Self-Consistent M\odeling of _Ultra Thin Body Doub le

Gate MO4SFET
M. K. Alam and Quazi D. M. Khosru
To calculate quantum mechanical charge distribution in
MOS devices incorporating the wave function penetration
effect within the oxide layer of MOS devices an open
boundary condition is a must for the solution
SchrtOdinger equation [6]. In the absence of suitable
boundary conditions, zero penetration of wave function
into gate oxide is assumed in the simulation of even deep
submicron MOSFETs. While an open boundary condition
can be neglected in devices with thick oxide layers, its
negligence in deep submicron MOSFETs cannot be
justified.

Abstract - An accurate and efficient one

dimensional self-consistent numerical model of double
gate MOS structure is presented based on finite
element method. The model is developed using
FEMLAB considering wave function penetration effect
into gate oxide. Hence, penetration effect is revealed
and presented for full depleted double gate MOSFET.
Accuracy of the model has been verified by comparing
with established results.

I. INTRODUCTION

II. NUMERICAL FORMULATION

As CMOS scaling is approaching its limit due to

processing as well as fundamental considerations,
double-gate (DG) MOSFET is becoming an intense
subject of VLSI research. Theoretically, DG MOSFETs
can be scaled to the shortest channel length possible for a
given oxide thickness [1]. Among the advantages
advocated for double-gate MOSFETs are: ideal
6Omv/decade sub-threshold slope, volume inversion [2],
setting of threshold voltage by the gate work function thus
avoiding dopants and associated number fluctuation
effects etc. As the gate length goes below deep submicron
tom nsoa

VG

y
suidd cbendings

callowedul envergy
consite

In this section, the solution procedure is portrayed

for a dual gate n-MOS structure shown in Fig. 1 in the
inversion region. FEMLAB with MATLAB has been used
as the Partial Differential Eqn. (PDE) solver (suitable for
linear or nearly linear problems), for the self-consistent
solution.

interface.Ithaslongbeensufficient
knownthatwithband
Source~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
bending,thepotentialwellcanbecomesufficientlynar ow~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
toquantizethemotionofinversionlayercarriers
the~in~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . ......... . . . . . . . . . . . . . . . . . . . . .
direction perpendicular to the interface [3]. This gives rise~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~...................... ......................
into
asplit ofintglehenverlgsy subbands
Gate t~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . . . . . . . . . . . . . . . . . . . . . .
(2-dimensionaldensity-of-states),suchthatlowestofthe~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . .
w
not................
levels formaelectrons fin ath
ldcoincideconduction
withthebotband.
ofothemhel diSoes
Due
to
VG~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . . . . . . . . . ...............
quantization,theelectrondensitydoesnotreachits~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . . . . . . . .

tchreshodine eutoinen sinThsolvedaude is usdfrteP i onad

ntly,ven3]. thethe
effective massapproximation.The boundary condition equation in coefficient form
is given in FEMLAB
as:.................................
M. K. Alam and Quazi D.M. Khosru are with the

Using relevant parameters of MOS

ch d greq to.P isn

.......................

electrostatic potential.....................
and....charge ....density......respectively.......
structures Eq. (1) can.................................
.............................................
angladesh,g

d2v(z)
dz2
dz

q[p(z) n(z) ND -NA]

-

(2)

Where n (z), p (z) are the electron, hole concentration and

ND, NA are the ionized donor, acceptor concentration
respectively. E is the relative dielectric constant of the
material and co is the permittivity of free space.
The electron concentration n(z) is obtained for n-MOS
structure according to the following expression:
(3)
n(z) = LNi, YJi, (z) 12

Where,

Where,

i+ e
Nu=fv_mikT

EF-

kT

(4)

(5)

Njinv=L N
And

EF= EFb11lk (bulk fermi level)+ Vgate (gate voltage)

(6)

