Politecnico di Torino

III Facoltà di Ingegneria

A Graphene nanoribbon field-effect transistor Modeling

Integrated system technology

GROUP: 21

Mohamed Zakarya Rashed Ardalan Lotfi

June 5, 2014
 A Graphene nanoribbon field-effect transistor Modeling 1

Contents

1 Abstract 2

2 Introduction 2

3 Theoretical analysis 4
3.1 GNR-FET MODEL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2 CALCULATION AND ANALYSIS OF CURRENT-VOLTAGE CHARACTERISTICS . . . . . . . . . . . . 8

4 Octave implemetation 10

5 Results 11

References 14
 A Graphene nanoribbon field-effect transistor Modeling 2

1 Abstract

We present an analytical device model for a field-effect transistor based on a heterostructure which consists of an array
of nanoribbons clad between the highly conducting substrate (the back gate) and the top gate controlling the source-drain
current. The equations of the model of a graphene nanoribbon field-effect transistor (GNR-FET) include the Poisson equation
in the weak nonlocality approximation. Using this model, we find explicit analytical formulas for the spatial distributions of
the electric potential along the channel and for the GNR-FET current-voltage characteristics (the dependences of the source-
drain current on the drain voltages as well as on the back gate and top gate voltages) for different geometric parameters of
the device.

2 Introduction

Graphene is a new carbon-based, purely two dimensional material discovered only six years ago[1]. It shows a multitude of
fascinating properties and the prospect of ultrahigh carrier mobilities exceeding those of the conventional semiconductors
[1, 2, 3].This has motivated intensive work focused on the development of graphene metal-oxide-semiconductor field-effect
transistors (MOSFETs)[5, 6]. If graphene is narrowed into slender graphene nanoribbon (GNRs), a sizeable band gap
opens[7, 8]. Graphene MOSFETs with GNR channels showing complete switch off and large on-off rations have been demon-
strated and are considered as a possible building block for future logic circuits. In the following, the abbreviation GFET will
be used for graphene MOSFETs with gapless large-area graphene channel, while graphene MOSFETs with semiconducting
GNR channels will be designated as GNR MOSFETs. To facilitate the wide applications of GFET, analytical models for
drain current, transconductance, threshold voltage, and cutoff frequency are highly needed. The drain current model is
one of the most important. Thiele et al[8] used a quasi-analytical modeling approach to calculate the currentvoltage (IV)
characteristics. Jimnez and Moldovan [9] used drift-diffusion carrier transport approach to obtain an explicit drain current
model. Wang et al [10] presented a virtual-source IV model. Though the aforementioned drain current models agreed well
with experiments, they all included many fitting parameters and were in a rather complicated form, thus it is inconvenient
for applications.

The focus of our model was supposed to merge the merits and advantages of all the pre-mentioned models but due to the
need for many fitting parameters in each of these models, which make it so difficult to model that approach for generalized
devices, each one is focusing on specific point of view of modeling the graphene FET behavior, for example the (drift-diffusion
carrier transport approach) band-gap engineering of graphene could be not needed which is a good point of that model, but
its so complicated to be modeled. Regarding the Virtual-source I-V model, the problem is that the model is iterative model
which need to be solved self consistently to reach the IDS value and so it takes a long time for computations. The operation
of G-FETs is accompanied by the formation of the lateral n-p-n (or p-n-p) junction under the controlling (top) gate and the
pertinent energy barrier. The current through this barrier can be associated by both thermionic and tunneling processes.
The utilization of the patterned graphene which constitutes an array of sufficiently narrow graphene strips (nanoribbons)
provides an opportunity to engineer the band structure to achieve the optimal device parameters. In particular, properly
choosing the width of the nanoribbons, one can fabricate the graphene structures with relatively wide band gap but rather
high electron (hole) mobility[11].

