Graphene Klein tunnel transistors for high speed analog RF applications

Yaohua Tan,∗ Mirza M. Elahi, Han-Yu Tsao, K. M. Masum Habib,† N. Scott Barker, and Avik W. Ghosh‡
Department of Electrical and Computer Engineering,
University of Virginia, Charlottesville, Virginia 22904, USA
(Dated: May 24, 2017)
We propose Graphene Klein tunnel (GKT) transistors as a way to enforce current saturation
while maintaining large mobility for high speed radio frequency (RF) applications. The GKT
transistor (GKTFET) consists of a sequence of angled graphene p-n junctions (GPNJs). Klein
arXiv:1705.08263v1 [physics.app-ph] 18 May 2017

tunneling creates a collimation of electrons across each junction, so that the lack of substantial
overlap between transmission lobes across successive junctions creates a gate-tunable transport
gap without significantly compromising the on current. Electron scattering at the device edge
tends to bleed parasitic states into the gap, but the resulting pseudogap is still sufficient to create
a saturated output (ID − VD ) characteristic even with wide area bulk graphene. Compared to
traditional graphene field effect transistors (GFETs) whose high mobility appears at the cost of
gaplessness, the pseudogap generates a high output resistance that can now exceed the contact
resistance of graphene. The modulated density of states generates a higher transconductance (gm )
and unity current gain cut-off frequency (fT ) than GFETs. More significantly the high output
resistance makes the unity power gain cut-off frequency (fmax ) of GKTFETs considerably larger
than GFETs, making analog GKT switches potentially useful for RF electronics. We estimate that a
GKTFET can reach an fT of 33 GHz for a 1µm channel, and scale up to 330GHz at 100nm channel
(assuming a single, scalable trapezoidal gate). Compared with GFETs, a GKTFET is expected to
reach a 10 times higher fmax , at 17GHz and 53GHz for devices at 1µm and 100nm channel lengths
respectively. The fmax of a GKTFET maybe further enhanced by 3 times by reducing the gate
resistance, such as with T-Gate.

I. INTRODUCTION in graphene nanoribbons and nanotubes [12]. Further-

Graphene-based devices have long promised exciting
applications, from interconnects and transparent elec-
trodes to gas sensing. However, their gaplessness com-
promises our ability to gate these devices as an efficient
electronic switch. For instance, graphene is a promis-
ing channel material for radio frequency (RF) applica-
tions [1–3] due to its intrinsic high carrier mobility and
long mean free path [4–7]. In fact, graphene RF devices
have been reported to achieve fT ’s larger than 300GHz
for sub-100nm channels [8, 9] (Fig 9 (a)). However, the
gaplessness of graphene makes its output resistance low,
arising from the lack of any current saturation. Conse-
quently, the power gain cut-off frequency fmax of most of
the reported GFETs are much lower than their fT ,and
does not scale with channel length[1–3] (Fig 9 (b)) be-
cause of the non-scalability of the dominant contact re-
sistances.
Efforts to improve the fmax of GFETs have focused
on reducing the input resistance and introducing current
saturation. Recent work by Guo et al. [10] showed an
improved fmax in GFETs by significantly reducing the
gate resistance using a T-shaped gate. To obtain current
saturation in GFETs, an energy bandgap can in principle
be introduced in graphene, such as by applying symme-
try breaking strain [11] or using quantum confinement
∗ yt5x@virginia.edu
† Present address: Intel Corp., Santa Clara CA 95054, USA.
‡ ag7rq@eservices.virginia.edu
more, scattering process in a long graphene channel can
also introduce natural current saturation [13]. However,
those band gap opening mechanisms significantly reduce
the carrier mobility due to the distorted bandstructure or
carrier scattering events [14]. Thus a technique which in-
troduces a transport gap in graphene without degrading
the carrier mobility [15, 16] would be quite unique and
highly desirable for graphene-based RF applications.
Over several publications in the past, we have proposed
an alternative way of introducing gaps into graphene,
through the employment of p-n junctions as collimator-
filter pairs. A GPNJ is an angle dependent momentum
filter driven by the Klein tunneling of Dirac electrons
[17–19]. As a result, the GPNJ creates gate-tunable
transmission gaps instead of energy band gaps, so that
states are available for conduction in the ON state but
removed for the OFF state. Multiple papers in the past
have suggested using the GKTFET as a digital switch
[15, 16, 20, 21]. Initial calculations [15] treating the an-
gled junctions as independent transmitters estimated ul-
trahigh ON-OFF ratios in excess of 104 , while the gate
tunability of the transport gap predicted a subthresh-
old swing that beats the fundamental Boltzmann limit.
However, in practice the ON-OFF ratio of these devices is
seen to be compromised by recurrent momentum scatter-
ing of rejected electrons at the device edges, which typ-
ically redirects leakage states into the transmission gap
and limits the experimentally measured ON-OFF ratio
to anywhere between 1.5 [21] to 12. In other words, the
predicted gap readily turns into a pseudogap because of
parasitic scattering events.
In this work, we propose to use the pseudogap in GK-

