Você está na página 1de 9

2008 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 53, NO.

9, SEPTEMBER 2006
A Physics-Based Analytic Solution to the MOSFET
Surface Potential From Accumulation to
Strong-Inversion Region
Jin He, Member, IEEE, Mansun Chan, Senior Member, IEEE, Xing Zhang, Member, IEEE,
and Yangyuan Wang, Fellow, IEEE
AbstractA physics-based analytic solution to the surface
potential from the accumulation to the strong-inversion region
has been derived from the complete MOSFET surface potential
equation in this paper without any need for smooth functions or
simplication by dropping some second-order related terms. Its
high accuracy in predicting the surface potential and the tran-
scapacitance under various bias conditions has also been veried
by a comparison with the numerical results. The explicit surface-
potential solution not only leads to a more clear understanding
of MOSFET device physics but also provides a better platform to
develop the advanced surface potential-based model for the circuit
simulation.
Index TermsAnalytic solution, compact modeling, MOSFETs,
surface potential.
I. INTRODUCTION
I
N RECENT compact model formulations, a considerable
attention has been focused on developing the surface
potential-based models [1][6]. This is due to some drawbacks
in the threshold-voltage-based models, such as the disconti-
nuity across different operation regions, negative conductance,
and negative capacitance [3]. A number of surface potential-
based compact models such as MISNAN , MOS-11, HiSIM,
Philips and Penn State (PSP), and Peking University Nano-
Scale IGFET Model (PUNSIM) [1], [7][10] have been devel-
oped based on Brewss charge-sheet model [11] with the claim
that such a model retains the fundamental physics with high
accuracy and continuity.
The core of these models is the boundary surface potential,
which is a key physics-based variable in the development of
various surface potential-based models. The boundary surface
potential cannot be explicitly expressed in terms of the input
Manuscript received January 27, 2006; revised June 1, 2006. This work
was supported in part by the Special Funds for Major State Basic Research
Project and Nature Science Funds of China under Grant 973 and the Nature
Science Funds of China under Grant 90607017 and in part by the competitive
Earmarked Grant HKUST6289/04E from the Research Grant Council of Hong
Kong, SAR. The review of this paper was arranged by Editor M. Miura-
Mattausch.
J. He, X. Zhang, and Y. Wang are with the Nanoscale Device and Cir-
cuit Group, Multi-Project-Wafer (MPW) Center, Institute of Microelectronics,
School of Electronic Engineering and Computer Sciences (EECS), Peking
University, Beijing 100871, China (e-mail: jinhe@ime.pku.edu.cn).
M. Chan is with the Department of Electrical and Electronics Engineering,
Hong Kong University of Science and Technology, Kowloon, Hong Kong.
Color versions of Figs. 1, 2, and 7 are available online at http://ieeexplore.
ieee.org.
Digital Object Identier 10.1109/TED.2006.880364
voltages, but must be found empirically, approximately, or
numerically. For example, the MANSIM and HiSIM used the
NewtonRaphson to iteratively compute the surface potential
that forms the model framework. Most iteration methods,
however, require a special care in programming and coding
to achieve reasonable computation efciency and simulation
speed [4], [11], which are very critical in the circuit computer
aided-design (CAD) application. On the other hand, various an-
alytical approximations to the solution of the surface potential
problem have also been developed in parallel in the compact
modeling community to simplify the compact model structure.
The MOS-11 model from Phillips used several smooth func-
tions to link the different operation regions of the MOSFET
[5]. The recent PSP model not only uses smooth functions to
get the initial value of the surface potential but also adopts an
iterative-like second-order perturbation method to enhance the
accuracy of the initial [6]. The most important features of these
conventional analytic approximations are the following.
1) The noncomplete surface potential equation has to be
used to get the simple initial values [5], [6]. Such sim-
plication discarded many device physics, and its result
also does not match perfectly with that of the complete
surface potential equation [12].
2) Different smoothing functions have to be used in the
different MOSFET operation regions. For example, PSP
uses at least three smooth functions from the accumu-
lation region to the strong-inversion regions [13]. These
smoothing functions not only degrade the model compu-
tation efciency [14] but also result in high-error [4] and
nonphysics prediction under some forward-bias operation
[15]. Thus, an explicit analytic solution to the MOSFET
surface potential equation with the detail device physics
is highly preferable for the compact model development
and understanding of the detail device physics.
In contrast with the recent PSP model and the previous
MOS-11 model, a physics-based analytic solution to the
MOSFET surface potential from the accumulation to strong-
inversion region has been derived in this paper. The derivation
is based on the complete surface potential equation of the bulk
MOSFETs without replying on any smoothing functions or re-
gional separation of MOSFET operation regions. This analytic
solution has been widely tested and veried with numerical
methods for its correctness and accuracy in the surface potential
and transcapacitance predictions from the accumulation to the
0018-9383/$20.00 2006 IEEE
HE et al.: PHYSICS-BASED ANALYTIC SOLUTION TO THE MOSFET SURFACE POTENTIAL 2009
strong-inversion region. This explicit solution can be adopted
as a basis to understand the fundamental device physics of
advanced MOSFET surface potential model development.
II. DEVELOPMENT OF ANALYTIC SOLUTION OF THE
SURFACE POTENTIAL
In a bulk MOSFET with the p-type uniformly doped sub-
strate, the complete surface potential equation based on Pois-
sons equation solution can be derived from the rst principle
as was done in [16] if the body contact is treated as the zero
potential reference, and the hole quasi-Fermi potential is treated
as the zero quasi-Fermi potential reference. The expression is
given in (1), which is shown at the bottom of the page, where
V
FB
= atband voltage;