In this report, we present an analytical device model for a graphene-nanoribbon FET (GNR-FET). The GNR-FET under
consideration is based on a patterned graphene layer which constitutes a dense array of parallel nanoribbons of width d
with the spacing between the nanoribbons ds d. The nanoribbon edges are connected to the conducting pads serving as
the transistor source and drain. A highly con-ducting substrate plays the role of the back gate, whereas the top gate serves
to control the source-drain current. The device structure is schematically shown in Fig. 1. Using the developed model, we
 A Graphene nanoribbon field-effect transistor Modeling 3

calculate the potential distributions in the GNR-FET as a function of the back gate, top gate, and drain voltages, Vb , Vg , and
Vd (reckoned from the potential of the source contact), respectively, and the GNR-FET dc characteristics. This corresponds
to a GNR-FET in the common-source circuit. For the sake of definiteness,the back gate voltage Vb and the top gate voltage
Vg are assumed to be, respectively, positive and negative (Vb > 0 Vg < 0) with respect to the potential of the source contact,
so we consider GNR-FETs with the channel of n-type.

Figure 1: Schematic side (a) and top (b) views of a GNR-FET structure.
 A Graphene nanoribbon field-effect transistor Modeling 4

3 Theoretical analysis
3.1 GNR-FET MODEL

Graphene strips (nanoribbons) exhibit the energy spectrum with a gap between the valence and conduction bands depending
on the nanoribbon width d:
p
ε−1
p,n = ±ν p2 + (π~/d)2 n2 (1)

Here ν = 108cm/s is the characteristic velocity of the electron (upper sign) and hole (lower sign) spectra, P is the momentum
in along the nanoribbon, ~ is the reduced Planck constant, and n = 1, 2, 3, ... is the subband index. The quantization corre-
sponding to Eq. (1) of the electron and hole energy spectra in nanoribbons due to the electron and hole confinement in one
of the lateral directions results in the appearance of the band gap between the valence and conduction bands and in a specific
the density of states (DOS) as a function of the energy. In contrast to graphene with zero energy gap, the electron (hole)
gas in relatively narrow nanoribbons with quantized energy spectrum and the pertinent energy gap between the subbands in
the valence and conduction band, can become degenerate at fairly high back gate voltages.in our model we restrict ourselves
by the consideration of the GNR-FET operation under the condition that the electron gas in the channel is nondegenerate.
Thus, the back gate voltage is assumed to be not excessively high, so that the electron density is moderate, the electron gas in
the channel is nondegenerate, and the electrons occupy only the lowest (n = 1) subband in the conduction band nanoribbons.
The channel sections adjacent to the source and drain contacts are highly conducting, so that these section are equipotential.
The potentials are equal to the potentials of the source and drain contacts, i.e., equal to φ = 0 and φ = Vd , respectively. The
GNR-FET region under the top gate (the device active region) defined as follows: Lg ≤ x ≤ Lg/2, W b ≤ z ≤ W g, where
Lg is the length of the top gate, and Wb and Wg are the thicknesses of the layers between graphene and the back and top
gates, respectively. To find the potential distribution along the channel, we use the following equation

(W b + W g) ∂ 2 φ φ − Vb φ−Vg 4πe
2
− − = (Σ− − Σ+) (2)
3 ∂x Wb Wg

with the boundary conditions

φ |x = −Lg/2 = 0, φ |x = −Lg/2 = Vd (3)

