Blondy

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 46

RF-MEMS Switches

Pierre Blondy

XLIM
CNRS - Universite de Limoges
123, Avenue Albert Thomas
87060 Limoges Cedex France

Contents

1 RF Switches and Switched varactors 3


1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Ohmic Contact switches . . . . . . . . . . . . . . . . . . . . 5
1.4 Microwave Capacitive Switches . . . . . . . . . . . . . . . . 8
1.5 Switched Varactors . . . . . . . . . . . . . . . . . . . . . . . 14

2 Thin Film and MEMS Structure Mechanics 19


2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Thin Film Properties . . . . . . . . . . . . . . . . . . . . . . 19
2.3 Fixed-fixed beams . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 Cantilevers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

1
2 CONTENTS

3 The Parallel Plate Electrostatic Actuator 25


3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2 Forces on a parallel plate actuator . . . . . . . . . . . . . . . 26
3.3 Triple Plate Systems . . . . . . . . . . . . . . . . . . . . . . 33
3.4 Scaling Rules . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.5 Problems Chapter 3 . . . . . . . . . . . . . . . . . . . . . . . 36

4 Contact Physics 37
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Contact forces on a parallel plate actuator . . . . . . . . . . 37
4.3 Pressure on a contact point . . . . . . . . . . . . . . . . . . 40
4.4 The Hertz contact model . . . . . . . . . . . . . . . . . . . . 41
4.5 Multiple Contact Points Model . . . . . . . . . . . . . . . . 43
4.6 Contact Supertemperature . . . . . . . . . . . . . . . . . . . 44
Chapter 1

RF Switches and Switched


varactors

1.1 Introduction
RF switches are very common components used for routing and controlling
RF signals in most Rf systems. They can be found in most microwave and
RF systems from Vector Network Analyzers to Cellular Phones, including
satellite payloads, radar systems and antennas. RF switches are very basic
components

1.2 Definitions
The perfect switch
An electrical switch is shown below:
It is basically a device that connects two wires together, upon an external
actuation. The connection in the closed state is perfect and the switch
isolation in the open state is infinite.
In real life however, the switch is not perfect, and in the open state,
there exists a series capacitance, that we will call the up state capacitance,
Cof f , leading to a series impedance of the switch Zof f .
In the closed state, that we will call the down state, there is series
resistance, Ron , due to the physical contact between the two metallic parts
of the switch. This resistance gives a series impedance Zon . There exists a
frequency, for which the moduli of Zof f and Zon are equal. This frequency is
called the cutoff frequency, fc of a switch, and corresponds to the frequency

3
4 CHAPTER 1. RF SWITCHES AND SWITCHED VARACTORS

Up State Down State

Cup Ron

Zup=1/(jCup!) Zdown=Ron

Figure 1.1: Operation principle of an ideal electric switch

where the switch has no contrast anymore. This should be taken as a


figure of merit (FoM) for a given switch technology, and not as a practical
limitation of a switch. The frequency is determined using the following
equations:

1
mod(Zof f ) = = Ron (1.1)
2πCof f f

The cutoff frequency fc is given by:

1
fc = (1.2)
2πCof f Ron

The best semi conductor switches have cutoff frequencies about 1-2 THz,
while MEMS switches have cutoff frequencies between 50-60 THz. It means
that the only figure that matters to define switch performances is the con-
trast between the up and down states. Then, if one needs a specific contact
resistance or a given isolation, then series or parallel arrangements are giv-
ing the same cutoff frequency. Specifically, for a given switch technology, a
designer will trade isolation for low loss and vice-versa.
In the case of capacitive switches, the contrast between the two states
is more depending on the contrast that exists between the up and down
states of the capacitance. The series resistance, Rs , is still there, and does
not change between the up and down states.
The ratio between the two impedances Zof f and Zon , is linked to the
capacitance ratio, Cof f and Con , since Rs is usually very small (a few mil-
liOhms) compared to the capacitance impedance.
1.3. OHMIC CONTACT SWITCHES 5

Up State Down State

Rs Cup Rs Con

Zup=Rs+1/(jCup!) Zdown=Rs+1/(jCon!)

Figure 1.2: Operation principle of an ideal capacitive MEMS switch

1.3 Ohmic Contact switches


An Ohmic switch is using a metal contact and is very similar to the device
presented in figure 1.1. This device physically turns itself from a capacitor
to a resistor when a bias voltage is applied on the switch. In a similar
manner to capacitive switches, such device can be implemented in a shunt
circuit or a series circuit.

Series Switches

V>Vp

Figure 1.3: Series Ohmic contact switch.

Series switches are among the most popular components in the area of
RF-MEMS, since they can replace diode and also conventional relays. The
6 CHAPTER 1. RF SWITCHES AND SWITCHED VARACTORS

devices are similar in many ways to the ideal relays that are shown in figure
1.4.
The insertion loss caused by a resistance on a 50 Ω line in series on a
MEMS switch is given by:

2Z0
S21 = (1.3)
2Z0 + Rs
The insertion loss can be computed using a very simple rule of thumb:
on a 50 Ohms line, the insertion loss of a MEMS switch is given by the
following formula:

S21 = 0.1Rs dB (1.4)


For instance, a 1 Ohm resistance switch gives about 0,1 dB of insertion loss
on a 50 Ω line.
In the up state, the capacitance can be computed from the surfaces of
the switch facing each other, and the gap height from the beam height. The
resulting capacitance, Cof f gives the isolation. The isolation in the up state
is given by:

S21 = 2ωCof f Z0 (1.5)

Application example 1
Let us consider the series switch depicted below:

Contact points, Surface S!

Input! Output!

Figure 1.4: Isometric view of a series ohmic contact switch.

The switch is made using a cantilever, in a very similar fashion to the


one depicted in figure 1.4. The gap height is 2 µm, and each contact surface
is 20x20 µm.

• Compute the off state capacitance. (Use the parallel plate approxi-
mation to compute the capacitance)

• Compute the isolation of the switch at 2 GHz.