s
= surface potential;

t
= (kT/q). Thermal voltage;

f
=
t
ln[(N
a
/n
i
)]. Fermi potential;

n
= quasi-Fermi potential;
= ((t
ox

2q
si
N
a
)/
ox
). Bulk factor.
The sgn() is the signum function that returns 1 if the band
bending is above or at the atband region and 1 when the band
bending is below the atband region.
We would like to stress that (1) is continuous numerically
and correct physically in all MOSFET operation regions. It
produces correct surface potential derivatives with respect to
the gate voltage near and at the atband point for any quasi-
Fermi potential, as shown in the Section III. Thus, there is no
need to make articial modication and addition as in the PSP
model [13], and [17], [18].
To simplify derivation of the analytic surface potential so-
lution, we rst focus on the surface potential above or at the
atband. The surface potential solution of the accumulation
region will be derived using a similar method, which will
be described in the last part of this section. In this case, the
normalized form of (1) can be written as

GB
x
I
= G
_
x
I
+ e
x
I
1 + e
2x
f
x
n
[e
x
I
x
I
1]
(2)

GB
= (V
GB
V
FB
)/
t
. Normalized effective gate
voltage;
x
I
=
s
/
t
. Normalized surface potential above the
atband region;
x
f
=
f
/
t
. Normalized Fermi potential;
x
n
=
n
/
t
. Normalized quasi-Fermi potential;
G = /

t
. Normalized bulk factor.
Following a simple mathematical procedure to obtain the
solution of a quadratic equation, a solution of the sur-
face potential with implicit exponential terms in (2) can be
written as
x
I
=
_

GB
+
G
2
2
_
G
_

GB
+ G
2
/4

1 +
e
x
I
1

GB
+ G
2
/4
+
e
2x
f
x
n
[e
x
I
x
I
1]

GB
+ G
2
/4
. (3)
Please note that our derivation starts from the complete
surface potential equation (2) and includes all related terms
without any simplication. In contrast, MOS-11 and PSP
models have to make necessary simplications to obtain the
initial guess and further corrections. Thus, they used the non-
complete surface potential equation in their surface potential
calculation. Such a simplication may affect the accuracy and
continuity of the surface potential and its derivative calculation.
Moreover, we would like to note that the square-root argu-
ment of (3) is always positive above or at the atband point.
In the accumulation region, if we invert the surface potential
and the effective gate-voltage polarity for convenience in the
mathematical formulation, the square-root argument of (3) is
also always positive even in the accumulation region when (3)
is tested. Such a treatment ensures that the conditions for the
following mathematical transformation can be performed more
conveniently.
The Taylor expansion of the argument inside the second
square root in (3) with only rst-order terms allows (3) to be
rewritten as
x
I
=
_

GB
+
G
2
2
_
G
_

GB
+ G
2
/4
[e
x
I
1]f
A
[e
x
I
x
I
1]f
I
(4)
where f
A
= G/(2
_

GB
sgn() + G
2
/4) and f
I
=
f
A
e
2x
f
x
n
.
The signum function in f
A
makes f
A
also available in the ac-
cumulation region. In fact, f
A
= G/(2
_

GB
sgn() + G
2
/4)
will become f
A
= G/(2
_

GB
+ G
2
/4) above the atband
point, which also comes from the exact result of the Taylor
expansion of (3).
The validity of using the Taylor expansion to approximate
the argument inside the square-root term of (3) will be further
demonstrated in the next section from the surface potential
calculation. In fact, the second term of argument in the second
square root of (3) is always a small value (for example, it is
less than unity once above or at the atband point), thus the
Taylor-expansion condition is justied. The third term of the
argument in the second square root of (3) will become larger
than unity in the strong inversion region, although it is also
less than unity in other operation regions. In such a case, the
Taylor expansion is available in other operation regions besides
V
GB
V
FB

s
= sgn()