Here, Σ− and Σ+ are the electron and hole sheet densities in the channel, e is the electron charge, and is the dielectric constant
of the material separating the channel from the gates. Equation (2), which governs the electric potential, φ = φ(x) = ψ(x, 0),
in the channel is a consequence of the two-dimensional Poisson equation for the electric potential ψ = ψ(x, z) for the
active region under consideration in the weak non-locality approximation [12, 13] this equation provides solutions, which
can be obtained from the exact solution of the two-dimensional Poisson by the expansion in powers of the parameter
δ = [Wb3 + Wg3 /45(W b + W g)L2 ] , which is much smaller than unity in the situation under consideration. The lowest
approximation in such an expansion corresponds to the so-called gradual channel approximation proposed by W. Shockley,
in which the first term in the left-hand side of Eq. (2) is neglected, so that the relationship between the potential in the channel
and the electron and hole charge becomes local. Thus, Eq. (2 ) can be used when δ < 1 and, hence, when L ≥ max(W b, W g)
.The harnessing of the approximation under consideration makes, in particular, possible to study essentially nonuniform
potential distributions in the GNR-FET channel and the short-gate effects analytically. Since the spacing, ds between the
nanoribbons is small, we disregard a small scale nonuniformity of the electric potential distribution in the in-plane direction
y.
 A Graphene nanoribbon field-effect transistor Modeling 5

III. POTENTIAL DISTRIBUTIONS

Considering that the electron and hole gases are non- degenerate, and, hence, the electron and hole distribution func-
tions in the subbands with n = 1 near the source and drain is given by

A. Electron and hole densities

The application of negative top gate voltage leads to the formation of a potential barrier in the channel under this gate.
This barrier determined by the gate voltage controls the source-drain current and, hence, is responsible of the device opera-
tion as a transistor.
p
e ± φ ± εf − v 2 P 2 + ∆2 /4
fb∓ ' exp ( ) (4)
KB T

−
Where ”F is the Fermi energy reckoned from the middle of the energy band gap and ∆ = ε+ 0,1 − ε0,1 = 4π~/d is the value
of the energy band gap, the electric potential, the electron densities and the Fermi energies in the source and drain regions
(marked by superscripts s and d, respectively) are related to each other as

εsf + eφ ∞ εsf + eφ
Z
4KB T ξexp(−ξ) 2∆ 2∆
Σs∓ = exp(± ) √ 2 dξ = exp(± )K1 ( )
π~dυ KB T ξm 2
ξ − ξm π~dυ KB T 2KB T

εsf + e(φ − V d) ∞ εsf + e(φ − V d)

Z
4KB T ξexp(−ξ) 2∆ 2∆
Σd∓ = exp(± ) √ 2 dξ = exp(± )K1 ( )
π~dυ KB T ξm 2
ξ − ξm π~dυ KB T 2KB T

Here, ξm = ∆/2KB T and k1 (ξ) is the modified Bessel function. Taking into account the asymptotic behavior of the Bessel
function atξ  1, namely, k1 (ξ) = (π/2ξ)1/2exp(ξ), for ∆  2KB T , we obtain
√ √
2 2∆KB T εsf + eφ ∆ 2 2∆KB T εsf + e(φ − V d) ∆
Σs∓ = √ exp(± KB T ), Σd∓ = √ exp(± − KB T ) (7)
π~dυ KB T − 2 π~dυ KB T 2

Considering that when φ = 0 and φ = Vd , Σs−,0 )s − Σs+,0 = æVb /4πeW b and ,Σs−,0 )s − Σs+,0 = æ(Vb Vd )/4πeW b, respectively,
where the quantities with the index 0 are the electron and hole densities in the immediate vicinity of the source and drain
contacts, we arrive at the following equation:

εsf ∆ æ eV d ~υ
sinh( ) = exp( )×[ √ b ( ( )]. (8)
KB t 2KB T 16π ∆KB T W b e2

εdf ∆ − 2eVd æ eVb − Vd d ~υ

sinh( ) = exp( )×[ √ ( ( )]. (9)
KB T 2KB T 16π ∆KB W b e2

When Vb = V g = Vd = 0 , one obtains Σsf = Σ = 0df = 0 so that the electron and hole densities are equal to their thermal value
√
2 ∆KB T ∆
ΣT = √ exp(− ). (10)
πν 2KB T
This requires the application of a sufficiently high back gate voltage. If