Where N/ is the carrier concentration in the jth

large

value (e.g. 4000 eV) that practically can be

considered as infinity.Gauss-Seidel iteration method has
been used for the numerical convergence. The weighting
factor should be greater than 0.9. With the increase of
weighting factor the assurance of convergence increases
but the simulation become slower and more iteration is
needed.
II. RESULTS AND DISCUSSION
In this section, the detail self-consistent results are
presented for NA=108 /cm3, Tsi=25 nm, Tox=1.5 nm and
VG=1.2875 volts. Fig. 2 presents the device structure taken
for calculation. 0.95 is used as the Gauss-Seidel update
coefficient. The boundary conditions used for Poisson
equation are-Dirichlet boundary condition for boundary
points 1, 2, 5, 6 and Neumann boundary condition for 3, 4
interfaces. A discontinuous electric field boundary
condition can be easily set with Neumann boundary
condition in FEMLAB.
ID double gate MOS geometry

subband of the ith valley, nvi and mdi are the ith valley
degeneracy and the ith density-of-states effective mass in
Si. Ninv is the total inversion carrier concentration and EF is
the Fermi energy. Eij and Yf are the eigenvalue and the
eigenfunction of an electron in the jth energy level of the
ith valley, which are obtained as a solution of the
one-dimensional Schrodinger equation.
Schrodinger equation in coefficient form is defined in
FEMLAB as:
- V. (cVu) +au
Au
(7)
In Eq. (7) a, u and i are electrostatic potential, eigen
function and eigen energies respectively Using relevant
parameters of effective mass Schrodinger equation for one
dimensional MOS structure quantum well, Eq. (7) can be
written ash2 d2

Front

Back

gate

metal oxide

Semiconductor

tN___T

-0 to .4

-----

gate
oxide metal
r

[2m*

dz2

+v(z)]

V/,(z)=E, V, (z)

(8)

At first, the device geometry (the fully depleted double

-1

-05

05

15

25

Z(xlO-8m)

3.5

Fig. 2 ID geometry of Double Gate MOS structure

gate structure) is defined in FEMLAB and the trial

potential is obtained by solving Eq. (1) using the full
depletion approximation assuming zero mobile charge
density i.e. n(z)=O in Eq. (2) with appropriate boundary
condition at each interface through FEMLAB linear PDE
solver.

Electrostatic Potential Vs Depth (front gate to back gate)

1.3

.25
1.2

Then the charge density profile n(z) is determined from Eq.

(3) by solving Eq. (7) using FEMLAB eigen-value solver
with Neumann boundary condition (needed for open
boundary condition to consider penetration effect).

This charge density profile is added to the source term of

Poisson equation and then Eq. (1) and Eq. (7) are solved
iteratively until the given convergence criteria [for
successive iteration, change in electrostatic potential at any
node point should be less than 10-8 volts] are fulfilled.
Hence we get the actual band profile and electrostatic
potential profile incorporating wave function penetration.
If the solution is to run without penetration effect, we just
increase the barrier height in the oxide region to a very

l\/l

15\
1.1\

05___,___. __.__.___. __.__.____

-

1
1
z (xl108 m)

Fi.3Tilptnilfo-fl elto

602

prxmto

Inversion Charge Vs Depth

Total

To verify numerical calculation, C-V characteristic of the

structure is simulated and results have been compared with
'Schred' (a well known established numerical solver that
doesn't consider penetration effect developed at Purdue
University available through www.nanohub.org) which is

Inlersion Charge

LongitudinalValley

TransverseVall

presented in Fig. 7 and

3p
= 21

1 ll
.o1 Us\

=1

\I

O1-1

-0.5.

00

1l |
2
lli
; # 8t\
17l

0
0.5

11 \

\/8<)':/
1

1
1.5

2.5

Fig. 8. Quantum simulation shows

classical analysis overestimate gate capacitance to a great

extent. Our simulated results without penetration have an
excellent match with 'Schred as the simulator doesn't
include penetration effect as shown in Fig. 8. Due to
penetration, the gate capacitance actually increases slightly
from that of without penetration. C-V study so far has been
done on DGMOS assuming the wave function to be zero at
the dielectric-gate electrode boundary [7, 8]. The C-V
calculated assuming such a boundary condition differs by
an insignificant amount from the C-V calculated by letting
the wave function penetrate into the gate electrode. This is
understood because the probability density function is
reduced by many orders of magnitude near the gate
electrode and hhence forcing it to
to zero does not introduce
and
ence in
very
very significant difference in C-V characteristics.

3.5

Z (X1O-8 m)
Fig. 4 Inversion charge density as a function of depth, z
Trial potential obtained from full depletion approximation
is shown in Fig. 3. Schrodinger equation is then solved for
the trial potential and inversion charge density is
calculated as shown in Fig. 4. The self consistent loop is
4
being run until the error is less than the given limit and
hence, actual band profile and electrostatic potential aren3e5
obtained as shown in Fig. 5 and 6.

zeractestics.