S
+
t
exp
_

t
_

t
+ exp
_

2
f
+
n

t
__

t
exp
_

t
_

t
_
(1)
2010 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 53, NO. 9, SEPTEMBER 2006
Fig. 1. Comparison of the surface potential calculation between the analytic
surface potential solution (16) and the numerical results for different quasi-
Fermi potential.
the strong inversion and results in some errors in the strong-
inversion region compared with the numerical simulation result.
The error in the strong-inversion region has an amplitude of a
few millivolts, as shown in Fig. 1. This result is good enough to
understand the general MOSFET device physics and predict the
device performance although it is not sufcient for the distor-
tion analysis and the transcapacitance prediction. Corrections
based on the same exact formulation in the inversion region and
accumulation will be further derived to enhance the accuracy of
the analytic solution coming from this Taylor expansion for the
compact model development.
One can nd from a simple observation of (4) that the
rst two terms in (4) are the exact solution of the normalized
surface potential of Poissons equation neglecting the mobile
concentration terms under the depletion approximation
x
D
=
_

GB
sgn() +
G
2
2
_

GB
sgn() + G
2
/4. (5)
Equation (5) matches well with the numerical solution of (1)
in the depletion region. Please note that the signum function
in (5) makes x
D
also available for the explicit solution of the
MOSFET surface potential in the accumulation region.
Substituting (5) into (4) results in
x
D
+ f
I
+ f
A
= [1 f
I
]x
I
+ f
I
e
x
I
+ f
A
e
x
I
. (6)
Since the hole term has only a little effect above the depletion
region, it can be accurately approximated by the depletion
surface potential. In this case, (6) is further simplied to be
x
D
+ f
I
+ f
A
(1 e
x
D
) = [1 f
I
]x
I
+ f
I
e
x
I
. (7)
Performing the identication transformation of (7) results in
x
D
+ f
I
+ f
A
(1 e
x
D
)
1 f
I
+ ln
_
f
I
1 f
I
_
= ln
_
f
I
1 f
I
e
x
I
_
+
f
I
1 f
I
e
x
I
. (8)
It is evident that (8) is a W-Lambert function with the
principal branch, which has the explicit analytic solution
f
I
e
x
I
(1 f
I
)
= W
0
_
f
I
1 f
I
e
x
D
+f
I
+f
A
(1e
x
D)
1f
I
_
(9)
where W
0
[x] is the notation of the principal-branch solu-
tion of the W-Lambert function that is widely available in
the mathematical packages and semiconductor device physics
[19][22]. As a result, the physics-based analytic solution of
the normalized surface potential is obtained in terms of the
W-Lambert function from (8) and (9)

I
=
x
D
+ f
I
+ f
A
(1 e
x
D
)
1 f
I
W
0
_
_
f
I
1 f
I
e
x
D
+f
I
+f
A
(1e
x
D)
1f
I
_
_
. (10)
Following a similar derivation, the analytic solution of the
surface potential in the accumulation region is also obtained
when the effective gate voltage is negative. In this case, the
surface potential solution is derived when we only need to
invert the effective gate voltage into the positive regime and
the exponential surface potential terms in (2) into the opposite
sign for the convenience in the mathematics. As a result, (2)
can also be used in the accumulation region, and an analytical
solution of the surface potential with implicit exponential terms
is written as
x
A
=
_

GB
sgn()+
G
2
2
_
G
_

GB
sgn()+G
2
/4

1+
e
x
A
1

GB
sgn()+G
2
/4
+
e
2x
f
x
n
[e
x
A
+x
A
1]

GB
sgn()+G
2
/4
(11)
where x
A
is the normalized absolute surface potential in the
accumulation region.
A Taylor expansion of the argument in the second square-
root term in (11) up to the rst-order terms in the accumulation
region simplies (11) and it becomes
x
D
+ f
I
+ f
A
= [1 f
I
]x
A
+ f
I
e
x
A
+ f
A
e
x
A
. (12)
Similarly, the electron-related term has a little effect in
the accumulation region, thus it is commonly neglected in
the conventional surface potential model. Here, we also use the
depletion approximation to handle it. As a result, we have
x
D
+ f
I
(1 e
x
D
) + f
A
= [1 f
I
]x
A
+ f
A
e
x
A
. (13a)
Identication transformation of (13a) results in
x
D
+ f
A
+ f
I
(1 e
x
D
)
1 f
I
+ ln
_
f
A
1 f
I
_
= ln
_
f
A
1 f
I
exp(x
A
)
_
+
f
A
1 f
I
exp(x
A
). (13b)
HE et al.: PHYSICS-BASED ANALYTIC SOLUTION TO THE MOSFET SURFACE POTENTIAL 2011
Equation (13) is another W-Lambert function with respect
to the normalized surface potential in the accumulation region.
Transforming (13b) into the standard W-Lambert function, we
obtain the exact analytical solution given as
f
A
e
x
A
(1 f
I
)
= W
0
_
f
A
1 f
I
e
x
D
+f
A
+f
I
(1e
x
D)
1f
I
_
. (14)
The explicit solution of the normalized surface potential
in the accumulation region is also expressed in terms of the
W-Lambert function by considering the nal sign of the surface
potential

