You are on page 1of 7

Home Search Collections Journals About Contact us My IOPscience

Analysis of electromechanical boundary effects on the pull-in of micromachined fixed–fixed

beams

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

2003 J. Micromech. Microeng. 13 S75

(http://iopscience.iop.org/0960-1317/13/4/312)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 139.80.2.185
The article was downloaded on 18/07/2013 at 05:09

Please note that terms and conditions apply.


INSTITUTE OF PHYSICS PUBLISHING JOURNAL OF MICROMECHANICS AND MICROENGINEERING
J. Micromech. Microeng. 13 (2003) S75–S80 PII: S0960-1317(03)59922-3

Analysis of electromechanical boundary


effects on the pull-in of micromachined
fixed–fixed beams
C O’Mahony1, M Hill2, R Duane1 and A Mathewson1
1
National Microelectronics Research Centre, Cork, Ireland
2
Cork Institute of Technology, Cork, Ireland
E-mail: comahony@nmrc.ie

Received 24 February 2003, in final form 23 April 2003


Published 13 June 2003
Online at stacks.iop.org/JMM/13/S75

Abstract
Using a commercial finite-element simulation tool, this work considers
some of the electromechanical effects commonly neglected during the
analysis of electrostatically actuated fixed–fixed beams. These structures are
used in many applications of micromechanical systems, from relay switches
and RF resonators to thin film characterization tests, but much of the
analytical modelling of the device behaviour disregards the effects of
electrostatic field fringing, plane-strain conditions and anchor compliance.
It is shown that the cumulative total of these errors can be substantial, and
may lead to large discrepancies in the expected operational characteristics of
the device. We quantify the influence of these effects on the electrostatic
pull-in of fixed–fixed beams, and illustrate some of the limitations of ideal
pull-in theory. In order to more accurately predict the pull-in voltage for a
real structure, a model is developed that combines ideal case theory with
anchor compliance correction factors extracted using finite-element
analysis. Three common anchor types (ideal, step-up and cup-style) are
characterized. The final model takes account of the compliance of the beam
anchors, electrostatic field fringing and plane-strain effects, and agrees well
with simulated results.

1. Introduction and these can cause substantial discrepancies between the


anticipated and measured operational characteristics of the
Micromachined fixed–fixed beams (figure 1) are commonly micromechanical system. In the case of thin film test
used in microelectromechanical applications such as relay structures, large errors in the extracted material properties may
switches [1, 2], HF frequency filters [3], mass flow sensors occur if the beam ends are mistakenly assumed to be ideally
[4], accelerometers [5] and resonators [6]. The beams are also clamped.
used to measure material properties of thin films, in particular This work concentrates on the use of fixed–fixed beams
Young’s modulus and residual stress, in conjunction with the for material property extraction, and uses a commercially-
electrostatic pull-in technique [7–9]. available FEM software tool to investigate three different types
The theory of electrostatic pull-in has been well of beam anchor. It is shown how the ideal (analytical) model
documented [10, 11], but much of this analytical analysis may be adapted to incorporate fringing fields, plane-strain
neglects electrostatic fringing fields and plane-strain effects. effects and anchor non-idealities.
More importantly, it is often assumed that the beam anchors
behave ideally, i.e. that the beam ends are held rigidly fixed in
all three degrees of freedom. Interferometric measurements 2. Electrostatic pull-in of ideally fixed–fixed beams
and FEM analysis indicate that this is often not the case
[12, 13]. Among the mechanical effects which may occur at Electrostatic pull-in is a well-known sharp instability in the
beam anchors are stress absorption and support deformation, behaviour of an elastically-supported structure subjected to