∆ Vb Vb − Vd
exp(− ),  , < 1. (11)
2KB T VF VF
 A Graphene nanoribbon field-effect transistor Modeling 6

where
√ √
∆KB T 8 π Wb e2
VF = [ ( )( ), (12)
e æ d ~ν

∆ Vb ∆ Vb − Vd
εsF ' + KB T ln( ), εdF ' + KB T ln( ), (13)
2 VF 2 VF

Under consideration, the electron density in the pertinent sections of the channel, as follows from Eqs. (6) and (12), are
given by

æVb eφ æVb eφ − Vd
εs− ' exp( ), εd− ' exp( ) (14)
4πeWb KB T 4πeWb KB T

Since at Vb > 0 , the electron density markedly exceeds the hole density, we can neglect Σh In the right-hand side of
Eq(2).Considering the relationships between the electron density and the electric potential, from Eq.(2) one can arrive at
the following equations governing the potential distribution in the active region:

∂2φ 3 3(Vb /Wb + Vg /Wg ) 3Vb eφ

=− φ=− + exp( ) (15)
∂x2 Wb Wg (Wb + Wg ) (Wb + Wg )Wb KB T

∂2φ 3 3(Vb /Wb + Vg /Wg ) 3Vb eφ − Vd

=− (φ − Vd ) = − + exp( ) (16)
∂x2 Wb Wg (Wb + Wg ) (Wb + Wg )Wb KB T

The exponential dependences in the right-hand sides of Eqs. (15) And (16) are valid provided that inequality (11) is satisfied,
in particular, if the electron gas is nondegenerate. The threshold value of the back gate voltage, at which the degeneration
of the electron gas occurs, can be estimated as Vb .VF .

B. Potential distributions at low top-gate voltages

When the top gate is negative (Vg < 0) and its absolute value |Vg | is sufficiently small, the modulus of the potential |φ|
can be not that large. In this case, we can linearize Eqs. (15) and (16) and present these equations in the following form:

∂2φ 3 Wg eVb 3vg

− [1 + ]φ = − . (17)
∂x2 Wb Wg Wb + Wg KB T (Wb + Wg )Wg

∂2φ 3 Wg e(Vb − Vd ) 3(vg /Wg − Vd /Wb )

− [1 + ]φ = − . (18)
∂x2 Wb Wg Wb + Wg KB T (Wb + Wg )

V g < 0. At V d = 0, Eqs. (17) and (18) yield

cosh(x/λ)
[1 − ]
V g Wb cosh(Lg /2λ)
φ' (19)
Wb + Wg Wg eVb
[1 + ]
(Wb + Wg )KB T
 A Graphene nanoribbon field-effect transistor Modeling 7

where
s s
Wg eVb Wg eVb
λ − Λ/ 1+ = Λ/ 1 + (20)
Wb + Wg K B T Wb + Wg KB T
p
Is the effective screening length andΛ = (W bW g/3) At V d = 0, as follows from Eq. (18), the function exhibits a minimum
φ = φm0 atx = 0 with

1
[1 −
V g Wg cosh(Lg /2λ) Vg Wb KB T
φm0 ' s ' (21)
Wb + Wg Wg eVb Vb Wg e
1+
Wb + Wg KB T

Here, we have taken into account that normally Vb  kBT /eandLg  (with < ).

Vg Wb KB T
φm ' φm0 ' (22)
Vb Wg e

Comparing e|φm |given by Eq. (20) with kBT, we find that Eqs. (16) - (19) are valid when |Vg | . Vb Wg /Wb

C. Potential distributions at high top-gate voltages

At high back gate voltages, the quantity playing the role of the screening length is rather small. In this case, the length
of the regions near the points x = Lg/2, in which the potential changes from φ = 0 to |φ| > KB T and from φ = Vd to
|0φV d| > kBT ,is small in comparison with the top gate length Lg .Insuchshortregionsnear x = ±Lg/2, the potential
distribution can still be describe by Eq. (18). However, in a significant part of the active region the electron charge, which
is in such a situation exponentially small, can be disregarded and the last (exponential) terms in Eqs. (15) and (16) can be
omitted. Taking this into account, at high top gate Voltages, Eqs. (15) and (16) in the central region can be presented in
the following form:

