports the above assignment consistently.
Therefore, we unambiguously assign the peak at ∼37 cm1 to
the E1g mode, the other three peaks located at ∼72, ∼131, and
∼174 cm1 are assigned to the A11g, E2g, and A21g vibrational
modes, respectively.37 To the best of our knowledge, this is the
first time that all of the four peaks were observed, and the
previous literature only refers to less than four phonon modes,
while the lowest frequency E1g mode was absent in experimental
observation.37 Figure 4d shows the phonon dispersion curve
along the ΓZ direction of the Brillouin zone, where Z corre-
sponds to (0, 0, 3π/c) in reciprocal space.39
We next discuss the Raman scattering spectroscopy of indivi-
dual Bi2Se3 NPs as a function of the NP thickness at room
temperature. Figure 5a shows the Stokes and anti-Stokes spectra
of NPs with various thicknesses as excited by a 632.8 nm
HeNe laser. Panels b and c of Figure 5 display a zoom-in view
of the A11g (b) and E2g (c) modes excited by 632.8 nm (left
column) and 532 nm (right column), respectively. It can be seen
that the A11g mode develops a pronounced red shift (about
5 cm1 for 4 QL) as the thickness decreases below 15 nm; other
modes exhibit negligible red shifts considering the systematic
error. The accurate peak position is plotted in Figure 5d versus
thickness after careful line shape analysis. According to the
normal modes outlined in Figure 4b, it is worth noting that the
A11g mode at ∼72 cm1 is more sensitive to thickness because it
reflects the out-of-plane vibrations of the Se and Bi atoms,37 and
the interlayer van der Waals interactions influence the effective
restoring forces acting on these atoms. This red shift will be
discussed further in more detail below, after analysis of the subtle
2410
line shape change. We also observed that the Raman intensity of
NPs is much higher than that of bulk single crystal Bi2Se3 under
the same excitation laser and power density; we believe this
phenomenon can be attributed to optical interference enhance-
ments occurring for both the excitation laser and the emitted
Raman radiation in the layered TI/SiO2/Si system. A similar
enhancement in the Raman intensity has been reported for
Bi2Se3 and Bi2Te3,24,25 and graphene deposited on SiO2/Si
substrates.40
In the present situation, it is difficult to evaluate which specific
thickness warrants a maximum enhancement due to the polar-
ization dependence, the relative direction between the laser
polarization and sample crystalline directions, the sample size,
and the focus optimization. However, it is worth noting that the
relative intensity of the same symmetry mode evolves as a
function of the sample thickness (as shown in Figure 5e). This
evolution reflects how the thickness influences the relative
oscillator strength of in-phase vibrations of the in-plane mode
(E1g) and the out-of-plane mode (A11g) of the outer BiSe1 pairs,
and the out-of-phase vibrations of the outer BiSe2 pairs. As
shown in Figure 5e, the ratio of the Raman intensities for the in-
phase shear mode of the adjacent layers (IE1g/IE2g) increases as the
NP thickness decreases, while the trend is opposite for the out-of-
plane modes (IA11g/IA1g2). It is not clear at the moment why the
oscillator strengths of the different modes (in-phase vs out-of-
phase) exhibit a different dependence on thickness.
