Você está na página 1de 16

Nonlinear Dyn (2011) 63: 719734

DOI 10.1007/s11071-010-9833-0
ORI GI NAL PAPER
A double microbeam MEMS ohmic switch
for RF-applications with low actuation voltage
Hatem Samaali Fehmi Najar Slim Choura
Ali H. Nayfeh Mohamed Masmoudi
Received: 21 December 2009 / Accepted: 1 September 2010 / Published online: 21 September 2010
Springer Science+Business Media B.V. 2010
Abstract In this paper, we propose the design of an
ohmic contact RF microswitch with low voltage ac-
tuation, where the upper and lower microplates are
displaceable. We develop a mathematical model for
the RF microswitch made up of two electrostatically
actuated microplates; each microplate is attached to
the end of a microcantilever. We assume that the mi-
crobeams are exible and that the microplates are
rigid. The electrostatic force applied between the two
microplates is a nonlinear function of the displace-
ments and applied voltage. We formulate and solve
the static and eigenvalue problems associated with the
proposed microsystem. We also examine the dynamic
behavior of the microswitch by calculating the limit-
cycle solutions. We discretize the equations of motion
by considering the rst few dominant modes in the mi-
crosystem dynamics. We show that only two modes
are sufcient to accurately simulate the response of the
H. Samaali () S. Choura M. Masmoudi
Micro Electro Thermal Systems Research Unit, National
Engineering School of Sfax, BP 3038, Sfax, Tunisia
e-mail: hatem.samaali@ept.rnu.tn
F. Najar
Applied Mechanics and Systems Research Laboratory,
Tunisia Polytechnic School, BP 743, La Marsa 2078,
Tunisia
A.H. Nayfeh
Department of Engineering Science and Mechanics,
MC 0219, Virginia Polytechnic Institute and State
University, Blacksburg, VA 24061, USA
microsystemunder DCand harmonic ACvoltages. We
demonstrate that the resulting static pull-in voltage and
switching time are reduced by 30 and 45%, respec-
tively, as compared to those of a single microbeam-
microplate RF-microswitch. Finally, we investigate
the global stability of the microswitch using different
excitation conditions.
Keywords MEMS Microswitch Pull-in Pull-out
Reduced-order model
1 Introduction
In recent years, the application of micro-electro-
mechanical systems (MEMS) to radio frequency (RF)
components has been developed, especially for RF-
MEMS switches. Traditional microelectronic switch-
es, such as silicon FETs (eld effect transistors)
and PIN (positive-intrinsic-negative) diode switches
present inadequate switching characteristics when the
signal frequency is greater than 1 GHz [1]. These
switches present high insertion loss and poor isolation
during the ON and OFF switching, whereas switches
fabricated by MEMS technology can overcome these
limitations [2, 3]. For instance, RF-MEMS switches
provide improved insertion loss, isolation, and linear-
ity, but they are limited because of their high actu-
ation voltage (up to 30 V) and slow switching time
(near 300 microseconds). Recent studies focused on
720 H. Samaali et al.
improving the switching time, minimizing the actua-
tion voltage, and integrating RF-MEMS switches with
integrated circuits (IC) [4].
In the literature, different types of RF-microswitch-
es with a variety of actuation mechanisms (electro-
static, magnetostatic, piezoelectric, or thermal), con-
tact modes (capacitive or ohmic contact), and cir-
cuit implementation (shunt or series) are identied.
Electrostatic actuation is the most used type in RF-
microswitches. However, this type requires relatively
high DC voltage (up to 30 V), and thus requires an
additional CMOS integrated up-converter to raise the
typical 5 Volt control voltage to the required level.
All electrostatically actuated RF-microswitches are
based on an out-of-plane suspension bridge or can-
tilever type [5]. The cantilever type, which is char-
acterized by less rigidity as compared to the suspen-
sion bridge, yields a reduced actuation voltage but in-
creases the switching time. The ohmic contact mode
is suitable for this type of switches since it leads to
very low ON state insertion loss and very high OFF
state isolation. However, they are highly susceptible
to corrosion, stiction, and microscopic bonding of the
contact electrode metal surfaces [5].
Integration of RF-microswitches with IC has been
one of the important trends in the last decades [68].
This integration requires that the RF-microswitch be
(a) very small size, (b) of low actuation voltage, and
(c) of low power consumption.
RF MEMS switches generate lower intermodula-
tion (IM) as compared to their equivalents semicon-
ductors [9]. In particular for ohmic switches, the gen-
erated IM is signicantly low because of the very
small capacitance at the OFF-state and the linearity of
the contact at the ON-state [3, 10].
Static and dynamic pull-in instabilities of MEMS
devices have been key issues in the literature. Sta-
tic pull-in, identied by Nathanson et al. [11], occurs
when the DC voltage exceeds a threshold value. Stud-
ies on static pull-in reveal that the maximum static sta-
ble deection varies from 33% to 41% of the original
electrode gap distance [12, 13].
The dynamic pull-in phenomenon have been also of
major interest in the literature. Dynamic pull-in takes
place when the system is excited using DC and/or AC
voltages. In fact, The transient behavior of MEMS de-
vices is important for RF switch and for optical ap-
plications [3]. Gupta et al. [14] and Krylov and Mai-
mon [15] showed that pull-in occurs at voltages below
the static pull-in value due to transient effects for mi-
crobeams actuated by a step voltage. Both studies in-
dicate that the dynamic pull-in voltage can be as low
as 91% of the static pull-in voltage. In the presence
of squeeze-lm damping, the dynamic pull-in voltage
is shown to approach the static pull-in voltage [15].
A recent study by Krylov [16] shows that, using Lya-
punov exponents, the systemmay become unstable be-
fore reaching the static pull-in voltage due to dynamic
effects when the system is excited using a DC volt-
age. Under the same excitation Nielson and Barbas-
tathis [17] concluded that dynamic pull-in occurs at
half of the electrostatic gap and 92% of the static pull-
in voltage.
Nayfeh and coworkers [1820] generated frequency
and force-response curves for electrostatic microactu-
ators whose main component is a clamped-clamped
microbeam. They showed that dynamic pull-in occurs
under voltages lower than the static pull-in voltage, as
low as 25%, when the frequency of the AC compo-
nent is in the neighborhood of a resonant frequency.
On the other hand, Najar et al. [21] and Lenci and
Rega [22] studied the basin of attraction of bounded
motions and showed that the erosion of the basin of
attraction is the principle reason for the occurrence
of dynamic pull-in by homoclinic bifurcation. They
showed that smoothness of the boundary of the basin
of attraction of bounded motions can be lost and re-
placed with fractal tongues as the excitation amplitude
is increased. Similar results were reported by Nayfeh
et al. [23] for a microcantilever with a rigid plate at-
tached to its free end.
The decrease of actuation voltage of electrostatic
RF-MEMS switches can be accomplished by (i) us-
ing different materials and hinges to reduce the mi-
crobeam rigidity, (ii) increasing the area of the elec-
trostatic eld, and/or (iii) decreasing the gap. These
variations degrade the principal parameters of the RF
MEMS switches, such as isolation. Abbaspour-Sani
and Afrang [24] proposed to decrease the equivalent
rigidity of the microswitch by using a structure com-
posed of two displaceable microplates This structure
preserves the microswitch parameters while increas-
ing its lifetime. Similarly, Chaffey and Austin [25] de-
creased the equivalent rigidity of the microsystem and
concluded that the use of two cantilever microbeams,
when compared to a single microbeam structure, re-
duces signicantly the pull-in voltage.
A double microbeam MEMS ohmic switch for RF-applications with low actuation voltage 721
The present paper examines the static and dynamic
behaviors of an electrostatically actuated ohmic con-
tact RF microswitch. The proposed design consists of
a pair of cantilevered microbeams. At the free end
of each microbeam, a rigid microplate is clamped.
An electrostatic force is applied between the two mi-
croplates (electrodes) causing the deections of both
microplatemicrobeam subsystems. These deections
continue to grow up to a point where the electrosta-
tic force exceeds the elastic force of the microbeams.
This leads to the collapse of the upper microplate onto
the lower one when the pull-in voltage is reached. The
means by which this instability occurs and the selec-
tion of the microsystem parameters that affect this in-
stability are of paramount importance in the design
of MEMS electrostatic devices. In practice, pull-in in-
stability of microswitches is suitable for changing the
state of an electric circuit from open to close or vice
versa. In this study, we develop an accurate mathemat-
ical model that accounts for signicant nonlinearities
of the microswitch. We rst investigate the static and
transient responses of the RF microswitch, and then
we study its openclosed cycle and compare it with
other designs. Here, stiction and Casimir forces will
be neglected due to the dominance of the electrosta-
tic force. To gain more insight into its dynamical be-
havior, we also simulate the frequency response of the
proposed RF microswitch. Finally, we investigate the
global stability of the microsystem by estimating the
size of the resulting basin of attraction of bounded mo-
tions.
2 Problem formulation
We propose the design of RF microswitches with
ohmic contact, as shown in Fig. 1. The proposed de-
sign consists of two cantilevered microbeams; each
one is attached to a rigid microplate (electrode) at its
free end, and clamped to the substrate at the other
end. The transmitted signal is applied to a transmis-
sion line located between the two electrodes. On top of
this transmission line an insulator layer is deposited in
order to provide separation between the actuation and
the transmitted voltages. The thickness of the trans-
mission line with its insulator layer is 0.8 m. An elec-
trical voltage, composed of DCand ACcomponents, is
applied between the two electrodes. Sufciently large
voltages cause the pull-in instability and, thus, the ON
Fig. 1 Ohmic contact RF-microswitch
722 H. Samaali et al.
state of the microswitch, whereas zero or low voltages
release the microbeams to establish the OFF state. The
microbeams are modeled as EulerBernoulli beams of
density , modulus of elasticity E, width

