Research Collection

Doctoral Thesis

Modeling dielectric elastomer actuators

Author(s):
Wissler, Tobias

Publication Date:
2007

Permanent Link:
[Link]

Rights / License:
In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For more
information please consult the Terms of use.

ETH Library
DISS. ETH NO. 17142

MODELING DIELECTRIC ELASTOMER

ACTUATORS

A dissertation submitted to the

SWISS FEDERAL INSTITUTE OF TECHNOLOGY in ZURICH

for the degree of

Doctor of Sciences

presented by

MICHAEL TOBIAS WISSLER

Dipl. Werkstoff.-Ing. ETH

born 25th October 1976

citizen of Sumiswald (BE)

accepted on the recommendation of

Prof. Dr. E. Mazza, examiner
Prof. Dr. P. Ermanni, co-examiner
Prof. Dr. M. Farshad, co-examiner
Dr. G. Kovacs, co-examiner

2007
Acknowledgments

I wish to express my sincere gratitude to

¾ My supervisor Prof. Dr. Edoardo Mazza, for his kind support and encouragement
during this work, the many fruitful discussions and his interest in my research
work.

¾ My co-examiners, Prof. Dr. Paolo Ermanni, Prof. Dr. Mehdi Farshad and Dr.
Gabor Kovacs for carefully reviewing this thesis.

¾ All my colleagues and supervisors at the laboratory for Materials and

Engineering at Empa for the good atmosphere, especially: Silvain Michel, Patrick
Lochmatter, Rui Zhang, Christian Dürager, Claudio Iseli and Lukas Kessler for
the interesting and helpful discussions concerning dielectric elastomers; Alfred
Schmidlin, Urs Hintermüller, Bernd Jähne and Kurt Gantner for technical
support; Dr. Luc Wullschleger, Christian Affolter and Bernhard Weisse for
assisting me with finite element simulations.

¾ Jonas Wyrsch, Riya Bhattacharya, Patrick Grau, Marcel Birrer and Julien Egger
for writing their semester thesis on topics directly related to my research.

¾ Dr. Alessandro Nava for performing the aspiration experiments.

¾ The Swiss National Science Foundation (SNF) for the financial support of this
thesis (Project number: 200021-107661/1).

¾ My family and my friends for their encouragement and support. This work is
dedicated to my parents.

Michael Wissler
Zürich, April 2007
Table of contents

Abstract v

Zusammenfassung vii

1 Introduction 1
1.1 General remarks ....................................................................................... 1
1.2 Design of dielectric elastomer actuators .................................................. 2
1.3 Constitutive equations .............................................................................. 4
1.4 Literature review on modeling of DE-actuators ...................................... 5
1.5 Objectives of the present work................................................................. 8

2 Theoretical Background and Material Details 11

2.1 Nonlinear continuum mechanics............................................................ 11
2.1.1 Finite strains........................................................................................ 11
2.1.2 Stresses ............................................................................................... 13
2.1.3 Strain energy functions and nearly incompressible materials ............ 14
2.1.4 Quasilinear viscoelasticity.................................................................. 15
2.2 Optimization procedure.......................................................................... 16
2.3 Material details ....................................................................................... 17
2.3.1 Elastomers........................................................................................... 17
2.3.2 Electrodes ........................................................................................... 18

3 Modeling Dielectric Elastomer Actuators 19

3.1 Article 1 .................................................................................................. 19
3.2 Article 2 .................................................................................................. 37
3.3 Article 3 .................................................................................................. 59
3.4 Article 4 .................................................................................................. 88

i
4 Further Verification of the Actuator Model 111
4.1 Introduction .......................................................................................... 111
4.2 Correction of material parameters........................................................ 112
4.3 Aspiration tests ..................................................................................... 114
4.3.1 Introduction....................................................................................... 114
4.3.2 Working principle of the aspiration device ...................................... 114
4.3.3 Description of the experiments......................................................... 115
4.3.4 Experimental results ......................................................................... 117
4.3.5 Finite element calculations ............................................................... 117
4.3.6 Results............................................................................................... 118
4.4 Compression tests................................................................................. 120
4.4.1 Introduction....................................................................................... 120
4.4.2 Experimental details ......................................................................... 121
4.4.3 Finite element simulations................................................................ 122
4.4.4 Results............................................................................................... 123
4.5 Deformation state in the experiments .................................................. 124
4.5.1 Introduction....................................................................................... 124
4.5.2 Uniaxial tensile tests ......................................................................... 126
4.5.3 Circular strain test............................................................................. 126
4.5.4 Aspiration test................................................................................... 127
4.5.5 Compression test............................................................................... 128
4.5.6 Discussion......................................................................................... 129
4.6 Influence of the passive part of a spring roll activation....................... 130
4.6.1 Introduction....................................................................................... 130
4.6.2 Analytical model............................................................................... 131
[Link] Kinematics ................................................................................... 132
[Link] Kinetics ........................................................................................ 134
[Link] Electromechanical coupling......................................................... 135
4.6.3 Numerical considerations ................................................................. 136
4.6.4 Results............................................................................................... 138

ii
5 Conclusions and Outlook 140
5.1 Main contributions of the present work ............................................... 140
5.2 General design considerations and suggestions for future work ......... 142
5.2.1 Material behavior of the elastomer................................................... 142
5.2.2 Geometrical design of the actuator................................................... 143
5.2.3 Reliability ......................................................................................... 144