1.3. OHMIC CONTACT SWITCHES 7

• The switch is now being used at 10 GHz, where 30 dB isolation is


needed. Propose a circuit to achieve such isolation.

• The series on state resistance of the switch is 1 Ω. Compute the cutoff


frequency of the switch. What is the insertion loss of the switch in
the previous configuration.

Application example 2
A MEMS switch has 0.4 dB insertion loss, and a measured isolation of 30 dB
at 2 GHz.

• Compute the off state capacitance.

• The contact surface is 10x10 µm. Compute the contact gap height.

Shunt Switches
Shunt contact switches can have in theory a very large isolation, since the
off state is dominated by the value of the resistance in the down state. Such
device is inherently wideband, but limited by the inductance of the device.

Cup

Ron

Figure 1.5: Equivalent circuit of an ohmic switch in a shunt configuration

A shunt switch in the down state isolation is shown below, for various
values of the resistances. The isolation is given by the equation below:

2Ron
S21 = (1.6)
Z0
We can see that the isolation is independent of frequency, if only the
contact resistance is considered.
When the series inductance is considered, the isolation is dramatically
reduced, and even a small inductance strongly reduces the isolation.
8 CHAPTER 1. RF SWITCHES AND SWITCHED VARACTORS

'"#
16 !"
8 !"
'$# 4 !"
2 !"

3 /4 5+56,7,+8
'%# 1 !"
0.5 !"
'&#

'!#

'(#

')#
# ! "# "! $# $! %# %! &#

*+,-./0 12

Figure 1.6: Simulated isolation of the switch in the down state

Cup

Ron

Figure 1.7: Equivalent scheme the switch when a small inductance is con-
sidered

In the on state, the parasitic shunt capacitance has little or no influence


on the return loss of the component.
It can be seen that inductances even with extremely small values (0.1 nH)
have a large influence on the isolation. Therefore, ohmic shunt switches are
not used in practice.

1.4 Microwave Capacitive Switches


Shunt Switches
A top view of a capacitive shunt switch is shown on the figure below. Such
a switch is made from a suspended metal membrane, standing above the
1.4. MICROWAVE CAPACITIVE SWITCHES 9

#
L=0.1 nH!
'"#

'$# L=0.01 nH!


3 /4 5+56,7,+8
'%# L=0 nH!

'&#
Ron= 1"!
'!#

'(#

')#
# ! "# "! $# $! %# %! &#

*+,-./0 12

Figure 1.8: Simulated isolation of the switch in the down state, when a
small inductance is considered

signal line of a CPW transmission line. A thin dielectric film is deposited


on top of this line, that prevents the metal membrane to be short circuited
when the beam is actuated using a static voltage, between the air bridge
and the signal line.

Air bridge!

Ground!

Signal!

Ground!

Figure 1.9: Top view of a capacitive MEMS bridge

Its principle of operation is based on the contrast of impedance between


the up state (when the membrane is standing above the CPW line) and the
down state (when the membrane is in contact with the dielectric layer)
A view of the switch actuated and not actuated is shown below:
A top view of a capacitive shunt switch is shown on
10 CHAPTER 1. RF SWITCHES AND SWITCHED VARACTORS

Dielectric film Ground

V
Substrate

Signal line

V>Vp

Figure 1.10: Cross section of a capacitive MEMS bridge un-actuated (top)


and actuated (bottom).

The equivalent scheme of the switch is given below

Zo, !, l Zo, !, l

L
Cup in the up state
Cmems
Cdown in the down state
Rs

Figure 1.11: Equivalent scheme of a capacitive MEMS bridge.

The switch response is shown below. When the switch is in te up state,


the signal passes through the bridge with little or no attenuation and only
parasitic remaining is given by the on state capacitance in parallel on the
transmission line. The main parameter in the reflection coefficient of the
transmission line, resulting from the perturbation on the short CPW line
section.
Thus, the reflection coeffcient S11 is increasing with frequency and fol-
lows the following equation:
Z0 Cof f .ω
S11 = (1.7)
2
1.4. MICROWAVE CAPACITIVE SWITCHES 11

Frequency Frequency
S11 dB

S11 dB
Up State Down State

Figure 1.12: Evolution of the reflexion coefficient S11 vs frequency.

In the down state, the switch shunts microwave signals to the ground,
and the isolation is increased with frequency. The transmission coefficient
S21 evolution versus frequency is shown below:

fr Frequency fr Frequency
S21 dB

S21 dB

S21 min

Up State Down State

Figure 1.13: Evolution of the transmission coefficient S21 vs frequency when


the capacitive switch is in the down state.

Since there is a small inductance in series with the shunt capacitance,


the isolation reaches a maximum for a frequency fr , that corresponds to
the maximum isolation attainable by this device. The value of the isolation
S21 min is limited by the resistance Rs on fig. 1.11.

fr is given by:

1
fr = √ (1.8)
2π LCdown
And the maximum isolation is given by:

2Rs
S21 min = (1.9)
Z0
12 CHAPTER 1. RF SWITCHES AND SWITCHED VARACTORS

Where Z0 is the characaeristic impedance of the transmission line.


Far from resonance, that is to say when the isolation is considered to be
dominated by the shunt capacitance, the isolation can be written as:

2
S21 (f ) = (1.10)
2πf Cdown Z0
This equation is valid if f << fr .
The resonance is commonly being used to design switches with a large
isolation at a given frequency with a specific capacitance. It is therefore
possible to design the device so that L and Cmems are giving a maximum
isolation for the proposed frequency.

Series Switches
Series switches are the most common implementation in electric circuits,
however, using a variable capacitance to allow signals to pass through the
transmission lines requires high contrasts and low loss.
An isometric 3D view of the switch can be seen on fig. 1.14

Cantilever beam!

Input! Output!

Figure 1.14: Isometric view of a cantilever MEMS series switch.

