CARBON 5 2 ( 2 0 1 3 ) 2 0 1 –2 0 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/carbon

Electronic thermal conductivity and thermopower

of armchair graphene nanoribbons

A.S. Nissimagoudar, N.S. Sankeshwar *

Department of Physics, Karnatak University, Dharwad 580 003, Karnataka, India

A R T I C L E I N F O A B S T R A C T

Article history:
Received 24 May 2012
Accepted 10 September 2012
Available online 17 September 2012
A theoretical investigation of the diffusion contribution to thermopower, Sd, and the elec-
tronic thermal conductivity, je, of semiconducting armchair graphene nanoribbons (GNRs)
is made for T 6 300 K. Considering the electrons to be scattered by edge roughness, impu-
rities and deformation-potential coupled acoustic phonons and optical phonons, expres-
sions for Sd and je are obtained. Numerical calculations of Sd and je, as functions of
temperature and linear carrier density, bring out the relative importance of the contribut-
ing scattering mechanisms. A GNR of width 5 nm, supporting an electron density
2 · 108 m1, is found to exhibit room temperature values of Sd and je as 42 lV/K and
26.5 W/mK, respectively. A decrease in armchair GNR width, is found to enhance Sd and
reduce je. The effect of varying the electron density is to increase their magnitude when
Fermi energy moves into the second subband. An analysis of thermopower and thermal
conductivity data in clean armchair GNR samples will enable better understanding of the
electron–phonon interaction.
 2012 Elsevier Ltd. All rights reserved.

1. Introduction armchair terminated GNR system can be either metallic or

Graphene has, in the recent past, attracted great interest be-
cause of its unique electronic properties and vast potential
for device applications [1]. In addition to the thinness,
mechanical strength and material flexibility graphene, a sin-
gle layer of carbon atoms arranged in a hexagonal honeycomb
lattice, has some outstanding physical properties such as
high carrier mobility (in excess of 105 cm2/Vs) and high ther-
mal conductivity (up to 5000 W/mK) [2–4]. The patterning of
graphene into graphene nanoribbons (GNRs), which are thin
strips of graphene of large length L and finite(a few nanome-
ters) width, W, confines carriers creating quasi-one-dimen-
sional (Q1D) transport system. The resulting subband
structure and the effective energy gap depend on the width
of the ribbon and the chirality of the ribbon edges. There
are two basic shapes for graphene edges, namely armchair
and zigzag. A zigzag GNR is metallic in nature where as an
semiconducting depending on its width [5]. Although GNR de-
vices are usually fabricated from single graphene sheets exfo-
liated onto thick SiO2/Si substrates [6,7], there have been
recent reports of fabrication of suspended GNRs [8,9]. Lin
et al. [7] have reported experimental observation of subband
formation in supported GNRs, where electrical conductance
becomes quantized due to lateral quantum confinement.
The possibility of tuning the bandgap by controlling almost
precisely the width and shape of graphene ribbons by litho-
graphic and chemical processes [7–9] makes GNRs potentially
useful structures for various nanostructured electronic and
thermal management applications.
The thermoelectric properties, being sensitive to the com-
position and structure of a system and to the external fields,
are important and interesting for study in GNRs. A study of
thermal conductivity is important for understanding thermal
transport in the system. In the presence of a temperature

