Você está na página 1de 4

Journal of the Korean Physical Society, Vol. 45, December 2004, pp.

S547S550

Quantum Mechanical Simulation of Charge Distribution in Schottky Barrier


MOSFETs
Mincheol Shin
School of Engineering, Information and Communications University, Daejeon 305-714

Moongyu Jang and Seongjae Lee


Nano-Electronics Device Team, ETRI, Daejeon 305-330
We have performed numerical simulations on Schottky barrier (SB) MOSFETs by solving the twodimensional Poisson equation self-consistently with the Schr
odinger equation. We have investigated
the equilibrium charge distribution and the conduction energy bending in the inversion layer of the
SB-MOSFETs. We have found that the quantum mechanical consideration is crucial for accurate
estimation of the current in nano-scale SB-MOSFETs. We have also found that the threshold
voltage for channel inversion increases when the channel length becomes shorter than about 30 nm,
in contrast to conventional MOSFETs.

PACS numbers: 85.30.T, 85.30.H


Keywords: Schottky barrier MOSFET, Device simulation, Quantum effect

I. INTRODUCTION
A Schottky barrier metal-oxide-semiconductor fieldeffect transistor (SB-MOSFET) has source and drain regions composed of silicide rather than heavily doped silicon. The device offers several important advantages over
a conventional MOSFET at nanometer scale, such as
ease of ultra-shallow junction formation, the possibility
of using metal as gate electrode, and elimination of complicated channel doping steps. Therefore, SB-MOSFETs
have been proposed as an alternative to the conventional
MOSFETs for decananometer-scale applications [13].
Most of the previous numerical simulations on SBMOSFETs have been carried out in the classical framework. The feasibility of SB-MOSFETs in reducing the
short-channel effect has been studied by using a commercial simulator in the work of Ref. 1. The Poisson
and continuity equations were solved self-consistently in
Ref. 4 for relatively long channel. By using a similar
approach, Huang et al. investigated the transistor characteristics of SB-MOSFETs with channel length down
to 10 nm [5]. Monte Carlo simulations with modeling
on Schottky barrier tunneling currents have also been
reported [6, 7]. Recently, a compact model using the
current continuity condition between the tunneling and
channel currents has been proposed [8]. On the other
hand, double-gate, thin body SB-MOSFETs were simulated by solving the two-dimensional (2D) Poisson equa E-mail:

mcshin@icu.ac.kr

tion self-consistently with the Schrodinger equation [9].


In this work, SB-MOSFETs are investigated by solving the 2D Poisson equation self-consistently with the
Schr
odinger equation. Our approach is similar to that
of Guo et al. [9], except that we simulated single-gate
SB-MOSFET in the equilibrium state. We have focused
on the equilibrium charge distributions in the inversion
layer : the charge distributions were compared in the
classical and quantum cases and the dependence of the
channel threshold voltages on channel length and doping
concentration was investigated.

II. SIMULATION APPROACH

Fig. 1. SB-MOSFET. The thick solid lines represent the


boundaries of the simulated device.
-S547-

-S548-

Journal of the Korean Physical Society, Vol. 45, December 2004

1. Poisson Equation

The 2D Poisson equation in the silicon region of the


SB-MOSFET shown in Figure 1 is given by
2 (x, z) =

q0
(p(x, z) n(x, z) NA ) ,
si

(1)

where (x, z) is the potential, q0 is the electron charge,


si is the dielectric constant of silicon, and NA is the
acceptor concentration. Classically, the charge densities
are given by
p(x, z) = p0 e(x,z)/T ,

(2)

n(x, z) = ncl (x, z) = n0 e(x,z)/T ,

(3)

where p0 is the bulk hole concentration (p0 =


ni e(Ei EF )/kB T ), ni is the intrinsic carrier concentration,
Ei is the intrinsic Fermi level, EF is the Fermi energy of
silicon, kB is Boltzmanns constant, T is the temperature, n0 is the bulk electron concentration (n0 = n2i /p0 ),
and T = kB T /q0 .
In the oxide region, the Poisson equation reduces to
the Laplace equation :
2 (x, z) = 0.

(4)

At the interface between the oxide and the silicon, the


relationship




ox
= si
(5)
z z=0
z z=0+

holds, where ox is the dielectric constant of gate oxide.

Solving Eqs. (6) and (7), we obtain quantum electron


density


X
EF El,i (x)
|l,i (x, z)|2
nq (x, z) =
Nl F0
kB T
i


X
EF Eh,i (x)
|h,i (x, z)|2 , (9)
+
Nh F0
k
T
B
i
where
Nl =

4 ml mt kB T
2mt kB T
,
N
=
,
h
~2
~2

(10)

and
F0 (x) = ln(1 + exp(x)).

(11)

In this work, the Poisson equation in Eq. (1) with the


electron density n(x, z) given by Eqs. (9)-(10) and the
Schr
odinger equations in Eqs. (6)-(8) were solved selfconsistently.