A
= W
0
_
f
A
1 f
I
e
x
D
+f
A
+f
I
(1e
x
D)
1f
I
_

_
x
D
+ f
A
+ f
I
(1 e
x
D
)
1 f
I
_
. (15)
Finally, the complete analytic surface potential solution from
the accumulation region to the strong-inversion region is ob-
tained by (16), which is shown at the bottom of the page.
The surface potential predicted by (16) matches well with
the numerical solution in the depletion region while giving
some millivolt errors in the strong accumulation and strong-
inversion region due to the application of Taylor expansion of
the argument inside the second square-root term in (3) and (11),
as shown in Fig. 1. This result is good enough for understanding
the device physics and predicting the MOSFET performance,
although it is not sufcient for distortion analysis and tran-
scapacitance calculation. In order to improve the accuracy of
(16), corrections based on the same exact formulation in both
inversion region and accumulation region are used as described
in the Appendix. In such case, the complete solution of the
physics-based normalized surface potential with the precise in
the nanovolt range can be expressed as

s
=
_

I
+ [f
I
, f
I
,
2
f
I
]
+[a
I
, b
I
, c
I
] for V
GB
V
FB
0

A
+ [f
A
, f
A
,
2
f
A
]
+[a
A
, b
A
, c
A
] for V
GB
V
FB
0
(17)
where
=
f/f
1 0.5 f
2
f/f/f
(18)
with
f
I,A
=
_

gb

I,A
G
_
2
[
I,A
+ e

i,A
1]
+ e
x
n
2x
f
[e

i,A

I,A
1] (19)
f
I,A
=
_
e

i,A
2(
gb

I,A
)/G
2
_
e
x
n
2x
f
(e

I,A
1
) 1 (20)

2
f
I,A
=2/G
2
e

i,A
e

I,A
x
n
2x
f
(21)
and
=
c
sgn() +
b
a

(1b
2
/2a)c
asgn()+b
. (22)
Above atband, we have
a
I
=(
gb

I

I
)
2
G
2
(
I
+
I
1 + e

I
(
I
+
I
)e
x
n
2x
f
e
x
n
2x
f
) (23)
b
I
=2(
gb

I

I
)
G
2
(e

I
+ e
x
n
2x
f
1) (24)
c
I
= ln
_
a
I
/G
2
_
(
I
+
I
) + x
n
+ 2x
f
(25)
and below atband, we have
a
A
=(
gb

A
)
2
G
2
(
A
+
A
1 + e

A
+
A
x
n
2x
f
(
A
+
A
)e
x
n
2
f
e
x
n
2x
f
) (26)
b
A
=2(
gb

A
)
G
2
_
(
A
+
A
)e
x
n
2x
f
1
_
(27)
c
A
= ln
_
a
A
/G
2
_
+ (
A
+
A
). (28)
III. RESULT AND DISCUSSION
The surface potential and its derivatives in different operation
regions under various bias and different geometrical dimen-
sions can be predicted from the analytic solution in (17). The
comparison of the proposed analytic solution of the MOSFET
surface potential with the numerical result of (1) from the accu-
mulation, through the weak inversion, and nally to the strong-
inversion region is shown in Fig. 2. An excellent agreement
is achieved between the analytic solution and the numerical
results from the weak inversion to the strong accumulation
and the strong inversion over a wide range of the quasi-Fermi
potential.
The preciseness of the surface potential calculation is highly
desirable in the weak and moderate inversion regions where
the current expression of the charge-sheet model easily leads
to the numerical issue. The error analysis between the analytic
solution and the numerical results demonstrates that the analytic

s
=
_

I
= sgn()
_
x
D
+f
I
+f
I
(1e
x
D)
1f
I
W
0
_
f
I
1f
I
e
x
D
+f
I
x
D
+f
I
+f
I
(1e
x
D)
1f
I
__
for V
GB
V
FB
0

