1494 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 61, NO.

5, MAY 2014

A 2-D Analytical Model for Double-Gate

Tunnel FETs
Mahdi Gholizadeh and Seyed Ebrahim Hosseini

Abstract— This paper presents a 2-D analytic potential model

for double-gate (DG) tunnel field effect transistors (TFETs) by
solving the 2-D Poisson’s equation. From the potential profile, the
electric field is derived and then the drain current expression is
extracted by analytically integrating the band-to-band tunneling
generation rate over the tunneling region. The model well predicts
the potential, subthreshold swing (SS), and transfer and output
characteristics of DG TFETs. We analyze the dependence of the
tunneling current on the device parameters by varying the gate
oxide dielectric constant, gate oxide thickness, body thickness,
channel length and channel material and also demonstrate its
agreement with TCAD simulation results. The SS which describes
the switching behavior of TFETs, is derived from the current
expression. The comparisons show that the SS of our model well Fig. 1. (a) Schematic diagram of a symmetric DG n-TFET with x-y
coincides with that of simulations. coordinates. (b) Energy band profile of an n-TFET in the OFF and ON state.
E is the energy pass window between E V,SOURCE and E C,CHANNEL .
Index Terms— Analytical model, band-to-band tunneling
(BTBT), BTBT generation rate, double-gate (DG) tunnel field
effect transistor (TFET), electric field, mobile charge, Poisson’s
equation, subthreshold swing (SS).
The performance of TFETs has been simulated using 2-D
TCAD simulations [1]–[6]. However, a physics-based analyt-
I. I NTRODUCTION ical model is essential for better understanding of the device
operation and facilitates compact modeling for circuit-level
I T HAS been highlighted that increasing power density is a
challenge for continued MOSFET scaling, due to nonscal-
ability of subthreshold swing (SS). The SS of a MOSFET is
studies. It is also useful to obtain fast results.
Several analytical studies on TFETs have been carried out
in [7]–[19]. Some 1-D analytical models without considering
limited to 60 mV/decade, which causes the leakage current to
the impact of the drain voltage have derived the drain current
increase. One of the promising devices to replace the MOSFET
[7]–[9]. Many 2-D studies on TFET modeling analytically
for lowpower applications is the tunnel field effect transistor
calculate the tunneling generation rate using 2-D electric
(TFET), which has demonstrated the potential to surmount
field, but the tunneling current is computed by numerically
the SS limit of MOSFETs [1]–[5]. The operation principle of
integrating over the volume of the device [10]–[13]. In some
the TFETs relies on the band-to-band tunneling (BTBT) of
pseudo-2-D analytical models [14], [15], the tunneling current
electrons, so that they are able to operate as steeper switches
has analytically been derived. They have assumed that the
at lower supply voltages [6].
electric field is constant over both the tunneling distance (along
The cross-sectional view of an n-type double-gate (DG)
the channel) and depth of the device (perpendicular to the
TFET is shown in Fig. 1(a). The energy band profile in the
channel) in their models, whereas simulations and different
OFF and ON -state is shown in Fig. 1(b). When a positive gate
analytical models such as [10] and [12] demonstrate that the
voltage is applied, the conduction band of the channel goes
distribution of the electric field is nonuniform in the channel.
down and a sufficiently high lateral electric field is created
In a different method in [16], Landauer approach is used to
at the source-channel junction. This electric field forces the
derive the dc characteristics of TFETs. In this model, the
electrons to tunnel from the occupied valence-band states of
width of the energy pass window E [Fig. 1(b)] is considered
the source to the unoccupied conduction-band states of the
to have a linear relationship with the gate voltage, but in
channel.
fact increasing the gate voltage leads to a rise in the voltage
Manuscript received May 11, 2013; revised December 19, 2013, February 4, drop across the gate oxide. Therefore, increasing of the gate
2014, and March 12, 2014; accepted March 18, 2014. Date of current version voltage leads to less increase in the width of the energy pass
April 18, 2014. The review of this paper was arranged by Editor M. Ieong.
The authors are with the Department of Electrical Engineering, window (E).
Ferdowsi University of Mashhad, Mashhad 91779-48974, Iran (e-mail: In this paper, we develop a 2-D analytical model to derive
mahdi.gholizadeh@stu.um.ac.ir; ehosseini@um.ac.ir). analytical expressions for different electrical parameters of
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. DG TFETs i.e., potential profile, SS, and transfer and output
Digital Object Identifier 10.1109/TED.2014.2313037 characteristics. In our calculations, the influences of the mobile
0018-9383 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
GHOLIZADEH AND HOSSEINI: 2-D ANALYTICAL MODEL FOR DG TUNNEL FETs 1495