The operating principle is shown on below, in the up and down states. In


the up state, the device is preventing the microwave signal to pass through
the lines, since is has a very large impedance. The up state isolation is
given by the following equation:

S21 = 2ωCup Z0 (1.11)


This equation is valid only if S21 is small (< 10−1 ).
The typical response of such a switch in the up and down states can be
seen on the figures below.
The reflexion coefficient of the switch in the down state is given by:

1
S11 = (1.12)
2Z0 Cdown ω
1.4. MICROWAVE CAPACITIVE SWITCHES 13

V>Vp

Figure 1.15: Cross section of a capacitive series switch in the up (above)


and down (below) states.

RF In! RF Out!

Rs! Cmems! Cup in the up state

Cdown in the down state

Figure 1.16: Equivalent scheme of a series capacitive switch.

Frequency Frequency
S11 dB

S11 dB

Up State Down State

Figure 1.17: Typical response of a series capacitive switch in the up and


down states.
14 CHAPTER 1. RF SWITCHES AND SWITCHED VARACTORS

Frequency Frequency
S21 dB

S21 dB
Up State Down State

Figure 1.18: Typical response of a series capacitive switch in the up and


down states.

These two equations allow to design switches with appropriate proper-


ties. The main characteristic of such a switch is its on/off ratio between
the two capacitance values.

1.5 Switched Varactors


Introduction
Switched varactors are among the most commonly used RF-MEMS devices.
Their main interest is high Q and low power consumption.

N Bits arrangements
Analog variable devices are difficult to fabricate and integrate in practice
inside a microwave circuit. The preferred approach is to use RF-MEMS ar-
rays of devices that can be used as building blocks for multibit arrangements
of variable capacitance. The array is based on the use of a single switched
device, very similar to the series capacitive switch shown above. For an par-
allel array arragement, the total value of the capacitance is nCmems , Cmems
being either Cup or Cdown , n the number of capacitors. The values of such
an array are bounded by the on/off ratio of the elementary capacitance.

Q factor
Let us consider the electrical scheme below:

Stored Energy
Q=ω (1.13)
Average Dissipated P ower
1.5. SWITCHED VARACTORS 15

V
I
C R

Figure 1.19: Equivalent scheme of a capacitor in series with a resistor

The stored energy in the capacitor is 12 CV 2


The power is dissipated in the series resistor, and is equal to 12 RI 2 .
If the series resistance is small compared to the modulus of the impedance
1
of the capacitor, R << Cω , then the modulus of the current I flowing
through the capacitor is I = V Cω. The average dissipated power is there-
fore 21 RV 2 C 2 ω 2 .

Q becomes:

1
2
CV 2
Q=ω1 (1.14)
2
RV 2 C 2 ω 2
Which simplifies to the well-known formula:

1
Q= (1.15)
RCω
The case of two lossy capacitors in parallel can be studied following the
same approach:

The stored energy in the capacitor array is 12 (C1 + C2 )V 2 . The power


dissipated in each branch is P1 = 21 R1 V 2 C12 ω 2 and P2 = 21 R2 V 2 C22 ω 2 .

In this case the Q factor can be rewritten as:

1
2
(C1 + C2 )V 2
Q=ω1 2C 2ω2 + 1 R V 2C 2ω2
(1.16)
2
R1 V 1 2 2 2
16 CHAPTER 1. RF SWITCHES AND SWITCHED VARACTORS

V
I1
C1 R1
I2
C2 R2

Figure 1.20: Equivalent scheme of two capacitors C1 , C2 each in series with


a resistor R1 and R2 .

This general equation simplifies to:

(C1 + C2 )
Q= (1.17)
ω(R1 C12 + R2 C22 )

This case can be used to analyse most 2-State switched capacitor Digi-
tally Controlled Capacitors Arrays. If we consider the array in the figure
below:

The quality factor of the array in the state 100 can be written as:

(4Con + 3Cof f )
Q= (1.18)
ω( R4 Con
2 + R C2 )
3 of f
1.5. SWITCHED VARACTORS 17

Con R

Con R 4 capacitances On

Con R
4 x Con R/4
Con R
3 x Coff R/3
Coff R

3 capacitances Off
Coff R

100 State of a 3-bit capacitor


Coff R

Figure 1.21: Equivalent scheme of a 3-bit Digitally Variable capacitor. 3


capacitors are left in the Off-State and 4 capacitors are left in the On-State.
Chapter 2

Thin Film and MEMS


Structure Mechanics

2.1 Introduction
2.2 Thin Film Properties
Definitions
Let s consider the beam presented on fig. 2.1.

ux(x) ux(x+!x)

x i+!x

Figure 2.1: Cross section of a solid, subjected to an external pressure along


the x axis

The relative elongation x , is defined as:

ux (x + ∆x) − u(x) ∂ux


x = = (2.1)
∆x ∂x
Then, the axial stress σx , which is the applied force per unit surface of
cross section, can be written as:

σx = E.x (2.2)

19
20CHAPTER 2. THIN FILM AND MEMS STRUCTURE MECHANICS

σx is expressed in P a
E is called the Young’s modulus of the material, and this is a constant of
the material.
x is the relative elongation, dimensionless.

Another phenomenon arises from the conservation of matter, as ex-


pressed below:

!x

!y

Figure 2.2: Cross section of a solid, subjected to an external pressure along


the x axis

The solid is compressed in the x direction while it is expanded in the


y direction. The ratio between the two elongations is called the P oisson
coefficient, defined as:

y = −γx (2.3)
Where γ is the Poisson coefficient of the material (0.3 is typical). The
minus sign expresses that the solid expands in direction orthogonal from
the applied compression.