TFETs for RF applications to overcome the lack of cur-
rent saturation in traditional GFETs. To understand the
characteristics of GKTFETs, we performed semiclassical
ray tracing calculations coupled with analytical models
for Klein tunneling to model electron transport in GK-
TFETs [22]. The critical device parameters for a given
geometry are extracted from finite element electrostatic
calculations in order to estimate the cut off frequencies.
According to our calculations, we argue that even a pseu-
dogap suffices to allow GKTFETs to have distinct cur-
rent saturation [23] and considerably larger output re-
sistance r0 than conventional GFETs, in fact, in excess
of their contact resistances. In the process, the mobil-
ity in a GKTFET is not significantly degraded because
the transmission gap dominates only for the OFF state
and is kept just small enough in the ON state to still
allow saturation. We expect the GKTFET can reach a
fT of 33 GHz in a 1µm channel device, and scale up to
330 GHz at 100nm channel length assuming ideal single
gate scaling. The fmax of GKTFET can reach 17 GHz
in a device with a 1µm channel and 53 GHz at 100nm
length, which is more than 10 times higher than that of
GFETs at a comparable channel length. Higher fmax of
49 GHz(1µm) and 158 GHz (100nm) can be reached if
the gate resistance of GKTFET can be significantly re-
duced by reducing the gate input resistance, such as with
a T-Gate.
II. DEVICE CHARACTERISTICS
Klein tunneling across a graphene p-n junction is
driven by the conservation of pseudospin, which in turn
is set by the phase coherent superposition of the dimer
pz basis sets. The GPNJ acts as an efficient momentum
filter, transmitting electrons that are injecting perpen-
dicularly to the junction regardless of the barrier height
across the junction. A GPNJ with a graded junction
potential further filters the non-normal electrons aggres-
sively, as those electrons see an angle dependent tunnel
barrier akin to the cut-off modes in a rectangular waveg-
uide [22, 24]. The device considered in this work con-
tains two back-to-back GPNJs controlled by three gates
including a trapezoidal one, as shown in Fig.1. (a). The
GPNJ on the right has a tilted angle δ(= π/4) with re-
spect to the left one [16]. In the OFF state, n-p-n re-
gions are formed in the channel by applying proper gate
biasing. These two back-to-back p-n junctions will col-
lectively turn off the current. The left GPNJ serves as
a collimator which blocks most of the incoming electrons
except for those incident perpendicular to it; the second
tilted GPNJ further blocks the electrons coming from
the collimator allowing in turn only electrons perpendic-
ular to itself to pass. An OFF state is thereby achieved
through sequential momentum filtering when the angle
of the second junction exceeds the critical angle at the
first junction. The average gate length considered in this
work is 1µm with a split length of d1,2 = 80 nm. For
2
(a) Top view
Graphene
(d)
Biasing
FIG. 1. Top view (a) and side views (b),(c) of the concep-
tional device schematic and (d) Gate biasing scheme. The
gate controls the charge concentration in the central green re-
gion. The side views in (b) and (c) suggest the N regions in (a)
can be created by either gate doping (b) or chemical doping
(c). The essential part of the devices is shown in dashed boxes,
where the Klein Tunneling effect near GPNJs dominates. In
the OFF state, two back-to-back GPNJs are formed. The left
GPNJ acts as a collimator and right GPNJ acts as a filter. In
the on state, n-n− -n regions are formed in the channel. Gray
regions in (d) correspond to the energy range of the transmis-
sion gap. The GPNJ on the right is tilted by angle δ = 45◦
with respect to the left one. Here the d is the split length
between gates. In this work, the device has an average gate
length of (L1 + L2 )/2 = 1µm (L1 = 1.5µm, L2 = 0.5µm) and
a width of W = 1µm. The gate oxide is equivalent to a 5nm
SiO2 .
this paper, we assume a gate oxide with an equivalent
SiO2 thickness of 5nm. In the ON state, the three wedge-
shaped gates are held at nearly the same voltage polarity
(n-n− -n), so there is no angular filtering of electrons in
the Fermi window between µS and µD . However, a small
transmission gap still exists because of the slight differen-
tial doping. This small transmission gap leads to current
saturation in the ON state. In the OFF state, we move
the polarity of the central gate to n-p-n where the gap
increases substantially and the current drops.
It should be noted that the GKTFET proposed here is
designed to establish proof-of-concept. In practice the
geometry needs to be optimized keeping in mind the
fabrication techniques, considering different approaches
such as electrical gating (Fig.1 (b)) or contact-induced
doping[25]/chemical doping (Fig.1 (c)) [26, 27] to create
the side gated regions (blue n-doped regions in Fig.1 (a)).
According to our finite element electrostatic calculations

