Phonon dispersions and vibrational properties of monolayer, bilayer, and trilayer

graphene
Jia-An Yan, W. Y. Ruan, and M. Y. Chou
School of Physics, Georgia Institute of Technology, Atlanta, Georgia 30332-0430, U.S.A.
(Dated: October 29, 2018)
The phonon dispersions of monolayer and few-layer graphene (AB bilayer, ABA and ABC trilay-
ers) are investigated using the density-functional perturbation theory (DFPT). Compared with the
monolayer, the optical phonon E2g mode at Γ splits into two and three doubly degenerate branches
arXiv:0901.3093v1 [cond-mat.mtrl-sci] 20 Jan 2009

for bilayer and trilayer graphene, respectively, due to the weak interlayer coupling. These modes are
of various symmetry and exhibit different sensitivity to either Raman or infrared (IR) measurements
(or both). The splitting is found to be 5 cm−1 for bilayer and 2 to 5 cm−1 for trilayer graphene.
The interlayer coupling is estimated to be about 2 cm−1 . We found that the highest optical modes
at K move up by about 12 cm−1 for bilayer and 18 cm−1 for trilayer relative to monolayer graphene.
The atomic displacements of these optical eigenmodes are analyzed.

PACS numbers:

I. INTRODUCTION surements are very close. In contrast to bulk graphite,

Raman scattering has been widely used for probing the
In recent years, monolayer and few-layer graphene G-band in graphene layers that corresponds to the Γ
have attracted great attention due to the unique proper- phonons.21,22,23,24 Recently reported Raman spectra for
ties observed experimentally.1,2,3 Many intriguing trans- graphene layers show that the intensity and position of
port phenomena, such as ballistic transport at room the first-order G-band as well as the second order D-
temperature,4,5 the anomalous quantum Hall effect,6,7 band (historically named the G’ band) are modified with
and novel many-body couplings8 have been reported. an increasing number of layers.21,22
In addition to being a physical system exhibiting novel On the theoretical side, Grüneis et al.25 presented the
properties, graphene and graphene layers have been pro- phonon dispersions of graphite using the 4th-nearest-
posed as promising candidates for future nanoelectron- neighbor force constant (4NNFC) approach. However,
ics. The epitaxial graphene grown on SiC is of particu- it has been argued that due to the Kohn anomaly at Γ
lar interest due to the compatibility with current silicon and K, it is not possible to obtain the correct phonon dis-
technology.1,9 persions near Γ and K from the force constant method.13
Besides their unusual electronic structure,10 vibra- Dubay and Kresse14 performed density-functional theory
tional properties and phonon spectra are also of fun- (DFT) calculations of the phonon dispersions in graphite
damental interest from which many physical properties within the local density approximation (LDA). Their re-
(such as thermal conductivity and heat capacity) can sults are in good agreement with phonon-measurements
be derived. Furthermore, phonons are crucial for study- by HREELS. Using the LDA and the generalized-
ing the quasiparticle dynamics8 and electrical transport gradient approximation (GGA), Wirtz and Rubio26 cal-
properties. Electrons excited by optical methods can culated the phonon dispersions of graphite and obtained
be scattered into another states by electron-phonon cou- results close to the vast majority of the experimen-
pling (EPC). It has been suggested that the scattering tal data-points. At the GGA-PBE level, Mounet and
between electrons and the optical phonon modes greatly Marzari27 also presented a detailed calculation of the
affects the high-field ballistic transport properties in car- phonon dispersions of graphene and graphite.
bon nanotubes.11 In graphene and metallic single-wall With regard to graphene layers, it is unclear how the
nanotubes, the EPC strongly affects the phonon frequen- phonon properties are affected by the stacking order and
cies, giving rise to Kohn anomalies12,13 and possible soft the weak interlayer coupling. This effect is important for
modes or Peierls distortions.14,15 understanding the EPC in multilayer graphene as well as
Many experimental methods have been used to mea- the interpretation of the Raman spectra. For example,
sure the phonon dispersions of graphite, such as inelas- the phonon dispersion around K is crucial for the correct
tic neutron scattering (INS),16 electron-energy loss spec- interpretation of the Raman second order D peak. It has
troscopy (EELS),17 high-resolution electron energy-loss been shown that in few-layer graphene, the electronic
spectroscopy (HREELS),18 and inelastic x-ray scatter- dispersions near the Fermi level exhibit various features
ing (IXS).19,20 These measurements require large enough depending on the stacking order.10 In this work, the vi-
samples of crystalline quality and are limited to specific brational properties of one- and few-layer graphenes are
directions or phonon modes. More recently, Mohr et al.20 calculated using density-functional perturbation theory
have presented complete measurements of both the opti- (DFPT).28 The monolayer, bilayer (AB stacking), and
cal and the acoustic phonon modes along the directions trilayer (ABA and ABC stackings) are considered in or-
Γ-K-M-Γ of graphite using IXS. The results in these mea- der to illustrate the effects of stacking order and inter-
 2