Another significant thickness dependence is considerable
broadening of the E2g mode. As shown in Figure 5f, the fwhm
of the E2g mode broadens to 24 cm1 for a 4 QL NP from 8 cm1
for the bulk sample. Raman band broadening in the finite size
regime has previously been explained using the following pheno-
menological exponential format41
Γ0 ðdÞ ¼ M þ N expð d=ηÞ ð2Þ
where the M, N, and η are the fitting parameters and d is the
thickness of the NPs. Equation 2 with M = 5.77, N = 40, and η = 5
gives a good fit to the fwhm of the E2g mode thickness dependence
of the nanoplatelets shown as a blue curve in Figure 5f. This
observation suggests that the layer-to-layer stacking also affects
the intralayer bonding, and therefore a significant decrease of the
phonon lifetime of the in-plane vibrational modes is probably due
dx.doi.org/10.1021/nl200773n |Nano Lett. 2011, 11, 2407–2414
Nano Letters LETTER

Figure 5. Raman spectra of few QL Bi2Se3 NPs and analysis. (a) Stokes and anti-Stokes Raman shifts of Bi2Se3 nanoplatelets for different thicknesses
excited by a 632.8 nm laser. (b, c) Zoom-in view of the A11g (b) and E2g (c) modes for various thicknesses of QLs excited by 532 nm (left panel) and
632.8 nm (right panel) lasers. (d) A11g mode frequency versus thickness. (e) The intensity ratio of the A1 g branch (blue circle line) and the Eg symmetry
modes (pink square line). The corresponding intensity ratio of the bulk sample is also shown as a bold line. (f) The fwhm of the E2g mode dependence on
thickness.

to an enhanced electronphonon coupling in the few QL
regime. The high sensitivity of the broadening to thickness also
suggests additional measure of determining the thickness of the
nanoplatelets by Raman spectroscopy.
More detailed and careful line shape analysis indicated that a
Lorentzian line shape function fits all four Raman peaks except for
A11g, which shows a pronounced misfit at both the low and high
energy side of the peak (Supporting Information Figure S-4). The
red shift of A11g is possibly an outcome of the confined phonon
states in 2D Bi2Se3 crystals. Does a confined phonon state in a 2D
model explain the line shape? On the basis of the phonon
confinement phenomenological theory proposed by Richter et al.42
for spherical nanocrystals (i.e., quantum dots) and extended by
Fauchet and Campbell to include nanowires and thin films,43 the
bulk Raman selection rule q = 0 is relaxed such that a range of
phonons extended from the Brillouin zone will be activated in
Raman scattering, therefore leading to red shifting and asymmetric
broadening toward the low energy side. On the basis of this model,
the intensity of the first-order Raman spectrum I0(ω) of a single
NP is given by the RFC (RichterFauchetCampbell) model
Z qmax
jCð0, q^ Þj2
I0 ðωÞ = A0 "  2 # dq^
0 2 Γ ð3Þ
½ω  ωðq^ Þ þ
2 the line shape evolves as a function of thickness according to the
where A0 is a constant amplitude, q^ is the wavevector along the c
axis, C(0,q^) is the Fourier coefficient of the spatial confinement
2411
function, ω(q^) is the phonon dispersion, and Γ is the fwhm of the
natural Raman peak. This model has been successfully used to
interpret the confined phonon states in various nanostructures.4446
More details of the explanation are given in the Supporting
Information. For the thin film geometry, Fauchet and Campbell
proposed the coefficient43
2 
2 
 2
d  iq^ d 
jCð0, q^ Þj ¼ exp4q^
2 2
1  erf pffiffiffiffiffiffiffiffi 
16π2  32π 
where d is the thickness of the NPs. The dispersion relationship
ω0(q) for bulk Bi2Se3 can be described by ω0(q) = [A þ B
cos(aq^)1/2], with A = 4.0808  103 cm2, B = 1.1272 
103 cm2, a = 2.7301 with units of cBi2Se3/3π (here cBi2Se3 =
2.864 nm is the lattice constant of the c axis of Bi2Se334) derived by
fitting the phonon dispersion curve of A11g shown in Figure 4d and
documented in ref 39. Therefore, only one free fitting parameter
A0 needs to be obtained for A11g. Panels a and b of Figure 6 display
plots of both the experimental data and the RFC model line shape
for 16 QL and 7 QL NPs. In the fitting procedure, we only adjust
A0 to make the intensity agree with the data for a certain d. It can be
seen that the 2D confinement model significantly deviates from
the experimental data, especially in thinner NPs. To elucidate how
RFC model (eq 3) for Bi2Se3 NPs, we plotted the calculated line
shape versus thickness as shown in Figure 6c. Below 10 QLs, a
significant asymmetric and broadened line shape is observed.