using Ansoft Maxwell, the side gate in (Fig.1 (b)) at the
drain end (Gate 3) introduces a large parasitic capaci-
tance. This extra capacitance will possibly compromise
the cut-off frequencies if it is AC connected to the ground
directly. Extra care should be taken to get rid of the
effect of this capacitance, as we discuss later in this pa-
per. Compared with electrostatic side gate doping, the
chemical doping shown in Fig.1 (c) does not suffer from
these large gate capacitances. However chemically doped
graphene has lower carrier mobility. In Fig.1, we showed
a proposed device with buried gates that was the basis
for the calculations in this paper ; however, it is worth
looking at alternate geometries, such as top gates [10, 28],
with the associated design trade-offs.
Graphene transistor
0.2
(a)
0 0
−0.2
E [eV]
Perfect Sample
−0.4 Imperfect Sample
−0.6
−0.8
0 100 200 300
2
G(E) [4q /h]
FIG. 2. Energy resolved conductance of the on-state for (a)
GFET and (b) GKTFET. In GFET, the G(E) ∝ |E| corre-
sponds to thep Dirac cone-like band structure of clean sample,
and G(E) ∝ E 2 + 2σ 2 /π in imperfect samples[29]. In the
GKTFET, a clear transmission gap can be observed. The
G(E) in the transmission gap is slightly non-zero due to the
edge reflection in the GKTFET.
In our proposed device, the total current IDS across the
GKTFET can be estimated by the Landauer equation.
2q
Z h i
IDS = Tav M f (µS ) − f (µD ) dE (1)
h cal transmission probability for each electron across the
where M is the number of modes, Tav is their mode-
averaged transmission, q is the charge of the electron, h is
Planck’s constant, f is the Fermi-Dirac distribution, and
µS,D are the bias-separated electrochemical potentials in
the source and drain. The mode-averaged transmission
Tav and number of modes M at energy E are controlled
by the potential drops on the channel VGS . The resulting
transconductance gm can be written as
∂IDS ∂ (Tav M ) h
Z i
gm = ∝ f (µS ) − f (µD ) dE (2)
∂VGS ∂VGS
≈ Tav M |µµD (3)
S
In the ON state, the GKTFET has a small transmission
gap around the Dirac point so that its mobility and gm
are expected to resemble a pristine GFET with the same
dimensions. The presence of a transmission gap will,
however, cause the current to saturate when the drain
electrochemical potential µD moves towards the Dirac
3
point and enters the transmission gap. In contrast, the
gm of ultra-clean GFETs has just a single point satura-
tion precisely when µD hits the Dirac point since there
is no gap in pristine graphene. This feature can be seen
later in Fig.4 and Fig.5. The output resistance rout can
be estimated by
Z −1
∂VDS ∂f (µD )
r0 = ∝ M Tav dE . (4)
∂IDS ∂µ
In the ballistic limit, the output resistance depends on
the modes inside the band gap. A perfect energy gap
in principle leads to infinite output resistance because
M Tav = 0 in the gap, while any states inside the gap due
to imperfections (such as scattering, defects) will lead
to a finite output resistance. In our proposed devices,
GKT transistor
0.2 the output resistance is limited primarily by the edge
(b)
reflection and carrier scattering.
In this work, we simulate the GKTFET using a
−0.2 semiclassical ray-tracing method coupled with analyti-
cal equations for chiral tunneling across the junctions
−0.4 Transmission gap
[22]. Standard quantum transport methods like the non-
−0.6
Perfect edge equilibrium Green’s function formalism (NEGF)[29, 30],
Rough edge
are computationally quite expensive for GKTFETs with
−0.8
0 100 200 300 sizes between few hundreds of nanometers to micro-
2
G(E) [4q /h] meters, and moreover, bring in spurious interference ef-
fects that are irrelevant at room temperature. In con-
trast, the ray-tracing method [31–33] coupled with a well-
benchmarked quantum tunneling model across junctions
can be applied readily to devices at high voltage bias
with large areas and complicated geometries including
a sequence of multiple reflections, and has shown ex-
cellent agreement with recent experiments on GPNJs
[22]. In our approach, electrons are thrown from the
source randomly with injection angles following a co-
sine distribution (the angular distribution of the quan-
tized transverse wave-vectors). The average transmission
(Tav ) across the junctions is then calculated by counting
electrons that successfully reach the drain. The analyti-
junction is a simple generalization of Gaussian filtering
2
T ∼ exp−πkF d sin θ established in [24] and extended now
to an asymmetrically doped junction [16]. Here, d is the
split length of the junction, θ is the incident angle of the
electron, and kF is the Fermi vector on both sides.
In our calculations, we considered the cases of GPNJ
with perfect edges as well as rough edges. Fig. 2 shows
the integrated transmission of (a) bulk graphene and
(b) GKTFET, both for pristine vs imperfect geometries
(charge puddles for bulk and edge roughness for GKTs).
We see that pristine graphene has no band gap and its
density of pstates D(E) ∝ |E| for a perfect sample, while
D(E) ∝ E 2 + 2σ 2 /π for a dirty sample [29], with
2 2 2
σ ≈ 2~ vF nimp +C describing the contribution of charge
puddles in washing out the Dirac point through spatial
averaging. We use a typical impurity density in a dirty
sample with nimp = 1 × 1012 cm−2 in this work [29]. In
contrast with GFETs, GKTFETs have a distinct trans-
mission gap, as indicated in Fig. 2 (b). Indeed, very few
 4