b, thick-
ness

h, cross-section area A =

b

h, and second mo-


ment of area I =

b

h
3
12
. The microplates are modeled
as rigid bodies of masses

M
1
and

M
2
and moments of
inertia

J
1
=
1
3

M
1

L
2
C
and

J
2
=
1
3

M
2

L
2
C
about the y
1
-
axis. Due to their small thicknesses, only the dominant
terms of the moments of inertia of the microplates are
considered in this study.
We derive the equations of motion using Hamil-
tons principle, which states that
_
t
2
t
1
(T V
D
V
E
) d

t = 0. (1)
The total kinetic energy of the microswitch is given
by
T =
1
2
A
_

L
1
0
_

w
1
_
x
1
,

t
__
2
d x
1
+
1
2
A
_

L
2
0
_

w
2
_
x
2
,

t
__
2
d x
2
+
1
2

M
1
_

w
1
_

L
1
,

t
_
+

L
c

w

1
_

L
1
,

t
__
2
+
1
2

M
2
_

w
2
_

L
2
,

t
_
+

L
c

w

2
_

L
2
,

t
__
2
+
1
2

J
1
_

w

1
_

L
1
,

t
__
2
+
1
2

J
2
_

w

2
_

L
2
,

t
__
2
(2)
where w
1
( x
1
,

t ) and w
2
( x
2
,

t ) are, respectively, the de-


ections of microbeams 1 and 2 at time

t about the z
1
and z
2
axes and at locations x
1
and x
2
, respectively.
The dot denotes the derivative with respect to time

t
and the prime designates the derivatives with respect
to the spatial variables x
1
and x
2
for microbeams 1
and 2, respectively. The total elastic energy of the mi-
crosystem is given by
V
D
=
1
2
EI
_

L
1
0
_
w

1
_
x
1
,

t
__
2
d x
1
+
1
2
EI
_

L
2
0
_
w

2
_
x
2
,

t
__
2
d x
2
. (3)
Finally, the potential energy due to the electrostatic
eld between the microplates is
V
E
=
b
p
2
(V
DC
+V
AC
)
2

_
2

L
c
0
ds

d
n
w

1
(

L
1
,

t )s w

2
(

L
2
,

t )(2

L
c
s)
=
b
p
(V
DC
+V
AC
)
2
2( w

2
(

L
2
,

t ) w

1
(

L
1
,

t ))
ln
_

d
n
2

L
c
w

1
(

L
1
,

t )

d
n
2

L
c
w

2
(

L
2
,

t )
_
(4)
where

d
n
=

d w
1
(

L
1
,

t ) w
2
(

L
2
,

t ), is the permit-
tivity of air, V
DC
and V
AC
are, respectively, the DC and
AC voltage differences between the microplates and s
is the local coordinate attached to the microplate in the
x directions.
Substituting (2)(4) into (1), incorporating viscous
damping, and using the following nondimensional
variables:
w
1
=
w
1

d
, w
2
=
w
2

d
, x
1
=
x
1

L
1
,
x
2
=
x
2

L
2
, t =

t

T
,

T =
_
A

L
4
1
EI
,
=

L
2

L
1
we obtain the nondimensional equations of motion
w
iv
1
(x
1
, t ) +c
1
w
1
(x
1
, t ) + w
1
(x
1
, t ) = 0, (5)
w
iv
2
(x
2
, t ) +c
2
w
2
(x
2
, t ) +
4
w
2
(x
2
, t ) = 0 (6)
where c
1
= c
1