dx.doi.org/10.1021/nl200773n |Nano Lett. 2011, 11, 2407–2414
Nano Letters
Figure 6. Line shape analysis of Bi2Se3 NPs excited by the 632.8 nm
laser at room temperature: (a, b) comparison between experimental data
(dots) and confinement line shape predicted by the RFC model (eq 3)
for two different thicknesses; (c) line shape evolution as a dependence
on the thickness of Bi2Se3 NPs, calculated by the RFC 2D confinement
model (eq 3); (d, e), Fano line shape fitting to experimental data showed
improved fitting compared with (a) and (b).
Nevertheless, the red shift due to confinement is only less than
2 cm1 for 1 QL, which suggests that the RFC confinement model
cannot explain both the red shift (∼5 cm1 for a 4 QL sample) and
the line shape observed in our experiment.
The phonon confinement model was based on the assumption
that the size of the crystal is still much larger than the lattice
constants (at least in one direction) such that the translational
symmetry and Bloch theorem are still applicable (Supporting
Information eqs S-4 and S-5). It is important to note that Bi2Se3
has an extremely large lattice constant along the c axis (consisting
of 3 QLs), such that the long-range translational symmetry is
readily broken in the regime of the NPs that we studied (e.g., 16
QL and 7 QL corresponds to ∼5 and 2 unit cells along the c axis,
respectively). Therefore the assumption of a classical confinement
model is no longer satisfied (the Supporting Information eq S-5).
Nevertheless, we believe that the uncertainty rule is still
applicable for estimating the frequency of the phonons that
participate in the scattering in finite thickness NPs. At present,
no detailed neutron scattering data or theoretical calculations of
the lattice dynamics exist for finite thickness samples of Bi2Se3
NPs down to the nanometer regime. Thereby, the phonon dis-
persion obtained from a bulk sample will be used for the purpose
of estimation. The uncertainty rule relates the position and mom-
entum, Δx•Δp ≈ h. Here Δp ≈ (h/2π) Δq, Δx = Δd ≈ 2π/Δq.
Using the phonon dispersion curve documented in ref 39, we
obtain the relation of ωA11g = 72.3  2.74(2π/Δd)1.71. Here Δd is
the thickness of NPs in units of nanometers. The blue line in
Figure 5d is the plot of the phenomenological relationship, while
the red circles and pink squares correspond to the experimental
Raman spectra excited by 532 and 632.8 nm lasers, respectively.
2412
LETTER
We can see that the experimental values are in good agreement
with the calculation.
Next, let us discuss the experimental line shape of the
A11g mode. We found that the BreitWignerFano (BWF)
resonance line shape (in short Fano resonance) fits the A11g data
better than the Lorentzian or confinement line shape. A Fano
resonance effect describes a quantum interference effect between
a discrete state (in this case phonon states) and a continuum
transition, in which the excited eigenstates are a mixture of the
discrete and continuum states.47 Historically, the Fano line shape
was observed in degenerete silicon systems,48 metallic carbon
nanotubes,5 graphene,19 and nanowires.49 The Fano resonance is
highly asymmetric, in which the asymmetry depends on the
electronphonon coupling strength. Hence, in those systems
Raman scattering spectroscopy is extensively used to determine
the electronphonon coupling without the need of electrical
transport measurements. The Fano line shape is described by the
following function5
½q þ ðω  ω0 Þ=Γ2
IðωÞ ¼ A ð4Þ
1 þ ½ðω  ω0 Þ=Γ2
where ω0 is the renormalized phonon frequency in the presence
of the coupled scattering, q is the asymmetry parameter, and Γ is
the line-width parameter which is related to the phonon lifetime.