Graphene transistor GKT transistor 10

0
2
(a) (b)
3
8 V GS = 1.83 V
V GS = 0.88 V V = 1.83 V 1.5
GS

rout [kΩ µm]

V GS = 0.88 V

gm [mS/ µm]
6 V GS = 0.27 V

I DS [mA/µm]
I DS [mA/ µm]

2 V GS = 0.27 V −1
1 10
4
V GS = 1.83 V
1 0.5
V GS = 0.88 V
2
V GS = 0.27 V
−2
(a) (b) 0 10
0 0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4
0 V [V]
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5 V [V] DS
DS
V DS [V] V DS [V]

FIG. 4. gm and rout vs. VDS in a GFET. (a) gm vs VDS

FIG. 3. IDS vs VDS in the GFET (a) and the GKTFETs (b).
in GFET; (b) rout vs. VDS in GFET. The solid lines are
In (a), the solid lines are IDS in imperfect graphene samples
imperfect GFET and the dashed lines are in perfect GFET.
which are linear functions of VDS ; and the dashed lines in
The gm shows 0.5 to 1.5mS/µm at saturation points. The
(a) are IDS in perfect graphene which show one point sat-
rout shows a value around 0.1kΩµm for dirty sample.
uration. In (b), the solid lines are in GKTFETs with edge
roughness considered, and the dashed lines are GKTFETs
with perfect edges. The GKTFETs shows obvious current 2.5
(a) (b)
saturation in both cases. The GKTFETs with edge rough- 10
2

ness show slightly larger slopes in the saturation region. The 2

quantum capacitance is considered in all cases. 0.1V, 0.2V

g [mS/ µm]

[kΩ µm]
1.5
and 0.3V are dropped in the channel.
0
10

out
1

r
V GS = 1.83 V V GS = 1.83 V
modes appear in the transmission gap. These gap states 0.5 V GS = 0.88 V V GS = 0.88 V
V GS = 0.27 V
arise from edge reflection of electrons rejected by the sec- −2
V GS = 0.27 V
0 10
ond junction, a process that redirects them towards the 0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4
V DS [V] V DS [V]
drain. Ultimately these states contribute to the leakage
current in the off-state and lead to a finite rout . Our
FIG. 5. gm and rout vs. VDS in a GKTFET. (a) gm vs
calculation indicates that perfect edges in the GKTFET
VDS in GKTFET with/without edge roughness; (b) rout vs
can reduce the leakage current by 20 to 40 times com- VDS in GKTFET with/without edge roughness. The solid
pared with GKTFET with rough edges in a 1µm wide lines correspond to GKTFET with edge roughness and dashed
device. In our calculation, the edge roughness introduces lines correspond to GKTFET with perfect edges. The edge
a random reflection angle with an variance of σ = 18◦ . roughness in GKTFET change the gm sightly. While the edge
roughness reduces rout from 10 to 100kΩµm to about 1kΩµm.
In both cases the rout are obviously larger than the rout of
III. RF PARAMETERS GFET.

