Università di Pisa
!"# $%&'%(# !"# )*++*,,&+-(# !"#$%&'"()* (+* ,-&./0)0* 1&)(-"22()* 3"0%456++07'* 8-&)9"9'(-9ƞ(# )...# ./-,0'&+#
1-2%,-#3-00-'4(#!"(#5(#66"789:78;#<;997="##
#
Simulation of Graphene Nanoribbon
Field-Effect Transistors
ƞ
#$%&'()%*+$,-$*
1%6*'0%>-+0&#?%#)+@-@+-'%*#?-//A)+B&'>*C%&+-D#./-00'&+%,*(#)+B&'>*0%,*(#E-/-,&>F+%,*C%&+%(#
G+%2-'4%0H#?%#I%4*#
#$(./00/*1%&&%)),&/*
1%6*'0%>-+0&#?%#)+@-@+-'%*#?-//A)+B&'>*C%&+-D#./-00'&+%,*(#)+B&'>*0%,*(#E-/-,&>F+%,*C%&+%(#
G+%2-'4%0H#?%#I%4*#
760 IEEE ELECTRON DEVICE LETTERS, VOL. 28, NO. 8, AUGUST 2007
Simulation of Graphene Nanoribbon
Field-Effect Transistors
Gianluca Fiori and Giuseppe Iannaccone
Abstract—We present an atomistic 3-D simulation of graphene
nanoribbon field-effect transistors (GNR-FETs), based on the self-
consistent solution of the 3-D Poisson and Schrödinger equations
with open boundary conditions within the nonequilibrium Green’s
function formalism and a tight- binding Hamiltonian. With respect
to carbon nanotube FETs, GNR-FETs exhibit comparable perfor-
mance, reduced sensitivity to the variability of channel chirality,
and similar leakage problems due to band-to-band tunneling.
Acceptable transistor performance requires prohibitive effective
nanoribbon width of 1–2 nm and atomistic precision that could
in principle be obtained with periodic etch patterns or stress
patterns.
Index Terms—Atomistic tight-binding Hamiltonian, graphene,
nanoribbon, nonequilibrium Green’s function formalism (NEGF),
3-D Poisson.
I. INTRODUCTION
I
NTHElastdecade,carbonnanotubes(CNT)haveattracted
extraordinary interest for their extremely interesting phys-
ical and electrical properties [1] and their potential as an
alternative to silicon as channel material for transistors be-
yond complimentary metal–oxide–semiconductor technology
[2]. Recent experiments by Novoselov et al. [3] demonstrated
the possibility of fabricating stable single atomic layer graphene
sheets, with remarkable electrical properties, that have brought
new excitation to the field of carbon electronics.
Two-dimensional graphene is a zero-gap material, which
makes it not suitable for transistor applications.
Energy gap can however be induced by means of lateral
confinement [4], realized for example by lithography defini-
tion of narrow graphene stripes, i.e., the so-called graphene
nanoribbons.
Experiments on graphene-based devices [5] and graphene
nanoribbon field-effect transistors [6] (GNR-FETs) have ap-
peared only very recently and demonstrate limited capability to
modulate the conductance of a graphene channel at room tem-
perature. The main problem is the need to fabricate extremely
narrow nanowires (on the order of 1 nm) with atomic preci-
sion to obtain an energy gap adequate for room temperature
operation.
Manuscript received April 19, 2007; revised May 30, 2007. This work
was supported by the European Science Foundation EUROCORES Pro-
gramme Fundamentals of Nanoelectronics, through funding from the Consiglio
Nazionale delle Ricerche (awarded to IEEIIT-PISA) and the European Com-
mission Sixth Framework Programme, under Project Dewint (Contract ERAS-
CT-2003-980409). The review of this letter was arranged by Editor B. Yu.
The authors are with the Dipartimento di Ingegneria dell’Informazione:
Elettronica, Informatica, Telecomunicazioni, Università di Pisa, 56126 Pisa,
Italy (e-mail: g.fiori@iet.unipi.it).