20
a) Graphene c) ABA trilayer d) ABC trilayer

Phonon energy (meV)

A
A C 15

b) AB bilayer
B B B 10

5
A A A
0
Γ A
FIG. 1: Stacking structure for (a) monolayer, (b) AB bilayer,
(c) ABA trilayer and (d) ABC trilayer graphene. FIG. 2: (Color online) Phonon dispersions of graphite along
the Γ-A direction. Solid lines are present calculational results.
Circles are IXS data from Ref. 20, and squares are neutron
scattering data from Ref. 16. The dotted lines are smooth
curves through the measured points.
layer coupling. The van der Waals corrections in graphite
have been shown to be important in order to correctly de-
scribe the long-range binding properties.29,30,31 However,
previous theoretical calculations based on DFT with both
LDA14 and GGA20,27 have indicated that rather reason-
able vibrational properties of graphite can be obtained III. RESULTS AND DISCUSSIONS
within DFT as compared with experiments. We find that
the phonon dispersions for graphene and graphene layers
exhibit somewhat different characteristics, especially at
Γ and K. Detailed analysis of the phonon modes is also The optimized LDA lattice constant in the graphene
presented. plane is 2.45 Å, in good agreement with the previous cal-
culated result,10 which is also close to the experimental
value of 2.46 Å for graphite.35 For the bilayer and tri-
layer systems, the lattice constant in the plane remains
almost the same as in graphene. The optimized interlayer
spacing is 3.33 Å, slightly smaller than the experimen-
tal value of 3.35 Å in bulk graphite.35 In comparison,
II. COMPUTATIONAL DETAILS
we obtained a theoretical value of 3.32 Å for graphite,
which is close to previous LDA results.36 The interlayer
Density-functional calculations are performed us- binding energy (defined as the total energy difference be-
ing the ESPRESSO code32 with the LDA. Troullier- tween the coupled and uncoupled graphene layers) of bulk
Martin (TM) norm-conserving pseudopotentials33 gen- graphite is calculated to be 25.2 meV/atom, while this
erated from the valence configuration of 2s2 2p2 for C are energy falls to 12.3 meV/atom for an AB bilayer. For
employed. The wavefunction and the charge density are the ABA and ABC trilayer, the interlayer binding en-
expanded using energy cutoffs of 110 and 440 Ryd, re- ergies are both 16.5 meV/atom. Our result of graphite
spectively. Methfessel-Paxton smearing34 with an energy is comparable to previous calculations using a combined
width of 0.03 Ryd is adopted for the self-consistent cal- density-functional and intermolecular perturbation the-
culations. The dynamical matrices are calculated based ory approach.31 To validate the DFPT phonon calcula-
on DFPT within the linear response. For the integra- tions, we also calculated the low energy phonon disper-
tion over electronic states in the calculations, we use a sions in bulk graphite along the Γ-A direction (perpendic-
48 × 48 × 1 uniform k-point mesh. A 6 × 6 × 1 grid is ular to the layers), as shown in Fig. 2. Except for a small
used for the phonon calculation to obtain the dynami- frequency shift, our LDA dispersions agree with experi-
cal matrices. We have carefully tested these parameters mental data rather well. Therefore, we believe our cal-
and the phonon frequencies are converged to be within 1 culations yield reliable descriptions of the phonon prop-
cm−1 . erties in graphite and graphene layers. This agreement
between LDA and experimental results indicate an error
Figure 1 shows the two-dimensional (2D) primitive cancellation for energy variations near the equilibrium
cells for monolayer, bilayer (AB stacking), and trilayer layer separation, even though the state-of-the-art local
graphene (ABA and ABC stacking). A large vacuum re- or semilocal exchange-correlation functionals are not able
gion of more than 10 Å along the z direction is used to properly describe the long-range interlayer interactions
to minimize the interactions between graphene layers in dominated by van der Waals dispersion forces (see Refs.
different supercells. [37] and [38] for details).
 3

1600 LO LO (a) (b)

TO TO
Phonon frequency (cm )
-1 1400
1200
1000 ZO ZO
β β
800
600 α α
LA
400 LA
TA TA
200
ZA ZA FIG. 4: (Color online) Pattern of atomic displacements for
0
Γ K M Γ the TO/LO modes at Γ in monolayer graphene.