* Corresponding author.
E-mail address: n_s_sankeshwar@hotmail.com (N.S. Sankeshwar).
0008-6223/$ - see front matter  2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.carbon.2012.09.021
202 CARBON 5 2 ( 2 0 1 3 ) 2 0 1 –2 0 8
gradient in a GNR sample, there will be a heat flow deter-
mined by its thermal conductivity. Besides the contribution
from the lattice, the transport of the electrons due to temper-
ature gradient also causes transport of energy. There are, in
general, two contributions to thermal conductivity of a mate-
rial, namely, lattice thermal conductivity, jp, and electronic
thermal conductivity, je: j = jp + je. When the thermal energy
is transported by the electrons, which also carry charge, there
is, under open circuit conditions, an electric field set up to
counteract the carrier velocity in the direction of the temper-
ature gradient. This is the thermoelectric field. In general,
there are two contributions to thermoelectric power, namely
the diffusion thermopower, Sd, and the phonon drag
thermopower, Sg: S = Sd + Sg. The diffusion contribution to
thermopower, Sd, reflects the energy-dependence of a scatter-
ing mechanism and is determined not only by the magnitude
of the scattering, but also by details concerning the distribu-
tion of the scatterers and their type. Sd and je are therefore,
sensitive probes of the scattering mechanisms operative in
a system [10]. In GNRs electrons are subject to a variety of
scattering interactions, including scattering by edge rough-
ness (ER), phonons and through long–range interactions with
charged impurities (IMPs) on the nanoribbon or more likely at
the supporting insulator substrate.
Although there have been investigations of thermoelectric
properties of graphene [11–14] there exist only a few studies
of thermopower and thermal conductivity of GNRs. The theo-
retical efforts, to understand these properties have been
based on different techniques. Divari and Kliros [15] have
studied thermopower of ballistic wide graphene ribbons with
aspect ratio W/L P 3 using linear response theory and the
Landauer formalism. Xing et al. [16] have studied thermopow-
er of GNRs in zero and non-zero magnetic field using
non-equilibrium Green’s function technique and shown that
thermopower depends on the chirality of the GNR. The ther-
moelectric properties have also been investigated by solving
the electron and phonon transport equations in the non-
equilibrium Green’s function formalism [17–20]. There exist
calculations of lattice thermal conductivity using Boltzmann
transport equations [21–23], molecular dynamics and non-
equilibrium Green’s function formalism [24,25]. These studies
find thermal conductivity to be a strong function of ribbon
width and ER. Seol et al. [26] have reported room temperature
values of 80 lV/K and 780 W/mK, respectively, for thermo-
power and thermal conductivity measured on supported
graphene samples. Liao et al. [27] have investigated high-field
transport in GNRs on SiO2, limited by self heating and ex-
tracted a thermal conductivity of 80 W/mK at room tempera-
ture and estimated electronic thermal conductivity je < 10 W/
mK. Recent studies of thermal conductivity of GNRs indicate
that the roughness of the line edges can degrade the lattice
component jp drastically and, for larger channel lengths, jp
can become even smaller than electronic component je
[18,21,28]. With view to achieve a high thermoelectric figure
of merit, Mazzamoto et al. [29] in their recent study of the
thermoelectric properties of GNRs have proposed optimized
patterning of the ribbons with regard to their width and edge
orientations. It therefore becomes necessary to undertake a
detailed study of the thermoelectric properties of GNRs.
Among the different formalisms employed, the semi-classical
Boltzmann approach is known to be appropriate for struc-
tures in which the potentials vary slowly on both the spatial
scale of the electron thermal wavelength and the temporal
scale of the scattering processes [30]. Further, it is known to
describe well the diffusive carrier transport in graphene sys-
tems, away from the Dirac point [31]. In this work, employing
Boltzmann formalism, we present a systematic study of the
electronic contribution to thermal conductivity, je, and the
diffusion thermopower, Sd, of semiconducting armchair GNR
systems. The purpose of the paper is first, to study for the first
time the influence of subband structure and of the ribbon
width of armchair GNRs on Sd and je, and second, to illustrate
the relative importance of electron scattering mechanisms on
these properties. In Section 2 the relevant theory of thermo-
power and thermal conductivity is presented. The results of
calculations are presented and discussed in Section 3.
2. Theory
2.1. Electron eigenvalues and wavefunctions
We consider a semiconducting armchair GNR of width W and
length L along x- and y-directions respectively, with electrons
confined to move in the plane of the ribbon. The electron
wave functions and energy Eigen values, derived from admix-
ing the states in the K and K 0 valleys and satisfying hard wall
boundary conditions, are expressed as [32–34]:
pffiffiffi
Wn;ky ðrÞ ¼ ½wn;ky ðrÞ  w0n;ky ðrÞ= 2 ð1Þ
with
 
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ik y i½ðDK=2Þk x 1
wn;ky ðrÞ ¼ 1=2LWe y e n
;
eihn;ky
 
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ik y i½ðDK=2Þk x 1
w0n;ky ðrÞ ¼ 1=2LWe y e n
eihn;ky
and
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
En;ky ¼ hvF
2
kn þ ky :
2
ð2Þ
Here r = (x, y), is the two-dimensional electron position vector,
hn;ky ¼ tan1 ðky =kn Þ, DK = 4p/3a, a being the lattice constant, mF
is the Fermi velocity, ky is electron wave vector along y direc-
tion and kn = ± np/3W = (En/hvF) is the allowed transverse
wave vector with the subband index n = 1, 2, 4, 5, 7, 8. . . The
resulting band gap is Eg = 2phmF/3W and the density of states
for an armchair GNR is given by [32],
hvF Þ½En;ky =fE2n;ky  ðEn Þ2 g1=2 
qGNR ðEn;ky Þ ¼ ð2=p ð3Þ
It may be noted that the density of states as given by (3)
shows singular behavior, whenever the energy passes
through the edge En of an armchair GNR subband.
2.2. Thermopower and electronic thermal conductivity
Thermopower, S, and thermal conductivity, j, which are in
general tensors, are defined by the relations [10]
E ¼ S$T ð4Þ
and
 CARBON 5 2 (2 0 1 3) 2 0 1–20 8 203

2 3
U ¼ jð$TÞ ð5Þ 2
p2 kB T 6 1 7
Sd ¼ 4p þ  2 5 ð14Þ
under open circuit conditions. Here U is the heat current 3eEF EF
En
1
density and E is the effective electric field produced by the
temperature gradient $T applied along the plane of the where the first term reflects the scattering mechanisms. The
GNR. We evaluate the electric current density, J, and heat cur- parameter p can also be expressed in the form
rent density, U, by solving Boltzmann transport equation in " #
the relaxation time approximation. The electron distribution EF dsðEn;ky Þ
p¼ ; ð15Þ
function f(En;ky ) can be written as sðEF Þ dEn;ky
En;ky ¼EF