Digital Object Identifier 10.1109/LED.2007.901680
Since at the moment the fabrication technology is at its
very first steps, computer simulations can be very useful to
provide physical insights of GNR-FETs and to estimate the
attainable performance. Recent theoretical works have shown
that graphene nanoribbons have an energy gap that has an os-
cillating behavior as a f unction of width, with average roughly
proportional to the inverse width, and that edge states play a
very important role in inhibiting the existence of fully metallic
nanoribbons [7]. Such behavior cannot be reproduced if one
does not consider edge effects [8].
Likewise, from the simulation point of view, research on
GNR-FETs is at an embryonic stage: the only works available
in the literature [9], [10] are based on a semiclassical analytical
top-of-the-barrier model. For short-channel transistors, only a
3-D simulation is suitable for an accurate evaluation of the
electrostatics and of intraband and interband tunneling.
To this purpose, we have developed a code for the simulation
of GNR-FETs, based on the nonequilibrium Green’s function
formalism (NEGF), with a tight-binding Hamiltonian built
from a p
z
orbital basis set in the real space, which has been
included in our in-house 3-D device simulator NANOTCAD
ViDES [11]. We will show that GNR-FETs have performance
comparable with CNT-FETs and that can be greatly affected by
the channel width and edge roughness.
II. P
HYSICAL MODEL AND RESULTS
Our approach is based on the self-consistent solution of the
3-D Poisson and Schrödinger equations with open boundary
conditions [12], which is able to take into account fully ballistic
transport, in order to outline the higher limits of device perfor-
mance, as well as elastic scattering due to line edge roughness.
The Hamiltonian is taken from [7], in which edge states at the
nanoribbon lateral ends have been considered. In this letter, we
refer to (N, 0) armchair graphene nanoribbons, which consist
of an unrolled (n, 0) zigzag nanotube with N = 2n.
The considered double-gate GNR-FETs have the structure
depicted in the inset of Fig. 1. The gates are metallic, the oxide
thickness t
ox
is equal to 1 nm, the channel is 15 nm long, and
W is the channel width. The source and drain extensions are
10 nm long and are doped with a molar fraction of fully ionized
donors f = 5 × 10
−3
.ThespacingbetweenparallelGNRs
is 4 nm.
In Fig. 1, the transfer characteristics of a (12,0) GNR-FET
(W = 1.37 nm) for drain-to-source voltage V
DS
of 0.1 and
0.5 V are shown and compared to those of a (16,0) CNT-FET
with the same geometry (same t
ox
, L,anddevicespacing),
whose energy gap E
gap
is close to that of the GNR-FET and
equal to 0.6 eV.
0741-3106/$25.00 © 2007 IEEE
FIORI AND IANNACCONE: SIMULATION OF GRAPHENE NANORIBBON FETs 761
Fig. 1. Transfer characteristics of double-gate CNT and GNR-FETs, with
doped source and drain reservoirs, with channel length equal to 15 nm, oxide
thickness t
ox
equal to 1 nm, and channel width W = 1.37 nm. The lateral
space is equal to 2 nm. Inset: Sketch of the GNR-FET.
Good control of the channel by the gate potential is shown
at V
DS
= 0.1 V, since the subthreshold swing S for the GNR-
FET and the CNT-FET are 64 and 68 mV/dec, respectively. For
V
DS
= 0.5 V, we observe a pronounced degradation of S,with
S = 191 mV/dec for the GNR-FET and almost 250 mV/dec
for the CNT-FET. This has to be imputed to hole-induced
barrier lowering (HIBL) [12]: in the subthreshold regime, when
sufficiently high V
DS
is applied, confined states in the valence
band of the channel align with the occupied states in the drain,
leading to band-to-band injection of holes in the channel.
If only elastic band-to-band tunneling can occur (as assumed
in our simulation), the excess of holes in the channel lowers
the channel potential, increasing the
OFF current and S,as
shown in Fig. 1: the lower the energy gap and the higher the
V
DS
,thehighertheHIBLeffect.HIBLismorepronouncedin
CNT-FETs than in GNR-FETs, because the conduction band of
CNTs is double degenerated and therefore CNTs have twice the
density of states of GNRs with the same gap.