Fig. 3 shows the IDS − VDS characteristics of (a)

GFETs and (b) GKTFETs. In each case, the dashed rout of 0.1kΩµm for imperfect samples, and only around
lines are for perfect samples while the solid lines include rout ∼ 0.3kΩµm for perfect samples at saturation VDS
imperfections. It can be seen clearly that an ultraclean for all gate biases, dropping rapidly for other VDS val-
GFET shows IDS − VDS with single-point saturation, ues. At zero temperature, ∂f∂µ(µ2 )
= δ(E − µ2 ) in eq (4),
while a GFET with a dirty sample shows a quasi-linear 2
rout = ∞ as M Tav = 0 at the Dirac point. At finite tem-
IDS − VDS due to spatial averaging that washes out the
perature rout drops to a finite value because ∂f∂µ (µ2 )
has
Dirac point. In contrast, GKTFETs with both perfect 2

and rough edges show a clear current saturation due to a non-vanishing spread of kT . Compared with GFETs,
the presence of a pseudogap. The rough edges in GKT- the GKTFET shows much higher rout of 1kΩµm. The
FETs lead to only a marginally smaller rout because of GKTFET with perfect edges shows even higher rout val-
the increase of M T in the transmission gap, which can ues that can reach 50 to 100kΩµm. Furthermore, the
be understood using eq(4). saturation region corresponds to a VDS in the range of
0.1 to 0.3V instead of one point saturation.
Figs. 4 and 5 show the gm and rout of GFET and GK- To estimate the RF properties such as fT and fmax of
TFETs. The gm of GFETs reach 0.5 to 1.5 mS/µm graphene-based RF devices, we consider the equivalent
(each gated region is 1 µm long in our simulation with circuit for AC signals, shown in Fig.6. This structure as-
splits of 80 nm each), while the GKTFET turns out to sumes that the dominant capacitance is from the central
have a slightly higher gm of 1 to 2mS/µm. The out- gate that is swung between p and n polarities, while the
put characteristic however proves more dramatic than side gated regions have lower capacitance (we discuss this
the transfer characteristic. The GFET shows a very low point later in the paper) For this equivalent circuit, the
 5

Cgd length of 100nm leads to an expected max fT = 100

RG RD
to 300GHz for a conventional GFET, which agrees with
Gate Drain
the published literature[9, 10, 28, 35–37]. The fmax of
Cgs gm vgs rout the GFET reaches only 1.3GHz, i.e., 14% of fT because
of its small output resistance. Compared with the ideal
case where RS/D = 0, the fmax of GFET with larger
RS/D = 0.1kΩµm is reduced by 5%, while the peak fT
RS
is reduced by 69%.
Compared with GFETs, the fT of the GKTFET is
larger due to a larger gm arising from the opening of
Source the transport gap and the resulting variation in density
of states over the finite temperature window. More no-
FIG. 6. Equivalent small signal circuit of the GKTFET. ticeably, the fmax and fmax /fT ratio in GKTFET are
significantly higher due to the current saturation aris-
30
(a)
2
V GS =1.83V, R S/D =0.1k Ω µm
(b)
ing from the engineered pseudogap. Figure8 shows that
25
V GS =0.88V, R S/D =0.1k Ω µm
V GS =0.27V, R S/D =0.1k Ω µm
the GKTFET with 1µm channel length reaches a fT of
1.5
V GS =3.11V, R S/D =0.1kΩ µm
34GHz and fmax of 17GHz. The fmax is 13 times larger
20 V GS =3.11V, R S/D =0
than that of GFET. Furthermore, the contact resistance
[GHz]
fT [GHz]

15 1 has a much weaker impact on fT in GKTFET - in fact,

max

0.1kΩµm RS and RD reduces the fT by only ∼ 10−20%.

f

10
0.5 The impact of RS and RD to fT in both GFETs and
5 GKTFET is determined by the factor gds (RS + RD ),
0 0 gds = 1/rout , as we see in the denominator of equation
0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4
V DS [V] V
DS
[V] (5). The large output resistance rout of GKTFET weak-
ens the influence of RS and RD on the fT . In Fig.8 , we
FIG. 7. The fT (a) and fmax (b) of a GFET vs. VDS . The fT also show the fT and fmax in the limit of Cgd = 0 by the
and fmax reach the max value at saturation points and the dashed lines. It can be seen that the small Cgd leads to
value drop rapidly for other VDS . The fmax is significantly a 30% increase of the max fT and 10% increase of max
smaller than fT because of the small rout . fmax .