Bibliography 147

Curriculum Vitae 153

iii
iv
Abstract
Dielectric elastomer actuators belong to the group of so called “electroactive
polymers” and are used in active structures where large strains are required. Dielectric
elastomers have many potential applications as actuators. Models and simulation
techniques are required for the design and optimization of actuator applications in
order to determine states of stress and strain as functions of the applied voltage.
Several aspects have to be considered in the modeling and simulation of dielectric
elastomer actuators, including the constitutive model of the elastomer, the
electromechanical coupling, numerical implementation and experimental validation of
the models. So far no adequate models exist which are able to predict the actuator
behavior over a large deformation and time range.
In the present work, actuators made of an acrylic elastomer (the widely used material
VHB 4910 by 3M) are investigated. The passive mechanical response of the elastomer
is decoupled from the electromechanical problem. Electrostatic forces arise only at the
interface between the electrodes and the elastomer. The elastomer behaves as a passive
layer and there is no direct interaction between electric field and its mechanical
properties. The electromechanical coupling, an equation proposed by Pelrine et al.
(Sensors and Actuators A, 64, 1998), is evaluated by analytical and numerical
methods. The dielectric constant of VHB 4910 is determined by LCR measurements
for various prestrain levels as well as by spring roll experiments. The passive
mechanical response of the elastomer is described using a ‘quasilinear visco-
hyperelastic model’ suitable for large strains and viscoelasticity. The constitutive
model and the electromechanical coupling are implemented in finite element models
for simulating the behavior of circular actuators. A systematic experimental
characterization of circular actuators under various prestrain and voltage conditions
has been carried out. Over 40 actuators have been tested at 11 different
prestrain/voltage levels. These experimental data are essential for model definition and
validation purposes. The constitutive model is further evaluated using different
deformation configurations. Uniaxial tensile and relaxation tests, aspiration tests and
compression tests have been carried out and compared to corresponding simulations.
The mechanical model (with an optimized parameter set) is valid over a wide
deformation and time range, as shown for circular actuator tests at different prestrain
and voltage levels as well as for uniaxial, aspiration and compression tests. The
concept of quasilinear viscoelasticity was validated by uniaxial relaxation tests which
showed that the normalized stress-relaxation function is independent of the magnitude
of the deformation. For the simulation of the circular actuators, a novel finite element

v
technique was developed which allows direct simulation of the actuator activation by
the applied voltage. An analytical model was derived for circular actuators. An
equation was obtained that links the voltage, the prestrain and the active strain. The
analytical model is valid for hyperelastic materials (time-dependent effects are
neglected) and can be used for the design of silicone actuators, with less pronounced
viscoelastic behavior. It has been demonstrated that fitting the strain energy forms of
Yeoh, Ogden and Mooney-Rivlin to uniaxial data leads to different simulation results
in the VHB 4910 actuator behavior. This illustrates the importance of characterizing
the biaxial response of the elastomer. This step was omitted in previous work on
dielectric elastomers, where material models were based only on uniaxial test data.
The numerical analysis for the evaluation of the electromechanical coupling provides
the charge, electric field and electrostatic force distribution for a circular actuator. It is
found that the electromechanical pressure acts in thickness direction and with the same
magnitude in lateral direction. By considering a superimposed hydrostatic stress state,
the resulting out-of-plane pressure corresponds to the equation proposed by Pelrine et
al. This analysis provides a new physical interpretation for the electromechanical
coupling. Measurements of the dielectric constant demonstrated that its value
decreases by increasing prestrain. The dielectric constant of VHB 4910 is about 3.2 in
the relevant prestrain range and not 4.7 as was proposed by Kofod et al. (Journal of
Intelligent Material Systems and Structures, 14, 2003) and used by many researchers.
This result was also confirmed by spring roll experiments.
The results of the present work represent fundamental contributions to the modeling
and characterization of dielectric elastomer actuators. On the basis of these findings
suitable numerical models can be defined for the design and optimization of general
actuator applications. The next challenge is to describe the failure behavior of these
material systems. These investigations are necessary in order to achieve the
commercialization of actuator applications.