0960-1317/03/040075+06$30.00 © 2003 IOP Publishing Ltd Printed in the UK S75


C O’Mahony et al

where σx is the original (undeflected) uniaxial stress in the


beam, and σ is the additional stress induced by stretching
[15]. According to Hooke’s law, this is σ = Eε =
EL/L, where E is the Young’s modulus of the material,
and ε is the additional strain induced in the deflected
beam. The beam deflection profile is assumed to be w(x) =
w1 cos2 (π x/L ), where w1 is the maximum (central) deflection
[7]. This satisfies (1) and provides a reasonable approximation
to measured deflection profiles (figure 3).
The deflection has been analysed using fixed–fixed beams
etched from a layer of 1 µm sputtered titanium over a
polyimide sacrificial layer (figure 1). There is a slight deviation
in the beam profile just beside each beam anchor, visible
under interferometric inspection. Not all devices are fully
planarized, and this effect is due to the polyimide sacrificial
layer conforming to the gap between the beam anchor and
Figure 1. Titanium fixed–fixed beams. The device at the centre bottom electrode. Since the dip is small (∼0.15 µm) and the
measures 81 µm × 20 µm × 1 µm.
moment of the electrostatic force is small here, its effects are
neglected during simulation analysis.
The increase in beamlength due to stretching, L, may be
approximated as [15]
  
1 L dw 2 π2 2
L = dx = w (3)
2 0 dx 4L2 1
and so
bhLσx2 π 2 σx hbw12 π 4 hbEw14
Um (w) = + + . (4)
2E 4L 32L3
The first term represents the initial potential energy of the
Figure 2. Ideally fixed–fixed beam deflected due to applied voltage undeflected beam due to residual stress in the thin film from
across the bottom electrode and movable top plate. The vertical which the beam is etched, and remains constant for a given
deflection of the beam is exaggerated for clarity. structure. The second and third terms represent the additional
energy generated because of beam stretching, also known as
parallel-plate actuation. It is conveniently analysed using a the stress-stiffening effect. For the range of stresses (σx >
clamped–clamped beam of dimensions length L × b × h, 100 MPa) and dimensions commonly encountered in our
subject to an axial stress σx , suspended a distance g over a process, (4) is dominated by the first two terms, and the
fixed electrode (figure 2). contribution made by the third term is negligible. For
It is assumed that all deflections are small, and the simplicity, it is discarded throughout the remainder of this
electrode and beam are concentric and of the same size. analysis.
When a dc voltage, V, is applied across the two plates of The bending energy of the beam is [15]
the device, an attractive electrostatic force is created, causing   
EI L d2 w 2 π 4 EI w12
the upper movable electrode (beam) to move downwards Ub (w) = dx = , (5)
2 0 dx 2 L3
towards the lower fixed electrode. At a certain critical ‘pull-
in’ voltage, VPI , the downward electrostatic force is no longer where I = bh3/12 is the second moment of area of the beam.
balanced by the elastic restoring force provided by the beam Since the electrostatic energy stored in a capacitor is 12 CV 2 ,
anchors. The structure thus becomes mechanically unstable, the potential electrostatic energy of this system is
and spontaneously collapses or ‘pulls-in’ to the ground plane.  L/2
1 1
It is assumed that the beam ends are ideally fixed–fixed Uel (w) = − εo bV 2 dx
2 −L/2 g − w(x)
(also referred to as clamped or built-in), i.e. the ends of the
beam are fully constrained in all three degrees of freedom. Co V 2
=− , (6)
Mathematically, this may be expressed as 2(1 − w1 /g)1/2
∂w(x, y) where Co and g represent the initial (undeflected, zero voltage)
w(x, y) = =0 at x = 0, L, (1) capacitance and airgap between the plates, respectively. The
∂x
total potential energy of this system, , is therefore
where w(x, y) is the deformation of the beam in the vertical
z-direction from its unloaded position as a function of the  = Um + Ub + Uel . (7)
horizontal x- and y-coordinates [14]. The beam has a potential For a stable equilibrium to exist at a given displacement, the
energy due to stretching of total force acting on the beam must be zero, i.e. ddw1 = 0. The
2
σ 2 hbL (σx + σ )2 hbL system becomes unstable at ddw2 = 0, when V = VPI. Solving
Um = = , (2) 1
the second of these conditions, we find that this instability
2E 2E

S76
Analysis of electromechanical boundary effects on the pull-in of micromachined fixed–fixed beams

Height (microns)
3.5

2.5

2 Measured
1.5 w(x)

1
0 50 100 150 200
Length (microns)
Figure 3. Comparison between measured and assumed deflection profiles for a 200 × 20 µm2 fixed–fixed beam under a bias voltage of
48.8 V. The assumed deflection profile is w(x) = w 1cos2(π x/L), which provides a reasonable fit to the measured profile.