A
= sgn()
_
x
D
+f
A
+f
I
(1e
x
D)
1f
I
W
0
_
f
A
1f
I
e
x
D
+f
A
+f
I
(1e
x
D)
1f
I
__
for V
GB
V
FB
0
(16)
2012 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 53, NO. 9, SEPTEMBER 2006
Fig. 2. Comparison of the surface potential calculation between the analytic
surface potential solution (17) and the numerical results for different quasi-
Fermi potential.
Fig. 3. Error of the analytic surface potential solution (17) compared with the
numerical results.
solution is accurate enough in the whole MOSFET operation
regions, as shown in Fig. 3, where the error is in the nanovolt
range. Such a high precision is good enough for the compact
modeling development to avoid the numerical issue. Compared
with the recent PSP surface potential analytic approximation,
the new analytic solution does not have any need for the smooth
function and different region algorithms from the subthreshold
to the inversion regions. More importantly, the new analytic
solution is more stable and accurate than the PSP code that
may encounter the complex and breakpoint issues close to the
atband due to the numerical overow, as shown in Figs. 811.
A key issue in the MOSFET device physics and the surface
potential-based models is the continuity of the various charges.
Fig. 4 shows the predicted gate charge, inversion charge, and
bulk charge versus the gate voltage for a constant quasi-Fermi
potential. The numerical results are also plotted in these curves
for comparison. It is evident that all charges are not only smooth
and continuous in all MOSFET operation regions, but also
match the numerical results very well.
One important test of MOSFET surface potential solution is
the high-precision requirement to display the transcapacitance
characteristics. Fig. 5 shows the comparison of the values of
Fig. 4. Comparison of the gate charge, bulk charge, and the inversion-charge
calculations between the analytic surface potential solution method and the
numerical results.
Fig. 5. Comparison of the transcapacitances based on both analytic surface
potential solution (17) and numerical method for the different quasi-Fermi-
potential conditions.
the transcapacitance (e.g., C
gg
and C
bg
) between the analytic
solution and the numerical result. These results indicate that
the accuracy of the new analytic solution of the surface po-
tential and its partial derivative is satisfactory for the com-
pact modeling and general device-performance calculation. The
conventional surface potential approximations such as that used
in MOS-11 and SP may result in errors in the calculation of
the surface potential, and nonphysical results may be obtained
in predicting the surface potential derivatives for the forward
bulk bias conditions [4], [15]. These issues come from the fact
that MOS-11 and SP have to ignore the hole-related terms off
above the atband while neglecting the electron-related terms
below the atband. The use of the incomplete surface potential
equation to get the closed-form approximations gives rise to
these errors. This kind of simplication does not only degrade
the accuracy of the surface potential, but also results in the poor
prediction of the transcapacitances and the incorrect result for
the surface potential at the source and drain ends under forward
bias [4], [14], [15]. In contrast, the physics-based analytic
HE et al.: PHYSICS-BASED ANALYTIC SOLUTION TO THE MOSFET SURFACE POTENTIAL 2013
Fig. 6. Comparison of the surface potential calculations between the analytic
solution (17) and the numerical results for reverse and forward bulk bias
conditions.
Fig. 7. Surface potential derivative calculations based on the analytic solution
(17) for reverse and forward bulk bias conditions.
solution of the surface potential obtained in this paper results
in the correct surface potential and its derivatives, as shown in
Figs. 6 and 7.
One interesting result in this paper is that the analytic
solution predicts the correct surface potential and its derivative
scaling characteristic for the different substrate doping con-
centration, as shown in Figs. 811, respectively. For the zero
quasi-Fermi potential and atband voltage, the predicted sur-
face potential curve and the corresponding derivative for the
intrinsic body region show a perfect symmetry with respect
to the gate voltage as expected from the device physics. This
model behavior is very important for nonclassical CMOS de-
vices such as ultrathin body (UTB) and SOI modeling, where
the lightly doped or undoped body is used for the mobility
improvement and reduction of the uctuation of statistical
dopant distribution. We would like to point out that even the
latest PSP algorithm [13] cannot pass this test, and the surface
potential derivative shows some splits and peaks near the at-
band regions. Such a result of the PSP model comes from the
fact that the smoothing functions and the regional evaluation of
the surface potential are used in the PSP surface potential codes.
The results of the PSP model for the same conditions are also
shown in Figs. 10 and 11 for comparison.
Fig. 8. Surface potential scaling calculation based on the new analytic
solution (17) for different substrate doping concentrations.
Fig. 9. Scaling characteristic of the surface potential derivative with respect to
the gate voltage based on the new analytic solution (17) for different substrate
doping concentrations.
Fig. 10. Surface potential scaling calculation based on the latest PSP code for
different substrate doping concentrations.
IV. CONCLUSION
The physics-based analytic solution to the MOSFET sur-
face potential from the accumulation region to the strong-
inversion region has been derived in this paper, and its
accuracy has been widely veried from the surface potential,
charge, and transcapacitance calculation. The analytic solution
eliminates the need for the complex iterative computation and
tedious smooth-function-based analytical approximation of the
2014 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 53, NO. 9, SEPTEMBER 2006
Fig. 11. Scaling characteristic of the surface potential derivative with respect
to the gate voltage based on the PSP code for different substrate doping
concentrations.
MOSFET surface potential. This explicit solution is thus useful
for the understating of the MOSFET device physics and the
development of the next-generation advanced surface potential
model.
APPENDIX
A. Derivation of the Correction Function and
Its Related Coefcients [23]
In order to derive the correction function available in both
the accumulation and the inversion region, an identication
transformation has to be performed to (2), and the result is
given as
f =
_

GB
x
G
_
2

_
x + e
x
1 + e
2x
f
x
n
(e
x
x 1)

.
(A1)
Assuming that is the initial value of the normalized surface
potential, the Schroder series of (A1) results in the surface
potential expression
x=
f
f