0
f ðEn;ky Þ ¼ f ðEn;ky Þ þ sðEn;ky Þvg which brings out the feature that Sd is determined by not only
    ! the magnitude of scattering but also by the energy depen-
En;ky  EF @f 0 ðEn;ky Þ
 eE  rT  ð6Þ dence of s(En;ky ) at En;ky ¼ EF . It may be noted that provided
T @En;ky
the energy dependence of s(En;ky ) does not vary with tempera-
where vg = (2/p⁄vF)/qGNR (En;ky ) is the electron group velocity, ture (though its magnitude will), Sd should be a linear func-
s(En;ky ) is the electron momentum relaxation time, f0(En;ky ) is tion of temperature.
the Fermi–Dirac distribution function and EF is the Fermi en-
ergy which is determined by the one dimensional carrier 2.2.1. Momentum relaxation rates
density For the range of temperatures considered (T 6 300 K), the
Z scattering mechanisms influencing carrier transport in a de-
nl ¼ qGNR ðEn;ky Þf 0 ðEn;ky ÞdEn;ky : ð7Þ fect free armchair GNR include scattering by ribbon ER, IMPs
and deformation-potential coupled acoustic phonons (APs)
The electric current density, J, and heat current density, U, and optical phonons (OPs). Using Eqs. (1)–(3) we obtain the fol-
can be evaluated from the expressions: lowing expression for the momentum relaxation rate:
X XX
J ¼ ð2=LWtÞ evg f ðEn;ky Þ ð8Þ s1 2p
jcj ðqy Þj2 ð1 þ cosðhn;ky  hn0 ;k0y ÞÞð1  cosðaÞÞ
j ðEn;ky Þ ¼ 
h
n;ky
n0 ;k0y qy

1f ðEn0 ;k0 Þ

and  1f ðEn;k yÞ DðEn;ky ; En0 ;k0y Þdqy ;ky k0y
X y

U ¼ ð2=LWtÞ ðEn;ky  EF Þvg f ðEn;ky Þ: ð9Þ ð16Þ

n;ky

where a is the angle between initial and final states, and

Here t is layer thickness of GNR. Using Eqs. (4)–(9), we obtain
|cj(qy)|2, with j  AP, OP, ER, IMP are the matrix elements for
expressions for the diffusion contribution to thermopower, Sd,
interaction of electrons with deformation-potential coupled
and for electronic thermal conductivity, je as
APs and OPs, ER and IMPs, respectively. We obtain
1 1
Sd ¼ K K21 ð10Þ D2AP hjqy j
eT 11 jcAP ðqy Þj2 ¼ jFn;n0 j2 ; ð17Þ
qLW3 vs
and
1 D2OP h
je ¼ ½K31  ððK21 Þ2 =K11 Þ; ð11Þ jcOP ðqy Þj2 ¼ jFn;n0 j2 ; ð18Þ
T qLW3 xLO
respectively, where the transport coefficients Krs, are defined as and
Z !
2 X 1
@f 0 ðEn;ky Þ 2E2n0 H2 K
Krs ¼ vg ðEn;ky  EF Þr1 ss ðEn;ky Þ  dEn;ky : jcER ðqy Þj2 ¼ jFn;n0 j2 ð19Þ
Wtp
h n @En;ky L W4 1 þ ðqy KÞ2
En

ð12Þ h i
sinðkn0 kn ÞðW=2Þ sinðDKkn0 kn ÞðW=2Þ
with jFn;n0 j2 ¼ j kn0 kn
 DKkn0 kn
j2 .
Eqs. (10) and (11) can be used to evaluate Sd and je taking Here, DAP and DOP are respectively the deformation-poten-
into account the relevant scattering mechanism(s). The over- tial constants for acoustic and optical phonons, q is mass
all contributions to Sd and je can be evaluated assuming the density of graphene, xac = vsqy is the frequency of APs of wave
overall relaxation time s to be given by Matthiessen’s rule: 0
vector qy ¼ jky  ky j, vs is the sound velocity in graphene and
s1 ¼ Ri s1
i , where the sum is over all the relevant scattering hxLO is OP energy. The factor D(En;ky , En0 ;k0y ) in Eq. (16) repre-
mechanisms, i. sents ðNqy þ 1=2  1=2dðEn;ky  En0 ;k0y hxqy Þ for the electron–
Often in literature, limiting forms of Sd are used in the phonon scatterings, Nqy being the Bose–Einstein distribution
analysis of data. In the degenerate limit a good approximation function and dðEn;ky  En0 ;k0y Þ for ER and IMP scatterings. In Eq.
to Eq. (10) is the well-known Mott expression [10]: (19) ER is described by two parameters H, the amplitude and
" # K, the correlation length of the roughness.
2
p2 kB T d ln rðEn;ky Þ In the case of IMP scattering, the relaxation rate given by Eq.
Sd ¼ ð13Þ
3e dEn;ky (16) takes into account an appropriate average over their distri-
En;ky ¼EF

bution, over the layer and/or the substrate. We assume that the
where r(En;ky ), is the energy dependent conductivity. If the en- carriers in the armchair GNRs, located at (x, y, 0) are scattered by
ergy dependence of relaxation time is taken as Ep, Eq. (13) can the unscreened coulomb potential of an ionized impurity charge
be more simply expressed as located at a distance d = [(x  x0)2 + (y  y0)2 + (z  z0)2]1/2 away
204 CARBON 5 2 ( 2 0 1 3 ) 2 0 1 –2 0 8

from the center of the ribbon and that the distribution NI(x0,
y0, z0) covers the IMPs in the bulk of the sample, that is, the
10
4 W = 5 nm (a)
10
3 T = 300 K
2 LI
GNR layer and the substrate. The momentum relaxation rate 10
2

τ-1 (ps-1)
1
due to the bulk impurities (BIs) can be written as 10
1

Z 0
10 ER
s1
BI ðEn;ky Þ ¼ dx0 dy0 dzo NI ðx0 ; y0 ; z0 Þs1
IMP ðEn;ky Þ ð20Þ 10
-1
BI AP
-2
10 OP
with the associated matrix element expressed as 10
-3