If, on the other hand, inelastic band-to-band tunneling or
Schockley–Read–Hall mechanisms are relevant, holes can re-
combine with electrons at the source and, instead of HIBL, we
observe a leakage current from source to drain due to gate-
induced drain leakage (GIDL) [13].
In strong inversion, the transconductance g
m
at V
DS
= 0.1 V
is 3600 and 6100 µS/µmfortheGNR-FETandtheCNT-
FET, respectively, whereas at V
DS
= 0.5 V, we obtain g
m
=
4800 µS/µmfortheGNRandag
m
= 8760 µS/µmforthe
CNT. The advantage of CNT-FETs is due to the double degen-
eracy of the conduction band in CNTs.
It is known that a variability of the chirality of fabricated
CNTs yields metallic nanotubes useless for transistor applica-
tions. For GNRs, this problem is mitigated, since all GNRs
are semiconducting. In order to investigate quantitatively the
effect of a finite fabrication tolerance on the width of GNRs, we
have computed the transfer characteristics of GNR-FETs with
different chiralities: (12,0), (14,0), and (16,0).
As can be seen in Fig. 2(a), the three devices behave as
transistors but show very different behavior, even if they differ
Fig. 2. Transfer characteristics in the (a) logarithmic and sensitivity on
(b) linear scale of GNR-FETs with different chiralities: (12,0), (14,0),
and (16,0) (channel width W equal to 1.37, 1.62, and 1.87 nm, respec-
tively), for V
DS
= 0.1 V. (Dashed line) Transfer characteristic for the (16,0)
GNR-FET when roughness at the lateral edge of the GNR is considered.
Inset: Sketch of the graphene nanoribbon, where randomly decoupled atoms
have been highlighted by thick lines.
by only one carbon atom along the channel width. The problem
is that the gap is still largely dependent on the chirality: the
(16,0) GNR (W = 1.87 nm) has the largest gap (E
gap
=
0.71 eV),whereasthe(14,0)(W = 1.62 nm) has the smallest
gap (E
gap
= 0.13 eV).Asaconsequence,the(16,0)device
shows the best gate control over the channel potential, whereas
the (14,0) shows the worst: the energy gap is so small that
elastic band-to-band tunneling occurs at the source and current
is dominated by GIDL.
Such problem is reduced if rough edges are considered. We
have considered the impact of line edge roughness in a (16,0)
GNR-FET device, by randomly decoupling carbon atoms on
the lateral boundaries of the GNR. The transfer characteristic
for one example is shown in Fig. 2 (dashed line). Since the
channel consists of several hundreds of rings, the rough GNR
behaves as a GNR with an intermediate effective gap. More
statistical simulations would be needed to assess the dispersion
of the electrical characteristics, although the typical GNRs
are probably long enough to provide sufficient averaging to
suppress interdevice dispersion. Rough edge scattering strongly
affects the
ON current and the transconductance suppressing
it by about 30% with respect to fully ballistic transistors.
Additional suppression in realistic GNR-FETs can be due to
defects, ionized impurities, and phonon scattering.
From the aforementioned simulations, it is clear that lateral
confinement way beyond state-of-the-art etching capabilities
would be needed to obtain adequate E
gap
.Wealsofoundthat
electrostatic periodic potential modulation with a peak-to-peak
value of a few volts is not sufficient to induce the required gap
of a few hundred millivolts.
In order to evaluate whether a periodic strain pattern can
allow to engineer the GNR gap, we have computed the energy
gap in a (24,0) GNR (W = 5.86 nm),multiplyingtheoverlap
integral of the element of the Hamiltonian in correspondence
to the couple of atoms in the middle of the GNR by a “strain
factor” σ : σ is larger than 1 for compressive strain, and
762 IEEE ELECTRON DEVICE LETTERS, VOL. 28, NO. 8, AUGUST 2007
Fig. 3. Energy gap of a (24,0) GNR, when tensile and compressive strains are
considered in correspondence to the middle of the nanoribbon, as a function
of the strain factor by which the Hamiltonian elements of the strained carbon
atoms are multiplied.
smaller than 1 for tensile strain. As can be seen in Fig. 3,
compressive strain seems to be able to increase the energy gap
of the nanowire by a significant amount. Of course, we can only
suggest to experimentalists to evaluate the option.