In plane stress
The thermal expansion coefficient, αT is given by
∂
αT = (2.4)
∂T
Where T is the temperature and  is the relative. The order of magnitude
for αT is 10−6 -10−7 . αT is expressed in Kelvin−1 .
The thermal expansion induced by a change in temperature ∆T is given
by:

T h = (αTf ilm − αTsubst )∆T (2.5)


The residual stress in the film is given by the following expression:
2.3. FIXED-FIXED BEAMS 21

E
σf ilm = ( )T h (2.6)
1−γ
E
Where ( 1−γ ) is the Biaxial modulus 6= Young’s modulus

2.3 Fixed-fixed beams


Fixed-Fixed beams are very common in RF-MEMS, since they can be rel-
atively easily implemented in RF circuits such as coplanar waveguides, or
other transmission lines. The easiest way for analyzing electrostatic actua-
tors is reducing a complex beam problem to a simple parallel plate actuator.
The spring constant can be derived using beam theory, and the character-
istics like Young’s modulus, E, the moment of inertia I. Let consider the
fixed-fixed beam depicted below, where a force F is applied in the middle
of the beam:

t
w
l

dy

Figure 2.3: Illustration of the deflection dy of a beam under a static force


F distributed around its center

A schematic illustration of a fixed-fixed beam, subjected to a force F


distributed at the center of the beam is shown on Fig. 2.4. The device
consist in a metallic beam, where l is the physical length of the beam, w is
the width, and t the thickness of the metallic beam.
The beam deflection, y, is given by the following equation:

d2 y
EI = MA + RA x (2.7)
dx2
The moment MA is given by:

Fa
MA = − (l − a)2 (2.8)
l2
22CHAPTER 2. THIN FILM AND MEMS STRUCTURE MECHANICS

y F
a
x
MA MB
l
RA RB

Figure 2.4: Cross section of a fixed-fifed beam, with a load applied at a


distance a from left anchorage

The vertical reaction, RA , is given by:

F
RA = (l − a)2 (l + 2a) (2.9)
l3
And the deflection is given by:

MA x2 RA x3
y= + (2.10)
2EI 6EI
A schematic illustration of a fixed-fixed beam, subjected to a force F
distributed at the center of the beam is shown on Fig. 2.4.

Membranes vs beams
A fixed-fixed beam may have two types of mechanical behaviours. If the
beam has a built in tensile stress, then the stiffness and the spring constant
2.4. CANTILEVERS 23

Figure 2.5: Cross section of a fixed-fixed beam with an actuation electrode


with a width x centered on the beam

may arises from the stress in the beam, and results in a spring constant
expression that can be expressed as:
t 1
kstress = 8.σ.(1 − γ).w.( ) (2.11)
l 3 − 2( xl )
When the geometrical factor is being considered, the spring constant
can be written as:
t 1
kgeom = 32.E.w( )3 . x 3 (2.12)
l 8.( l ) − 20.( l ) + 14.( xl ) − 1
x 2

In general, the spring constant of the beam is the sum of the two:

k = kstress + kgeom (2.13)


It can be noticed that the stress component of the spring constant is
linear with respect to the beam geometry, and that the geometrical com-
ponent of the spring constant is proportionnal to the power 3 with respect
to the geometrical cross section of the beam.

Effects of temperature
Since the beams are fixed at both ends, any temperature change will modify
its internal axial stress and reduce the beam stiffness. This is especially true
for membrane type of switches.

2.4 Cantilevers
There is a single spring constant is this case, given by:

t 1 − ( xl )
kgeom = 2.E.w.( )3 (2.14)
l 3 − 4( xl )3 + ( xl )4
24CHAPTER 2. THIN FILM AND MEMS STRUCTURE MECHANICS

l w

dy

Figure 2.6: Cross section of a cantilever beam with an actuation electrode


with a width x centered on the beam

Figure 2.7: Cross section of a cantilever beam with an actuation electrode


with a width x centered on the beam

And we approximate the surface of the actuator to the surfaces facing


each other:

S = (l − x).w (2.15)
Then we can apply the following methodology for the computation of
an electrostatic actuator:

• Compute the spring constant of the mechanical beam

• Compute pull in pull out voltages of the MEMS structure using the
surface
Chapter 3

The Parallel Plate


Electrostatic Actuator

3.1 Introduction
The mechanical structures that are being used for the fabrication of RF-
MEMS structures are usually simple. They consist in suspended bridges
and cantilevers (levier in french, or poutre console), that can be studied in
a simple and concise manner. The purpose of this chapter will be to provide
the reader with equivalencies between the geometry of the considered struc-
ture and the equivalent parameters (spring constant, mechanical resonance
of frequency of the first order flexion mode of the beam). The methodology
that we will apply is outlined below: We will use analytical expression of the
beam stiffness and resonance frequency to reduce the mechanical structures
to an equivalent spring/mass system.

F
k

-Spring Constant
-Fundemental
flexion mode

MEMS Beam
Parallel plate equivalent structure

Figure 3.1: Mechanical model reduction

25
CHAPTER 3. THE PARALLEL PLATE ELECTROSTATIC
26 ACTUATOR
3.2 Forces on a parallel plate actuator

Suspension spring!

Moveable plate!
Electrostatic gap!
Applied !
voltage!
Dielectric layer!

Fixed plate!

Figure 3.2: Schematic view of a parallel plate electromechanical actuator

A parallel plate actuator is shown on figure 3.2. It is a planar capacitor,


with one its electrodes suspended to a spring, allowing its vertical move-
ment. The second electrode is fixed, and covered with a thin dielectric
layer. When a voltage is applied on this structure, an electrostatic force
is generated, causing the plates to be attracted towards each other. The
applied voltage on the structure is V and the resulting electrostatic force is
given by:
1
Fe = Q.E (3.1)
2
where Q is the charge in the parallel plate capacitance and E is the electric
field between the plates.
The restoring force is given by the elongation of the spring, caused by the

k!
Fr!
V!

hup! Fe!

Figure 3.3: Cross section of a parallel plate electromechanical actuator

plate moving from its initial height hup down to a height h. The restoring
force Fr is then given by:
Fr = k.(hup − h) (3.2)
At equilibrium, the restoring force is equal to the electrostatic force:
Fr = Fe (3.3)
3.2. FORCES ON A PARALLEL PLATE ACTUATOR 27

1
k.(hup − h) = Q.E (3.4)
2
1 V
k.(hup − h) = C.V. (3.5)
2 h
C.V 2
k.(hup − h) = (3.6)
2.h

By expressing C from its geometry h in the value of the capacitance, and


neglecting the dielectric thickness, gives:

o .S.V 2
k.(hup − h) = (3.7)
2.h2
Finally, V can be written as a function of h:
s
2.k.h2 (hup − h)
V = (3.8)
o .S

Where S is the surface of the parallel plates.