occurs when w1 = 13 g, i.e. pull-in occurs when the beam 55

Pull-in Voltage, Volts


centre has travelled through 13 of the original gap spacing.
50
Substitution of w1 = 13 g into ddw1 = 0 allows evaluation of
the voltage at that point (VPI), which for an ideal beam is
45
 Simulated
Eh3 g 2 g 3 bhσx
VPI = 11.7 3
+ 3.6 . (8) 40 Eqn (8)
Co L Co L Corrected
35
0 10 20 30 40 50 60
Beamwidth, microns
3. Corrections to ideal model
Figure 4. Pull-in voltages for a 200 µm long, 1 µm thick beam as
predicted by equation 10 (—), corrected for plane-strain effects
3.1. Electrostatic field fringing
(E ∗ = plate modulus) and field fringing effects ( ) and simulated ◦
The increase in device capacitance because of electrostatic (). Biaxial stress = 100 MPa, airgap = 1.5 µm. Error in
simulations is estimated to be 5%.
field fringing is often ignored during analysis of
microelectromechanical systems. This increase may be
1.05
substantial when the beamwidth becomes small in comparison 1
to the airgap, i.e. when the airgap/width ratio becomes 0.95
large. The parallel-plate approximation (Co = εLb/g) is no
VPI/VPI (ideal)

0.9
0.85
longer valid in this case. The larger capacitance causes a
0.8
corresponding increase in the electrostatic energy between the 0.75
Ideal
plates, reducing the voltage needed to achieve pull-in. To first 0.7
Step-up
0.65
order, fringing fields may be approximately compensated for Cupped
0.6
by using an ‘effective width’,
 beff,in calculations such as those 0 100 200 300 400 500 600
above, where beff = b 1 + 0.65 gb [8]. Beamlength, microns

Figure 5. Example of simulated VPIV(ideal)


PI
as a function of length for
3.2. Plane-strain effects cupped and step-up anchors. Material properties are σo = 225 MPa,
E = 110 GPa, v = 0.33 MPa. All beams are 10 µm wide.
It is customary to compensate for width effects when dealing
with wide beams by replacing the Young’s modulus of the
material by the effective modulus E ∗, where E ∗ = E for narrow 3.3. Anchor compliance
beams and E ∗ = E/(1−v) for wide beams. Beams are usually
considered ‘wide’ when b > 5h and L  b, L  h. The uniaxial No beams are ideally fixed–fixed, and the increased
stress σx is replaced by σo (1 − v), where v is the Poisson’s compliance and flexibility of real anchors indicate that
ratio of the beam material, and σo is the original biaxial stress expression (8) should be taken as an upper bound for the
in the thin film from which the beam is etched [8]. pull-in of fixed–fixed beams, since real anchors allow more
The influence of these field fringing and plane-strain rotation and stress absorption than the ideal ones. This leads
correction factors on the pull-in model (8) is illustrated in to lower pull-in voltages (figure 5).
figure 4. The corrected model is in reasonable agreement with To complete the model, we use CoventorWare 2001.3,3
simulated results, except where the beam width to thickness to analyse three common types of beam anchor: the ideal
and airgap ratios approach unity. The results indicate that care case (figure 2), the simple step-up support (figure 6) and the
should be taken to ensure that, where possible, microstructures cup-shaped anchor (figure 7).
should be designed in the ‘wide beam’ regime in order to 3CoventorWare version 2001.3, Coventor Inc., 4001 Weston Parkway, Cary,
eliminate these problems. NC 27513.

S77
C O’Mahony et al

Figure 8. Mesh detail of a cupped anchor. Each model was meshed


with an average of 6000 panels.

Having first determined c1, the second compliance factor


c2 is evaluated from the slope of a straight line fitted to a plot
of VPI2 versus σo , through an offset determined by c1. A typical
Figure 6. Step-up beam anchor. In this case, the beam is fabricated plot is shown in figure 10. Although both factors may be
from 0.9 µm thick polysilicon.
simultaneously evaluated from a curve fitted to equation (9),
the errors in both may be reduced by decoupling the two sets
of simulations.
The simulations detailed above yield the following anchor
compliance factors, completing equation (9).