50 20 V GS = 0.88V, R S/D =0.1k Ω µm

V GS =1.83V, R S/D=0.1k Ω µm

fT and fmax can be estimated by [34]. V GS =1.83V, R S/D =0

V GS =1.83V, C gd =0
V GS = 0.27V, R S/D =0.1kΩ µm

40
15
gm 1
fT = 30
fmax [GHz]
fT [GHz]

2π (Cgs + Cgd )
q
2
[1 + g0 (RS + RD )] − (g0 RS )2 10
20
(5)
with g0 = gds + gm [Cgd /(Cgs + Cgd )] and 10
5

(a) (b)
fT 0 0
fmax = q . (6) 0 0.1 0.2
V
0.3
[V]
0.4 0 0.1 0.2 0.3 0.4
V [V]
2 RGr+R
ds
S
+ 2πfT RG Cgd DS DS

To illustrate the impact on fT and fmax from improved FIG. 8. The fT (a) and fmax (b) of a GKTFET vs Vds .
For given VG , the fT and fmax reach their maximum at the
gm and rout of GKTFET in comparison with GFETs,
saturation region which range over 0.1V to 0.3V . Due to high
we use Cgs = 6.9pF/mm, Cgd = 0.7pF/mm and RG = output resistance,the max fmax in GKTFETs is about 50%
1kΩµm for both transistors. However, it is worth re- of the max fT .
emphasizing that the parameters are strongly dependent
on the device geometry, for instance, the Cgd of the GK-
TFET in Fig.1 (b) is in fact negligible by finite element
electrostatic calculation using Ansoft Maxwell. We ac-
cordingly choose an experimentally achievable ratio of IV. DISCUSSION
Cgd /Cgs = 0.1 [10] in the following calculations of fT
and fmax . We have shown that a 1µm long GKTFET shows much
Figs.7 and 8 show the fT and fmax of the GFETs and better rout and cutoff frequencies fmax than GFETs due
GKTFET. The fT reaches 9.3 to 29.3GHz in GFETs to the transmission gap engineered in pristine graphene
with a channel length of Lchannel = 1µm (better con- using gate geometry. Perfect edges in a GKTFET would
tact resistances and smaller Cgd gives higher fT ). It is further reduce the leaked density of states in the trans-
known that the fT in pristine GFETs is inversely propor- mission gap, leading to larger rout and fmax . The small
tional to Lchannel [34]. Projecting accordingly, a channel contact resistances RS /RD also have significant impacts