Equation 3.8 is plotted on figure 3.4. It can be seen that there are
two possible heights for a single value of V . In practice, the behavior of a
parallel plate actuator can be described as follows. The initial height of the
gap is hup , and this gap is decreasing as the applied voltage is increased.
As the voltage is increased, the gap comes to a height equal to 2/3 of the
initial gap.
At this point, the parallel plate system becomes unstable, and the plate
comes down until its comes into contact with the dielectric layer on the
bottom electrode. The voltage for which the plate reaches 2/3 of its height
is called the pull − in voltage, denoted Vp .
Then, the voltage needs to be reduced down to a value called Vr for the
plate to come up again. This is a highly non linear behavior of the parallel
plate, and only the first third of the height permits linear movement of the
plate.
The pull-in voltage is given by:
s
8.k.h3up
Vp = (3.9)
27.0 .S

To compute the release voltage of the system, we first need to know the
CHAPTER 3. THE PARALLEL PLATE ELECTROSTATIC
28 ACTUATOR

V!

Vp!

Vr!

hdown! hup! h!

Figure 3.4: Applied voltage V versus gap height h

capacitances in the up and down states. The capacitance in the up state


between the plates is given by:
o S
Cup = (3.10)
hup

and

o r S
Cdown = (3.11)
hdown

where o is the permittivity of vacuum (8, 82.10−12 F m−1 )


r is the relative permittivity of the dielectric layer.

The electrostatic force generated can be written as:

1 V2
Fe = C (3.12)
2 h

In the down state, the contact between the upper plate and the dielectric is
not perfect and a small air gap remains between the plate and the dielectric,
as shown on figure 3.5.
The capacitance in the down state is given by:

o ef f S
Cdown = (3.13)
hdown
1 o r SV 2
Fe = (3.14)
2 h2down
3.2. FORCES ON A PARALLEL PLATE ACTUATOR 29

C2!
C2C1!
Cdown=!
!eff! C1 + C2 !
!!" C1!

Figure 3.5: Effects of the dielectric roughness on the actuator in the down
state, and equivalent modeling using two capacitances, C1 and C2 , in series.

The release voltage, Vr , is given by equalling the electrostatic force in the


down state with the restoring force Fr from the spring elongation.

Fr = k(hup − hdown ) = kgo (3.15)


and
1 o ef f SVr2
Fr = kgo = (3.16)
2 h2down
leading to
2kgo h2down
Vr2 = (3.17)
o ef f S
then v
u 2kgo h2down
u
Vr = t (3.18)
o ef f S

From equation 3.9, the pull-in voltage, Vp is increasing with the initial gap
h to the power 3/2, while the release voltage is increasing with the square
root of the initial gap, and linearly with the dielectric thickness.
It means that the mechanical restoring forces have more importance com-
pared to electrostatic forces when the initial gap is small. As we will see
later in the chapter, parallel plate electrostatic actuators are more reliable
for a given pull-in voltage if the initial electrostatic gap is small, which is
very counter intuitive.

Vr also depends on the capacitance in the down state and the height of
the plate in the down state. Modelling is difficult, since the discrepancy
between the theoretical value of the capacitance and the measured one may
come from a poor contact between the dielectric and the plate, but also
from a change in the dielectric layer relative permittivity. Overall, the best
possible approach is to determine the permittivity of the layer from on-wafer
CHAPTER 3. THE PARALLEL PLATE ELECTROSTATIC
30 ACTUATOR
MIM capacitors, and then compute the air gap giving the correct value for
C2 on figure 3.5.

Cdown!

1,5 Cup!

Cup!
-Vp! -Vr! Vr! Vp!
V!
O!

Figure 3.6: Capacitance to voltage characteristics of the parallel plate ac-


tuator

Finally, the first third of the movement of the parallel plate can be used
as a linearly varying capacitance. The maximum value that can be reached
by the capacitance, Cmax,linear from the capacitance at 23 of the initial height,
hup :
o S 3
Cmax,linear = 2 = Cup (3.19)
h
3 up
2

This type of operation must be used with caution, since the instability oc-
curs well before this height, and usually one can expect 10 percent variation
and not 50 percent theoretically as shown above.
The capacitance to voltage characteristics is shown on figure 3.6.
It is shown clearly that there is an hysteresis, and that the voltage needs
to be strongly reduced in order to release the gap.

Example 1
Let us consider a parallel plate actuator with square dimensions of 200x200µm.
The measured up state capacitance Cup is 100 fF, and the down state ca-
pacitance is Cdown is 3 pF. The dielectric layer is td =0,4µm thick, with a
relative permittivity r =10. The up state capacitance is:
o S
Cup = (3.20)
go
o r S 8, 82.10−12 .4.10−8
go = = = 3, 6µm (3.21)
Cup 10−13
3.2. FORCES ON A PARALLEL PLATE ACTUATOR 31

The measured pul-in voltage Vp is 35 V. The corresponding spring constant


k is:

27 2 S
k= V o = 32N.m−1 (3.22)
8 p g3
C1 is computed from the dielectric film characteristics:

o r .S 10.8, 82.10−12 .4.10−8


C1 = = = 8, 82pF (3.23)
td 0, 4.10−6
Next, C2 is given by:

Cdown C1
C2 = = 4pF (3.24)
C1 − Cdown

Then, the air gap value is extracted from C2

0 S
hair = = 0, 1µm (3.25)
C2
Then, the total height of the capacitance hdown is hair +hdiel =0.5 µm
ef f can then be computed from Cdown

Cdown .hdown
ef f = = 4, 25 (3.26)
o .S

The theoretical value for Vr can then be computed:


v
u 2kgo h2down
u
Vr = t = 9, 4V (3.27)
o ef f S

Example 2
Let s consider the parallel plate actuator shown below. The initial gap g0
is 2 µm, and the pull-in voltage Vp is 20 Volts. The surface of the actuator
is 100x100 µm.