charges on the potential profile and the drain bias on the ψ1 (x, y) is the solution to the 2-D Laplace equation
current are considered. The model also predicts the impacts
∂ 2 ψ1 ∂ 2 ψ1
of structural parameters which is useful to provide a design + = 0· (6)
insight. This paper is organized as follows. In Section II the ∂x 2 ∂y 2
potential profile and lateral electric field are derived using The solution of the Laplace equation is given by
fully 2-D solution of the Poisson’s equation. In Section III
ψ1 (x, y) = u L (x, y) + u R (x, y)· (7)
using Kane’s model, an analytical expression for the current
∞
is extracted by integrating the BTBT generation rate over the where u R  and u L can be written as u R = n=1 u Rn (x, y)
tunneling region. Then, from the current expression, the SS is and u L = ∞ n=1 u Ln (x, y) where u Rn (x, y) and u Ln (x, y) are
derived. The analytic model is validated by comparing it with Eigen functions that are obtained as follows [23]:
TCAD simulation results for different sets of parameters in  
sinh(π y/λn ) nπ πx
Section IV. u Rn (x, y) = cn sin + (8)
sinh(π L/λn ) 2 λ
 n 
sinh (π(L − y)/λn ) nπ πx
II. 2-D P OISSON ’ S E QUATION S OLUTION u Ln (x, y) = bn sin + · (9)
sinh(π L/λn ) 2 λn
In this section, an accurate solution to the Poisson’s equation where λn are Eigen values which are obtained from
with considering the mobile charge term is presented. The    
structure of the DG TFET is shown in Fig. 1(a). The Poisson’s π Ti nπ π Tsi
εsi tan = εi tan − · (10)
equation in the channel is written as λn 2 2λn
  According to [23], we obtain the first-order coefficients as
∂ 2ψ ∂ 2ψ q ψ −V
+ = n i exp (1)
∂x2 ∂y 2 εsi Vt 2λ21 tan(π Ti /λ1 ) sin(π Tsi /2λ1 )
b1 =   (ϕSC − VGS + φ)
Tsi sin(π Tsi /λ1 )
where ψ(x, y) is the electrostatic potential in the intrinsic π Ti
2 + Ti
channel, εsi is the permittivity of the silicon, n i is the intrinsic 2 sin(2π Ti /λ1 )
carrier density, Vt is the thermal voltage and V is the elec- (11)
tron quasi-Fermi potential. According to [2], the electrostatic sin(π Tsi /2λ1 )
c1 =  
potential using superposition principle will be written as Tsi sin(π Tsi /λ1 )
+ Ti
2 sin(2π Ti /λ1 )
ψ(x, y) = υ(x) + ψ1 (x, y) (2) 

−4Vt λ1 2λ21 υ T2si − ϕDC
× ln(cos(βd )) − 2 tan(π Ti /λ1 )
where v(x) is the solution of 1-D Poisson’s equation π π Ti
  (12)
∂ 2υ q υ−V
= n i exp · (3)
∂x2 εsi Vt where ϕSC and ϕDC are built-in potentials at the source and
As described in [21], v(x) can be obtained by twice drain junctions, respectively. For a given VGS , βd is obtained
integrating (3) from (5). For symmetric DG structures, even order coefficients
    are zero [23]. The expression for the potential in the channel
Tsi qn i 2β considering the first-order Eigen function is
υ(x) = V − 2Vt ln cos x . (4)
2β 2εsi Vt Tsi
ψ(x, y) = υ(x) + cos(π x/λ1 )
 