FIG. 3: Phonon dispersions for monolayer graphene.

~ m − ωqν t + φj ), (1)
X
= ês |ǫjqν,s |cos(~q · R qν,s
s=x,y,z
A. Phonon Properties of Graphene
where R ~ m is the lattice vector for unit cell m, φj
qν,s de-
Figure 3 shows the phonon dispersions for monolayer notes the phase factor of the complex eigenvector ǫjqν,s ,
graphene calculated at the theoretical lattice constant, and ωqν is the phonon frequency.
which will be compared with multilayer results in the For the degenerate TO/LO phonon modes at Γ,
next section. In contrast to the linear dispersion near the Figs. 4(a) and (b) schematically show the atomic dis-
Γ point for the in-plane TA and LA modes, the out-of- placements associated with the two eigenmodes. Clearly,
plane ZA mode shows a q 2 dispersion, which is a charac- two neighboring atoms vibrate opposite to one another.
teristic feature of the phonon dispersions in layered crys- This gives rise to a large bond distortion and couples to
tals as observed experimentally.27,39,40 The same feature electronic states near the Dirac point (which can be pro-
also appears in bilayer and trilayer phonon dispersions, jected into two states localized at atom α and β, respec-
as will be discussed in Section III B. tively) through an intravalley scattering (with phonon
The calculated frequency (1586 cm−1 ) of the degen- q≈0). Therefore, a strong EPC is expected, which has
erate LO and TO modes at Γ is slightly smaller than also been demonstrated by the effective mass theory42 as
the previous value of 1595 cm−1 obtained by Dubay well as the tight-binding model.43
et al.,14 but is in excellent agreement with the exper- In contrast, for the highest TO A′1 mode at K the clas-
imental result of 1587 cm−1 by inelastic x-ray scatter- sical displacements of neighboring atoms α and β follow
ing measurements.19 At the Brillouin zone corner K, the the pattern
phonon energy of the symmetric TO A′1 mode (1306
cm−1 ) is close to the frequency (1326 cm−1 ) calculated π π
by Wirtz et al.26 Our result is also consistent with the es- ~uα = u0 [êx cos( − ωt) − êy sin( − ωt)], (2)
2 2
timate by Yao et al.41 from their high-voltage transport π π
measurements for graphite. They suggested that scatter- ~uβ = u0 [êx cos( − ωt) + êy sin( − ωt)] (3)
2 2
ing by phonons with an energy of about 1300 cm−1 gives
rise to the dramatic conductance drop at a high bias. In and atoms α and β move circularly. In particular, one
general, our calculated phonon dispersions for monolayer moves counterclockwise, while the other clockwise, as
graphene are comparable with those obtained in previous shown in Fig. 5. Accordingly, each atom approaches its
calculations14,26 and agree very well with experimental three nearest neighbors successively during one period.
results.19,41 Fig. 5 shows three snapshots of the atomic displacements
A previous study by Piscanec et al.13 showed that the in one period.
degenerate E2g modes at Γ and the highest TO mode at Since the degenerate electronic states at the Dirac
K have a strong EPC, leading to Kohn anomalies in the point can be projected into two states localized at atom
phonon dispersions. A detailed analysis of the origin of α and β, respectively, the above mode of ionic vibra-
the strong EPC for these modes is presented below. tion facilitates the transition of an electron from atoms
For a specific phonon mode ν with wave vector q, the α to β or vice versa, resulting in the electronic interval-
displacement of atom j (j=α,β) in unit cell m will oscil- ley scattering via a phonon with momentum K. ~ There-
late according to the following expression in the classical fore, a strong EPC is expected for this mode compared
picture: with other modes at K.44 Based on the detailed analyses
of these modes, we anticipate a distinct electron-phonon
~
X
~umj
qν = ês Re{ǫjqν,s ei(~q·Rm −ωqν t) } interactions for these modes in few-layer graphene. The
s=x,y,z results will be presented elsewhere.
 4