2
f
2f
_

f
f
_
2
+
(3
2
f)
2
f
3
f
6(f)
2
_

f
f
_
3
.
(A2)
The rst- and second-order derivative terms in (A2) are
considered in order to simplify the nal correction function. In
such a case, (A2) is further simplied as
= x =
f
f
_
1 +

2
f
2f
f
f
_
(A3)
Since = x = f/f is the rst-order correction and
[1 + (
2
f/2f)(f/f)] is another correction over the rst-
order correction, it is evident that
_
1 +

2
f
2f
f
f
_
1. (A4)
Making use of (A4), a more accurate and more robust
correction is obtained from the approximated equivalent
variation of (A3)
=
f/f
1 0.5 f
2
f/f/f
. (A5)
The correction function is available in both inversion
region and accumulation region; the same coefcient formu-
lations, e.g., f, f, and
2
f can also be obtained from (A1) via
the initial value and the derivative rule. For example
f =
_

GB

G
_
2

_
+ e

1 + e
2x
f
x
n
(e

1)

(A6)
f =
_
e

2(
gb
)/G
2

e
x
n
2x
f
(e

1) 1 (A7)

2
f =2/G
2
e

e
x
n
2x
f
. (A8)
B. Derivation of the Correction Function and
Its Related Coefcients [23]
Equation (2) is transformed into the following form for the
correction-function derivation.
(
GB
x)
2
G
2
_
x + e
x
1 + e
2x
f
x
n
(e
x
x 1)
_
.
(B1)
Assuming that + is the initial value of the surface
potential and substituting this initial into (B1) by setting
x = + + . (B2)
Above or at the atband point, some symbols are used to
simplify the mathematical result and the electron contribution
term taking the logarithm-function form, we obtain
= c
I
+ ln
_
1
b
I

I
+

2

I
_
(B3)
where the symbol notations have the following expressions
a
I
=(
gb
)
2
G
2
( + 1 + e

( + )e
x
n
2x
f
e
x
n
2x
f
) (B4)
b
I
=2(
gb
)
G
2
(e

+ e
x
n
2x
f
1) (B5)
c
I
= ln
_
a
I
/G
2
_
( + ) + x
n
+ 2x
f
. (B6)
A Taylor expansion of (B3) up to the second-order terms
results in
= c
I

b
I

I
+
b
2
I

2
2a
2
I
+

2

I
. (B7)
HE et al.: PHYSICS-BASED ANALYTIC SOLUTION TO THE MOSFET SURFACE POTENTIAL 2015
From (B7), we get formulation solution with the implicit
terms
=
c
I
b
I

b
2
I

2a
2
I
+

I
+ 1
. (B8)
The in the implicit terms of (B8) is given by the rst-
order Taylor expansion of (B3)
=
c
I
b
I

I
+ 1
. (B9)
Thus, the closed-form expression of is nally obtained
=
c
I
b
I
a
I

(1b
2
I
/2a
I)c
I
a
I
+b
I
+ 1
. (B10)
Similarly, we can get the closed-form equation of the correc-
tion function beyond and at the atband point following the
same mathematical method. With some symbols being used to
simplify the mathematical result and the hole contribution term
in (B1) taking the logarithm-function form, we obtain
= c
A
+ ln
_
1
b
A

A
+

2

A
_
(B11)
where the symbol notations have the following expressions
a
A
=(
gb

A
)
2
G
2
(
A
+
A
1 + e

A
+
A
x
n
2x
f
(
A
+
A
)e
x
n
2
f
e
x
n
2x
f
) (B12)
b
A
=2(
gb

A
)
G
2
_
(
A
+
A
)e
x
n
2x
f
1
_
(B13)
c
A
= ln
_
a
A
/G
2
_
+ (
A
+
A
). (B14)
A Taylor expansion of (B11) up to the second-order terms
results in
= c
A

b
A

A
+
b
2
A

2
2a
2
A
+

2

A
. (B15)
From (B15), we get formulation solution with the implicit
terms
=
c
A
b
A

b
2
A

2a
2
A
+

A
1
. (B16)
The in the implicit terms of (B16) is given by the rst-
order expansion of (B11)
=
c
A
b
A

A
1
. (B17)
Thus, the closed-form expression of in the accumulation
region is nally obtained
=
c
A
b
A
a
A

(1b
2
A
/2a
A)c
A
a
A
+b
A
1
. (B18)
Since (B10) and (B18) have the exact function formulation
with only different signs and coefcients, we can use the unied
equation in both inversion and accumulation region for the
correction function
=
c
sgn() +
b
a