Usually it is convenient to work with 1/q, which is referred to as
the coupling strength. The larger the |1/q| value is, the stronger is
the coupling strength. In the limit of |1/q| f 0, the Fano
resonance line shape reduces to the Lorentzian line shape. In
panels d and e of Figure 6, we display the BWF fitting to the
Stokes Raman spectra of 16 QL and 7 QL thick Bi2Se3 NPs,
respectively. Compared with the RFC model (Figure 6a,b) and
the Lorentzian linshape (Figure S-2), an improved fitting to the
experimental data is observed using the RFC model. Furthermore,
the fitting result revealed that 1/q of 7 QL is much larger than
that for 16 QLs (the value of q is shown in Figure 6d,e).
The positive value of 1/q implies that there is a constructive
interference between the discrete phonon and electronic con-
tinuum on the high-energy side and a destructive interference on
the low-energy side. Therefore, the electronphonon coupling
of 7 QL NPs is much stronger than that for the 16 QL NPs.
Nevertheless we do not know at the moment which continuum
level participates in this coupling to the A11g mode. The details
need to be further studied. Recently, an asymmetry in the Fano
absorption feature positioned at 61 cm1 (which corresponds to
the infrared actived Eu1 mode of Bi2Se3) has been identified in
the infrared spectra of bulk single crystal Bi2Se3 in a magnetic
field.50 Their results indicate that there is an interaction of the
phonon with the continuum free-carrier spectrum and a signifi-
cant magnetoelectric coupling results from their interaction.50
Another report also showed a significant feature with an electron
phonon interaction at the A11g phonon frequency position probed
by ultrafast time-resolved reflection spectra of bulk crystal
Bi2Se3.51
When the coupling between the discrete mode and the
continuum is significant, the Fano resonance peak positon may
blue shift or red shift significantly due to the frequency renormali-
zation, as reported for carbon nanotubes5 and graphite intercala-
tion compounds.52 However, the electronphonon coupling is
weak in our case because the largest value of 1/q is only about 0.1.
This coupling strength changes the line shape of the A11g mode
but will not shift the peak position significantly. Thus we believe
dx.doi.org/10.1021/nl200773n |Nano Lett. 2011, 11, 2407–2414
Nano Letters
that this frequency renormalization is negligible in our experi-
ments, and the main contribution to the red shift of the A11g mode
is still the breaking of the bulk Raman selection rule in the
nanoscale regime and the phonon softening due to the decrease
of the interlayer van der Waals forces.
In conclusion, few QL high crystalline quality topological
insulator Bi2Se3 nanoplatelets were prepared by a solution-based
polyol method. All of the four Raman modes predicted by group
theory were observed and assigned for the first time by their
polarized Stokes and anti-Stokes Raman spectra. The out-of-
plane A11g mode located at ∼72 cm1 in bulk Bi2Se3 exhibits a
sensitive red shift as the thickness decreases. The red shift of the
A11g mode can be explained by a phonon softening because the
interlayer van der Waals forces decrease the effective restoring
force acting on the atoms with decreasing number of quintuple
layers. We found that the 2D RFC phonon confinement model
cannot quantitatively explain the red shift and line shape, and the
spectra can actually be better fit with a BWF resonance line shape.
A significant broadening of the in-plane E2g mode was also identified
as the thickness decreases, resulting from a decrease of the phonon
lifetime of the intralayer modes, possibly due to an enhanced
electronphonon coupling in the few QL regime. Both the
A11g mode frequency and E2g mode line width provide effective and
accurate determinations of the thickness of TI nanoplatelets by
Raman spectroscopy. Our studies provide key evidence that the
electronphonon coupling in the few QL Bi2Se3 system is
remarkably different from that in a bulk sample. Further investiga-
tions are needed to understand how the electronphonon cou-
pling affects the electronic properties in atomically thin Bi2Se3.
’ ASSOCIATED CONTENT
bS Supporting Information. Additional information on
Bi2Se3 NP synthesis, the refractive index of Bi2Se3 by the Kramers
Kronig analysis, laser power dependence of Raman scattering,
and A11g line shape analysis. This material is available free of
charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION
Corresponding Author
*E-mail: Qihua@ntu.edu.sg.
Author Contributions
r
These authors contributed equally to this work.