III. C
ONCLUSION
In this letter, a simulation study of GNR-FETs has been
performed by means of the self-consistent solution of the 3-D
Poisson and Schrödinger equations with open boundary con-
ditions within the NEGF formalism. Edge states have been
considered at the lateral ends of the nanoribbon using the model
proposed in [7]. GNR-FETs exhibit performance similar to
CNT-FETs, also showing significant band-to-band tunneling
when small gap devices are considered and large V
DS
is applied.
GNR-FETs are more robust than CNT-FETs with respect to
variability of the channel chirality, and edge roughness seems to
play a useful averaging effect. Finally, we suggest that periodic
strain could in principle represent an alternative to etching for
inducing an energy gap in graphene.
ACKNOWLEDGMENT
The authors would like to thank Prof. M. Macucci for sug-
gestions and fruitful discussions.
R
EFERENCES
[1] A. Javey, J. Guo, Q. Wang, M. Lundstrom, and H. Dai, “Ballistic carbon
nanotube field-effect transistors,” Nature,vol.424,no.9593,pp.654–657,
2003.
[2] International Technology Roadmap for Semiconductor 2005. [Online].
Available: http://public.itrs.net
[3] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang,
S. V. Dubonos, I. V. Grigorieva, and A. A. Firsov, “Electric field effect in
atomically thin carbon films,” Science,vol.306,no.5696,pp.666–669,
2004.
[4] M. Y. Han, B. Özyilmaz, Y. Zhang, and P. Kim, “Energy Band Gap
Engineering of Graphene Nanoribbons,” Phys. Rev. Lett. ,vol.98,no.20,
p. 206 805, May 2007.
[5] M. C. Lemme, T. J. Echtermeyer, M. Baus, and H. Kurz, “A graphene
field-effect device,” IEEE Electron Device Lett.,vol.28,no.4,pp.282–
284, Apr. 2007.
[6] Z. Chen, Y.-M. Lin, M. J. Rooks, and P. Avouris, Graphene Nano-Ribbon
Electronics. [Online]. Available: http://arxiv.org/abs/cond-mat/0701599
[7] Y. W. Son, M. L. Cohen, and S. G. Louie, “Energy gaps in
graphene nanoribbons,” Phys. Rev. Lett.,vol.97,no.21,p.216803,
Nov. 2006.
[8] S. Reich, J. Maultzsch, C. Thomsen, and P. Ordejon, “Tight-binding
description of graphene,” Phys. Rev. B, Condens. Matter,vol.66,no.3,
p. 035 412, Jul. 2002.
[9] Y. Ouyang, Y. Yoon, J. K. Fodor, and J. Guo, “Comparison of per-
formance limits for carbon nanoribbon and carbon nanotube tran-
sistors,” Appl. Phys. Lett.,vol.89,no.20,pp.203107–203109,
Nov. 2006.
[10] G. Liang, N. Neophytou, D. E. Nikonov, and M. S. Lundstrom, “Per-
formance projections for ballistic graphene nanoribbon field-effect tran-
sistors,” IEEE Trans. Electron Devices,vol.54,no.4,pp.677–682,
Apr. 2007.
[11] G. Fiori and G. Iannaccone, “Code for the 3D simulation of nanoscale
semiconductor devices, including drift-diffusion and ballistic transport in
1D and 2D subbands, and 3D tunneling,” J. Comput. Electron.,vol.4,
no. 1/2, pp. 63–66, Apr. 2005.
[12] G. Fiori, G. Iannaccone, and G. Klimeck, “A three-dimensional simulation
study of the performance of carbon nanotube field-effect transistors with
doped reservoirs and realistic geometry,” IEEE Trans. Electron Devices,
vol. 53, no. 8, pp. 1782–1788, Aug. 2006.
[13] J. Chen, T. Y. Chan, I. C. Chen, P. K. Ko, and C. Hu, “Subbreakdown drain
leakage current in MOSFET,” IEEE Electron Device Lett.,vol.EDL-8,
no. 11, pp. 515–517, Nov. 1987.