• Compute the spring constant of the system.

• Compute the restoring force of the beam


CHAPTER 3. THE PARALLEL PLATE ELECTROSTATIC
32 ACTUATOR

k S=100x100 µm

g0

Figure 3.7: Sketch of the parallel plate actuator

The spring constant can be computed from eq. 3.9.

27 Vp2 .0 .S
k= 3
= 17N.m−1 (3.28)
8 g0

The restoring force can be computed from the spring elongation.

Fr = k.g = 36.2.10−6 µN = 34µN (3.29)

Next, the gap is reduced to 1 µm. The pull in voltage is the same.
The new spring constant is:

27 Vp2 .0 .S
k= = 136N.m−1 (3.30)
8 g03

The new restoring force is:

Fr = k.g = 136.10−6 µN = 136µN (3.31)

We can see that the restoring force scales as 1/g02 for a given pull in
voltage. Therefore, devices with small gaps have more reliability than de-
vices with large gaps. The downside is the reduced contrast between the
two states.

Example 3
We will consider the previous example, with an initial component with
Vpinitial =20 Volts, k = 17N.m−1 , and g0 = 2µm. The restoring force is 34
3.3. TRIPLE PLATE SYSTEMS 33

µN . The initial gap g0 is reduced to 1 µm.


Compute the new spring constant that gives the same restoring force with
this new gap. Compute the new pull-voltage,Vpnew .

The new spring constant is twice the initial value k = 34 N.m−1 .


s
8kg03
Vpinitial = (3.32)
270 S
Since the spring constant is twice the initial value, Vp is multiplied by
2 , but since the gap is twice less, Vp is also divided by 23/2 . Vp is therefore
1/2

divided by 2 if the gap is divided by 2, if the restoring force is maintained


constant.The new Vp is:
Vpinitial
Vpnew = (3.33)
2

Therefore, if the restoring force is maintained constant, the pull-in volt-


age scales linearly with the inverse of the initial gap. At the same time, the
spring constant is increasing linearly with the inverse of the initial gap.

3.3 Triple Plate Systems


Triple plate systems are of special importance for Radio Frequency applica-
tions. This type of structure indeed permits the fabrication of high power,
ultra linear RF variable capacitors. A cross section of the device is shown
in the figure below:

tdiel, εr

V h2

h1 tdiel, εr

Figure 3.8: Cross section of a triple parallel-plate electrostatic system

In the above figure, a moveable plate is inserted between two fixed par-
allel plates. The center plate is connected to ground, and bias voltages can
be applied to the two fixed plates. In the example above, structure is made
with two separation gaps h1 and h2 . The spring constant k is given by the
two springs on the sides of the moveable plate. Actuation of the parallel
CHAPTER 3. THE PARALLEL PLATE ELECTROSTATIC
34 ACTUATOR
plate by one of the two fixed plates is not different from a conventional two-
plate system, and separate actuation upwards or downwards can be easily
described by the equations given above. Vp1 and Vp2 can be computed from
the following equations:
s
8kh31
Vp1 = (3.34)
270 S
s
8kh32
Vp2 = (3.35)
270 S
Triple-plate systems differ when the moveable plate is actuated down to
one of the fixed plates. This situation is depicted below:

V1
h1+h2

Figure 3.9: Cross section of a triple parallel-plate actuator in an actuated


state

When V1 is applied, the electrostatic force that holding the plate in


contact with the bottom plate is given by the following equation:

1 o r SV12
Fe = (3.36)
2 t2diel

V1 V2
h1+h2

Figure 3.10: Cross section of a triple parallel-plate actuator in an actuated


state, with another voltage applied on the top layer

In a first approach, the restoring force generated by the suspension


springs will be neglected.
When a voltage is applied on the upper plate, the force generated is
given by the following equation:

1 o SV22
Fup = (3.37)
2 (h1 + h2 )2
3.3. TRIPLE PLATE SYSTEMS 35

The ”Pull-up” voltage is obtained from the following calculus:

Fup = Fe (3.38)

1 o SV22 1 o r SV12
= (3.39)
2 (h1 + h2 )2 2 t2diel

V22 r V12
= (3.40)
(h1 + h2 )2 t2diel

V22 r (h1 + h2 )2
= (3.41)
V12 t2diel

V2 √ (h1 + h2 )
= r (3.42)
V1 tdiel
Therefore, the voltage V2 required to overcome the force holding the
lower plate in contact, is generally much larger than the applied static
voltage V1 .

Example
A design example is given below:
tdiel = 0.6µm, h1 = h2 = 0.3µm, r = 10

V2 √ 0.6
= 10 = 3.33 (3.43)
V1 0.6
There is one final case to be studied, is when an RF power is applied on
the upper plate, like depicted above:

V1
VRF

Figure 3.11: Cross section of a triple parallel-plate actuator in an actuated


state, with an RF voltage applied on the top layer

We consider that the capacitance impedance is high, at the considered


frequency and the capacitance is mounted in parallel on a 50Ω system.

The RF voltage can be expressed from the incident power VRF = PRF .Z0
CHAPTER 3. THE PARALLEL PLATE ELECTROSTATIC
36 ACTUATOR
3.4 Scaling Rules
The restoring force is constant
Mechanical parameters Vp Restoring force k
3/2
Initial gap, g0 1 g0 k 1/2
Restoring force 1 g0 k

The pull-down voltage is constant


Mechanical parameters Vp Restoring force k
3/2 1/2
Initial gap, g0 1 g0 k
Restoring force 1 g0 k

Summary
• Small electrostatic gaps are the most efficient approach for the fabri-
cation of reliable components.

• The restoring force and pressure are the most important parameters
to consider.