’ ACKNOWLEDGMENT
Q.X. is grateful for the strong support from the Singapore
National Research Foundation through Singapore 2009 NRF
fellowship grant (NRF-RF2009-06), the Singapore Ministry of
Education via a Tier2 grant (M45110025), very generous start-
up grant support (M58110061) and the New Initiative Fund
(M58110100) from Nanyang Technological University. M.S.D.
acknowledges support from US NSF/DMR-1004147. Q.X. also
acknowledges valuble discussions with Professor Apparao Rao
and Professor Pingheng Tan.
’ REFERENCES
(1) Qi, X. L.; Zhang, S. C. arXiv:1008.2026v1, 2010.
(2) Hasan, M. Z.; Kane, C. L. Rev. Mod. Phys. 2010, 82 (4),
3045–3067.
2413
LETTER
(3) Konig, M.; Wiedmann, S.; Brune, C.; Roth, A.; Buhmann, H.;
Molenkamp, L. W.; Qi, X.-L.; Zhang, S.-C. Science 2007, 318 (5851),
766–770.
(4) Chadov, S.; Qi, X. L.; Kubler, J.; Fecher, G. H.; Felser, C.; Zhang,
S. C. Nat. Mater. 2010, 9 (7), 541–545.
(5) Lin, H.; Wray, L. A.; Xia, Y. Q.; Xu, S. Y.; Jia, S. A.; Cava, R. J.;
Bansil, A.; Hasan, M. Z. Nat. Mater. 2010, 9 (7), 546–549.
(6) Xia, Y.; Qian, D.; Hsieh, D.; Wray, L.; Pal, A.; Lin, H.; Bansil, A.;
Grauer, D.; Hor, Y. S.; Cava, R. J.; Hasan, M. Z. Nat. Phys. 2009, 5 (6),
398–402.
(7) Zhang, H.; Liu, C.-X.; Qi, X.-L.; Dai, X.; Fang, Z.; Zhang, S.-C.
Nat. Phys. 2009, 5 (6), 438–442.
(8) Majumdar, A. Science 2004, 303 (5659), 777–778.
(9) Yu, R.; Zhang, W.; Zhang, H.-J.; Zhang, S.-C.; Dai, X.; Fang, Z.
Science 2010, 329 (5987), 61–64.
(10) Chen, J.; Qin, H. J.; Yang, F.; Liu, J.; Guan, T.; Qu, F. M.; Zhang,
G. H.; Shi, J. R.; Xie, X. C.; Yang, C. L.; Wu, K. H.; Li, Y. Q.; Lu, L. Phys.
Rev. Lett. 2010, 105 (17), 176602.
(11) Peng, H.; Lai, K.; Kong, D.; Meister, S.; Chen, Y.; Qi, X.-L.;
Zhang, S.-C.; Shen, Z.-X.; Cui, Y. Nat. Mater. 2010, 9 (3), 225–229.
(12) Zhang, Y.; He, K.; Chang, C.-Z.; Song, C.-L.; Wang, Li-Li;
Chen, Xi; Jia, J.-F.; Fang, Z.; Dai, X.; Shan, W.-Y.; Shen, S.-Q.; Niu, Q.;
Qi, X.-L.; Zhang, S.-C.; Ma, X.-C.; Xue, Q.-K. Nat. Phys. 2010, 6 (8),
584–588.
(13) Fu, L.; Kane, C. L.; Mele, E. J. Phys. Rev. Lett. 2007, 98 (10),
106803.
(14) Hsieh, D.; Qian, D.; Wray, L.; Xia, Y.; Hor, Y. S.; Cava, R. J.;
Hasan, M. Z. Nature 2008, 452 (7190), 970–974.
(15) Qi, X.-L.; Li, R.; Zang, J.; Zhang, S.-C. Science 2009, 323 (5918),
1184–1187.
(16) Zhang, T.; Cheng, P.; Chen, X.; Jia, J.-F.; Ma, X.; He, K.; Wang,
L.; Zhang, H.; Dai, X.; Fang, Z.; Xie, X.; Xue, Q.-K. Phys. Rev. Lett. 2009,
103 (26), 266803.