 2 2 Z W=2 10
4
2
jcIMP ðqy Þj2 ¼ 2 2
e
½cosðkn0  kn Þx 10
3
W = 3 nm (b)
L W 4pe0 j W=2 10
2
ER 2 1 T = 300 K
2 1

τ-1 (ps-1)
10
 cosðDK  kn0  kn ÞxK0 ðjqy jdÞdx ð21Þ 0 AP
10
-1 LI
10
where K0(x) is the modified Bessel function of the second kind -2 OP
10
of order zero. Here NI(x0, y0, z0), is volume density of the IMPs, 10
-3 BI
-4
assumed to be distributed uniformly and which can vary with 10
distance from the GNR plane. The general formalism given by 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1
Eqs. (20) and (21) allows us to evaluate at the same time, the E/EF
effect of scattering by volume-distributed BIs as well as layer
impurities (LIs). To study the influence of scattering of elec- Fig. 1 – Energy dependence of relaxation rates for armchair
trons by only LIs, that are located in the GNR layer or at the GNRs of widths (a) 5 nm, and (b) 3 nm, with nl = 5 · 108 m1
layer/substrate interface alone, one may obtain their inde- at T = 300 K. Dash-double-dotted, dash-dotted, short-dash-
pendent contribution s1 dotted, dashed and dotted curves, respectively, represent
LI ðEn;ky Þ by taking NI ! n2D d(z), where
n2D is the LI concentration. For the evaluation of the relaxa- contributions from ER, LI, BI, AP and OP scatterings. Curves
tion rates, we consider BIs and LIs separately to highlight 1 and 2, respectively denote overall relaxation rates s1
1 and

their relative importance [32]. s1

In the quantum limit, considering scattering to be re-
stricted to the lowest subband (n = 1), and neglecting interval-
ley scattering, the expressions for the relaxation rates given
above can be shown to reduce to those given by Fang et al.
[32].
3. Results and discussion
We have performed numerical calculations of diffusion ther-
mopower, Sd, and of electronic thermal conductivity, je, of
armchair GNR using Eqs. (10), (11), (12), and (16) for parame-
ters characteristic of graphene [32]: DAP = 16 eV, D OP =
1.4 · 1011 eV/m,  hxLO = 160 meV, ts = 2 · 104 m/s, vF = 1 · 106 m/s,
q = 7.6 · 107 kg/m2. The other parameters chosen are typi-
cally used in the analyses of transport in GNRs: W = 5 nm,
H = 0.5 nm, K = 3 nm and n2D = 1014 m2 [32]. To bring out the
effect of subband structure on Sd and je in a simple way, we
choose to present our calculations neglecting intervalley
and intersubband scatterings and assuming the AP scattering
to be quasi-elastic [32]. The range of electron concentrations
(nl = 1–7 · 108 m1) considered ensures that Fermi energy EF
can move into the second subband (n = 2). For a 5 nm wide
armchair GNR, a value nl = 5 · 108 m1 gives EF = 346 meV,
the energies of the first two subband minima being
E1 = 138 meV and E2 = 276 meV.
In order to understand the relative importance of various
scattering mechanisms considered operative in the armchair
GNR, we first study the energy dependence of relaxation rates
s1 ðEn;ky Þ. Fig. 1a shows the energy dependences of the individ-
ual momentum relaxation-rates calculated using Eqs. (16)–
(21) for an armchair GNR of width W = 5 nm and
nl = 5 · 108 m1. The dash-double-dotted, dash-dotted,
short-dash-dotted, dashed and dotted curves, respectively,
represent the rates due to ER, LI, BI, AP and OP scatterings.
2 , given in the text.
Curve 1 denotes overall relaxation rate calculated assuming
Matthiessen’s rule: s1 1 1 1 1
1 ðEn;ky Þ ¼ sER þ sLI þ sAP þ sOP . Here
1
sLI ðEn;ky Þ represents scattering only by LIs. For the range of
energies considered and the parameters used the following
features are noticed. The relaxation rates decrease with en-
ergy from the band edge E1 until the next subband
(E2 = 0.884EF) is reached where they show a sudden rise. This
is a result of the nature of the density of states; at each new
subband minimum there is a large density of states for elec-
trons to scatter to. When the energy of a state is
E1 + hxLO(= 0.966EF), the rates again show a sudden rise. We
find that the dominant contribution to the overall rate
s1
1 ðEn;ky Þ (Curve 1) is from LI scattering, the contribution from
AP scattering being less. Dominance of LI scattering in room
temperature mobility has been reported for n2D = 1016 m2
[32]. In order to estimate the influence of IMPs in the sub-
strate, we have also performed calculations of the overall rate
s1 1 1 1 1 1
2 ðEn;ky Þ ¼ sER þ sBI þ sAP þ sOP , where the rate sBI ðEn;ky Þ, calcu-
21 3
lated assuming NI = 10 m , corresponds to IMPs distributed
uniformly not only in the layer but also in the substrate of
thickness ts taken to be 300 nm [7]. Curve 2 in Fig. 1a shows
the results. We find that the dominant contribution to overall
rate s1
2 ðEn;ky Þ is now from ER scattering and the contribution
from BIs is relatively weak compared to that from ER and
LIs, indicating that the contribution from the IMPs in the sub-
strate is small [32].
With a view to study the dependence of relaxation rates on
the GNR width, we have also performed calculations of relax-
ation rates for an armchair GNR of width 3 nm, for which the
energies of the first two subband minima for nl = 5 · 108 m1
are E1 = 230 meV and E2 = 460 meV with EF = 470 meV. Fig. 1b
shows the results. The rates are found to exhibit the same
features as seen for W = 5 nm (Fig. 1a), however, with a