For a given VGS , β can be solved from b1 sinh (π(L − y)/λ1 ) + c1 sinh(π y/λ1 )
× .
   sinh(π L/λ1 )
VGS − φ − V 2 2εsi Vt (13)
− ln
2Vt Tsi qn i
Fig. 2 demonstrates the surface potential under different gate
2εsi Ti
= ln β − ln(cos(β)) − β tan(β) (5) voltages. At high VGS due to the inversion of the channel, the
εi Tsi potential is pinned to the drain voltage. As Fig. 2 shows, the
where φ is work function difference between the gate agreement of the model and simulation can be easily seen and
electrode and the semiconductor and εi is the permittivity the model well captures the effect of the mobile charges.
of the insulator. Since the current flows mainly along the The lateral electric field E y (x, y) is found by differentiating
channel from the source to the drain, the electron quasi- the electrostatic potential expression
Fermi potential is almost constant in x-direction and varies 
π cosh (π(L − y)/λ1 )
only in y-direction [21]. For DG-MOSFETs, V is assumed E y (x, y) = − b1
λ1 sinh(π L/λ1 )
to be constant in the channel direction except at the end of 
π cosh (π y/λ1 )
channel where it reaches VDS [22]. Similar to MOSFETs, + c1 × cos (π x/λ1 ) · (14)
TFETs simulations demonstrate that V in the channel length λ1 sinh(π L/λ1 )
direction stays constant (equal to VDS ) except at the beginning As mentioned earlier, v(x) almost stays constant in the
of the channel. This causes v(x) to be nearly constant in the y-direction, therefore its derivative is zero. Indeed, for well-
y-direction except near the source junction. scaled devices, the dominant tunneling paths are lateral.
1496 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 61, NO. 5, MAY 2014

and substituting (16) and (17) into (15), we obtain

E y (x, y) = (k1 − k3 + k5 − · · ·)

x2
+ −k1 (π/λ1 )2 + k3 (π/λ3 )2 − k5 (π/λ5 )2 + · · ·
2!

x 4
+ k1 (π/λ1 )4 − k3 (π/λ3 )4 + k5 (π/λ5 )4 − · · · + ···
4!
(18)

Solving (10) results in λ1 /λn ∼ n [23] Therefore (18) becomes

  2 2
π x
E y (x, y) = (k1 − k3 + k5 − · · ·) × 1 + ×
Fig. 2. Surface potential obtained from the model (symbols) and simulations λ1 2!
(lines) at different gate voltages. The threshold voltage VTH according to the    
definition in [16] is 0.6 V. (εsi = 11.8 ε0 , φ = 0, ϕSC = −0.6 V, and −k1 + 9k3 − 25k5 + · · · k1 − 81k3 + 625k5 − · · ·
ϕDC = 0.6 V). +
k1 − k3 + k5 − · · · k1 − k3 + k5 − · · ·
 4 4 
π x
× + · · · · (19)
Therefore, it is reasonable to assume tunneling takes place λ1 4!
primarily along the channel length direction, as discussed in
[11] and [14]. Therefore, the lateral electric field has only Using a proper fitting parameter N, (19) can be approximated
been taken into account in the calculations. The lateral electric as the following expression:
field shows a nonuniform behavior along the source-channel  
ω12 x 2 ω24 x 4
junction. It reaches its maximum value at the channel surface E y (x, y) = Nk1 1+ + +· · · ∼= Nk1 cosh(ω1 x)
and gradually decreases to its minimum value at the middle 2! 4!
of channel. This is because the channel underneath the gate is (20)
more influenced by the gate voltage than depths of the channel.
On the contrary, the highest value of E y (x, y) in (14) is at the where ω1 can be approximated as ω1 = π/λ1 Therefore the
middle of the channel. The reason behind this disagreement is lateral electric field is expressed as
due to considering only the first-order Eigenfunction. Although 
π cosh (π(L − y)/λ1 )
the first-order Eigenfunction works well to describe the chan- E y (x, y) = N − b1
λ1 sinh(π L/λ1 )
nel potential, as shown in Fig. 2, the higher order terms play   
π cosh (π y/λ1 ) π
a more important role on capturing a sharper potential profile + c1 cosh x · (21)
at the source-channel junction, which strongly influence the λ1 sinh(π L/λ1 ) λ1
electric field profile [11]. On the other hand, considering the As (21) indicates getting closer to the channel surface, the
higher order terms extremely complicates the analytic solution lateral electric field increases.
and it is difficult to derive a straightforward expression for the
current. Therefore, there is a tradeoff between the precision
and simplicity. In order to overcome the tradeoff, we apply III. D RAIN C URRENT D ERIVATION
the impacts of the higher order terms to the first-order term In this section, we obtain analytical expressions for the drain
through Taylor series expansion. current and SS. The most widely used model to calculate the
Based on the potential solution, the full expression for the tunneling current is the Kane model [24], which determines
lateral electric field is written as the BTBT generation rate of carrier tunneling from the valence
∞  band of the source to the conduction band of the channel, as
π cosh (π(L − y)/λ1 )
E y (x, y) = − bn  
λn sinh(π L/λn ) B
n=1
   G BTBT = AE exp −
D
· (22)
π cosh (π y/λ1 ) nπ πx E
+ cn sin + ·
λn sinh(π L/λn ) 2 λn where E is the electric field; D is 2.5 for the indirect
(15) and 2 for the direct tunneling processes; A and B are the
tunneling process-dependent parameters. This paper develops
Because of vanishing even order coefficients, sin(nπ/2 + an analytical model for indirect tunneling process; however,
π x/λn ) is equivalent to ±cos(π x/λn ). The term cos(π x/λn ) it can be easily extended to the direct tunneling process. The
can be expanded using Taylor series expansion as drain current can be computed as [18]
(π x/λn )2 (π x/λn )4 
cos(π x/λn ) = 1 − + − ··· (16)