(17) Cheng, P.; Song, C.; Zhang, T.; Zhang, Y.; Wang, Y.; Jia, J.-F.;
Wang, J.; Wang, Y.; Zhu, B.-F.; Chen, X.; Ma, X.; He, K.; Wang, L.; Dai,
X.; Fang, Z.; Xie, X.; Qi, X.-L.; Liu, C.-X.; Zhang, S.-C.; Xue, Q.-K. Phys.
Rev. Lett. 2010, 105 (7), 076801.
(18) Dresselhaus, M. S.; Jorio, A.; Hofmann, M.; Dresselhaus, G.;
Saito, R. Nano Lett. 2010, 10 (3), 751–758.
(19) Malard, L. M.; Pimenta, M. A.; Dresselhaus, G.; Dresselhaus,
M. S. Phys. Rep.—Rev. Sec. Phys. Lett. 2009, 473 (56), 51–87.
(20) Castro Neto, A. H.; N. M. R. Peres, F. G.; Novoselov, K. S.;
Geim., A. K. Rev. Mod. Phys. 2009, 81.
(21) Ferrari, A. C.; Meyer, J. C.; Scardaci, V.; Casiraghi, C.; Lazzeri,
M.; Mauri, F.; Piscanec, S.; Jiang, D.; Novoselov, K. S.; Roth, S.; Geim,
A. K. Phys. Rev. Lett. 2006, 97, 18.
(22) Pisana, S.; Lazzeri, M.; Casiraghi, C.; Novoselov, K. S.; Geim,
A. K.; Ferrari, A. C.; Mauri, F. Nat. Mater. 2007, 6 (3), 198–201.
(23) Das, A; Pisana, S; Chakraborty, B; Piscanec, S; Saha, S. K.;
Waghmare, U. V.; Novoselov, K. S.; Krishnamurthy, H. R.; Geim, A. K.;
Ferrari, A. C.; Sood, A. K. Nat. Nanotechnol. 2008, 3 (4), 210–215.
(24) Teweldebrhan, D.; Goyal, V.; Balandin, A. A. Nano Lett. 2010,
10 (4), 1209–1218.
(25) Dang, W.; Peng, H.; Li, H.; Wang, P.; Liu, Z. Nano Lett. 2010,
10 (8), 2870–2876.
(26) Zhao, S. Y. F.; Beekman, C.; Luke, J. Sandilands; Kwok, D.; Lee,
N.; Sang.W. Cheong; Burch, K. S. arXiv:1008.0396v1 [cond-mat.
mtrl-sci], 2010.
(27) Kong, D.; Randel, J. C.; Peng, H.; Cha, J. J.; Meister, S.; Lai, K.;
Chen, Y.; Shen, Z.-X.; Manoharan, H. C.; Cui, Y. Nano Lett. 2009, 10
(1), 329–333.
(28) Hong, S. S.; Kundhikanjana, W.; Cha, J. J.; Lai, K. J.; Kong, D. S.;
Meister, S.; Kelly, M. A.; Shen, Z. X.; Cui, Y. Nano Lett. 2010, 10 (8),
3118–3122.
(29) Zhang, G.; Qin, H.; Teng, J.; Guo, J.; Guo, Q.; Dai, X.; Fang, Z.;
Wu, K. Appl. Phys. Lett. 2009, 95 (5), 053114.
(30) Fievet, F.; Lagier, J. P.; Blin, B.; Beaudoin, B.; Figlarz, M. Solid
State Ionics 1989, 323, 198–205.
dx.doi.org/10.1021/nl200773n |Nano Lett. 2011, 11, 2407–2414
Nano Letters LETTER

(31) Toneguzzo, P.; Viau, G.; Acher, O.; Fievet-Vincent, F.; Fievet,
F. Adv. Mater. 1998, 10 (13), 1032.
(32) Peng, Z. P.; Liu, M. Z.; Yu, C.; Chai, Z. L.; Zhang, H. J.; Wang,
C. Nanoscale 2010, 2 (5), 697–699.
(33) Feldmann, C.; Metzmacher, C. J. Mater. Chem. 2001, 11 (10),
2603–2606.