L
4
1
EI

T
and c
2
= c
2

4

L
4
1
EI

T
are the nondimen-
sional damping coefcients of the microbeams. Now,
the prime and dot denote the derivatives with respect
to x and t , respectively.
Equations (5) and (6) are subject to the following
nondimensional boundary conditions:
w
1
(0, t ) = 0, (7)
w

1
(0, t ) = 0, (8)
w
2
(0, t ) = 0, (9)
w

2
(0, t ) = 0, (10)
w

1
(1, t ) =M
1
_
w
1
(1, t ) +L
c1
w

1
(1, t )
_

2
Q
w
,
(11)
w

2
(1, t ) =
4
M
2
_
w
2
(1, t ) +L
c2
w

2
(1, t )
_

22
Q
w
,
(12)
A double microbeam MEMS ohmic switch for RF-applications with low actuation voltage 723
w

1
(1, t ) = M
1
L
c1
_
w
1
(1, t ) +
4
3
L
c1
w

1
(1, t )
_

1
Q
w

1
, (13)
w

2
(1, t ) =
4
M
2
L
c2
_
w
2
(1, t ) +
4
3
L
c2
w

2
(1, t )
_

12
Q
w

2
, (14)
where
Q
w
=
(V
DC
+V
AC
)
2
(d
n
2L
c1
w

1
(1, t ))(d
n
2L
c2
w

2
(1, t ))
Q
w

1
=
(V
DC
+V
AC
)
2
(
w

2
(1,t )

1
(1, t ))
2

_
ln
_
d
n
2L
c1
w

1
(1, t )
d
n
2L
c2
w

2
(1, t )
_
+
2L
c1
(
w

2
(1,t )

1
(1, t ))
d
n
2L
c1
w

1
(1, t )
_
,
Q
w

2
=
(V
DC
+V
AC
)
2
(
w

2
(1,t )

1
(1, t ))
2
,

_
ln
_
d
n
2L
c1
w

1
(1, t )
d
n
2L
c2
w

2
(1, t )
_

2L
c1
(
w

2
(1,t )

1
(1, t ))
d
n
2L
c2
w

2
(1, t )
_
,
d
n
= 1 w
1
(1, t ) w
2
(1, t ), L
c1
=

L
c

L
1
,
L
c2
=

L
c

L
2
,
1
=

b
p

L
4
1
2EI

d
3
,
12
=
1

4
,
M
1
=
2

b
p

L
c

b

L
1
, M
2
=
2

b
p

L
c

b

L
2
,

2
=

b
p

L
c

L
3
1
EI

d
3
,
22
=
2

3
.
3 Static analysis
The static problem can be formulated by setting the
time derivatives and the AC forcing terms in (5)(14)
equal to zero. The solutions of the static equations are
given by
w
s1
=Ax
3
1
+Bx
2
1
+Cx
1
+D, (15)
w
s2
=Gx
3
2
+Fx
2
2
+Ex
2
+H (16)
where w
s1
(x
1
) and w
s2
(x
2
) are the static deections
for microbeams 1 and 2, respectively.
Using boundary conditions (7)(10) yields C =
D = E = H = 0. Use of the remaining boundary
conditions leads to the following nonlinear algebraic
equations:
6A=
2
(V
DC
)
2
_
1
D
n
2L
c1
D
1
__
1
D
n
2L
c2
D
2
_
,
(17)
6G=
22
(V
DC
)
2
_
1
D
n
2L
c1
D
1
__
1
D
n
2L
c2
D
2
_
,
(18)
6A+2B =

1
(V
DC
)
2
(
D
2

D
1
)
2
_
ln
_
D
n
2L
c1
D
1
D
n
2L
c2
D
2
_

2L
c1
(
D
2

D
1
)
D
n
2L
c1
D
1
_
, (19)
6G+2F =

12
(V
DC
)
2
(
D
2

D
1
)
2
_
ln
_
D
n
2L
c1
D
1
D
n
2L
c2
D
2
_
+
2L
c1
(
D
2

D
1
)
D
n
2L
c2
D
2
_
, (20)
where D
n
= 1 AB GF, D
1
= 3A+2B and
D
2
= 3G+2F.
The microbeam-microplate subsystems deect un-
der an applied electric eld. We examine these de-
ections by developing closed-form solutions for the
static deections, in (15) and (16), whose coefcients
are determined using the NewtonRaphson method
in Mathematica from (17)(20). The geometric and
physical parameters of the microswitch are given in
Table 1. Figure 2 shows variation of the static de-
ections of microbeams 1 and 2 at x
1
= x
2
= 1 with
the applied DC voltage. For a given DC voltage each
microbeam has two equilibrium solutions; one is sta-
Table 1 Geometric and physical parameters of the microswitch

L
1

L
2

b

h

d
250 m 250 m 5 m 1.5 m 4 m

L
c

b
p
E
25 m 20 m 2300 kg/m
3
160 GPa 8.851 10
12
F/m
724 H. Samaali et al.
ble (lower branch) and the other is unstable (upper
branch). As the voltage is increased, these solutions
meet at the pull-in point which is characterized by a
voltage V
DC
= V
p
= 5.89 V and maximum deection
equal to 0.1536. Since the microbeam microplate sub-
systems are alike, the deections w
s1
(x
1
) and w
s2
(x
2
)
are identical. We also simulate, in Fig. 2, the static re-
sponse of the system using the commercial software
ANSYS and show very good agreement between both
solutions. The pull-in voltage obtained by ANSYS is
5.9 V at a maximum deection equal to 0.144. We
show in Fig. 3 the deected congurations of the mi-
Fig. 2 Microbeams deection under an applied DC Voltage
crobeams using ANSYS for V
DC
= 5.5 V, the simula-
tion validate the rigid plate assumption adopted in our
model even for a relatively high DC voltage.
Figure 4 shows the inuence of the length of the
second microbeam