corresponding shift in the positions of the two peaks. In this
case, it may be noted that the onset of OP emission (at
E1 + 
hxLO = 0.829EF) occurs before the occupation of the
second subband (at 0.977EF). Curve 2 represents the overall
rate s1
2 ðEn;ky Þ. This changed energy dependence of the rates
is expected to influence the behavior of and lead to distinctive
features in Sd and je.
Fig. 2 shows the temperature variation of diffusion ther-
mopower, Sd, for an armchair GNR of width 5 nm with linear
carrier density nl = 2 · 108 m1, for which value the carriers
are confined to the lowest subband only. The dash-double-
dotted, dash-dotted, dashed and dotted curves, respectively,
represent individual contributions from ER, LI, AP and OP
scattering mechanisms. For the range of temperatures and
scattering mechanisms considered, we find that the overall
Sd (curve 1) increases almost linearly with temperature up
to T  75 K, the dominant contribution being from LI scatter-
ing. At higher temperatures, the contribution due to OP scat-
tering becomes increasingly important and Sd attains a value
42 lV/K at T = 300 K. Further, a change in sign of Sd is observed
in the case of individual contributions from APs (at lower
temperatures) and from OPs (at higher temperatures). The
dominant contribution to Sd from IMPs is from LIs. Our calcu-
lations of Sd including the contribution from IMPs in the sub-
strate show that Sd (not depicted in Fig. 2) is underestimated
by 12% at 100 K and, due to increasing importance of OPs,
by 30% at 300 K. In the absence of experimental data on
thermopower of armchair GNRs, it may be noted that the
room temperature value of Seebeck coefficient measured for
supported single layer graphene of widths W  1.5–3.2 lm is
S  80 lV/K [12,26]. It may be mentioned here that the pho-
100 100
W = 5 nm
80
80
a
60
-Sd (μV/K)
40 b
ER
60 20
- Sd (μV/K)
0
LI
50 100 150 200 250 300
T (K)
40
1 1
20 AP
0 OP
50 100 150 200
T (K)
Fig. 2 – Variation of Sd with temperature for an armchair GNR
of width 5 nm for nl = 2 · 108 m1. Dash-double-dotted,
dash-dotted, dashed and dotted curves, respectively,
represent contributions from ER, LI, AP and OP scatterings.
Curve 1 denotes overall contribution to Sd calculated with
overall relaxation rate s1
1 given in the text. The inset shows
overall Sd calculated using s11 for two armchair GNR widths:
3 nm (Curve a) and 5 nm (Curve b). Dashed curves show the
overall Sd according to Mott expression.
CARBON 5 2 (2 0 1 3) 2 0 1–20 8 205
non-drag contribution, Sg, to thermopower is expected to be
dominant at lower temperatures and calculations, in the
boundary scattering limit, show Sg < 500 lV/K for an arm-
chair GNR of width 5 nm for T < 50 K [35]. Our results are also
in consonance with the results of Zheng et al. [20] who have
reported calculations of Seebeck coefficient using non-
equilibrium Green’s function techniques for armchair GNRs
of varying smaller widths (0.4–0.8 nm). The inset in Fig. 2
shows the temperature dependence of overall Sd calculated
for two values of armchair GNR widths and linear carrier den-
sity nl = 2 · 108 m1. Curves a and b show the behavior of Sd
for W = 3 and 5 nm, respectively. We find that for W = 3 nm,
not only is the magnitude of Sd enhanced but also its behavior
is altered due to the changed temperature and energy depen-
dences of the individual contributions. It may be mentioned
that ER scattering becomes more important with decreasing
width. We have also shown as dashed curves, the result of
our calculations using Mott expression (14), with p given by
equation (15). At lower temperatures Sd is found to follow
the Mott relation with p = 1.66 for W = 5 nm and p = 3.82 for
W = 3 nm, where as for T > 75 K, Mott expression overesti-
mates Sd.With a view to illustrate the influence of subband
structure on Sd we have performed calculations of Sd as a
function of linear carrier density, nl. Fig. 3 shows the variation
of Sd for an armchair GNR of width 5 nm at T = 300 K. The
dash-double-dotted, dash-dotted, dashed and dotted curves,
respectively, represent individual contributions to overall Sd
(curve 1) due to scattering from ER, LIs, APs and OPs. For the
parameters considered, we observe that, the overall contribu-
tion at first decreases with increase in nl, shows a step and
then increases with nl. The step like behavior seen around
150 T = 300 K
200 W = 5 nm
T = 300 K
100
-Sd (μV/K)
150
50
a
100
- Sd (μV/K)
0 b
c
ER
1 2 3 4 5 6
LI 8 -1
nl (x10 m )
50
AP
0
OP
-50
1 2 3 4 5 6
8 -1
250 300 nl (x10 m )
Fig. 3 – Variation of Sd with linear carrier density, nl, for
armchair GNR of width 5 nm at T = 300 K. The dash- double-
dotted, dash-dotted, dashed and dotted curves, respectively,
represent contributions from ER, LI, AP and OP scatterings.