(1b
2
/2a)c
asgn()+b
(B19)
where the coefcients have different values above and below
the atband point, as shown in (B4) and (B5) and (B12B14).
REFERENCES
[1] A. R. Boothroyd, S. W. Tarasewicz, and C. Slaby, MISNANA
physically based continuous MOSFET model for CAD applications,
IEEE Trans. Comput.-Aided Des. Integr. Circuits Syst., vol. 10, no. 12,
pp. 15121529, Dec. 1991.
[2] M. Miura Mattausch, U. Feldmann, A. Rahm, M. Bollu, and
D. Savignac, Unied complete MOSFET model for analysis of digital
and analog circuits, IEEE Trans. Comput.-Aided Des. Integr. Circuits
Syst., vol. 15, no. 1, pp. 17, Jan. 1996.
[3] K. Joardar, K. K. Gullapalli, C. McAndrew, M. E. Burnham, and
A. Wild, An improved MOSFET model for circuit simulation, IEEE
Trans. Electron Devices, vol. 45, no. 1, pp. 134148, Jan. 1998.
[4] R. Rios, S. Mudanai, W.-K. Shih, and P. Packan, An efcient surface
potential algorithm for compact MOSFET models, in IEDM Tech. Dig.,
2004, pp. 755758.
[5] R. Van Langevelde and F. M. Klaassen, An explicit surface potential
based MOSFET model for circuit simulation, Solid State Electron.,
vol. 44, no. 3, pp. 409418, Mar. 2000.
[6] T. L. Chen and G. Gildenblat, Analytical approximation for the MOSFET
surface potential, Solid State Electron., vol. 45, no. 2, pp. 335339,
Feb. 2001.
[7] MOS-11. [Online]. Available: http://www.semiconductors.philips.com/
Philips_Models
[8] HISIM. [Online]. Available: http://www.starc.jp/kaihatu/pdgr/hisim/
index-e.html
[9] PSP Model. [Online]. Available: http://pspmodel.ee.psu.edu
[10] J. He, Y. Song, X. Niu, G. Zhang, X. Zhang, M. Chan, and Y. Wang,
Benchmark tests on conventional surface potential based charge-sheet
models and the advanced PUNSIM development, in Proc. Nanotech,
WCM, Boston, MA, May 2006, vol. 3, pp. 616621.
[11] J. R. Brews, A charge-sheet model of the MOSFET, Solid State
Electron., vol. 21, no. 2, pp. 345355, Feb. 1978.
[12] J. He, X. Zhang, G. Zhang, M. Chan, and Y. Wang, Benchmark tests
on surface potential based charge-sheet models, Solid State Electron.,
vol. 50, no. 2, pp. 263267, Feb. 2006.
[13] G. Gildenblat, T.-L. Chen, X. Gu, H. Wang, and X. Cai, SP: An advanced
surface potential-based compact MOSFET model, in Proc. IEEE CICC,
San Jose, CA, Sep. 2003, pp. 233240.
[14] I. Pesicin Commercial SPICE vendor perspectives on what makes a good
model, Nov. 20, 2005. SILVACO report, Silvaco, CA, CMC website of
Next Generation MOSFET Model Standard Phase-III Evaluation Results.
[Online]. Available: http://www.eigroup.org/CMC
[15] HISIM team, Important Features of HISM and Comparison With PSP,
Nov. 10, 2005. HISIM team report, Hiroshima, JP, CMC website of
Next Generation MOSFET Model Standard Phase-III Evaluation Results.
[Online]. Available: http://www.eigroup.org/CMC
[16] J. He, X. Zhang, and Y. Wang, Comments on Modeling MOSFET
surface capacitance behavior under non-equilibrium, Solid State Elec-
tron., vol. 50, no. 2, pp. 259262, Feb. 2006.
[17] C. C. McAndrew and J. J. Victory, Accuracy of approximations in
MOSFET charge models, IEEE Trans. Electron Devices, vol. 49, no. 1,
pp. 7281, Jan. 2002.
[18] W. Wu, T. L. Chen, G. Gildenblat, and C. C. McAndrew, Physics based
mathematical conditioning of the MOSFET surface potential equation,
IEEE Trans. Electron Devices, vol. 49, no. 7, pp. 11961199, Jan. 2004.
[19] R. M. Corless, G. H. Gonnet, D. E. Hare, D. J. Je.rey, and
D. E. Knuth, On the Lambert W function, Adv. Comput. Math., vol. 5,
no. 1, pp. 329359, Dec. 1996.
[20] A. Ortiz-Conde, F. J. Garca-Snchez, and S. Malobabic, Analytic solu-
tion of the channel potential in undoped symmetric dual-gate MOSFETs,
IEEE Trans. Electron Devices, vol. 52, no. 7, pp. 16691672, Jul. 2005.
2016 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 53, NO. 9, SEPTEMBER 2006
[21] T. C. Banwell, Bipolar transistor circuit analysis using the Lambert
W-function, IEEE Trans. Circuits Syst. I, Fundam. Theory Appl., vol. 47,
no. 12, pp. 16211633, Dec. 2000.
[22] C. Banwell and A. Jayakurmar, Exact analytical solution for current
ow through diode with series resistance, Electron. Lett., vol. 36, no. 4,
pp. 291292, Feb. 2000.
[23] J. He, Y. Song, X. Zhang, M. Chan, and Y. Wang, PUNSIM Documents-2:
The Analytic Calculation of the Surface Potential in the PUNSIM Model.
Beijing, China: Electron. Eng. Comput. Sci., Peking Univ. (internal
document, to be provided by the requirement).
Jin He (M04) received the B.S. degree from Tianjin
University, Tianjin, China, in 1988, and the M.S.
and Ph.D. degrees from the University of Electron
Science and Technology of China, Chengdu, in 1993
and 1999, respectively.