10 3
L−1
330G/100nm
427G
300G
10
168G
f max (GHz)
f T (GHz)
10 2
10
GFET
GKT (a)
10 1 1
10 10 2 10 3
10
10 1
Channel length (nm)
FIG. 9. The fT (a) and fmax (b) of GKTFETs and reported
GFETs. The data are from ref [9, 10, 28, 35–37]. The fT of
reported GFETs is roughly inversely proportional to channel
length; while the fmax does not show this trend due to low
output resistance. For ideally scaled GKTFET with 100nm
channel length, the fT is expect to reach 330GHz, which is
comparable to the highest report fT in GFETs. The scaling of
fmax of GKTFET follows L−0.5 ideally. For a GKTFET with
100nm channel, the fmax is expect to reach 53GHz, and even
high fmax = 158GHz can be expected if T-gate technique is
used to reduce gate resistance.
to the cut off frequencyfT as they compete with rout .
Compared with GFET, the fT of the GKTFET is less
sensitive to RS and RD due to a larger rout .
The parasitic capacitances Cgd and gate resistance RG
are critical device parameters and significantly impact fT
and fmax . The Cgd and RG are strongly dependent on
the real design and geometry of the transistor. For in-
stance, recent experiments used T-shape gate to reduce
RG dramatically to get high fmax in GFETs [10]. Sim-
ply including a sizeable side gate would create a large
parasitic capacitance AC connected to the ground. To
mitigate this, we will need to reference the third gate to
the drain at a constant bias offset or include an induc-
tor between the two to filter out the high-frequency AC
signals. A more convenient choice would be to dope the
two end regions chemically and work with a trapezoidal
central gated region alone.
While our simulations were done for 1µm, the ulti-
mate advantage of the GKTFET for high-performance
RF depends on its overall scalability, since the Cgs and
−1
RG are proportional to the channel length. In the GK-
TFET, the gate width and length are related, as we used
a 45◦ tilted Junction. We show the fT and fmax of GK-
TFETs and GFETs in Fig.9. Ideally, the fT and fmax
−1 −1 −0.5
follow fT ∝ Cgs ∝ L−1 and fmax ∝ Cgs RG ∝ L−0.5 .
The scaling of the gate length of the GKTFET with from
1µm to 100nm is expected to increase the fT and fmax by
[1] F. Schwierz, Nature nanotechnology 5, 487 (2010).
[2] F. Schwierz, Proceedings of the IEEE 101, 1567 (2013).
[3] G. Fiori, F. Bonaccorso, G. Iannaccone, T. Palacios,
D. Neumaier, A. Seabaugh, S. K. Banerjee, and
6
10 and 3.2 times respectively, as indicated by the dashed
L−0.5 Single point
158G/100nm
saturation
34G
lines. In contrast, the fmax of GFETs does not scale with
2
58G channel length due to the low output resistance, as shown
in Fig.9 (b). To estimate the fT and fmax of 100nm GK-
53G/100nm
TFETs, we made the following assumptions: the scaling
1
down of the GKTFET does not change the electrostatics
GFET
GFET+T−gate
in the device (gate control is still dominant), the pseu-
0
GKT
GKT+T−gate (b)
dogap can be effectively created by GPNJs in a scaled
10 2 10 3 GKTFET, and device parameters such as Cgs and Cgd
Channel length (nm) scale properly with channel length. Detailed quantum
simulations coupled with numerical 3D electrostatics are
needed to test the performance of these devices at their
scaling limits.
V. CONCLUSION
To summarize, we propose a conceptual high-frequency
RF device in Fig.1. This device operates by geometry
engineering of a gate-tunable transport gap in pristine
graphene, using the physics of Klein tunneling. In con-
trast to conventional GFETs which suffer from weak cur-
rent saturation due to gaplessness, the engineering of
a transmission gap allows the GKTFET to enjoy both
high carrier lifetimes and current saturation. Our cal-
culation of the GKTFET shows a significant improve-
ment on fmax and a slightly higher fT compared with
GFETs. The device is expect to achieve an fT of 33GHz
and a comparable fmax of 17GHz in a device with
1µm gate length, and ramp up to fT = 330GHz and
fmax = 53GHz as we shrink the gate to 100nm. Higher
fmax of 49GHz for 1µm channel and 158GHz for 100nm
channel can be expected by reducing gate resistance with
the technique of T-Gate. In addition, the cut-off frequen-
cies of the GKTFET are seen to be much less sensitive
to the contact resistance than GFETs, once again due to
the significant increase in output resistance arising from
current saturation.
ACKNOWLEDGMENTS
This work is supported by the Semiconductor Research
Corporation’s NRI Center for Institute for Nanoelectron-
ics Discovery and Exploration (INDEX). The authors ac-
knowledge computational resources on UVa HPC System
Rivanna. We acknowledge Kurt Gaskill from Naval Re-
search Laboratory, Claire Berger from Georgia Institute
of Technology and Philip Kim from Harvard University
for helpful discussions.
L. Colombo, Nature nanotechnology 9, 768 (2014).
[4] K. I. Bolotin, K. Sikes, Z. Jiang, M. Klima, G. Fudenberg,
J. Hone, P. Kim, and H. Stormer, Solid State Commu-
nications 146, 351 (2008).
 7