~-3

2.5

0.2
0.4-ihu
0.6 penetrain
1
0.8
2~ ~ ~
~ ~ ~ ~~~~~~~crepnst 0 caactac

0.5

1.2

Gate Voltage, VG (Volts)

o0.5

-05

05

7S (X1 o m

.5

Fig. 7 Gate capacitance as a function of gate voltage. WP

corresponds to capacitance with penetration; WoutP
corsndtoapiacewhuteerto.

2.5

Fig. 5 Actual band diagram for the simulated structure

Electrostatic
Potential Vs Depth
1.3

1.25

Semiclssic_

GaeVlae,V Vls
-1 -0.5 0 0.5 1 1.51.52-2

0.5

Z (Xl

0-8

crresponds

M)

~ Vs Depth
~ ~~
~~~
Electrostatic Potential

0.

0.2

0.4

0.8
1
1.2
wit0.6 peerton;tout

-0.5-

0.95

-12

cpacitance

to0

60
603~~~~~~~~~~~~~~~~1

Seiclssca

E 8
Na=1017 cm3Tsi=1lnm Tox=1.5nm
f.

6-

5-

4
4-

32

-0.4

-0.2

0.2

0.4

0.6

0.8

This new model is used to study the effects of wave

\ function penetration on double gate n-MOS transistor by
comparing the numerical results with those of without
penetration. Self-consistent
solution is an important
tool for simulation of many devices where the QM effects
become significant. Our self-consistent model may be used
for simulating many systems, such as single gate, double
gate MOS structure, high electron mobility transistor,
resonant tunneling diodes and quantum well lasers etc.
where self-consistent calculation with open boundary
condition is necessary.

1.2

REFERENCES

Gate Voltage, VG (Volts)

in CG
error 1n
%
9
CGas
Fig. 00g.
9 /O
as aa function of gate voltage[1]
voltage
error
Effect of penetration is shown in Fig. 9. 00 error is
calculated as the criteria of comparison which is calculated

[2]

as:

CG (With Pen.)

(3 4)

[3]
[4]

Penetration effect i.e. 00 error is more severe in low

inversion regime although differential error is greater in
the strong inversion as justified from the Fig. 9. All the
physical parameters for this work is taken from ref. [3]

[5]

[6]

IV. CONCLUSION

[7]

% error

(Without Pen.)

=G
CG (With Pen.)

(ihPn)

x100 C

An improved and numerically efficient

self-consistent model has been developed for MOSFET
simulation

1ntothegate

which

rtegionand walsoebe

enetration into the gate oxide

oxide region and can also beused
used
penetration
i

[8

for simulation without penetration effects.

604

M.
"Monte Carlo
Frank,S.of aLaux, and
simulation
MOSFET:How
short can
30nm dual-gate Fischetti,

D.

silicon go?," IEDM Tech Dig., p. 553, 1992.

F. Balestra and S. Cristoloveanu et al., "Double-gate

silicon-on-insulator transistor with volume inversion: A

new device with greatly enhanced performance," IEEE

Electron Device Letter., vol.EDL-8, p. 410, 1987.
F. Stern, "Self-consistent results for n-type Si inversion
layers," Phys. Rev. B, vol. 5, pp. 4891-4899, 1972.
S. M. Sze, Physics of Semiconductor Devices, Wiley
Eastern Limited, pp. 362-366, 1987.

Y. Tsividis, Operation and Modeling ofMOS Transistor,

McGraw-Hill, Ch-2, 1999.
A. Haque and M. Z. Kauser, "A comparison of

wave-function penetration effects on gate capacitance in

deep submicron n- and p-MOSFETs," IEEE Trans.
Electron Devices, vol. 49, no. 9, pp. 1580-1587, Sep.
2002.
G. Baccarani and S. Reggiani, "A compact double-gate
MOSFET model comprising quantum-mechanical and
nonstatic effects," IEEE Trans. Electron Devices, vol. 46,
pp. 1656-1666, Aug. 1999.
Lixin Ge, Francisco Gamiz, 0. Workman, and Surya
Veeraraghavan, "On the Gate Capacitance Limits of

Nanoscale DG and FD SOI MOSFETs," IEEE Trans. on

Electron Device.,vol. 53,no.4, pp. 753-758, 2006.