(a) (b) (c) high-frequency phonons are derived from the superposi-
tions of intralayer optical modes in each graphene plane.
For the modes in the bilayer as shown in Fig. 7, the two
atoms on top of each other in two adjacent layers vibrate
β' β' β'
either in the opposite direction (Eg mode, 1587 cm−1 )
α α α or in the same direction (Eu mode, 1592 cm−1 ). Sim-
β" β β" β β" β
ilar atomic displacements can also be seen in ABA and
ABC trilayers, as shown in Figs. 8 and 9. In other words,
FIG. 5: (Color online) Three snapshots of atomic displace- the original intralayer modes couple to each other via in-
ments for the highest TO mode at K in monolayer graphene, terlayer interactions, giving rise to a small splitting in
where atom α approaches its three nearest neighbors β (a), the final frequencies. The upper and lower modes in the
β ′ (b), and β ′′ (c) successively.
bilayer correspond to the ‘in-phase’ and ‘out-of-phase’
superpositions of the two intralayer modes, respectively,
similar to the E1u and E2g modes in bulk graphite.
B. Phonon Dispersions for Graphene Layers This splitting of the phonon frequencies at Γ can be
illustrated using a simple model. Using the original in-
In this section, we focus on the optical phonon modes tralayer optical modes as the basis and assuming the in-
in multilayer graphene. The phonon dispersions for bi- teraction strength between adjacent layers is ǫ, the re-
layer and trilayer are shown in Fig. 6 (a). The detailed duced Hamiltonian for the bilayer and trilayer can be
dispersions for the high optical branches near Γ and K are expressed as:
enlarged in Figs. 6(b) and (c), respectively. The optical    
phonon frequencies are also listed in Table I. 0 ǫ E0 ǫ
H2 = E0 I + = (4)
ǫ 0 ǫ E0
Compared with the monolayer result, several distinct
features can be identified for graphene multilayers. First, and
there is one (two) additional low-frequency mode with en-
ergy of about 90 cm−1 at Γ in bilayer (trilayer) graphene.
 
E0 ǫ 0
These modes arise from interlayer movement (so-called H3 =  ǫ E0 + δ ǫ  , (5)
‘layer breathing’ modes). Second, at Γ the doubly de- 0 ǫ E0
generate E2g branch in the monolayer evolves into two
(three) doubly degenerate branches for bilayer (trilayer) respectively. Here, E0 is the energy of the intralayer
graphene, as shown in Fig. 6(b). These small splittings mode, and only the first nearest-neighbor layer-layer in-
are due to the weak interlayer coupling: about 5 cm−1 teraction is considered. For the trilayer, a small variant
for bilayer and no more than 5 cm−1 for trilayer (see Ta- of δ is introduced in Eq. (5) to account for the change
ble I). Moving away from Γ, each of these degenerate of the on-site energy in the middle layer due to the new
branches breaks into two different modes. Recent exper- geometry. (This is similar to the on-site energy variation
iments show that the Raman G-peak intensity enhances due to environmental changes in electronic tight-binding
almost linearly with respect to the layer number (up models.45,46 ) Solving the secular equation det(H−λI)=0,
to four layers).21,23 This phenomenon could be ascribed one obtains the eigenvalues and eigenvectors.
to the increased number of optical phonon modes at Γ For the bilayer, the eigenvalues are λ1,2 = E0 ±ǫ. From
within a small energy window for multilayer graphene. Fig. 7, the E0 and ǫ can be determined: E0 =1589.5 and
The stackings of graphene layers have various point |ǫ|=2.5 cm−1 . E0 shows a small shift compared with the
group symmetry for the Γ phonons. The monolayer value for a single-layer graphene (1586 cm−1 ) as a result
graphene possesses the D6h symmetry (Schönflies nota- of the environmental change mentioned above. With ǫ >
tion). It reduces to D3d for the AB bilayer and ABC 0, the corresponding eigenvectors are φ1,2 = (1, ±1)T .
trilayer, and D3h for the ABA trilayer. Correspondingly, This is consistent with the displacements we obtained in
their high optical zone-center modes are of different mode Fig. 7. The lower frequency corresponds to the out-of-
symmetry: E2g mode in graphene evolves into Eg and Eu phase superposition of the two intralayer modes (with
for the AB bilayer, 2E ′ +E ′′ for the ABA trilayer, and respect to the motion of the two atoms on top of each
2Eg +Eu for the ABC trilayer. The Eg and E ′′ modes other in two adjacent layers), while the higher one corre-
are Raman active, Eu is IR active, while the E ′ modes sponds to the in-phase superposition. In both cases, the
are both Raman and IR active. Therefore, a complete two intralayer modes have equal amplitudes.
picture of the zone-center modes can be obtained from For the
√ trilayer, the eigenvalues are λ1 =E0 , λ2,3 =E0 +
a combination of Raman and IR measurements. These (δ ± δ 2 + 8ǫ2 )/2, with corresponding
√ eigenvectors
mode splittings provide significant information about the φ1 =(1, 0, −1)T , φ2,3 =(ǫ, (δ ± δ 2 + 8ǫ2 )/2, ǫ)T . For the
layer number and the stacking geometry. latter two eigenvectors, the ratio of the √
mode amplitudes
In Figs. 7-9, we show the schematic atomic displace- in each layer is a1 : a2 : a3 =ǫ : (δ ± δ 2 + 8ǫ2 )/2 : ǫ.
ments of these optical eigenmodes at Γ for the AB bilayer, Using the frequencies for the ABA trilayer as shown in
ABA trilayer, and ABC trilayer, respectively. These Fig. 8, we obtain δ ≈ 3 cm−1 and ǫ ≈ 2.2 cm−1 , and the
 5