3. Numerical Methods

The boundary conditions used in solving the


Schr
odinger equations in Eqs. (6) and (7) are such that
wave functions vanish at the oxide-silicon interface and
at z = D, where D is the depth of the simulated device
in Figure 1. Namely,
(x, 0) = (x, D) = 0.

(12)

The boundary conditions used in solving the Poisson


equation in Eq. (1) are as follows. The Dirichlet boundary conditions were used for the source-silicon, drainsilicon and gate-oxide interfaces. Namely,

2. Schr
odinger Equation

The one-dimensional (1D) Schr


odinger equations in
the vertical direction (z-direction) were solved for each
x-position :


~2 2

+ q0 V (x, z) El,i l,i (x, z) = 0, (6)


2ml z 2


~2 2

+ q0 V (x, z) Eh,i h,i (x, z) = 0,(7)


2mt z 2
where l,i (h,i ) and El,i (Eh,i ) are the i-th wave function
and energy level in the lower (higher) subband ladder,
respectively, ml and mt are longitudinal and transverse
effective masses (ml =0.98 m0 and ml = 0.19 m0 , where
m0 is the free electron mass), and V (x, z) is the potential
profile defined as
V (x, z) = Ec /q0 (x, z),

(8)

where Ec is the conduction band energy in the silicon


bulk.

(0, z) = (L, z) = Vbi , for 0 < z < Dm ,


(x, tox ) = Vg , for 0 < x < L,

(13)
(14)

where L is the length of the channel, Dm is the depth


of the source and drain, tox is the oxide thickness, Vg is
the gate potential, and Vbi is the built-in potential at the
Schottky contacts, which is defined as
Vbi = (Ec EF )/q0 Bn ,

(15)

where Bn is the Schottky barrier height. The Neumann


boundary conditions

= 0,
~n

(16)

where ~n denotes the direction normal to the boundary


surface, were used for all the other boundaries [10].
The 1D Schr
odinger equations in Eqs. (6)-(8) with the
boundary conditions given by Eq. (12) were solved numerically by using the standard shooting method [11].
The Poisson equation in Eq. (1) was discretized by the
standard five-point finite difference method and solved

Quantum Mechanical Simulation of Charge Distribution Mincheol Shin and Moongyu Jang and Seongjae Lee

by using the Newton-Raphson method with the boundary conditions given by Eqs. (13)-(15). For inversion
of the resultant sparse matrix, the ILU-CG iteration
method provided by LASpack was employed [12].
The self-consistent calculations proceeded as follows.
We first solved the Poisson equation with an initial guess
for the potential (x, z). The Schr
odinger equation using
the potential obtained from solving the Poisson equation
was then solved to update the quantum electron density. The iteration between the Poisson equation and the
Schr
odinger equation was repeated until self-consistency
was achieved, namely, until the quantum charge density converged to a value within a pre-set error bound.
For better convergence of the self-consistent solution, the
Poisson equation modified for a predictor-corrector approach [13] was solved at the m-th iteration :

q0  (m)
2 (m) =
p
n
(m)
NA ,
(17)
q
si
(m)/T

where p(m) = p0 e
Eq. (9), except that

(m)

and n
q

-S549-

Fig. 2. Quantum electron concentration for NA = 1015


cm3 . The gate voltage was 1 V, L = 50 nm, Lm = 50 nm,
Dm = 50 nm, tox = 5 nm, and the Schottky barrier height
was Bn = 0.28 V. The unit for the electron concentration is
1017 cm3 .

is the same as nq of

EF EF + q0 ((m) (m1) ).

(18)

To save computational time in computing nq , the classical approximation was employed for energies above
EF + 12kB T , as described in Ref. 14.

III. RESULTS AND DISCUSSION


The quantum-mechanical equilibrium charge distributions in the inversion layer were obtained for various doping concentrations and channel lengths. Figures 2 and 3
show the quantum electron distributions near the oxide
interface for the doping concentrations NA = 1015 and
1018 cm3 , respectively. While the electron concentration profile for the former exhibits the peak at the center
of the channel as one may expect, the electron distribution for the latter case exhibits two symmetric peaks
close to source and drain contacts. We attribute this effect to the shortening of electron screening length as the
doping concentration increases.
Figures 4 and 5 show the conduction band energy at
the oxide-channel interface along the lateral (x) direction
for NA = 1015 and 1018 cm3 , respectively. For low gate
voltages, the conduction-band energy profiles from the
classical and quantum calculations are almost the same.
For higher gate voltages, however, the quantum results
deviate considerably from the classical results. In the
middle of the channel, especially, the difference between
the two becomes as much as 0.1 eV (in the strong inversion regime). More importantly, the lowering of the
conduction-band energy in the quantum case results in
the thinning of the Schottky barrier. Since the current
through the Schottky barrier is mainly due to the tunneling current which exponentially depends on the barrier thickness, the source-drain current is expected to

Fig. 3. Quantum electron concentration for NA = 1018


cm3 . The gate volt age was 2 V, L = 100 nm, Lm = 100
nm, Dm = 50 nm, tox = 5 nm, and the Schottky barrier height
was Bn = 0.28 V. The unit for the electron concentration is
1018 cm3 .