Curve 1 denotes overall contribution. The inset shows
overall Sd for an armchair GNR of width 5 nm for three
values of LI concentrations, n2D: 1014 m2 (curve a), 1013 m2
(curve b), and 1012 m2 (curve c). Dashed curve shows the
overall Sd for armchair GNR of width W = 3 nm, with
n2D = 1014 m2.
206 CARBON
nl
3·108 m1 (EF = 278 meV) is a reflection of the singular
nature of the density of states and corresponds to the second
subband occupation. We find that the major contribution to
overall Sd is due to LI scattering with OP scattering becoming
dominant around nl
2–3 · 108 m1 corresponding to carrier determined by the LI and ER contributions and for higher
energy (E1 + hxLO), and around this value OP contribution
is seen to show a change in sign. With proper control of the
parameters governing extrinsic scattering mechanisms, one
could therefore tune the behavior of Sd. The influence of the
OP scattering in determining Sd is found to be appreciable
especially in purer samples. The inset in Fig. 3 brings out
the relative importance of the OP and LI scattering mecha-
nisms in an armchair GNR of width 5 nm. Curves a, b and c
depict respectively, the behavior of overall Sd calculated for
three different values of LI concentration, n2D: 1014 m2,
1013 m2 and 1012 m2, the other parameters remaining the
same. We find that for a decrease in n2D from 1014 m2 by
one order, |Sd| decreases from a value 5.56 lV/K to
13.73 lV/K near the first subband minimum showing a
change in sign due to the dominance of OP interaction. A de-
crease in the value of n2D below 1012 m2 is not found to influ-
ence much the behavior of Sd. The dashed curve in the inset
in Fig. 3 shows the behavior of overall Sd for W = 3 nm and
n2D = 1014 m2. A comparison of the dashed curve with curve
a (corresponding to W = 5 nm) depicts the effect of ribbon
width on Sd. We find that, one, although the features of Sd
are similar, the step, corresponding to the occupation of the
3 · 108 m1, corresponding to the occupation of the
second subband, now occurs around nl
4.7 · 108 m1, and second subband. For nl < 4.5 · 108 m1, we find that je is deter-
second, the magnitude of Sd is larger for nl < 2.9 · 108 m1
and for nl > 6 · 108 m1.We now present our calculations of
je. Fig. 4 shows the variation of je as a function of temperature
for an armchair GNR of width 5 nm for nl = 2 · 108 m1. The
dash-double-dotted, dash-dotted, dashed and dotted curves,
3 · 108 m1; however, its contribution to
respectively, represent the individual contributions to je due
5
10
OP
4 4
10
3
10
κe (W/mK)
ER
2 AP
10
LI
1 10
1
10 2
0
10
-1
10
50 100 150 200 250
T (K)
Fig. 4 – Temperature dependence of electronic thermal
conductivity, je, of armchair GNR of widths 5 nm (curve 1)
and 3 nm (curve 2) for nl = 2 · 108 m1. Dash-double-dotted,
dash-dotted, dashed and dotted curves, respectively,
represent individual contributions from ER, LI, AP and OP
scatterings for an armchair GNR with W = 5 nm.
5 2 ( 2 0 1 3 ) 2 0 1 –2 0 8
to scattering from ER, LIs, APs and OPs. Curve 1 represents
the overall contribution to je. With increase in temperature
the overall je is found to increase at first and then almost sat-
urate at higher temperatures. We find that for T < 125 K, je is
temperatures AP scattering becomes important; the contribu-
tion from OP scattering is negligible. Curve 2 depicts the
behavior of overall contribution to je for an armchair GNR
of width 3 nm. From curves 1 and 2 it is seen that although
the behavior is similar, the magnitude of je is lesser for the
armchair GNR with W = 3 nm. An inclusion of the contribu-
tion from IMPs in the substrate is found to result in an over-
estimation of je by nearly 45% at T = 300 K (not show in
Fig. 4). This is due to the changed relative importance of scat-
tering mechanisms with the influence of impurities in the
substrate being less, and ER scattering being dominant for
T < 80 K and AP scattering for T > 80 K. Our calculated values
of je at room temperature are found to be comparable to
those extracted from experimental data of thermal conduc-
tivity of GNRs and estimated with Wiedemann–Franz law
[27].Fig. 5 shows the concentration dependence of je for an
armchair GNR of width 5 nm at T = 300 K. The dash-double-
dotted, dash-dotted, dashed and dotted curves, respectively,
represent the individual contributions to je from ER, LI, AP
and OP scatterings. Curve 1 represents the overall contribu-
tion. It is seen that overall je increases with nl, with a step
at nl
mined mainly by LIs and for higher values of concentration,
AP scattering becomes important. The contributions from
ER scattering to overall je is about 10%. The individual contri-
bution to je due to OPs is found to increase with nl with a step-
like increase at nl