(a) (b) (c)

1600
1620 1350
1200
1600 1300
AB bilayer
800
Phonon frequency (cm )
-1

1580 1250
400
1200
0 1560
1600
1620 1350
1200
1600 1300
800 ABA trilayer
1580 1250
400
1200
0 1560
1600
1620 1350
1200
1600 1300
800 ABC trilayer

1580 1250
400
1200
0 1560
Γ K M Γ 0.00 0.05 0.10 0.15 0.20 0.25 0.5 0.6 0.7 0.8 0.9
Γ K [2π/a0] Γ K [2π/a0]

FIG. 6: Phonon dispersions for graphene multilayers. From top to bottom: AB bilayer, ABA trilayer, and ABC trilayer.
Column (a): full phonon spectra; column (b): optical phonon dispersions near Γ; column (c): optical phonon dispersions near
K.

TABLE I: High optical phonon frequencies ω (in cm−1 ) at Γ and K for monolayer, bilayer, trilayer graphene, and bulk graphite.
The phonon frequencies at Γ and K from recent DFT calculations (with LDA and GGA) as well as experimental measurements
are listed for comparison. The point group symmetry at Γ (K) for the monolayer, AB bilayer, ABA trilayer, ABC trilayer, and
graphite is D6h (D3h ), D3d (C3v ), D3h (C3h ), D3d (C3v ), and D6h (D3h ), respectively. In parentheses are the mode symmetries.
Graphene AB ABA ABC Graphite Graphite exp.
Γ 1586 (E2g ) 1587 (Eg ) 1586 (E ′ ) 1586 (Eg ) 1586 (E2g ) 1582f ,1581h
1595a , 1597b c 1592 (Eu ) 1588 (E ′′ ) 1589 (Eu ) 1595 (E1u ) 1588g
1569d , 1581e 1593 (E ′ ) 1594 (Eg )

K 1306 (A′1 ) 1318 (E) 1316 (E1′ , E1′′ ) 1318 (E) 1322 (E)
1371a , 1326c 1324 (E2′ ) 1325 (A1 )
1289d , 1300e
1265h
a LDA, soft projector augmented wave (PAW), Ref. [14].
b LDA, Hard PAW, Ref. [14].
c LDA, TM potentials, Ref. [26].
d GGA, TM potentials, Ref. [26].
e GGA, Ref. [19].
f Expt. ω(E ), Refs.[47,48,49].
2g
g Expt. ω(E ), Ref. [47] and [50].
1u
h Inelastic X-ray data of Ref. [19] and [20].

mode amplitude ratios of 2.2:5.0:2.2 and 2.2:(-2.0):2.2 for one can easily estimate the optical phonon frequencies
these two modes, respectively. This result agrees with for more graphene layers using an interlayer interaction
the displacements from the direct first-principles calcu- of 2-3 cm−1 and the values of E0 and δ obtained above.
lations shown in Fig. 8. Similar results can be obtained The highest optical phonon branch at K becomes dou-
for the ABC trilayer. Based on these numerical results, bly degenerate at 1318 cm−1 in the bilayer system (see
 6

-1
Table I), nearly 12 cm−1 higher than that in graphene. (a) 1587 cm (b) 1592 cm
-1