∂2φ 3 3(vg /Wg + Vb /Wb )

2
− φ=− . (23)
∂x Wb Wg (Wb + Wg )

Solving Eqs. (22) still using boundary conditions given by Eq. (3), we arrive at

Wg Wb cosh(x/Λ) sinh(x + Lg /2)/Λ

φ = (Vg + Vb ) [1 − ] + Vd , (24)
Wg (Wb + Wg Lg /2Λ sinh(Lg /Λ)

At V d = 0, φ exhibits a minimum at x = 0 and

Wg Wb 1
φm0 ' (Vg + Vb ) [1 − )] (25)
Wg (Wb + Wg cosh(Lg /2Λ

At reasonable values of the drain voltage Vd (sufficiently small compared to Vb ), Eq(22) yields

Wg Wb 1 Vd
φm ' (Vg + Vb ) [1 − )] + (26)
Wg (Wb + Wg cosh(Lg /2Λ 2 cosh(Lg /2Λ)

Figures 2 and 3 show examples of the spatial distributions (along the channel, i.e., in the x-direction) of the electric potential
in the active region (under the top gate) calculated for a GNR-FET with Wb = 100nm, Wg = 30nm , and Lg = 300nm at
 A Graphene nanoribbon field-effect transistor Modeling 8

the back gate voltage Vb = 2.0V , assuming different values of the top gate voltage Vg and the drain voltage Vd .

As seen from Fig. 2, the potential distribution markedly sags and the height of the barrier for electrons near the center
increases with increasing absolute value of the top gate potential |Vg |. Figure 3 demonstrates, in particular, that in the
GNR-FET with chosen parameters the minimum value of the potential φm and, hence, the height of the barrier −eφm for the
electrons propagating from the source are virtually insensitive to the drain voltage. The approach used in this subsection is
valid when e|φm |  KB T , i.e., when |Vg |−Vb Wg /Wb KB T (Wb +Wg )/eWb ∼ KB T /e.In the limit Lg  Λ > λ the equations
for φm obtained above (in particular Eq. (25)) coincide with those obtained using the gradual channel approximation.

3.2 CALCULATION AND ANALYSIS OF CURRENT-VOLTAGE CHARACTERISTICS

Considering that the source-drain current is determined by the electrons overcoming the potential barrier under the top gate,
one can use the following formula for the density of this current (per unit length):
Z ∞ Z ∞
2e
J= ( υp fps dp − υp fpd dp)
π~d Pm s p
Pm

where

ε+
p,0 2 P
υp = υ p (28)
dP υ P + ∆2 /4
2 2

s d
Is the velocity of the electron with momentum p in the lowest subband of the nanoribbon conduction band and Pm And Pm
are the momenta of the electrons with the energies e|φm | and e|phim + Vd )| , respectively. We have taken into account that
the nanoribbon array is dense: ds  d. Integrating in Eq. (27), we arrive at
 r     
æ KB T eφm eVd
J =υ exp × Vb − (Vb − Vd )exp − (29)
2π 3/2 Wb ∆ KB T KB T
 A Graphene nanoribbon field-effect transistor Modeling 9