[5] J.-H. Chen, C. Jang, S. Xiao, M. Ishigami, and M. S. [31] C. Beenakker and H. Van Houten, Physical review letters
Fuhrer, Nature nanotechnology 3, 206 (2008). 63, 1857 (1989).
[6] C. R. Dean, A. F. Young, I. Meric, C. Lee, L. Wang, [32] S. Milovanović, M. Ramezani Masir, and F. Peeters,
S. Sorgenfrei, K. Watanabe, T. Taniguchi, P. Kim, Applied Physics Letters 103, 233502 (2013).
K. Shepard, et al., Nature nanotechnology 5, 722 (2010). [33] S. Milovanović, M. Ramezani Masir, and F. Peeters,
[7] S. Morozov, K. Novoselov, M. Katsnelson, F. Schedin, Journal of Applied Physics 115, 043719 (2014).
D. Elias, J. A. Jaszczak, and A. Geim, Physical review [34] C. Rutherglen, D. Jain, and P. Burke, Nature Nanotech-
letters 100, 016602 (2008). nology 4, 811 (2009).
[8] Y. Wu, K. A. Jenkins, A. Valdes-Garcia, D. B. Farmer, [35] Y.-M. Lin, K. Jenkins, D. Farmer, A. Valdes-Garcia,
Y. Zhu, A. A. Bol, C. Dimitrakopoulos, W. Zhu, F. Xia, P. Avouris, C.-Y. Sung, H.-Y. Chiu, and B. Ek, in Elec-
P. Avouris, et al., Nano letters 12, 3062 (2012). tron Devices Meeting (IEDM), 2009 IEEE International
[9] R. Cheng, J. Bai, L. Liao, H. Zhou, Y. Chen, L. Liu, Y.- (IEEE, 2009) pp. 1–4.
C. Lin, S. Jiang, Y. Huang, and X. Duan, Proceedings [36] Y.-M. Lin, C. Dimitrakopoulos, K. A. Jenkins, D. B.
of the National Academy of Sciences 109, 11588 (2012). Farmer, H.-Y. Chiu, A. Grill, and P. Avouris, Science
[10] Z. Guo, R. Dong, P. S. Chakraborty, N. Lourenco, 327, 662 (2010).
J. Palmer, Y. Hu, M. Ruan, J. Hankinson, J. Kunc, J. D. [37] I. Meric, C. R. Dean, S.-J. Han, L. Wang, K. A. Jenkins,
Cressler, et al., Nano letters 13, 942 (2013). J. Hone, and K. Shepard, in Electron Devices Meeting
[11] Z. H. Ni, T. Yu, Y. H. Lu, Y. Y. Wang, Y. P. Feng, and (IEDM), 2011 IEEE International (IEEE, 2011) pp. 2–1.
Z. X. Shen, ACS nano 2, 2301 (2008). [38] J. G. Champlain, Solid-State Electronics 67, 53 (2012).
[12] M. Y. Han, B. Özyilmaz, Y. Zhang, and P. Kim, Physical
review letters 98, 206805 (2007).
[13] V. Perebeinos and P. Avouris, Physical Review B 81,
195442 (2010).
[14] I. Meric, N. Baklitskaya, P. Kim, and K. L. Shepard,
in Electron Devices Meeting, 2008. IEDM 2008. IEEE
International (IEEE, 2008) pp. 1–4.
[15] R. N. Sajjad and A. W. Ghosh, Applied Physics Letters
99, 123101 (2011).
[16] R. N. Sajjad and A. W. Ghosh, ACS nano 7, 9808 (2013).
[17] M. Katsnelson, K. Novoselov, and A. Geim, Nature
physics 2, 620 (2006).
[18] T. Low, S. Hong, J. Appenzeller, S. Datta, and M. S.
Lundstrom, IEEE Transactions on Electron Devices 56,
1292 (2009).
[19] R. N. Sajjad, S. Sutar, J. Lee, and A. W. Ghosh, Physical
Review B 86, 155412 (2012).
[20] M. S. Jang, H. Kim, Y.-W. Son, H. A. Atwater, and
W. A. Goddard, Proceedings of the National Academy
of Sciences 110, 8786 (2013).
[21] Q. Wilmart, S. Berrada, D. Torrin, V. H. Nguyen,
G. Fève, J.-M. Berroir, P. Dollfus, and B. Plaçais, 2D
Materials 1, 011006 (2014).
[22] S. Chen, Z. Han, M. M. Elahi, K. M. Habib, L. Wang,
B. Wen, Y. Gao, T. Taniguchi, K. Watanabe, J. Hone,
et al., Science 353, 1522 (2016).
[23] M. M. Elahi and A. W. Ghosh, in 74th Device Research
Conference (IEEE, 2016) pp. 1–2.
[24] V. V. Cheianov and V. I. Falko, Physical review b 74,
041403 (2006).
[25] J. Cayssol, B. Huard, and D. Goldhaber-Gordon, Phys-
ical Review B 79, 075428 (2009).
[26] K. Brenner and R. Murali, Applied Physics Letters 96,
063104 (2010).
[27] B. Tang, H. Guoxin, and H. Gao, Applied Spectroscopy
Reviews 45, 369 (2010).
[28] L. Liao, Y.-C. Lin, M. Bao, R. Cheng, J. Bai, Y. Liu,
Y. Qu, K. L. Wang, Y. Huang, and X. Duan, Nature
467, 305 (2010).
[29] R. N. Sajjad, F. Tseng, K. M. Habib, and A. W. Ghosh,
Physical Review B 92, 205408 (2015).
[30] R. N. Sajjad, C. A. Polanco, and A. W. Ghosh, Journal
of Computational Electronics 12, 232 (2013).