overall je is negligible. Curve 2 depicts the behavior of overall
contribution to je for an armchair GNR of width 3 nm. We find
OP T = 300 K
10
3
10
κe (W/mK)
ER
2
AP
1
LI
2
1
10
0
10
300 1 2 3 4 5 6
8 -1
nl (x10 m )
Fig. 5 – Variation of electronic thermal conductivity, je, with
linear carrier density, nl, for armchair GNRs of widths 5 nm
(curve 1) and 3 nm(curve 2) at T = 300 K. Dash-double-
dotted, dash-dotted, dashed and dotted curves, respectively,
represent individual contributions from ER, LI, AP and OP
scatterings for an armchair GNR with width 5 nm.
 CARBON 5 2 (2 0 1 3) 2 0 1–20 8
that magnitude of je is reduced mainly due to increased ER
scattering [21,32].
It may not be out of context if we compare the values of je
obtained with those reported in literature in respect of jp, the
lattice contribution. Aksamija and Knezevic [21] have calcu-
lated jp for armchair GNRs of varying ER parameters and
widths. For a GNR of width 5 nm and rms roughness param-
eter 0.5 nm, they find jp
800 W/mK, at room temperature. Acknowledgments
For a GNR of similar dimension and values of parameters,
we obtain je
90 W/mK for nl = 7 · 107 m1; this implies This work is supported by UGC (India). One of the authors
je  0.12jp. However, since je can become significant in doped
materials [4], it would be of interest to estimate the ranges of
nl, W and T that could obtain values of je comparable with
that of jp. In this regard, we note the following features of
the operative scattering mechanisms determining je. First,
over a wide range of values of nl, W, T and low n2D, the behav-
ior and magnitude of overall je is determined by the intrinsic
acoustic and optical phonon-scattering mechanisms, and
second, for a given value of nl, the magnitude of je is influ-
enced by ER for smaller widths (W < 15 nm) and OP scatter-
ing becomes significant for higher values of W (W > 30 nm)
and T (>100 K). Keeping in view the dominance of the various
scattering mechanisms in the different ranges of nl, W and T,
we find, for instance, that for a pure armchair GNR of width
0.6 lm and nl = 20 · 108 m1 je may attain a value of
1000 W/mK at T = 200 K. In the case of graphene ribbons of
larger dimensions (lm) and rough edges with specularity
parameter 0.9, Nika and co-workers [22] obtain a value of
jp  500–6000 W/mK.
Our calculations open up interesting perspectives. With
the possibility of modulating the Fermi level and the number
of Q1D conduction modes contributing to the transport in
GNRs with a control on gate and bias voltages [7], it is to be
expected that the subband structure in Sd and je could be
detectable. With proper choice of parameters characterizing
the extrinsic scattering mechanisms, our calculations indi-
cate that it could be possible to tune the behaviors of overall
Sd and je, and better understand the scattering mechanisms
operative in armchair GNRs. We find that decreasing the rib-
bon width results in an enhancement of Sd and a reduction
in je, a feature promising applications in thermoelectric de-
vices. Further, decreasing the LI concentration is found to
bring out the influence of OPs which is seen to influence a
change in sign of Sd; in purer samples, therefore, an investiga-
tion of drag component, independent of diffusion component
of thermopower could be possible.
4. Conclusions
In conclusion, we have for the first time presented numerical
calculations of diffusion contribution to thermopower, Sd, and
electronic thermal conductivity, je, of defect free semicon-
ducting armchair GNRs. For a GNR of width 5 nm and
nl = 2 · 108 m1, the room temperature values of Sd and je
are found to be 42 lV/K and 26.5 W/mK, respectively. The
behaviors of Sd and je, as functions of temperature and linear
carrier density are studied. The relative importance of carrier
scattering mechanisms operative, the effect of ribbon width
and of subband structure on the behaviors are also studied.
207
In the present analysis we have ignored screening of scatter-
ing potential by electrons and confinement of phonons in the
armchair GNR which are expected to influence the properties
studied. An analysis of thermopower and thermal conductiv-
ity data in clean GNR samples will enable better understand-
ing of the intrinsic electron phonon interaction.
(ASN) holds an UGC-RFSMS award.
R E F E R E N C E S
[1] Castro Neto AH, Guinea F, Peres NMR, Novoselov KS, Geim
AK. The electronic properties of graphene. Rev Mod Phys
2009;81(1):109–62.
[2] Geim AK, Novoselov KS. The rise of graphene. Nat Mat
2007;6(3):183–91.
[3] Balandin AA, Ghosh S, Bao W, Calizo I, Teweldebrhan D, Miao
F, et al. Superior thermal conductivity of single-layer
graphene. Nano Lett 2008;8(3):902–7.
[4] Balandin AA. Thermal properties of graphene and
nanostructured carbon materials. Nat Mat 2011;10(8):569–81.