3.5 Problems Chapter 3


• Let s consider the following example: We will consider a gold (E=80
GPa,γ=0,3) fixed-fixed beam, with l=200µm, w=100µm and t=0.5µm.
The beam has a built in tensile stress of 100 MPa at 25 C. The bot-
tom electrode is x=100µm. The electrostatic gap is 2 µm. The beam
is fabricated on Quartz, that has a negligible CTE compared to gold.
The CTE of gold is α=14,2.10−6 .m−1 .K −1 .

• Compute the spring constant

• Compute pull in voltage of the beam

• Compute the up state capacitance

• Cdown /Cup is 20. The dielectric layer is 0.2 µm thick. Compute the
effective permittivity of the dielectric layer.

• Compute the release voltage of the switch.


Chapter 4

Contact Physics

4.1 Introduction
Together with charge trapping, contact physics have been the most chal-
lenging issues to achieve large ohmic switches reliability. The most difficult
part to understand was that MEMS contact physics derives mostly from
static contact macroscopic models, and not from conventional macroscopic
dynamic models. For instance, arcing plays a preeminent role in relay
physics. In MEMS, there is no arcing because of the Pashen law, that was
mentionned earlier in the course. Therefore the durability of an electrical
contact in MEMS relays mostly depend on the contact force applied on the
metals, and the nature of the contact metals at play. This is similar in
many points to static contact physics, that has been studied by connector
engineering for instance.

4.2 Contact forces on a parallel plate


actuator
Let us consider the modified parallel plate actuator on Fig. 4.1. This paral-
lel plate actuator is actuated when a voltage Vapp is applied, but instead of
being stopped on a dielectric sheet, the device stops on a contacting dimple,
where the electrostatic force is concentrated. The dimple height is h2 and
the remaining gap is h1 . In the up state, the total height is h1 +h2 . We will
neglect the surface of the contact in the computations below, and consider
this actuator to be identical to the previous models.

The electrostatic force applied to the plate is given by:

37
38 CHAPTER 4. CONTACT PHYSICS

k!
Fr!
V! k!
h2! Fc! Fr!
Fe!
h1! Vapp!
h2!
Fe!

Figure 4.1: Parallel plate electrostatic actuator with a contact point. h2 is


the height of the contact dimple, h1 is the height of the air gap between the
contact point and the electrode.

CV 2 o SV 2
Fe = = (4.1)
2h 2h2

When the plate is in contact, the remaining electrostatic gap is h2 , and


at equilibrium, one can write:

Fc + Fr = Fe (4.2)

Where Fc is the contact force, i.e. the force applied on the contact point,
and Fr is the restoring force provided by the mechanical spring, and Fe the
electrostatic force.
This equation can be written as:
2
0 SVapp
Fc + kh1 = (4.3)
2h22

2
0 SVapp
Fc = − kh1 (4.4)
2.h22

Let s consider the case where Vapp =Vp . Vp2 is given by:

8 k.(h1 + h2 )3
Vp2 = (4.5)
27 0 .S
4.2. CONTACT FORCES ON A PARALLEL PLATE ACTUATOR 39

Then, the contacting force Fc is given by:

4 k.(h1 + h2 )3
Fc = − k.h1 (4.6)
27 h22

2
We can express this quantity as a function of Vapp /Vp2 :
2
4 Vapp k.(h1 + h2 )3
Fc = − k.h1 (4.7)
27 Vp2 h22

One interesting quantities that will be useful for further computations,


is the Fc /Fr ratio. This can be expressed as:
2
4 Vapp (h1 + h2 )3 h1
Fc = k.h2 .[ − ] (4.8)
27 Vp2 h32 h2

2
Fc h2 4 Vapp (h1 + h2 )3 h1
= ( − ) (4.9)
Fr h1 27 Vp2 h32 h2

This ratio does not depend on the spring constant of the device, nor
to the pull in voltage. It only depends on the geometry of the contacting
dimple and the electrostatic gap (h2 /h1 ). This ratio is directly impacting
the reliability of contacting devices. Ohmic contacting devices should have
a an Fc /Fr less than 3-4 to be reliable.

Example 1
Let s consider the case where h2 /h1 is 1. Then (h1 + h2 )/(h2 )=2
If Vapp is Vp then the Fc /Fr is (32/27 − 1) = 0, 185
if Vapp is 1.4xVp then Fc /Fr is (64/27 − 1) = 1, 37

Let s consider the case where h2 /h1 is 0,5. Then (h1 + h2 )/(h2 )=3
If Vapp is Vp then the Fc /Fr is 0, 5.(108/27 − 2) = 1
40 CHAPTER 4. CONTACT PHYSICS

if Vapp is 1.4xVp then Fc /Fr is 0, 5.(216/27 − 2) = 3

The last interesting case is where h2 /h1 is 2. Then (h1 + h2 )/(h2 )=1, 5
If Vapp is Vp then the Fc /Fr is 2.(27/(2.27) − 0.5) = 0
if Vapp is 1.4xVp then Fc /Fr is 0, 5.(1 − 0.5) = 0, 25

This last case is interesting only if the applied voltage is sufficient to gen-
erate enough contact force.

Example 2
We want to design to design switch with a contact force of 500 µN . The
fabrication process allows to use a 2 step sacrificial layer of 0,5 µm each
(h1 =h2 =0,5µm). The switch is to be used at 1,4 Vp .The actuator is a
200x200 µm2 plate.
1. Compute the theoretical Fc /Fr on this process. Deduce the restoring
force Fr , and the spring constant k.
2. Compute the theoretical pull-in voltage of the switch, Vp

From the previous example, Fc /Fr =1,37. Fr is therefore 500/1,37= 365


µN . The spring constant k is given by:

365.10−6
k= = 730N.m−1 (4.10)
0, 5.10−6

Vp is deduced from eq. 3.9:


s
8.730.(10−6 )3
Vp = = 24, 7V olts (4.11)
27.8, 82.10−12 .4.10−8

The applied voltage will be 34,8 Volts.