L
2
on the pull-in voltage V
p
. If
= 0, which corresponds to

L
2
= 0 while the lower
microplate remains under the upper one, the system
becomes a conventional microswitch with a single
cantilevered microbeam. This corresponds to the mi-
croswitch studied by Nayfeh et al. [23] with the pull-in
voltage V
p
= 8.3 V, which is in agreement with Fig. 4
for = 0. As shown in Fig. 4, the pull-in voltage V
p
is
reduced as increases, and attains its lowest value at
= 1, which is adopted for the rest of the simulations.
4 Natural frequencies and mode shapes
The deections of the microbeams, subjected to an
electrostatic force, can be decomposed into the sum
of static components due to the DC voltage and dy-
namic components due to the AC voltage, denoted by
u
1
(x
1
, t ) and u
2
(x
2
, t ); that is,
w
1
(x
1
, t ) =w
s1
(x
1
) +u
1
(x
1
, t ), (21)
w
2
(x
2
, t ) =w
s2
(x
2
) +u
2
(x
2
, t ). (22)
Expanding the nonlinear electrostatic force terms us-
ing Taylor series about u
i
= 0 (i = 1, 2) yields the
problem describing the dynamics of the system about
its static equilibrium. We drop the nonlinear forcing
Fig. 3 Static solution using
Ansys for V
DC
= 5.5 V
A double microbeam MEMS ohmic switch for RF-applications with low actuation voltage 725
Fig. 4 Pull-in voltage variation with
and damping terms in (5)(14), use (21)(22), and let
u
i
(x
i
, t ) =
i
(x
i
)e
jt
(i = 1, 2), where
i
(x
i
) is the
mode shape and is the associated nondimensional
natural frequency. The linear eigenvalue problem is
given by

iv
1
(x
1
)
2

1
(x
1
) = 0, (23)

iv
2
(x
2
)
2

2
(x
2
) = 0 (24)
and the boundary conditions

1
(0) = 0, (25)

2
(0) = 0, (26)

1
(0) = 0, (27)

2
(0) = 0, (28)

1
(1) =M
1
L
c1

1
(1) +
4
3
M
1
L
2
c1

1
(1)

1
V
2
DC
_

1
(1) +

1
(1) +
2

2
(1)
+

2
(1)
_
, (29)

2
(1) =
4
M
2
L
c2

2
(1) +
4
3

4
M
2
L
2
c2

2
(1)

12
V
2
DC
_

1
(1) +

1
(1) +
2

2
(1)
+

2
(1)
_
, (30)

1
(1) = M
1

1
(1) M
1
L
c1

1
(1)

2
V
2
DC
_

1
(1) +

1
(1) +
1

2
(1)
+

2
(1)
_
, (31)

2
(1) =
4
M
2

2
(1)
4
M
2
L
c2

2
(1)

22
V
2
DC
_

1
(1) +

1
(1) +
1

2
(1)
+

2
(1)
_
, (32)
where

1
=
1

1
_

2
+
1

1
+
2L
c1

2
1
_
,

1
=
1

1
_

2ln(

2
)

2
1
+
4L
c1

1
+
4L
2
c1

2
1
_
,
2
=
1
,

2
=
1

1
_
2ln(

2
)

2
1

2L
c1

2L
c2

2
_
,

1
=
1

1
_
1

2L
c1

2
2
_
,

1
=
1

1
_
2ln(

2
)

2
1

2L
c1

2L
c1

1
_
,
2
=
1
,

2
=
1

1
_
2ln(

2
)

2
1
+
2L
c1

2
+
2L
c2

4L
c1
L
c2

2
2
_
,

1
=
_
1

2
1

2
+
1

2
2
_
,

1
=
2L
c1

2
1

2
,

2
=
2L
c2

2
2

1
,
1
=
w

s2
(1)

s1
(1),

1
= 1 2L
c1
w

s1
(1) w
s1
(1) w
s2
(1),

2
= 1 2L
c2
w

s2
(1) w
s1
(1) w
s2
(1).
The general solutions of the eigenvalue problem
can be expressed by

1
(x
1
) =
1
cos(x
1
) +
2
sin(x
1
)

1
cosh(x
1
)
2
sinh(x
1
), (33)

2
(x
2
) =
1
cos(x
2
) +
2
sin(x
2
)

1
cosh(x
2
)
2
sinh(x
2
), (34)
where =

and the boundary conditions (25)
(28) are taken into account. The rest of the coef-
cients are determined by solving the system of non-
linear algebraic equations that result from substitution
of (33)(34) into the boundary conditions (29)(32).
A NewtonRaphson technique is used in Mathemat-
ica to calculate three of these coefcients in terms of
the fourth coefcient, and thus to determine the rst
few natural frequencies and mode shapes. In Fig. 5,
we show the rst four mode shapes and corresponding
natural frequencies for V
DC
= 3 V. The mode shapes
show the in phase and out of phase motion of the two
microbeams for each natural frequency of the system.
726 H. Samaali et al.
Fig. 5 First four mode
shapes and corresponding
natural frequencies for
V
DC
= 3 V
Figure 6 shows variation of the rst natural fre-
quency
1
with the applied DC voltage. We observe
a signicant drop in
1
as the DC voltage approaches
the static pull-in voltage. In Table 2, we show the ef-
fect of varying the DC voltage on the rst ten natural
frequencies. In this case, the frequencies
2
to
10
are
relatively insensitive to the DC voltage.
5 Reduced-order dynamic model
A reduced-order model (ROM) is derived to simulate
the dynamic response of the microsystem. This ROM
is obtained by applying the Galerkin method by using
the rst two global mode shapes associated with the
rst two natural frequencies. In addition, we let = 1,
L
c1
= L
c2
= L
c
,

L
1
=

L
2
=

L and M
1
= M
2
= M.
The Lagrangian is used here to derive the discretized
equations of motion in nondimensional form; that is,
=
_
1
0
w
2
1
(x
1
, t ) dx
1
+
_
1
0
w
2
2
(x
2
, t ) dx
2

_
1
0
_
w

1
(x
1
, t )
_
2
dx
1

_
1
0
_
w

2
(x
2
, t )
_
2
dx
2
A double microbeam MEMS ohmic switch for RF-applications with low actuation voltage 727
Table 2 Variation of the
rst ten natural frequencies
with V
DC
V
DC
(V) 0 2 3 4 5.89