[5] Palacios JJ, Fernandez-Rossier J, Brey L, Fertig HA. Electronic
and magnetic structure of graphene nanoribbons. Semicond
Sci Technol 2010;25(3):033003-1–033003-10.
[6] Han MY, Ozyilmaz B, Zhang Y, Kim P. Energy band-gap
engineering of graphene nanoribbons. Phys Rev Lett
2007;98(20):206805-1–4.
[7] Lin Y, Perebeinos V, Chen Z, Avouris P. Electrical observation
of subband formation in graphene nanoribbons. Phys Rev B
2008;78(16):161409(R)-1–4.
[8] Li X, Wang X, Zhang L, Lee S, Dai H. Chemically derived,
ultrasmooth graphene nanoribbon semiconductors. Science
2009;319(5867):1229–32.
[9] Lin MW, Ling C, Agapito LA, Kioussis N, Zhang Y, Cheng MMC,
et al. Approaching the intrinsic band gap in suspended high-
mobility graphene nanoribbons. Phys Rev B
2011;84(12):125411-1–7.
[10] Gallagher BL, Butcher PN. Classical transport and
thermoelectric effects in low dimensional and mesoscopic
semiconductor structures. In: Landsberg P, editor. Handbook
on semiconductors, vol. 1. Amsterdam: Elsevier; 1992. p.
721–816.
[11] Nika DL, Balandin AA. Two-dimensional phonon transport in
graphene. J Phys Condens Matter 2012;24(23):233203-
1–233203-18.
[12] Zuev YM, Chang W, Kim P. Thermoelectric and
magnetothermoelectric transport measurements of
graphene. Phys Rev Lett 2009;102(9):096807-1–4.
[13] Vaidya RG, Kamatagi MD, Sankeshwar NS, Mulimani BG.
Diffusion thermopower in graphene. Semicond Sci Technol
2010;25(9):092001-1–5.
[14] Bao WS, Liu SY, Lie XL. Thermoelectric power in graphene. J
Phys Condens Matter 2010;22(31):315502-1–6.
[15] Divari PC, Kliros GS. Modeling the thermopower of ballistic
graphene ribbons. Physica E 2010;42(9):2431–5.
[16] Xing Y, Sun Q, Wang J. Nernst and Seebeck effects in a
graphene nanoribbon. Phys Rev B 2009;80(23):235411-1–8.
[17] Ouyang Y, Guo J. A theoretical study on thermoelectric
properties of graphene nanoribbons. Appl Phys Lett
2009;94(26):263107-1–3.
208 CARBON 5 2 ( 2 0 1 3 ) 2 0 1 –2 0 8
[18] Karamitaheri H, Neophytou N, Pourfath M, Faez R, Kosina H.
Engineering enhanced thermoelectric properties in zigzag
graphene nanoribbons. J Appl Phys 2012;111(5):054501-1–9.
[19] Sevincli H, Cuniberti G. Enhanced thermoelectric figure of
merit in edge-disordered zigzag graphene nanoribbons. Phys
Rev B 2010;81(11):113401-1–4.
[20] Zheng H, Liu HJ, Tan XJ, Lv HY, Pan L, Shi J, et al. Enhanced
thermoelectric performance of graphene nanoribbons. Appl
Phys Lett 2012;100(9):093104-1–5.
[21] Aksamija Z, Knezevic I. Lattice thermal conductivity of
graphene nanoribbons: anisotropy and edge roughness
scattering. Appl Phys Lett 2011;98(14):141919-1–3.
[22] Nika DL, Askerov AS, Balandin AA. Anomalous size
dependence of the thermal conductivity of graphene ribbons.
Nano Lett 2012;12(6):3238–44.
[23] Nika DL, Pokatilov EP, Askerov AS, Balandin AA. Phonon thermal
conduction in graphene: role of umklapp and edge roughness
scattering. Phy Rev B 2009;79(15):155413-1–155413-12.
[24] Haskins J, Kinaci A, Sevik C, Sevincli H, Cuniberti G, Cagin T.
Control of thermal and electronic transport in defect-
engineered graphene nanoribbons. ACS Nano
2011;5(5):3779–87.
[25] Xu Y, Chen X, Gu B, Duan W. Intrinsic anisotropy of thermal
conductance in graphene nanoribbons. Appl Phys Lett
2009;95(23):233116-1–3.
[26] Seol JH, Jo I, Moore AL, Lindsay L, Aitken ZH, Pettes MT, et al.
Two-dimensional phonon transport in supported graphene.
Science 2010;328(5975):213–6.
[27] Liao AD, Wu JZ, Wang X, Tahy K, Jena D, Dai H, et al.
Thermally limited current carrying ability of graphene
nanoribbons. Phys Rev Lett 2011;106(25):256801-1–4.
[28] Evans W, Hu L, Keblinski P. Thermal conductivity of graphene
ribbons from equilibrium molecular dynamics: effect of
ribbon width, edge roughness, and hydrogen termination.
Appl Phys Lett 2010;96(20):203112-1–3.
[29] Mazzamuto F, Hung-Nguyen V, Apertet Y, Caer C, Chassat C,
Saint-Martin J, et al. Enhanced thermoelectric properties in
graphene nanoribbons by resonant tunneling of electrons.
Phys Rev B 2011;83(23):235426-1–7.
[30] Ferry DK, Goodnick SM, Bird J. Transport in nanostructures.
2nd ed. Cambridge: New York; 2009.
[31] Das Sarma S, Adam S, Hwang EH, Rossi E. Electronic
transport in two-dimensional graphene. Rev Mod Phys
2011;83(2):407–70.
[32] Fang T, Konar A, Xing H, Jena D. Mobility in
semiconducting graphene nanoribbons: phonon, impurity,
and edge roughness scattering. Phys Rev B
2008;78(20):205403-1–8.
[33] Trauzettel B, Bulaev DV, Loss D, Burkard G. Spin qubits in
graphene quantum dots. Nat Phys 2007;3(3):192–6.
[34] Brey L, Fertig HA. Electronic states of graphene nanoribbons
studied with the Dirac equation. Phys Rev B
2006;73(23):235411-1–5.
[35] Bhargavi KS, Kubakaddi SS. Phonon-drag thermopower in an
armchair graphene nanoribbon. J Phys Condens Matter
2011;23(27):275303-1–5.