A.Current versus drain voltage

The dependence of the source-drain current on the drain voltage is associated with the dependence of φm on this volt-
age given by Eqs. (21) and (25) and the voltage dependence of the last factor in the right-hand side of Eq. (29). At low top
gate voltages (|Vg | ≤ vg Wg /Wb ), using Eqs. (21) and (29), we obtain
  
eVd
J ∝ Vb − (Vb − Vd )exp − , (30)
KB T

i.e J ∝ Vd If Vd . KB T /e , and J = const. If (KB T /e  Vd < vg Wg /Wb ). At moderate and high top gate voltages when
φm is given by Eq. (27), we arrive at the following dependence
    
eVd eVd
J ∝ exp × Vb − (Vb − Vd )exp − (31)
2KB T cosh(Lg /2Λ) KB T

At Vd . KB T /e , Eq.(32) yields the same linear dependence on the drain voltage as that described by Eq. (30).
WhenVd  KB T /e , one obtains
 
eVd
J ∝ exp (32)
2KB T cosh(Lg /2Λ)

B. Current versus gate voltages

As follows from Eqs. (21), (25), and (28), the source- drain current as a function of the top gate voltage is described
by the following relations:

 
V g Wb
J ∝ exp (33)
Vb W g

At |Vg | . Vb Wg /Wb .

  
eVg Wb 1
J ∝ exp 1− (34)
KB T (Wb + Wg ) cosh(Lg /2Λ)

At |Vg | > Vb Wg /Wb .

As follows from Eqs. (33) and (34), the dependence of the source-drain current on the top gate voltage is much steeper in the
range high top gate voltages |Vg | & Vb Wg /Wb than at |Vg | . Vb Wg /Wb , i.e., when the central region of the channel becomes
essentially depleted. The source-drain current versus the back gate dependence at high top gate and drain voltages is given by
  
eVg Wb 1
J ∝ Vb exp 1− (35)
KB T (Wb + Wg ) cosh(Lg /2Λ)

The source-drain current increases with increasing backgate voltage due to the pertinent increase in the electron density in
all regions of the channel.
 A Graphene nanoribbon field-effect transistor Modeling 10

4 Octave implemetation

Table 1: Description of the inputs of all the symbols used in the theoretical part
Inputs
Quantity name Description u.m.(S.I.) Variable name
Lg length of the top gate nm Lg
Wb Thicknesses of layers btwn graphene and the back gate nm Wb
Wg Thicknesses of layers btwn graphene and the top gate nm Wg
æ Dielectric const of materials btwn the channel and gates F/m ae
Vb The back gate voltage V olt Vb
Vd The potential of the drain contact, V olt Vd
Vg The top gate voltage V olt Vg
φ The electrostatic potential V Fm
e elementary charge coulombs e
Ψ The electric potential V Psi
υ the characteristic velocity of the electron Cm/s v
∆ Energy Bandgap eV delta
KB Boltzman Constant m kgs− 2K − 1
2
KB
T Temperature Kelvin T

Table 2: Description of the outputs of all the symbols used in the theoretical part
Ouputs
Quantity name Description u.m.(S.I.) Variable name
J Current density (mA/cm) J
Ion Current in On condition Amber Io n
Iof f Current in off condition Amber Io f f
 A Graphene nanoribbon field-effect transistor Modeling 11

5 Results
The J versus Vd dependences for a GNR-FET with Lg = 300nm at the back gate voltages Vg = 2 V and Vg = 3V calculated
for the different top gate voltages Vg are shown in Figs 4 (a) and (b).

Figure.4: The source-drain current density versus drain voltage dependencies at fixed back gate voltage (Vb = 2.0V ) and
different top gate voltages Vg .

Here, as in the previous and consequent figures, we assumed that = 0.4eV , æ = 4, Wb = 100nm, Wg = 30nm, and
T = 300K. Since the gate voltages were set to be relatively high (Vb , |vg |  kB T /e0.025V ), Eqs. (26) and (28) were used for
the calculations. As seen from Figs.4 a and b , the source-drain current as a function of the drain voltage in a GNR-FET with
relatively long gate (Lg = 300nm) exhibits saturation starting rather low drain voltages. This is a consequence of very weak
sensitivity of the potential barrier for the electrons propagating from the source to the drain voltage. The transformation of
the J vs Vd dependences with varying top gate voltage in the range of the latter |Vg | > Vb Wg /Wb . These figures confirm a
strong sensitivity of the source-drain current to the gate voltages.