(34) Sobolev, V. V.; Shutov, S. D.; Popov, Y. V.; Shestatskii, S. N.
Phys. Status Solidi 1968, 30 (1), 349.
(35) Ni, Z. H.; Wang, H. M.; Kasim, J.; Fan, H. M.; Yu, T.; Wu, Y. H.;
Feng, Y. P.; Shen, Z. X. Nano Lett. 2007, 7 (9), 2758–2763.
(36) Greenawa, D. L.; Harbeke, G. J. Phys. Chem. Solids 1965, 26
(10), 1585.
(37) Richter, W.; Kohler, H.; Becker, C. R. Phys. Status Solidi B 1977,
84 (2), 619–628.
(38) Kullmann, W.; Eichhorn, G.; Rauh, H.; Geick, R.; Eckold, G.;
Steigenberger, U. Phys. Status Solidi B 1990, 162 (1), 125–140.
(39) Semiconductors, Non-Tetrahedrally Bonded Elements and Binary
Compounds I; Madelung, O.; R€ossler, U., Schulz, M., Eds.; Springer-
Verlag: Berlin, 1998; Vol. 41C.
(40) Wang, Y. Y.; Ni, Z. H.; Shen, Z. X.; Wang, H. M.; Wu, Y. H.
Appl. Phys. Lett. 2008, 92 (4), 043121.
(41) Zhang, S. L.; Wu, S. N.; Yan, Y.; Hu, T.; Zhao, J.; Song, Y.; Qu,
Q.; Ding, W. J. Raman Spectrosc. 2008, 39 (11), 1578–1583.
(42) Richter, H.; Wang, Z. P.; Ley, L. Solid State Commun. 1981, 39
(5), 625–629.
(43) Campbell, I. H.; Fauchet, P. M. Solid State Commun. 1986, 58
(10), 739–741.
(44) Adu, K. W.; Gutierrez, H. R.; Kim, U. J.; Sumanasekera, G. U.;
Eklund, P. C. Nano Lett. 2005, 5 (3), 409–414.
(45) Xiong, Q. H.; Gupta, R.; Adu, K. W.; Dickey, E. C.; Lian, G. D.;
Tham, D.; Fischer, J. E.; Eklund, P. C. J. Nanosci. Nanotechnol. 2003, 3
(4), 335–339.
(46) Adu, K. W.; Xiong, Q.; Gutierrez, H. R.; Chen, G.; Eklund, P. C.
Appl. Phys. A: Mater. Sci. Process. 2006, 85 (3), 287–297.
(47) Fano, U. Phys. Rev. 1961, 124 (6), 1866.
(48) Cerdeira, F.; Fjeldly, T. A.; Cardona, M. Phys. Rev. B 1973, 8
(10), 4734–4745.
(49) Gupta, R.; Xiong, Q.; Adu, C. K.; Kim, U. J.; Eklund, P. C. Nano
Lett. 2003, 3 (5), 627–631.
(50) LaForge, A. D.; Frenzel, A.; Pursley, B. C.; Lin, T.; Liu, X. F.;
Shi, J.; Basov, D. N. Phys. Rev. B 2010, 81 (12), 125120.
(51) Qi, J.; Chen, X.; Yu, W.; Cadden-Zimansky, P.; Smirnov, D.;
Tolk, N. H.; Miotkowski, I.; Cao, H.; Chen, Y. P.; Wu, Y.; Qiao, S.; Jiang,
Z. Appl. Phys. Lett. 2010, 97, 18.
(52) Eklund, P. C.; Subbaswamy, K. R. Phys. Rev. B 1979, 20 (12),
5157–5161.

2414 dx.doi.org/10.1021/nl200773n |Nano Lett. 2011, 11, 2407–2414