1
1.51787 1.49077 1.45075 1.37900 0.46613

2
1.51787 1.51807 1.51830 1.51869 1.52198

3
13.18123 13.1814 13.1784 13.1795 13.1739

4
13.18123 13.2266 13.1817 13.1822 13.1864

5
39.53288 39.5330 39.5311 39.5314 39.5281

6
39.53288 39.5629 39.5330 39.5333 39.5350

7
80.56617 80.5642 80.5653 80.5647 80.5651

8
80.56617 80.584 80.5677 80.5684 80.5694

9
139.011 138.938 138.947 138.988 138.945

10
139.011 139.069 139.031 139.037 139.07
Fig. 6 Variation of the rst natural frequency
1
with V
DC
+
R(V
DC
+V
AC
)
2
(w

2
(1, t ) w

1
(1, t ))
ln
_
d
n
2L
c
w

1
(1, t )
d
n
2L
c
w

2
(1, t )
_
+M
_
w
2
(1, t ) +L
c
w

2
(1, t )
_
2
+
ML
2
c
3
_
w

1
(1, t )
_
2
+
ML
2
c
3
_
w

2
(1, t )
_
2
+M
_
w
1
(1, t ) +L
c
w

1
(1, t )
_
2
(35)
where R =

b
p

L
4
EI

d
3
.
Next, we approximate the microsystem deections
by
w
i
(x
i
, t ) w
si
(x
i
) +
2

j=1
q
i
(t )
ij
(x
i
) i = 1, 2
where the
ij
(x
i
) are the mode shapes and the q
i
(t )
are the associated modal amplitudes. Substituting
the above approximation into (35), using the Euler
Lagrange equation and adding viscous modal damping
terms, we obtain the following reduced-order model
described by two second order ODEs in time and rep-
resented in matrix form as:
[M][

Q] + [C][

Q] + [K
L
][Q]
+
R(V
DC
+V
AC
sint )
2
B
F
[K
N
][Q]
+ [F
1
] +
R(V
DC
+V
AC
sint )
2
A
F
[F
2
] = [0] (36)
where
[M] =
_
m
11
m
12
m
12
m
22
_
, [C] =
_
c
1
0
0 c
1
_
,
[K
L
] =
_
k
11
k
12
k
12
k
22
_
, [K
N
] =
_
0 K
12
K
12
0
_
,
[F
1
] =
_
f
11
f
12
_
,
f
11
=
_
1
0
w

s1
(x
1
)

11
(x
1
) dx
1
+
_
1
0
w

s2
(x
2
)

21
(x
2
) dx
2
,
f
12
=
_
1
0
w

s1
(x
1
)

12
(x
1
) dx
1
+
_
1
0
w

s2
(x
2
)

22
(x
2
) dx
2
,
[F
2
] =
_
f
21
f
22
_
=
_

21
(1)

11
(1)

22
(1)

12
(1)
_
,
728 H. Samaali et al.
[Q] =
_
q
1
(t )
q
2
(t )
_
and
m
11
=
_
1
0

2
11
(x
1
) dx
1
+
_
1
0

2
21
(x
2
) dx
2
+M
_

2
11
(1) +
2
21
(1)
+2L
c
_

11
(1)

11
(1) +
21
(1)

21
(1)
_
+
4
3
L
2
c
_

11
2
(1) +

21
2
(1)
_
_
,
m
12
=
_
1
0

11
(x
1
)
12
(x
1
) dx
1
+
_
1
0

21
(x
2
)
22
(x
2
) dx
2
+M
_

11
(1)
12
(1) +
21
(1)
22
(1)
+L
c
_

12
(1)

11
(1) +
11
(1)

12
(1)
+
22
(1)

21
(1) +
21
(1)

22
(1)
_
+
4
3
L
2
c
_

11
(1)

12
(1) +

21
(1)

22
(1)
_
_
,
m
22
=
_
1
0

2
12
(x
1
) dx
1
+
_
1
0

2
22
(x
2
) dx
2
+M
_

2
12
(1) +
2
22
(1)
+2L
c
_

12
(1)

12
(1) +
22
(1)

22
(1)
_
+
4
3
L
2
c
_

12
2
(1) +

22
2
(1)
_
_
,
k
11
=
_
1
0

2
11
(x
1
) dx
1
+
_
1
0

21
2
(x
2
) dx
2
,
k
12
=
_
1
0

11
(x
1
)

12
(x
1
) dx
1
+
_
1
0

21
(x
2
)

22
(x
2
) dx
2
,
k
22
=
_
1
0

2
12
(x
1
) dx
1
+
_
1
0

2
22
(x
2
) dx
2
,
K
12
=L
c
_

22
(1)

11
(1)
11
(1)

12
(1)

21
(1)

12
(1) +
12
(1)
_

11
(1)

21
(1)
_

22
(1)

21
(1) +
11
(1)

22
(1)
+
21
(1)

22
(1)
+2L
c
_

11
(1)

22
(1)

12
(1)

21
(1)
__
,
A
F
=
2(w

s1
(1) w

s2
(1) +

2
j=1
q
j
(t )(

1j
(1)

2j
(1)))
2
Ln[

1
+

2
j=1
q
j
(t )(
1j
(1)+
2j
(1)+2L
c

1j
(1))

2
+

2
j=1
q
j
(t )(
1j
(1)+
2j
(1)+2L
c

2j
(1))
]
,
B
F
=
_

1
+
2

j=1
q
j
(t )
_

1j
(1) +
2j
(1)
+2L
c

1j
(1)
_
__

2
+
2

j=1
q
j
(t )
_

1j
(1)
+
2j
(1) +2L
c

2j
(1)
_
__
w

s1
(1) w

s2
(1)
+
2

j=1
q
j
(t )
_

1j
(1)