4. Analysis

Note that for the ideal case, the value of c1 extracted using
FEA is higher than that predicted by the analytical case (8).
This is due to the influence of the higher-order stretching
π 4 hbEw4
energy term 32L3 1 in (4). For the micromachining process of
interest here, the error caused is small, as the pull-in voltage is
dominated by the residual stress terms (σo ∼ 200 MPa), and so
this term has been neglected throughout this work. However,
Figure 7. Cup-style anchor design. The titanium beam is 10 µm it serves as an indication that this effect may need to be more
wide and 1.0 µm thick. closely examined as the thin film stress tends towards 0 (σo ∼
10 MPa). For thin film characterization purposes, the Young’s
Since the principles of actuation remain the same, we modulus of a material remains reasonably constant despite
expect that the pull-in voltage for each case may be predicted variations in processing conditions. Because of this, and the
by an expression of the same form as (8), but will reflect influence of the residual stress on pull-in voltages, the first
the increased elasticity and compliance of the step-up and term in (9) is often of little interest. In contrast, the stress in a
cup-style beam anchors. This is done by re-evaluating the thin film often varies enormously with deposition parameters,
numerical compliance factors, c1 and c2, as outlined below. To anneals and back-end processing steps [16, 17], and accurate
verify this approach, the ideal case anchor is also investigated. determination of c2 is far more critical and of more interest to
The corrections for elecrostatic field fringing and plane-strain
microsystems designers.
effects discussed earlier are also included in (9),
  Beams that employ a cup-style anchor are subject to
1 E ∗ h3 g 3 g 3 hσ further effects caused by the two outside anchor walls, which
VPI = g c1 4
+ c2 (1 − υ) . (9)
1 + 0.65 d εL εL2 have the effect of stiffening the anchor. These effects
The compliance factor c1 was extracted by assuming arbitrary may become particularly pronounced as the beam becomes
mechanical properties (in this case, titanium; E = 110 GPa, narrower and the anchor walls make up a greater fraction of
v = 0.33), and zero residual stress, and fitting a straight line to the overall beamwidth (figure 11). In each case, the width of
a plot of simulated VPI2 versus 1/L4, repeated for each anchor the anchor walls remains constant, and the width of the anchor
type. All devices were meshed with approximately 6000 cup is equal to the beamwidth minus the width of the anchor
panels (figure 8) and from mesh studies, results are estimated walls (figures 7 and 8). This increased stiffness is reflected
to be accurate to ±5%. in an increase in pull-in voltage as the beamwidth decreases.
A typical c1 extraction plot is shown in figure 9. Structural In cases such as this, the compliance factors may need to be
dimensions are typical for MEMS devices; beam width is re-evaluated at each specific beamwidth in order to cater for
20 µm, thickness is 1 µm and gap height is 1.5 µm. the influence of the anchor walls.

S78
Analysis of electromechanical boundary effects on the pull-in of micromachined fixed–fixed beams

7.E+03 Ideal
6.E+03 y = 6.40E-13x
Stepup y = 5.96E-13x
5.E+03

VPI2 (V2)
4.E+03 Cupped
3.E+03 y = 4.70E-13x

2.E+03
1.E+03
0.E+00
0.E+00 2.E+15 4.E+15 6.E+15 8.E+15 1.E+16 1.E+16
4 -4
1/L (m )
Figure 9. Extraction data for c1. The constant is found from the slope of the line.

5000 V2 (L = 200 microns) ideally clamped device, and we have shown that substantial
V2 (L = 500 microns)
differences exist between the performances of different beam
4000
anchors. This is due to the stress absorption and increased
compliance that occur in these regions. Three common anchor
VPI2, V2

3000
y = 1.03E-05x + 2.93E+02
types (ideal, step-up and bucket-style) have been investigated,
2000 and it has been estimated that a step-up anchor design may
y = 1.64E-06x + 7.50E+00 reduce the pull-in voltage by as much as 25% when compared
1000
with an ‘ideal’ anchor.
0 In order to more accurately predict the pull-in voltage for
0.E+00 1.E+08 2.E+08 3.E+08 4.E+08 5.E+08 a real structure, an approach has been developed that combines
Stress, MPa ideal case theory with electrostatic field fringing compensation
and anchor compliance correction factors extracted using
Figure 10. Extraction plot for the compliance factor c2. This is data finite-element modelling. The final model takes into account
for the step-up anchor; other anchor designs are evaluated in the the compliance of the beam anchors, electrostatic field fringing
same way. Two beamlengths are used to verify the result. and plane-strain effects.
80
70 References
Pull-in voltage, Volts