Figure.5: a,b)The source-drain current density as a function of the top gate voltage at different back gate voltages.
 A Graphene nanoribbon field-effect transistor Modeling 12

In particular, as seen from Eqs. (31) and (33), in a GNR-FET with a long top gate, the source-drain current saturates
when vd becomes larger that KB T /e, whereas in a GNR-FET with Lg comparable with Λ, the source- drain current markedly
increases with increasing vd even at rather large values of the latter. This is confirmed by the current-voltage characteristics
calculated for GNR-FETs with a long (Lg = 300nm) and a short Lg = 100nm) top gates shown in Figs.5 a,b.

Figure.6 : The source-drain current density as a function of the drain voltage in GNR-FETs with different gate lengths at
different top gate voltages: solid lines −(Lg = 300nm) (long gate) and dashed lines −(Lg = 300nm) (short gate).

Figure.7 : a)Current in On condition Ion . b) Current in Off condition Iof f .

 A Graphene nanoribbon field-effect transistor Modeling 13

Figure.8 :The Ion /Iof f ratio.

References
[1] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva, and A. A. Firsov,
Science 306, 666 (2004).

[2] A. K. Geim and K. S. Novoselov, Nature Mater. 6, 183 (2007).

[3] J.-H. Chen, C. Jang, S. Xiao, M. Ishigami, and M. S. Fuhrer, Nat. Nanotechnol. 3, 206 (2008).

[4] F. Chen, J. Xia, D. K. Ferry, and N. Tao, Nano Lett. 9, 2571 (2009).

[5] M. C. Lemme, T. J. Echtermeyer, M. Baus, and H. Kurz, IEEE Electron Device Lett. 28, 282 (2007).

[6] I. Meric, N. Baklitskaya, P. Kim, and K. L. Shepard, Tech. Dig. - Int. Electron Devices Meet. 2008, 21.2.

[7] M. Han, B. zyilmaz, Y. Zhang, and P. Kim, Phys. Rev. Lett. 98, 206805 (2007).

[8] S. A. Thiele, J. A. Schaefer, and F. Schwierz, Modeling of graphene metal-oxide-semiconductor field-effect transistors
with gapless large-area graphene channels, J. Appl. Phys., vol. 107, no. 9, pp. 094505-1094505-8, May 2010.

[9] 9 D. Jimnez and O. Moldovan, Explicit drain-current model of graphene field-effect transistors targeting analog and
radio-frequency applications, IEEE Trans. Electron Devices, vol. 58, no. 11, pp. 40494052, Nov. 2011.

[10] H. Wang, A. Hsu, J. Kong, D. A. Antoniadis, and T. Palacios, Compact virtual-source current-voltage model for top-
and back-gated graphene field-effect transistors, IEEE Trans. Electron Devices, vol. 58, no. 5, pp. 15231533, May 2011.

[11] Z. Chen, Y.-M. Lin, M. J. Rooks, and P. Avouris, PhysicaE 40, 228 (2007).

[12] A. A. Sukhanov and Yu. Ya. Tkach, Sov. Phys. Semicond. 18, 797 (1984).

[13] V. I. Ryzhii and I. I. Khmyrova, Sov. Phys. Semicond. 22, 807 (1988).
 A Graphene nanoribbon field-effect transistor Modeling 14

[14] T. Manku and A. Nathan, “Electron Drift Mobility Model for Devices Based on Unstrained and Coherently Strained-
Si1−x Gex Grown on (001) Silicon Substrate,” IEEE Trans. Electron Devices, vol. 39, no. 9, pp. 2082-2089, 1992.

[15] I. Balslev, “Influence of Uniaxial Stress on the Indirect Absorption Edge in Silicon and germanium,” Phys. Rev., vol.
143, pp. 636-647, 1966.