The degeneracy is imposed by the symmetry of the bi-

layer. The two degenerate modes correspond to two in-
tralayer modes within individual layers with little cou-
pling between them. In contrast, the optical phonons in
the ABA split into three modes, with two of them be-
ing almost degenerate, while the degeneracy is imposed
by symmetry in the ABC trilayer. For the ABA and
ABC trilayers, the highest phonon (singlet) frequencies
FIG. 7: (Color online) Atomic displacements of the two split
are 1324 and 1325 cm−1 , respectively. This result is con-
optical branches (a) 1587, and (b) 1592 cm−1 at Γ for the AB
sistent with the Raman observation that the second-order bilayer. Only one mode for each degenerate pair is shown.
D mode at about 2700 cm−1 increases with an increasing
layer number.51 The second-order D mode in the Raman -1 -1 -1
spectrum of graphene and graphene layers, which is dou- (a) 1586 cm (b) 1588 cm (c) 1593 cm
ble of the highest optical phonon frequency at K, can be
well illustrated using a double-resonant model.51
Figure 10 shows the schematic atomic displacements
of the three eigenmodes at K for the ABA trilayer. Each′
eigenmode comprises a superposition of the intralayer A1
modes from each layer. As shown in Fig. 10, the two al-
most degenerate low-frequency modes (about 1316 cm−1 )
correspond to the combinations of the modes from the
top and bottom A layers, while the high branch is an
intralayer mode from the middle B layer almost exclu-
FIG. 8: (Color online) Atomic displacements of the three split
sively.
optical branches (a) 1586, (b) 1588, and (c) 1593 cm−1 at Γ
The splitting in frequencies at K can be analyzed in a for the ABA trilayer. The length of the arrow represents
similar way as before. In contrast to the Γ phonons, the the amplitude of the eigenvector. Only one mode for each
coupling between adjacent intralayer modes is zero due degenerate pair is shown.
to the mode symmetry, and there might only be a small
interaction between the top and the bottom layers. In
this case, the Hamiltonian can be expressed as: layer graphene using the density-functional perturbation
 ′
E0 0 η
 theory. Due to the weak coupling between layers, the
H =  0 E0′ + δ ′ 0  , (6) highest optical phonon branch at Γ in graphene splits
η 0 E0′ into two (three) doubly degenerate branches with small
yet unnegligible splittings for bilayer (trilayer) graphene.
with η the small interaction between second nearest- The splitting is about 5 cm−1 for the bilayer. In trilayer
neighbor layers. The eigenvalues are λ = E0′ ± η, and ABA and ABC graphene, these splittings are about 2
E0′ +δ ′ . According to the frequencies as shown in Fig. 10, cm−1 and 5 cm−1 , respectively, which are not equally
we obtain δ ′ =7.2 cm−1 and |η|=0.2 cm−1 . The ampli- spaced. These modes are of various mode symmetry and
tudes are a1 : a2 : a3 =±1 : 0 : 1 for λ = E0′ ± η, and exhibit different sensitivity to either Raman or IR mea-
0 : 1 : 0 for λ = E0′ + δ ′ . This is consistent with the surements and therefore a combination of Raman and IR
atomic displacements illustrated in Fig. 10 with η <0.
Note that the small interactions between second nearest- -1 -1 -1
(a) 1586 cm (b) 1589 cm (c) 1594 cm
neighbor layers have induced a small splitting of the two
low-frequency modes.
In the case of ABC stacking, the interlayer coupling
matrix elements are identically zero. Therefore, the three
eigenmodes at K are localized on each of the three lay-
ers, respectively. Due to the higher on-site energy in the
middle layer, one state will be higher than the other two
modes, with the lower two modes doubly degenerate due
to the symmetry.

FIG. 9: (Color online) Atomic displacements of the three split

IV. SUMMARY optical branches (a) 1586, (b) 1589, and (c) 1594 cm−1 at Γ
for the ABC trilayer. The length of the arrow represents
In summary, we have studied the phonon dispersions the amplitude of the eigenvector. Only one mode for each
and vibrational properties for monolayer, bilayer, and tri- degenerate pair is shown.
 7