2! 4! I = q AE E avg D−1
exp −B/E avg d V (23)
Defining kn as
where E is the local electric field and E avg is the average
π cosh (π(L − y)/λn ) π cosh (π y/λn ) electric field. In this paper we aim at analytically calculating
k n = − bn + cn (17)
λn sinh(π L/λn ) λn sinh(π L/λn ) the tunneling current.
GHOLIZADEH AND HOSSEINI: 2-D ANALYTICAL MODEL FOR DG TUNNEL FETs 1497

Due to exponential term in (28) and also y2 > y1 , it is

straightforward to show S(y1 )  S(y2 ). Ignoring S(y2 ), it is
sufficient to compute y1 . At y1, the channel potential reaches
(E g + E b )/q. Computing y1 requires inverting potential
distribution equation. Since the charge concentration near
the source-channel junction is negligible [5], the potential
distribution along the x-axis can be written as v(x) = VGS −
ϕ. Inserting this value in the potential distribution equation
and ignoring the second term (c1 sinh(π y/λ1)) due its small
contribution at the source-channel junction, y1 can be obtained
as
⎛ ⎞
Fig. 3. Energy band diagram along the y-direction during ON-state. l1 and E g+E b
λ1 +ψ(x, 0) − VGS + φ + Veff
sinh−1 ⎝ ⎠
l2 are the lengths of the shortest and longest tunnel paths, respectively. q
y1 = L −    
π π Tsi
b1 cos 2λ1 πL
sinh λ1
E is computed from (21). Since electron tunneling takes place (29)
mainly at the source-channel junction, the second term in (21)
is negligible and thereby E y is written as where Veff takes care of the error introduced with deriving
(29), 0 < Veff < 0.15.
E y (x, y)
    The effects of the drain bias appear at the higher order
πb1 π terms [11]; furthermore, we do not completely consider the
=N − exp (π(L − y)/λ1 ) exp x ·
4λ1 sinh(π L/λ1 ) λ1 impacts of the highorder harmonics in the calculations. Thus,
(24) a correction factor (F) is needed to compensate the impact of
The average electric field can be written as the drain bias. We employ the correction factor used in [15],
but the difference here is that we add VGS in the denominator
E avg = E g /(qlpath ) (25) of the exponential term in order to avoid the effect of the drain
where lpath is the length of tunneling path and varies from bias at low VGS
l1 to l2 [Fig. (3)]. Due to high doping concentration, the source 2
depletion region is ignored, therefore l1 = y1 and l2 = y2 . At F =1−    . (30)
y = y1, the difference between the source conduction band 1 + exp VVDS
GS
and channel conduction band is E g + E b , where E b =
E VS − E FS . Similarly, at y = y2 , the difference between the Note that when VGS tends to be zero, F tends to unity.
source conduction band and channel conduction band reaches Subthreshold swing is an important parameter to describe
E g + ϕ, where ϕ = E VS − E VC . the switching behavior of transistors. From the current expres-
Inserting the local and average electric fields in (23) yields sion it can be expressed as
Tsi  −1
 2 y2   d log(I )
πb1 exp (π L/λ1 ) E gD−1 SS = . (31)
I =q AN − d VGS
4λ1 sinh(π L/λ1 ) q D−1 y D−1
−Tsi y1
2