From 1999 to 2001, he worked as a Postdoctoral
and an Associate Professor with the Institute of Mi-
croelectronics, Peking University, China. From 2001
to 2005, he worked with the University of California,
Berkeley, as a Visiting Scholar and Research Engi-
neer in the Department of Electrical Engineering and
Computer Sciences (EECS). He was a main contributor to BSIM4.3.0, which is
an international industrial standard MOSFET model. He independently devel-
oped the core model for BSIM5, which is the next-generation BSIM MOSFET
compact model for mixed-signal application, including the hydrodynamic and
quasi-ballistic transport mechanisms. In August, he came back to the Institute
of Microelectronics, EECS, Peking University as a Full Professor and started
up the Nanoscale and Terahertz Device and Circuit Group. He is the most
main contributor and developer of Peking University Nano-Scale IGFET Model
(PUNSIM), which is an advanced surface potential-based compact MOSFET
model for the 6532-nm-technology very large scale integration (VLSI) circuit
simulation. He has published more than 120 technical papers in international
journals and conferences and coauthored three books.
Mansun Chan (S92M95SM01) received the
B.S. degree (highest honors) in electrical engineering
and the B.S. degree (highest honors) in computer
science from University of California at San Diego,
La Jolla, in 1990 and 1991, respectively, and the M.S.
and Ph.D. degrees from the University of California
at Berkeley, in 1994 and 1995, respectively.
During his undergraduate study, he was with
Rockwell International Laboratory, working on
heterojunction bipolar transistor (HBT) modeling,
where he developed the self-heating Simulation Pro-
gram with Integrated Circuits Emphasis (SPICE) model for HBTs. His research
at Berkeley covered a broad area in silicon devices ranging from process
development to device design, characterization, and modeling. A major part of
his work was on the development of record-breaking silicon-on-insulator (SOI)
technologies. He has also maintained a strong interest in device modeling and
circuit simulation. He is one of the major contributors to the unied BSIM
model for SPICE, which has been accepted by most U.S. companies and
the Compact Model Council (CMC) as the rst industrial standard MOSFET
model. In January 1996, he joined the Electrical and Electronics Engineering
(EEE) Faculty at the Hong Kong University of Science and Technology,
Kowloon, Hong Kong. Between July 2001 and December 2002, he was a
Visiting Professor at the University of California at Berkeley and the Codirector
of the BSIM program. He is still currently consulting on the development of
next-generation compact models. His research interests include nanodevice
technologies, image sensors, SOI technologies, high-performance IC, three-
dimensional (3-D) circuit technology, device modeling, and Nano biological
microelectromechanical systems (BIOMEMS) technology.
Dr. Chan was the recipient of the UC Regents Fellowship, the Golden Keys
Scholarship for Academic Excellence, the SRC Inventor Recognition Award,
a Rockwell Research Fellowship, the R&D 100 Award (for the BSIM3v3
project), the Teaching Excellence Appreciation Award in 1999, and the Dis-
tinguished Teaching Award in 2004, among others.
Xing Zhang (M95) was born in Shandong, China,
in 1965. He received the B.S. degree in physics from
Nanjing University, Nanjing, China, in 1986, and
the M.S. and Ph.D. degrees in microelectronics from
Shaanxi Microelectronics Institute, Xian, China, in
1989 and 1993, respectively.
He is a Professor with the Institute of Mi-
croelectronics, Peking University, Beijing, China.
His research elds include novel electron devices,
device physics, SOI technology, CMOS process,
and application-specied-integrated-circuit (ASIC)
design. He has authored or coauthored 3 books and more than 200 papers.
Yangyuan Wang (SM87F01) was born in
Zhejiang, China, in 1935. He received the degree
from Peking University, Beijing, China, in 1958.
Currently, he is an Academician with the Chinese
Academy of Science, Professor, and Director of the
Institute of Microelectronics, Peking University. He
is currently engaged in the investigation of the new
processes, new devices, and new structure of VLSI.
He conducted several research projects such as the
oxidation kinetics of polysilicon lm in VLSI/ULSI,
the device modeling and simulation of SOI/CMOS,
and the reliability of thin-oxide lm in deep submicrodevices and ICs. He has
published 7 books and more than 230 papers and is the holder of 9 invention
patents.

Você também pode gostar