-1 -1
(a) 1316.2 cm (b) 1316.5 cm (c) 1324 cm
-1
The frequency of the highest optical phonon mode at
K in bilayer (trilayer) graphene is about 12 (18) cm−1
higher than that in monolayer graphene. For trilayer
graphene, the K-A′1 mode splits into three branches in
the ABA trilayer, with the two lower modes nearly dou-
bly degenerate. It is found that the on-site energy varia-
tions for the middle layer in ABA and ABC are about 7-8
cm−1 , higher than that of Γ phonons. Due to the sym-
metry, the interlayer coupling between adjacent layers for
these intralayer modes is zero.
FIG. 10: (Color online) Atomic displacements of the three
optical phonon modes (a) 1316.2, (b) 1316.5, and (c) 1324
cm−1 at K for the ABA trilayer. The length of the arrow Acknowledgments
represents the amplitude of the eigenvector. Only one mode
for each degenerate pair is shown.
We acknowledge helpful discussions with M.
Wierzbowska, S. Piscanec, and A. C. Ferrari. We
measurements of the zone-center optical modes should thank M. Mohr for providing the experimental data
give a clear identification of the layer number as well as of bulk graphite. This work is supported by the De-
the stacking geometry. partment of Energy (Grant No. DE-FG02-97ER45632)
A simple interaction model is applied to illustrate the and by the National Science Foundation (Grants No.
frequency splitting and the characteristics of the eigen- DMR-02-05328). The computation used resources of the
modes at Γ. The interlayer coupling strength is identified National Energy Research Scientific Computing Center
as about 2 cm−1 . In the trilayer system, a shift of about (NERSC), which is supported by the U.S. Department
3 cm−1 in the on-site energy in the middle layer is deter- of Energy (Grant No. DE-AC03-76SF00098), and San
mined. Diego Supercomputer Center (SDSC) at UCSD.

1
W. A. de Heer, C. Berger, X. Wu, P. N. First, E. H. Con- Robertson, Phys. Rev. Lett. 93, 185503 (2004).
14
rad, X. Li, T. Li, M. Sprinkle, J. Hass, M. L. Sadowski, O. Dubay, G. Kresse, Phys. Rev. B 67, 035401 (2003).
15
M. Potemski, G. Martinez, Solid State Commun. 143, 92 S. Piscanec, M. Lazzeri, J. Robertson, A. C. Ferrari, and
(2007).. F. Mauri, Phys. Rev. B 75, 035427 (2007).
2 16
A. K. Geim, and K. S. Novoselov, Nature Materials 6, 183 R. Nicklow, N. Wakabayashi, and H. G. Smith, Phys. Rev.
(2007). B 5, 4951 (1972).
3 17
For a recent review, see M. I. Katsnelson, Mater. Today, C. Oshima, T. Aizawa, R. Souda, Y. Ishizawa, and Y.
10, 20 (2007), and references therein. Sumiyoshi, Solid State Commun. 65, 1601 (1988).
4 18
C. Berger, Z. M. Song, T. B. Li, X. B. Li, A. Y. Ogbazghi, S. Siebentritt, R. Pues, K.-H. Rieder, and A. M. Shikin,
R. Feng, Z. T. Dai, A. N. Marchenkov, E. H. Conrad, P. Phys. Rev. B 55, 7927 (1997).
19
N. First, and W. A. de Heer, J. Phys. Chem. B 108, 19912 J. Maultzsch, S. Reich, C. Thomsen, H. Requardt, P. Or-
(2004). dejón, Phys. Rev. Lett. 92, 075501 (2004).
5 20
C. Berger, Z. Song, T. Li, X. Li, A. Y. Ogbazghi, R. Feng, M. Mohr, J. Maultzsch, E. Dobardz̆ić, S. Reich, I.
Z. Dai, A. N. Marchenkov, E. H. Conrad, P. N. First, W. Milos̆ević, M. Damnjanović, A. Bosak, M. Krisch, and C.
A. de Heer, Science 312, 1191 (2006). Thomsen, Phys. Rev. B 76, 035439 (2007).
6 21
K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, D. Graf, F. Molitor, K. Ensslin, C. Stampfer, A. Jungen,
M. I. Katsnelson, I. V. Grigorieva, S. V. Dubonos, A. A. and C. Hierold, and L. Wirtz, Nano Lett. 7, 238 (2007).
22
Firsov, Nature 438, 197 (2005). A. C. Ferrari, J. C. Meyer, V. Scardaci, C. Casiraghi, M.
7
Y. Zhang, Y. W. Tan,, H. L. Stormer, and P. Kim, Nature Lazzeri, F. Mauri, S. Piscanec, D. Jiang, K. S. Novoselov,
438, 201 (2005). S. Roth, and A. K. Geim, Phys. Rev. Lett. 97, 187401
8
A. Bostwick, T. Ohta, T. Seyller, K. Horn and Eli Roten- (2006).
23
berg, Nature Physics 3, 36 (2007). A. Gupta, G. Chen, P. Joshi, S. Tadigadapa, and P.C.
9
J. Hass, R. Feng, J. E. Millan-Otoya, X. Li, M. Sprinkle, Eklund, Nano Lett. 6, 2667 (2006).
24
P. N. First, C. Berger, W. A. de Heer, and E. H. Conrad, J. Yan, Y. Zhang, P. Kim, and A. Pinczuk, Phys. Rev.
Physical Rev. B 75, 214109 (2007). Lett. 98, 166802 (2007).
10 25
S. Latil, and L. Henrard, Phys. Rev. Lett. 97, 036803 A. Grüneis, R. Saito, T. Kimura, L. G. Cancado, M. A.
(2006). Pimenta, A. Jorio, A. G. Souza Filho, G. Dresselhaus, M.
11
M. Lazzeri, S. Piscanec, F. Mauri, A. C. Ferrari, and J. S. Dresselhaus, Phys. Rev. B 65, 155405 (2002).
26
Robertson, Phys. Rev. Lett. 95, 236802 (2005). L. Wirtz, A. Rubio, Solid State Comm. 131, 141(2004).
12 27
W. Kohn, Phys. Rev. Lett. 2, 393 (1959). N. Mounet and N. Marzari, Phys. Rev. B 71, 205214
13
S. Piscanec, M. Lazzeri, F. Mauri, A. C. Ferrari, and J. (2005).
 8