   By rewriting (11) as b1 = h(ϕSC − VGS + ϕ) and neglecting
π −q B F because of its small variations at low gate voltages, the
exp (−π y/λ1 ) exp x exp y d yd x· (26)
λ1 Eg SS is obtained as
The channel width is 1 μm in the model and simulations.    
1 B π 1.5 d L 1 −1
In (26), y ranges from y1 to y2 , in this interval variation of SS = ln 10 − + +
VGS −φ − ϕSC E g λ1 L 1 d VGS
the exponential term is dominant compared with polynomial
term (1/y D−1) [18]. Therefore, we perform integration over (32)
exponential term to compute the drain current
  where d L 1 /d VGS is obtained as
q AN E gD−1 πb1 exp (π L/λ1 )  
I = −  − πL  
qB π 4λ1 sinh(π L/λ1 ) d L1 λ 1 sinh λ1 E g + E b + Veff
+ q D−1 =−  
Eg λ1 (ϕSC + φ − VGS )2
     d VGS πh cos π Tsi
2λ1 Tsi π 2λ1
exp −1 (S(y2 ) − S(y1 )) (27) 1
π 2λ1 × (33)
 2
where S(y) is defined as π(L−L 1 )
    1 + sinh λ1 .
qB π
exp − + y
Eg λ1 It is observed that unlike MOSFETs, the SS of TFETs is a
S= . (28)
y D−1 function of the gate voltage.
1498 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 61, NO. 5, MAY 2014

Fig. 4. I –VGS characteristic. The drain current is plotted on logarithmic Fig. 6. I –VGS characteristics for Si DG TFETs as function of gate oxide
(left) and linear (right) scales (L = 50 nm, N = 0.1, and φ = −0.5). thickness. The values of N used for comparisons are 0.1 for Ti = 1.5 nm,
0.1 for Ti = 2 nm and 4 for Ti = 3 nm (L = 50 nm and φ = −0.6).

Fig. 5. I –VDS characteristics of the DG TFET at different gate voltages Fig. 7. I –VGS characteristics for Si DG TFETs as function of gate oxide
(L = 50 nm, N = 0.1, and φ = −0.5 V). dielectric. The values of N used for comparisons are 4 for εi = 3.9ε0 , 4 for
εi = 6ε0 , 1 for εi = 8.3ε0 (L = 50 nm and φ = −0.7 V).

IV. M ODEL VALIDATION

In this section we validate the proposed model by comparing
it with 2-D TCAD simulations. We evaluate the efficacy of the
model by varying the gate oxide dielectric constant, gate oxide
thickness, body thickness channel length and channel mater-
ial. We employed the nonlocal tunneling model to simulate
TFET current. The source and drain doping concentrations
are 1020 cm−3 and 1018 cm−3 , respectively, and the channel
is intrinsic. The default parameters for the Kane model are
A = 4 × 1014 cm−1/2 V−5/2 s−1 and B = 1.9×107 V/cm [25].
Fig. 4 shows the I –VGS characteristic of a DG TFET calcu-
lated by the model and simulations. It is observed that our
model well captures both the subthresold and superthreshold
Fig. 8. I –VGS characteristics for Si DG TFETs as function of body thickness.
currents. Fig. 5 shows the output characteristic of the device The values of N used for comparisons are 1 for Tsi = 8 nm, 1 for Tsi = 10 nm
at different gate voltages. The model predicts the saturation and 0.1 for Tsi = 12 nm (L = 50 nm and φ = −0.6 V).
current very well and also it shows a qualitative agreement in
the linear regime. decreasing the body thickness [1], as shown in Fig. 8. We also
In order to improve the ON-current (ION ) and achieve high examined the gate length scaling with the analytical model.
ON to OFF current ratios (I ON /I OFF ) in TFETs, gate oxide The DG TFETs exhibit extremely lower shift in their I –VGS
thickness reduction and high-k dielectric materials are usually characteristics, due to short channel effects, than their DG
employed. We investigate the model with different gate oxide MOSFETs counterparts at the gate lengths longer than four
thicknesses and gate oxide dielectrics in Figs. 6 and 7. With times their natural scaling lengths (4λ1 /π) [11]. In Fig. 9 the
reducing Ti and increasing εi , higher ION , steeper subthresh- drain current for two different structures at the gate lengths
old slope, and higher ION /IOFF is achieved [1]. Our model longer than their 4λ1 /π have been plotted and the accuracy
well predicts the changes of I –VGS characteristics induced of our model can be easily seen.
by varying Ti and εi . The reduction of the bulk capacitive Although in this model we used silicon as the channel
effects also can increase the current which can be obtained by material in calculations, we can also apply the model for
GHOLIZADEH AND HOSSEINI: 2-D ANALYTICAL MODEL FOR DG TUNNEL FETs 1499