2j
(1)
_
_
.
6 Static and dynamic responses
6.1 Static response
We calculate the static response of the microsystem by
solving (15)(20) with different applied DC voltages.
If the microswitch is at its ON state, the DC voltage
must be lowered to recover its OFF state. The voltage,
at which the microswitch electrodes loose contact, is
known as the pull-out voltage. We estimate the pull-
out voltage by conducting a transient analysis. We as-
sume that the ON state (both electrodes are in contact)
is the initial state of the transient response. The elec-
trodes are in contact, with the transmission line and the
insulator layer, when the microbeam deections attain
0.4, where a distance of 0.2 is kept as a separation gap.
Thus, the pull-out voltage can be estimated by solving
the dynamic equations (36) by means of long-time in-
tegration using the RungeKutta technique in Mathe-
matica.
A comparison of the pull-in and pull-out voltages
associated with the proposed model and that of Nayfeh
et al. [23] ( = 0) is shown in Fig. 7. We remark that
the computed pull-in voltages associated with the sin-
gle beam (Nayfehs model) and the proposed double-
beam microstructure are 8.297 V and 5.89 V, respec-
tively. Consequently, a 30% reduction of the actuation
voltage is obtained.
A double microbeam MEMS ohmic switch for RF-applications with low actuation voltage 729
Fig. 7 The responses of the single microbeam (gray curve) and
double microbeam (black curve) designs subjected to a DC volt-
age
6.2 Dynamic response
We use the Finite Difference Method (FDM) [15] to
examine the limit cycles of the microsystem model by
two modes subject to DC and AC voltages. The AC
excitation is harmonic with period T = 2/, where
is the forcing frequency. We discretize the orbit
into m + 1 points and enforce the periodicity condi-
tion q
i0
= q
i
(t
0
) = q
i
(t
m
) = q
im
. This condition im-
plies that the rst and last points of the orbit (point 0
and m) are identical. At each of these points, we have
_
q
ip
=q
v
ip
,
q
v
ip
=f (q
ip
, q
v
ip
, V
DC
, V
AC
(t
p
))
(37)
where p = 1, 2, . . . , m, i = 1, 2, t
p
= pt , t =
T
m
,
q
ip
=q
i
(t
p
), and q
v
ip
=q
v
i
(t
p
) = q
i
(t
p
). The state func-
tion f is given by
f
_
q
ip
, q
v
ip
, V
DC
, V
AC
(t
p
)
_
= [M]
1
_
[C]
_
q
v
1p
q
v
2p
_
+R
_
V
DC
+V
AC
sin(2t
p
)
_
2

_
[K
N
]
B
F
_
q
1p
q
2p
_
+
[F
2
]
A
F
_
+ [K
L
]
_
q
1p
q
2p
_
+ [F
1
]
_
.
The FDM can now be applied to system (37) to
yield a set of nonlinear algebraic equations. In this
case, a two-step explicit central-difference scheme is
Table 3 Loading cases
Study cases V
DC

1

2
V
AC
Q
Case 1 3 V 1.450 1.518 1 V 100
Case 2 3 V 1.450 1.518 0.05 V 100
used to approximate the time derivatives. Therefore,
for an m+1 FDM discretized orbit, the microstructure
dynamics can be approximated by a set of 4m non-
linear algebraic equations in 4m unknown displace-
ments and velocities. These equations can be solved
for the unknowns using the NewtonRaphson method.
The stability of the orbits can then be analyzed by
means of long-time integration (LTI). Next, we exam-
ine the frequency-response curves of the microbeam-
microplate system, whose parameters are given in Ta-
ble 1 and simulate its time response subject to two
different loading cases described in Table 3. We sim-
ulate the frequency response of the rst microbeam
tip w
max
=w
1
(1, t ) in the neighborhood of the nondi-
mensional fundamental frequency
1
while xing the
number of FDM time steps per period to m= 100. For
simplication we use the same modal damping coef-
cient for both mode shapes by dening c =
1
/Q,
where Q is the quality factor. This assumption is ver-
ied by the fact that the two rst natural frequencies
are very close to each others.
In Fig. 8, we use the loading case 1 to validate the
convergence of the proposed two modes approxima-
tion by plotting the frequency-response curve using
two, three and four mode shapes in the approxima-
tion of the beams deections. Figures 9 and 10 display
the frequency-response curves of the system subject to
loading Cases 1 and 2, respectively. For loading Case 1
(Fig. 9), we nd two resonance regions associated with
the rst symmetric mode and the rst antisymmetric
mode. The symmetric and antisymmetric modes corre-
spond to the in-phase and out-of-phase motions of the
microbeams, respectively. For loading Case 2, the res-
onance corresponding to the antisymmetric mode al-
most disappears because the AC voltage being applied
is small. In both gures, the solid and dashed curves
denote stable and unstable limit cycles, respectively,
and the gray dashed line represents the limit of stabil-
ity given by the unstable equilibrium solution. Further
details on the stability of these branches can be found
in [18] and [20].
We also investigate the force-response curve from
which the minimum applied AC voltage is assigned to
730 H. Samaali et al.
Fig. 8 Convergence of the proposed solution using a two modes
approximation
Fig. 9 Frequency response curve for loading case 1
Fig. 10 Frequency response curve for loading case 2
the proposed microswitch design. The use of a combi-
nation of DC and AC voltages can lead to the phenom-
Fig. 11 Force response-curve of the microswitch for = 1
Fig. 12 Force response curve of the microswitch for = 1.38
enon of dynamic pull-in [18, 20], which corresponds
to setting up the microswitch to its ON state. Fig-
ures 11 and 12 show the force-response curves for dif-
ferent values of the ACvoltage V
AC
by solving (37) for
= 1 and = 1.38. The resulting bifurcation curves
showthat the minimumACvoltage to be applied to the
microplates for dynamic pull-in is, respectively, 2.9
and 1.013 V for = 1 and = 1.38. We conclude
that dynamic pull-in is better obtained for excitation
frequencies close to the fundamental frequency.
7 Switching time estimation using static
and dynamic pull-in
In the proposed design, as the electrostatically actu-