60
50 [1] Huang J-M, Liew K M, Wong C H, Rajendran S, Tan M J and
40 microns
40 Liu A Q 2001 Mechanical design and optimization of
20 microns
30 capacitive micromachined switch Sensors Actuators A 93
10 microns
20
273–85
10
[2] Muldavin J B and Rebeiz G M 2000 High-isolation CPW
MEMS shunt switches: Part 1. Modeling IEEE Trans.
0
Microw. Theory Tech. 48 1045–52
0 100 200 300 400 500 600
[3] Nguyen C T-C, Katehi L P B and Reibeiz G M 1998
Length, microns Micromachined devices for wireless communications Proc.
IEEE 86 1756–68
Figure 11. Measured pull-in voltages for titanium fixed–fixed [4] Howe R T and Muller R S 1986 Resonant-microbridge vapor
beams of different widths. The influence of the anchor walls is sensor IEEE Trans. Electron Devices ED-33 499–506
apparent. [5] Burns D W, Horning R D, Herb W R, Zook J D and Guckel H
1996 Sealed-cavity resonant microbeam accelerometer
Table 1. Extracted compliance correction factors. Sensors Actuators A 53 249–55
Ideal Cupped anchor Step-up Analytical [6] Ahn Y, Guckel H and Zook J D 2001 Capacitive microbeam
anchor (b = 20 µm) anchor case (equation (8)) resonator design J. Micromech. Microeng. 11 70–80
[7] Zou Q, Li Z and Liu L 1995 New methods for measuring
c1 14.1 13.1 10.3 11.7 mechanical properties of thin films in micromachining:
c2 3.4 2.3 1.7 3.6 beam pull-in voltage method and long beam deflection
method Sensors Actuators A 48 137–43
[8] Osterberg P M and Senturia S D 1997 M-Test: a test chip for
5. Conclusions MEMS material property measurement using
electrostatically actuated test structures
J. Microelectromech. Syst. 6 107–17
This work illustrates the manner in which several modifications [9] Gupta R K, Osterberg P M and Senturia S D 1996 Material
may be made to the ideal pull-in model in order to take property measurements of micromechanical polysilicon
into account effects such as electrostatic field fringing, finite beams Proc. SPIE 2880 39–45
beamwidth and anchor compliance. The data of table 1, and the [10] Ijntema D J and Tilmans H A C 1992 Static and dynamic
measurements of figure 11, highlight the dangers of assuming aspects of an air-gap capacitor Sensors Actuators A 35
121–8
an ideal anchor during analysis of pull-in tests. [11] Nemirovsky Y and Bochobza-Degani O 2001 A methodology
The results show that large errors in device performance and model for the pull-in parameters of electrostatic
or measurement data may result from the assumption of an actuators J. Microelectromech. Syst. 6 601–15

S79
C O’Mahony et al

[12] Kobrinsky M J, Deutsch E R and Senturia S D 2000 Effect of [15] Young W C 1989 Roark’s Formulas for Stress and Strain
compliance and residual stress on the shape of doubly 6th edn (New York: McGraw-Hill)
supported surface-micromachined beams [16] Berney H, Hill M, Hynes E, O’Neill M and Lane W A
J. Microelectromech. Syst. 9 361–9 2000 Investigation of the effect of processing steps
[13] Jensen B D, Bitsie F and de Boer M 1999 Interferometric on the stress in a polysilicon structural
measurements for improved understanding of boundary membrane J. Micromech. Microeng. 10
effects in micromachined beams Proc. SPIE 223–34
Micromachining and Microfabrication (Santa Clara, CA) [17] Windischmann H 1992 Ìntrinsic stress in sputter-deposited
[14] Gere J M and Timoshenko S P 1999 Mechanics of Materials thin films Crit. Rev. Solid State Mater. Sci. 17
4th edn (Cheltenham, UK: Stanley Thornes) 547–96

S80

You might also like