V. C ONCLUSION
In this paper, a 2-D analytical model for DG TFETs has
been developed. This model takes into account the influences
of all the structural parameters, i.e., Tsi , Ti , εi , εsi , and L
together with the biases in the calculations and predicts well
the effects of them. We included the mobile charge term in
the solution of the Poisson’s equation. Comparing the model
results for different electrical parameters, i.e., potential profile,
BTBT generation rate, SS, I –VGS and I –VDS characteristics,
with the simulation results shows a good agreement.
R EFERENCES
Fig. 9. I –VGS characteristics for various gate lengths. DG TFETs exhibit [1] K. Boucart and A. M. Ionescu, “Double-gate tunnel FET with high-
lower shift in their I –VGS characteristics at gate lengths longer than κ gate dielectric,” IEEE Trans. Electron Devices, vol. 54, no. 7,
(4λ1 /π )N = 1 for both plots (φ = −0.6 V). pp. 1725–1733, Jul. 2007.
[2] C.-H. Shih and N. D. Chien, “Sub-10-nm tunnel field-effect transistor
with graded Si/Geheterojunction,” IEEE Electron Device Lett., vol. 32,
no. 11, pp. 1498–1500, Nov. 2011.
[3] L. Lattanzio, L. D. Micheilis, and A. M. Ionescu, “Complementary
germanium electron-hole bilayer tunnel FET for sub-0.5-V operation,”
IEEE Electron Device Lett., vol. 33, no. 2, pp. 167–169, Feb. 2012.
[4] A. Mallik and A. Chattopadhyay, “Tunnel field-effect transistors for
analog/mixed-signal system-on-chip applications,” IEEE Trans. Electron
Devices, vol. 59, no. 4, pp. 888–894, Apr. 2012.
[5] A. Mallik and A. Chattopadhyay, “Drain-dependence of tunnel field-
effect transistor characteristics: The role of the channel,” IEEE Trans.
Electron Devices, vol. 58, no. 12, pp. 4250–4257, Dec. 2011.
[6] S. H. Kim, S. Agarwal, Z. A. Jacobson, P. Mathue, C. Hu, and
T. J. K. Liu, “Tunnel field-effect transistor with raised germanium
source,” IEEE Electron Device Lett., vol. 31, no. 10, pp. 1107–1109,
Oct. 2010.
[7] N. N. Mojumder and K. Roy, “Band-to-band tunneling ballistic nanowire
FET: Circuit-compatible device modeling and design of ultra-low-power
digital circuits and memories,” IEEE Trans. Electron Devices, vol. 56,
no. 10, pp. 2193–2201, Oct. 2009.
Fig. 10. I –VGS characteristic for different semiconductor materials. InAs [8] W. Vandenberghe, A. Verhulst, G. Groeseneken, B. Soree, and
(φ = −0.125 V, N = 10), Si0.7 Ge0.3 (φ = −0.09 V and N = 10), and W. Magnus, “Analytical model for a tunnel field-effect transistor,” in
Si (φ = −0.7 V and N = 4). Tsemi is the body thickness. Proc. MELECON, 2008, pp. 923–928.
[9] M. M. M. Elahi, K. Ahmed, and M. S. Islam, “An improved analytical
model of current in tunnel field effect transistor,” in Proc. ICECE, 2012,
TABLE I pp. 802–805.
S UBTHRESHOLD S WING VALUES AT D IFFERENT G ATE V OLTAGES [10] M. G. Bardon, H. P. Neves, R. Puers, and C. V. Hoof, “Pseudo-
FOR T WO D IFFERENT S TRUCTURES two dimensional model for double-gate tunnel FETs considering the
junctions depletion regions,” IEEE Trans. Electron Devices, vol. 57,
no. 4, pp. 827–834, Apr. 2010.
[11] L. Liu, D. Mohata, and S. Datta, “Scaling length theory of double-gate
interband tunnel field-effect transistors,” IEEE Trans. Electron Devices,
vol. 57, no. 4, pp. 827–834, Apr. 2012.
[12] M. J. Lee and W. Y. Choi, “Analytical model of single-gate silicon on
insulator (SOI) tunneling field effect transistors (TFETs),” Solid State
Electron., vol. 63, no. 1, pp. 110–114, Sep. 2011.
[13] A. Pan, S. Chen, and C. O. Chui, “Electrostatic modeling and insights
regarding multigate lateral tunneling transistors,” IEEE Trans. Electron
Devices, vol. 60, no. 9, pp. 2712–2720, Sep. 2013.
[14] A. Pan and C. O. Chui, “A quasi-analytical model for double-gate
tunneling field-effect transistors,” IEEE Electron Device Lett., vol. 33,
no. 10, pp. 1468–1470, Oct. 2012.
other materials such as III–V semiconductors. We used InAs [15] L. Zhang, J. He, and M. Chan, “A compact model for double-gate tun-
neling field-effect transistors and its implications on circuit behaviors,”
and Si0.7 Ge0.3 , which have direct and indirect transitions in Proc. IEDM, 2012, pp. 143–146.
respectively, as alternatives to Si in the model. Fig. 10 shows [16] B. Bhushan, K. Nayak, and V. R. Rao, “DC compact model for SOI
that the model can be apply for different direct and indirect- tunnel field-effect transistors,” IEEE Trans. Electron Devices, vol. 59,
no. 10, pp. 2635–2642, Oct. 2012.
bandgap semiconductor materials. When applying the model [17] L. Zhang, X. Lin, J. He, and M. Chan, “An analytical charge model
for direct-bandgap materials the values of D, A, and B are for double-gate tunnel FETs,” IEEE Trans. Electron Devices, vol. 59,
adjusted. no. 12, pp. 3217–3223, Dec. 2012.
[18] A. S. Verhulst, D. Leonelli, R. Rooyackers, and G. Groeseneken, “Drain-
Table I lists the SS calculated by the model and simulations voltage dependent analytical model of tunnel field-effect transistors,”
for two different structures. The model well predicts the low J. Appl. Phys., vol. 110, no. 2, pp. 024510-1–024510-10, Jul. 2011.
subthreshold values (>60 mV/decade) observed in TFETs as [19] A. S. Verhulst, B. Soree, D. Leonelli, W. G. Vandenberghe, and
G. Groeseneken, “Modeling the single-gate, double-gate, and gate-all-
well as the dependence of the SS on the gate voltage and around tunnel field effect transistor,” J. Appl. Phys., vol. 107, no. 2,
shows a good match with simulation results. pp. 024518-1–024518-8, Jul. 2010.
1500 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 61, NO. 5, MAY 2014