28 39
S. Baroni, S. de Gironcoli, and A. Dal Corso, Rev. Mod. H. Zabel, J. Phys.: Condens. Matter 13, 7679 (2001).
40
Phys. 73, 515 (2001). I. M. Lifshitz, Zh. Zksp. Teor. Fiz. 22, 475 (1952).
29 41
L. A. Girifalco and M. Hodak, Phys. Rev. B 65, 125404 Z. Yao, C. L. Kane, and C. Dekker, Phys. Rev. Lett. 84,
(2002). 2941 (2000).
30 42
S. D. Chakarova-Käck, E. Schröder, B. I. Lundqvist, and T. Ando, J. Phys. Soc. Jpn. 75, 124701 (2006).
43
D. C. Langreth, Phys. Rev. Lett. 96, 146107 (2006). J. Jiang, R. Saito, A. Grüneis, G. Dresselhaus, and M. S.
31
Y. J. Dappe, M. A. Basanta, F. Flores, and J. Ortega, Dresselhaus, Chem. Phys. Lett. 392, 383 (2006).
44
Phys. Rev. B 74, 205434 (2006). M. Lazzeri and F. Mauri, Phys. Rev. Lett. 97, 266407
32
S. Baroni, A. Dal Corso, S. de Gironcoli, P. Giannozzi, (2006).
45
C. Cavazzoni, G. Ballabio, S. Scandolo, G. Chiarotti, P. D. J. Chadi, in: V. Vitek and D. J. Srolovitz (Eds), Atom-
Focher, A. Pasquarello, K. Laasonen, A. Trave, R. Car, N. istic Simulations of Materials: Beyond Pair Potentials
Marzari, A. Kokalj, http://www.pwscf.org/ (Plenum Press, New York and London, 1989) p.309.
33 46
N. Troullier and J. L. Martins, Phys. Rev. B 43, 1993 J. L. Mercer, and M. Y. Chou, Phys. Rev. B 49, R8506
(1991). (1994).
34 47
M. Methfessel and A. T. Paxton, Phs. Rev. B 40, 3616 R. J. Nemanich, S. A. Solin, Phys. Rev.B 20, 392 (1979).
48
(1989). F. Touinstra, J. L. Koenig, J. Chem. Phys. 53, 1126 (1970).
35 49
R. W. G. Wyckoff, Crystal Structure, Vol. 1 (Interscience, L. J. Brillson, E. Burstein, A. A. Maradudin, T. Stark, in:
New York, 1963). D. L. Carter, R. T. Bate (Eds.), The Physics of Semimet-
36
N. Ooi, A. Rairkar, J. B. Adams, Carbon 44, 231 (2006). als and Narrow Gap Semiconductors (Pergamon, Oxford,
37
W. Kohn, Y. Meir, and D. E. Makarov, Phys. Rev. Lett. 1997), p. 187.
50
80, 4153 (1998). R. A. Friedel, G. C. Carlson, J. Phys. C 75, 1149 (1971).
38 51
H. Rydberg, M. Dion, N. Jacobson, E. Schroder, P. A. C. Ferrari, Solid State Commun. 143, 47 (2007).
Hyldgaard, S. I. Simak, D. C. Langreth, and B. I.
Lundqvist, Phys. Rev. Lett. 91, 126402 (2003).