ated microbeams deect, each microplate travels only
A double microbeam MEMS ohmic switch for RF-applications with low actuation voltage 731
Fig. 13 Inuence of DC voltage on the switching time for sim-
ple and double cantilever beam designs
50% of the gap distance to reach the ON state. We
expect that in the proposed design the switching time
is shorter when compared to the single beam design.
To conrm this, we solve the dynamic equations (36)
and estimate the transient time needed for the mi-
crobeams to travel from the equilibrium position to
the ON state under an applied DC or a DCAC volt-
ages. The switching time is an important parameter
in RF-MEMS since it represents the main limitation
for reaching high frequencies [26]. In this section, we
examine the effect of reducing the switching time by
increasing the applied voltage and its corresponding
power requirement for a given design using rst DC
voltage only and second using a combination of DC
and AC voltages.
7.1 Switching time using DC voltage: static pull-in
Here, we use the pull-in voltages calculated in Sect. 6.1
to measure the switching time for simple and double
cantilever beam designs.
Figure 13 shows the inuence of DC voltage on the
switching time for both single and double cantilever
beam designs. It is clear that, when using the DC volt-
age alone to actuate the microswitch, the double beam
design offers a signicant improvement of the switch-
ing time, especially when high actuation frequencies
are required. In Fig. 14, we simulate the corresponding
requirements on power characterizing the electrostatic
energy of the system given by (4). We note that the
double cantilever beam design requires a very low ac-
tuation power when compared to the single cantilever
beam design. The difference can reach one order of
magnitude.
Fig. 14 Inuence of DC voltage on the power consumptions
for simple and double cantilever beam designs
7.2 Switching time using DC and AC voltages:
dynamic pull-in
When we use a combination of DC and AC voltages
to actuate the microswitch from the OFF state to the
ON state, a voltage lower than the DC voltage can be
applied to force the collapse of the microplates. This
phenomenon is known as dynamic pull-in. This latter
is a homoclinic bifurcation that occurs as the AC volt-
age is increased [21] when stable and unstable mani-
folds of the saddle approach each other, touch and then
intersect transversely innitely many times as a result
of a homoclinic entanglement [27].
Figure 15 displays the inuence of the AC voltage
on the switching time associated with the simple and
double-beam microswitch designs with different actu-
ation frequencies for V
DC
= 3 V and Q = 100. We
vary the AC voltage amplitude and calculate the cor-
responding switching time. Figure 15 depicts that the
switching time is low at frequencies in the neighbor-
hood of the rst natural frequency for both designs.
In order to reduce the electrostatic voltage, we select
actuation frequencies that belong to the pull-in band,
within which no stable solutions exist (see Fig. 9), or
we apply an AC voltage higher than the smallest volt-
age corresponding to a stable solution in the corre-
sponding force-response curve. We conclude that dy-
namic pull-in is not an appropriate alternative to DC
voltage actuation since it results in large switching
times for both simple and double beam designs. How-
ever, compared to the simple beam design the pro-
posed design considerably reduces the actuation AC
voltage by 33% to 50%.
In Fig. 16, we simulate the electrostatic power asso-
ciated with Fig. 15 for different excitation frequencies.
732 H. Samaali et al.
Fig. 15 Inuence of AC voltage on the switching time for sim-
ple and double cantilever beam designs at different actuation
frequencies
For both designs, we show that actuation frequencies
closer to the rst natural frequency lower the electro-
static power. The double beam design depicts one or-
der of magnitude reduction in the electrostatic power.
However, the resulting electrostatic power using dy-
namic pull-in is larger as compared the case of static
pull-in for the same switching time. This conrms the
results obtained in Fig. 15.
In case of initial conditions corresponding to an
equilibrium position, the dynamic pull-in actuation for
a double beam design yields lower switching time and
electrostatic power. In the next section, we investigate
the inuence of other initial conditions on the global
stability of the microsystem and the corresponding
switching time for the unstable case.
8 Stability of the microswitch under small and
large perturbations
To study the global stability of the limit-cycle solu-
tions obtained by solving system (37) for a single
mode, we investigate the motion of the microswitch
for a set of initial conditions in the proximity of either
the stable or unstable xed points. In the beginning,
we study the behavior of the system for the undamped
and unforced case by examining the separatrices asso-
ciated with the microswitch dynamics. Then we incor-
porate damping and forcing and examine their inu-
ence on the region of stable motion in the phase space.
Fig. 16 Inuence of AC voltage on the power consumptions for
simple and double cantilever beam designs at different actuation
frequencies
Fig. 17 Separatrix for V
DC
= 3 V
8.1 Unforced and undamped case
In the absence of damping and forcing, we investigate
the region in the phase space that leads to bounded mo-
tion. In Fig. 17, we show the separatrix for V
DC
= 3 V
by integrating system (36) in forward and backward
time using LTI, for the undamped and unforced case
(c = 0 and V
AC
= 0), starting from the unstable static
solutions corresponding to displacements w
s1
(1) and
w
s2
(1). The dashed curves represent the separatrices
corresponding to both microbeams. These separatri-
ces are perfectly symmetric about half distance of the
electrostatic gap