[20] Y. Taur and T. H. Ning, Fundamentals of Modern VLSI Devices. Seyed Ebrahim Hosseini received the B.Sc. degree
Cambridge, U.K.: Cambridge Univ. Press, 1998. in electrical engineering from the Isfahan Univer-
[21] Y. Taur, X. Liang, W. Wang, and H. Lu, “A continuous, analytic drain- sity of Technology, Isfahan, Iran, the M.Sc. degree
current model for DG MOSFETs,” IEEE Electron Device Lett., vol. 25, from Tarbiat Modarres University, Tehran, Iran, and
no. 2, pp. 107–109, Feb. 2004. the Ph.D. degree in electrical engineering from the
[22] Q. Chen, E. M. Harrell, and J. D. Meindl, “A physical short- Sharif University of Technology, Tehran, in 2001.
channel threshold voltage model for undoped symmetric double- He is currently an Associate Professor of Elec-
gate MOSFETs,” IEEE Trans. Electron Devices, vol. 50, no. 7, tronics Engineering with the Ferdowsi University of
pp. 1631–1637, Jul. 2003. Mashhad, Mashhad, Iran.
[23] X. P. Liang and Y. Taur, “A 2-D analytical solution for SCEs
in DG MOSFETs,” IEEE Trans. Electron Devices, vol. 51, no. 9,
pp. 1385–1391, Sep. 2004.
[24] E. O. Kane, “Zener tunneling in semiconductors,” J. Phys. Chem. Solids,
vol. 12, no. 2, pp. 181–188, Jan. 1960.
[25] ATLAS User’s Manual, Silvaco Inc., Santa Clara, CA, USA, 2010.

Mahdi Gholizadeh received the B.Sc. degree in

electrical engineering from the Shahid Chamran
University of Ahvaz, Ahvaz, Iran, and the M.Sc.
degree in electrical and electronic engineering from
the Ferdowsi University of Mashhad, Mashhad, Iran,
in 2011 and 2014, respectively.
His current research interests include design, sim-
ulation, fabrication, and characterization of ultralow-
power devices.