Fig. 4. Conduction band energy at the oxide-channel interface along the lateral (x) direction for NA = 1015 cm3 .
See caption of Fig. 2 for simulation parameters. Dotted lines
are results from classical calculations, and solid lines from
quantum calculations. From the top, Vg = 0.65, 1.0, and 1.5
V, respectively.

be much different between the quantum and classical


cases. Our results therefore indicate that for accurate
estimation of the current in nano-scale SB-MOSFETs,
the quantum mechanical consideration is necessary.
Figure 6 shows the threshold voltage for various channel lengths and doping concentrations. Notice that, due
to the presence of the Schottky barrier in SB-MOSFETs,

-S550-

Journal of the Korean Physical Society, Vol. 45, December 2004

sion layer of SB-MOSFETs by solving the Poisson and


Schr
odinger equations self-consistently. We have found
that the quantum mechanical calculations led to a thinner Schottky barrier, compared to classical calculations,
which indicates that the quantum mechanical consideration is necessary for accurate estimation of the current in
nano-scale SB-MOSFETs. We have also found that the
threshold voltage for channel inversion first decreases as
the channel length shortens, but, contrary to conventional MOSFETs, it increases rapidly when the channel
length becomes shorter than about 30 nm.
Fig. 5. Conduction-band energy at the oxide-channel interface along the lateral (x) direction for NA = 1018 cm3 .
See caption of Fig. 3 for simulation parameters. Dotted lines
are results from classical calculations, and solid lines from
quantum calculations. From the top, Vg = 1.5 and 2.0 V,
respectively.

ACKNOWLEDGMENTS

This work has been supported by the Ministry of Information and Communications, Korea.

REFERENCES

Fig. 6. Dependence of threshold gate voltages on the channel length, for NA = (a) 1015 , (b) 1017 , and (c) 1018 cm3 ,
respectively.

the threshold voltage for source-drain conduction can be


different from that for channel inversion, while the two
are the same in conventional MOSFETs. The threshold
voltage for channel inversion is shown in Figure 6. In
sharp contrast to conventional MOSFETs, the threshold
voltage first decreases as the channel length decreases,
followed by rapid increase as the channel length further
decreases. We attribute this effect to the constant builtin potential of the source and drain contacts. For shorter
channel lengths, it becomes harder for the gate voltage
to overcome the built-in potentials clipped at both the
contacts. To verify the threshold voltage behavior, solution of the transport equation coupled with the Poisson
and the Schrodinger equations or using an approximated
method [15] is needed.

IV. CONCLUSION
We have investigated the equilibrium charge distribution and the conduction energy bending in the inver-

[1] J. R. Tucker, C. Wang and P. S. Carney, Appl. Phys.


Lett, 65, 618 (1994).
[2] J. P. Snyder, C. R. Helms and Y. Nishi, Appl. Phys. Lett,
67, 1420 (1995).
[3] M. Jang, W. Cho, W. Jung, S. Lee, K. Park, J. Lee and
J. Cha, J. Korean Phys. Soc. 42, S189 (2003).
[4] R. Hattori and J. Shirafuji, Jpn. J. Appl. Phys. 33, 612
(1994).
[5] C. K. Huang, W. E. Zhang and C. H. Yang, IEEE Trans.
Electron Devices 45, 842 (1998).
[6] B. Winstead and U. Ravaioli, IEEE Trans. Electron Devices 47, 1241 (2000).
[7] L. Sun, X. Y. Liu, M. Liu, G. Du and R. Q. Han, Semicond. Sci. Technol. 18, 576 (2003).
[8] M. Jang, K. Kang, S. Lee and K. Park, Appl. Phys. Lett.
82, 2718 (2003).
[9] J. Guo and M. S. Lundstrom, IEEE Trans. Electron Devices 49, 1897 (2002).
[10] S. Selberherr, Analysis and Simulation of Semiconductor
Devices. (New York, Springer-Verlag, 1984).
[11] W. H. Press, B. P. Flannery, S. A. Teukolsky and W. T.
Vetterling, Numerical Recipes in C (Cambridge, Cambridge University Press, 1988).
[12] Tomas Skalicky, LASPack Reference Manual,
http://www.tu-dresden.de/mwism/skalicky/laspack/
laspack.html.
[13] A. Trellakis, A. T. Gallick, A. Pacelli and U. Ravaioli, J.
Appl. Phys. 81, 7880 (1997).
[14] A. S. Spinelli, A. Benvenuti and A. Pacelli, IEEE Trans.
Electron Devices 45, 1342 (1998).
[15] S. Jin, Y. J. Park and H. S. Min, J. Korean Phys. Soc.
44, 87 (2004).

Você também pode gostar