d.
8.2 Forced and damped case
The integration of AC-forcing and damping in study-
ing the microsystem global stability requires a differ-
A double microbeam MEMS ohmic switch for RF-applications with low actuation voltage 733
Fig. 18 Basin of attraction of safe motion V
DC
= 3 V, = 1.38
and V
AC
= 0.05 V
ent approach. In fact, we use LTI to determine the sta-
bility of the system by assuming a set of initial con-
ditions belonging to the phase space of the system. To
determine these initial conditions, associated with the
rst microbeam, we divide the phase plane using a grid
composed of 500 500 lines. The grid points are cho-
sen as initial conditions to solve system (36).
In Figs. 18 and 19, we show the basin of attraction
of bounded solutions of the microswitch for = 1.38
and V
AC
= 0.05 V (Fig. 18) and V
AC
= 0.9 V (Fig. 19)
using a one-mode approximation. The use of only sin-
gle mode for approximating the dynamic solution is
related to the reduction of the computational time and
justied by the choice of the excitation frequencies,
which are far from the second resonance frequency of
the microsystem. In both gures, the red regions cor-
respond to initial conditions that lead to bounded mo-
tions. These regions correspond to the case in which
the pull-in dynamics fails to set the microswitch to
its ON state. Outside the basin of attraction, where
dynamic pull-in occurs, the color levels indicate the
magnitude of the switching time. More contrasted col-
ors correspond to smaller switching times. In case
V
AC
= 0.9 V, we note that the choice of the initial con-
ditions is crucial for determining the switching time.
This may be considered as a drawback of the proposed
design because its performance depends on both mi-
crobeams being able to recover their initial stable po-
sitions in order to initiate the next switching cycle.
9 Conclusions
We proposed the design of a RF microswitch with low-
voltage actuation. The microsystem is composed of
two displaceable microcantilever electrodes at which
Fig. 19 Basin of attraction of safe motion V
DC
= 3 V, = 1.38
and V
AC
= 0.9 V
two rigid microplates generate the nonlinear electro-
static force to actuate the microswitch. A mathemati-
cal model was developed to analyze the static, eigen-
value, and dynamic problems where the limit-cycle
solutions of the system are calculated under DC and
harmonic AC voltages. Using a static and transient
analysis, we demonstrated that the resulting static pull-
in voltage and switching time are reduced by 30 and
45%, respectively, as compared to the design made
of a single microbeam-microplate system. We also
showed that microswitches can pull-in at voltages be-
low the static pull-in voltage due to the transient ef-
fects when the system is excited using a combination
of AC and DC voltages. The frequency- and force-
response curves indicate the use of a set of AC voltage
frequency at which no stable positions exists.
This fact is suitable for switching applications.
Then we studied and compared the switching times for
DC and AC actuation. We showed that using a combi-
nation of DC and AC voltages minimize the switching
electrical power, however, in that case the switching
time is much higher than a standard DC forcing case.
This fact allows an optimization of the choice of the
actuation voltage with respect to the required switch-
ing performance. Finally, the global stability of the mi-
crosystem was studied in order to examine the varia-
tion of the switching time as the initial conditions are
varied. For small actuation voltages, we found a large
sensitivity to initial conditions. This fact inuences the
performance of the switching time.
Acknowledgements The authors would like to thank the re-
search team of Dr. Mohammad Younis of Binghamton Univer-
sity for their assistance in the ANSYS simulations.
734 H. Samaali et al.
References
1. Wang, L., Cui, Z., Hong, J.-S., McErlean, E.P., Greed,
R.B., Voyce, D.C.: Fabrication of high power RF MEMS
switches. Microelectron. Eng. 83, 14181420 (2006)
2. Yao, J.J.: RF MEMS from a device perspective. J. Mi-
cromech. Microeng. 10, 938 (2000)
3. Rebeiz, G.M.: RF MEMS: Theory, Design, and Technol-
ogy. Wiley, New York (2003)
4. Billard, C.: Micro et nanotechnologies mergentes pour ap-
plications radiofrquence. Rev. lectr. lectron. 8, 4257
(2007)
5. Lucyszyn, S.: Review of radio frequency microelectro-
mechanical systems technology. IEEE Proc. Sci. Meas.
Technol. 151, 93103 (2004)
6. Yao, Z.J., Chen, S., Eshelman, S., Deniston, D.: Microma-
chined low-loss microwave switches. IEEE J. Microelectro-
mech. Syst. 8, 129134 (1999)
7. Hah, D., Yoon, E.: A low-voltage actuated micromachined
microwave switch using torsion springs and leverage. IEEE
Trans. Microw. Theory Tech. 48, 25402545 (2000)
8. Park, J.Y., Kim, G.H., Chung, K.W., Bu, J.U.: Mono-
lithically integrated micromachined RF MEMS capacitive
switches. Sens. Actuators A 89, 8894 (2001)
9. Rebeiz, G.M.: RF MEMS switches: Status of the technol-
ogy. In: TRANSDUCERS03, 12th International Confer-
ence on Solid-State Sensors, Actuators and Microsystems,
vol. 2, pp. 17261729 (2003)
10. Dussopt, L., Rebeiz, G.M.: Intermodulation distortion and
power handling in RF MEMS switches, varactors, and
tunable lters. IEEE Trans. Microw. Theory Tech. 51(4)
(2003)
11. Nathanson, H.C., Newell, W.E., Wickstrom, R.A., Davis,
J.R.: The resonant gate transistor. IEEE Trans. Electron.
Devices 14, 117133 (1967)
12. Abdel-Rahman, E.M., Younis, M.I., Nayfeh, A.H.: Charac-
terization of the mechanical behavior of an electrically ac-
tuated microbeam. J. Micromech. Microeng. 12, 759766
(2002)
13. Najar, F., Choura, S., El-Borgi, S., Abdel-Rahman, E.M.,
Nayfeh, A.H.: Modeling and design of variable-geometry
electrostatic microactuators. J. Micromech. Microeng. 15,
419429 (2005)
14. Gupta, R.K., Hung, E.S., Yang, Y.J., Ananthasuresh, G.K.,
Sentura, S.D.: Pull-in dynamics of electrostatically actuated
beams. In: Technical Digest Solid State Sensor and Actua-
tor Workshop, pp. 12 (1996)
15. Krylov, S., Maimon, R.: Pull-in dynamics of an elas-
tic beam actuated by continuously distributed electrostatic
force. J. Vib. Acoust. 126, 332342 (2004)
16. Krylov, S.: Lyapunov exponents as a criterion for the dy-
namic pull-in instability of electrostatically actuated mi-
crostructures. Int. J. Non-Linear Mech. 42, 626642 (2007)
17. Nielson, G.N., Barbastathis, G.: Dynamic Pull-in of
parallel-plate and torsional electrostatic MEMS actuators.
J. Microelectromech. Syst. 15, 811821 (2006)
18. Nayfeh, A.H., Younis, M.I., Abdel-Rahman, E.M.: Dy-
namic pull-in phenomenon in MEMS resonators. Nonlinear
Dyn. 48, 153163 (2007)
19. Nayfeh, A.H., Younis, M.I.: Dynamics of MEMS res-
onators under super-harmonic and sub-harmonic excita-
tions. J. Micromech. Microeng. 15, 18401847 (2005)
20. Najar, F., Nayfeh, A.H., Abdel-Rahman, E.M., Choura, S.,
El-Borgi, S.: Nonlinear analysis of MEMS electrostatic mi-
croactuators: primary and secondary resonances of the rst
mode. J. Vib. Control 16(9), 13211349 (2010)
21. Najar, F., Nayfeh, A.H., Abdel-Rahman, E.M., Choura, S.,
El-Borgi, S.: Dynamics and global stability of beam-based
electrostatic microactuators. J. Vib. Control 16(5), 721748
(2010)
22. Lenci, S., Rega, G.: Control of pull-in dynamics in a non-
linear thermoelastic electrically actuated microbeam. J. Mi-
cromech. Microeng. 16, 390401 (2006)
23. Nayfeh, A.H., Ouakad, H.M., Najar, F., Choura, S., Abdel-
Rahman, E.M.: Nonlinear dynamics of a resonant gas sen-
sor. Nonlinear Dyn. 59(4), 607618 (2010)
24. Abbaspour-Sani, E., Afrang, S.: A low voltage MEMS
structure for RF capacitive switches. Prog. Electromagn.
Res. PIER 65, 157167 (2006)
25. Chaffey, J., Austin, M.: Analytical modelling of the electro-
mechanical coupling of cantilever beams. Proc. SPIE 4935,
8693 (2002)
26. De, S.K., Aluru, N.R.: Full-Lagrangian schemes for dy-
namic analysis of electrostatic MEMS. J. Microelectro-
mech. Syst. 13(5), 737758 (2004)
27. Nayfeh, A.H., Balachandran, B.: Applied Nonlinear Dy-
namics. Wiley, New York (1995)

Você também pode gostar