4.3 Pressure on a contact point


The force generated in such a system can be around 100µ N, which is
considerable when it is applied on a small surface like a contact point. For
instance, we will consider a contact point of 5x5 µ m, with 100µ N force
4.4. THE HERTZ CONTACT MODEL 41

applied, which is typical for an Ohmic switch. The contact pressure is given
by:

F
Pc = Pa (4.12)
S

The contact pressure is

100.10−6
Pc = Pa (4.13)
25.10−12

Pc = 4.106 P a = 4M P a (4.14)

gold to gold thermocompression:

1
Pc = P a = 100M P a (4.15)
10−8

4.4 The Hertz contact model


The Hertz contact model is the most simple contact model and it was
designed by Hertz in 1881, during a Christmas vacation. Let s consider two
metal contacting spheres, of radii R1 and R2 . The materials 1 and 2 have a
Young modulus E1 and E2 , and a Poisson coefficient ν1 and ν2 The contact
radius, a, is very small compared to the contacting spheres. The model
neglects friction and adhesion forces, and considers that the two materials
involved in the contact process are linear, and elastic. A contact scheme is
shown in Fig. :
δ is called the interf erence of the contact, and Fc is the force applied
on the contact.

3Fc
δ=( ∗ (1/2)
)(2/3) (4.16)
4E R
a, the contact radius is given by:
42 CHAPTER 4. CONTACT PHYSICS

Figure 4.2: Two contacting spheres.

E2,!2
R2
!

2a
E1,!1
R1

Figure 4.3: Cross section of two contacting spheres.

3Fc R (1/3)
a=( ) (4.17)
4E ∗
The effective radius, R, is given by:

1 1 1
= + (4.18)
R R1 R2
and E ∗ :

1 1 − ν12 1 − ν22
= + (4.19)
E∗ E1 E2
The resulting constriction resistance is given by:
ρ1 + ρ2
Rc = (4.20)
4a
where ρ is the arithmetic average of the two resistivities.
4.5. MULTIPLE CONTACT POINTS MODEL 43

Figure 4.4: Schematic description of electrical current constriction in a


metal-to-metal contact.

This is called the constriction resistance, which is often (and erroneously)referred


to as the contact resistance.
Since a is inversely proportional to the pressure applied on the contact,
the contact resistance is inversely proportional to the applied pressure on
the contact.

4.5 Multiple Contact Points Model


In practice, MEMS switches are based on multiple contact spots, and a
practical cross section can be seen in fig.

Figure 4.5: Two contacting surface with significant roughness.


44 CHAPTER 4. CONTACT PHYSICS

4.6 Contact Supertemperature


When two metals are brought in contact, the current that flows through the
contact causes the contact area to heat up, above the initial temperature of
the two metals brought in contact. This temperature is called the contact
supertemperature. Since the heat is produced by the current lines flowing
through the contact points, the temperature elevation is directly related to
the voltage drop in the contact.

The contact Voltage-Temperature relation


Upon current flow, an electrical contact is heated The maximum tempera-
ture of a contact, Tm , has been derived by Kohlrausch:
Z Tm
V = 2[2 λρdT ]1/2 (4.21)
T0
and Z Tm
2
V =8 λρdT (4.22)
T0
Where V is the voltage drop across a monometallic contact, T0 , is the metal
initial temperature, and Tm is the maximum temperature of the contact.
ρ and λ are the electrical resistivity and the thermal conductivity of the
materials, respectively.
If λ and ρ are independent of temperature, then (Tm −T0 ), the supertemperature
Ts of the metal contact, can be expressed as:
V2
Ts = (Tm − T0 ) = (4.23)
8λρ
It is noteworthy that this temperature depends only on the voltage drop
across the contact. Also, the supertemperature does not depend on the
contact geometry. When one wants to achieve a reliable contact, the tem-
perature of the contact must be maintained below the metal softening tem-
perature, and well below the melting point of the metal. This behavior
stems from the opposite variations of the metals thermal conductivity and
resistivity versus temperature, given by the Wiedeman-Franz law.

The Wiedeman-Franz Law


The Wiedeman-Franz Law states that the thermal, λ, and electrical re-
sistivity, ρ of a metal are changing with temperature accordingly to the
following relation:
λρ = LT (4.24)
4.6. CONTACT SUPERTEMPERATURE 45

Where L is the Lorentz constant (2.45 x 10−8 (V/K)2 ) and T is the absolute
temperature. From eq. 4.21, the temperature can be deduced from the
following rigorous relation:

V 2 = 4L(Tm2 − T02 ) (4.25)


These relations allow to compute the practical limitations of the use of
contact material for a given amount of current / or power. It is indeed
possible to compute a contact resistance from the constriction resistance,
and then the voltage drop across the contact from the power flowing through
the current. There is a critical voltage, called the softening voltage for a
given metal, that corresponds to the temperature point where the metal
starts to loose its mechanical properties, by reaching its softening point for
instance.

Examples
Gold softening voltage
The softening temperature for pure gold is 65o C. Compute the correspond-
ing softening voltage. λ=317 Wm−1−1 and ρ=1/(45,2.106 ) Ohm.m
The ambient temperature is 293K. Use eqs. 4.25 and 4.21.

Case study: the Radant MEMS switch


The radant MEMS Switch has an electrostatic actuator with dimensions
of about 80x20 um, separated by a 0.5 µm gap. The pull-in voltage is 60
Volts. The dimples are 0.3 µm deep. The switch is actuated at 90 Volts.
The dimple radius is 0.02 µm.

• Compute the spring constant of the switch.

• Compute contact force and the restoring force of the switch.

• Compute the constriction resistance at 90V, considering that the con-


tact metal is Gold (E=80 GPa, ν=0,3), and that the contact force is
equal on both contacts. Use a simple Hertz model.

• Compute the insertion loss of the switch.


46 CHAPTER 4. CONTACT PHYSICS

Figure 4.6: SEM view of the Radant MEMS switch.

You might also like