vi
Zusammenfassung
Dielektrische Elastomere gehören zur Gruppe der so genannten „elektroaktiven
Polymere“ und werden in aktiven Strukturen für die Erzeugung von grossen
Dehnungen verwendet. Dielektrische Elastomere verfügen über verschiedene
Anwendungen als Aktuatoren. Für die Auslegung und die Optimierung solcher
Aktuatoren sind Modelle und Simulationstechniken erforderlich, welche mechanische
Spannungs- und Dehnungszustände in Abhängigkeit der angelegten elektrischen
Spannung beschreiben. Verschiedene Aspekte sind bei der Modellierung und
Simulation von Aktuatoren aus dielektrischen Elastomeren zu berücksichtigen: Das
konstitutive Modell des Elastomers, die elektromechanische Kopplung, numerische
Implementierung der Modelle und ihre experimentelle Verifizierung. Bis anhin
existieren keine geeigneten Modelle, um das Aktuator-Verhalten über grosse
Deformations- und Zeitbereiche zu beschreiben.
In der vorliegenden Arbeit wird das Material VHB 4910 von 3M untersucht, welches
sehr häufig für dielektrische Elastomer-Aktuatoren verwendet wird. Das passive
mechanische Verhalten des Elastomers wird vom elektromechanischen Problem
getrennt betrachtet. Elektrostatische Kräfte treten nur beim Übergangsbereich
zwischen Elektroden und Elastomer auf. Das Elastomer verhält sich wie eine passive
Schicht und keine direkte Wechselwirkung zwischen elektrischem Feld und den
mechanischen Eigenschaften treten auf, deshalb ist diese getrennte Betrachtung
zulässig. Die elektromechanische Kopplung wurde mit der Gleichung von Pelrine et
al. (Sensors and Actuators A, 64, 1998) berücksichtigt. Die Gültigkeit dieser
Gleichung wurde hier mit analytischen und numerischen Methoden untersucht. Die
dielektrische Konstante von VHB 4910 wurde mit LCR-Messungen bei
unterschiedlichen Vordehnungen und mit Zylinderaktuator-Experimenten ermittelt.
Das passive mechanische Verhalten des Elastomers wird mit einem konstitutiven
Modell, dem ‚quasi-linearen visko-hyperelastischen Modell’, beschrieben. Dieses
Modell berücksichtigt grosse Dehnungen und Viskoelastizität. Für die Simulation von
Kreisaktuatoren wurde das konstitutive Modell und die elektromechanische Kopplung
in ein Finite Elemente Modell implementiert. Eine systematische, experimentelle
Charakterisierung von Kreisaktuatoren bei verschiedenen Vordehnungen und
elektrischen Spannungen wurde durchgeführt. Über 40 Aktuatoren wurden bei 11
unterschiedlichen Vordehnungs-/Spannungsbedingungen gestestet. Einachsige Zug-
und Relaxationsversuche, Aspirationsversuche und Kompressionsversuche wurden
durchgeführt und mit den entsprechenden Simulationen verglichen.

vii
Das mechanische Modell (einschliesslich den entsprechenden Parametern) ist über
einen weiten Deformations- und Zeitbereich gültig. Dies wurde sowohl für
Kreisaktuatoren bei unterschiedlichen Vordehnungen und elektrischen Spannungen als
auch für einachsige Versuche, Aspirationsversuche und Kompressionsversuche
gezeigt. Das Konzept der quasilinearen Viskoelastizität wurde mit einachsigen
Relaxationsversuchen verifiziert, in denen gezeigt wurde, dass die mechanische,
relative Spannungsrelaxationsfunktion unabhängig vom Ausmass der Deformation ist.
Für die Simulation der Kreisaktuatoren wurde eine neuartige Finite Elemente Technik
entwickelt, welche es erlaubt, eine direkte Simulation mit der elektrischen Spannung
als Input-Parameter durchzuführen.
Ein analytisches Modell wurde für Kreisaktuatoren hergeleitet, welches die elektrische
Spannung, die Vordehnung und die aktivierte Dehnung verknüpft. Das analytische
Modell ist für hyperelastische Materialien gültig (zeitabhängige Effekte wurden
vernachlässigt) und könnte sehr nützlich für das Design von Silikon-Aktuatoren sein,
welche kein ausgeprägtes viskoelastisches Verhalten aufweisen. Es wurde gezeigt,
dass die Kalibrierung des Modells anhand von einachsigen Daten für die
Dehnungsenergiepotentiale von Yeoh, Ogden und Mooney-Rivlin zu
unterschiedlichen Resultaten im Aktuatorverhalten führt. Dieses Ergebnis betont die
Wichtigkeit, das biaxiale Verhalten des Elastomers zu charakterisieren. Dies wurde in
vorhergehenden Arbeiten über dielektrische Elastomere vernachlässigt. Die
numerische Analyse des elektromechanischen Verhaltens liefert die Ladungs-,
elektrische Feld- und elektrostatische Kräfteverteilung für den Kreisaktuator. Es wurde
gezeigt, dass der elektromechanische Druck in Dickenrichtung mit demselben
Ausmass auch in seitlicher Richtung wirkt. Bei der Betrachtung eines überlagerten
hydrostatischen Spannungszustandes entspricht der resultierende Druck in
Dickenrichtung der Gleichung von Pelrine et al. Diese Analyse liefert eine neue
physikalische Interpretation der elektromechanischen Kopplung. Kapazitätsmessungen
zeigten eine Abnahme der dielektrischen Konstante bei zunehmender Vordehnung.
Die dielektrische Konstante von VHB 4910 hat einen Wert von ungefähr 3.2 im
relevanten Vordehnungsbereich und nicht 4.7 (wie von Kofod et al. publiziert (Journal
of Intelligent Material Systems and Structures, 14, 2003) und von vielen Forschern
verwendet). Dieses Ergebnis wurde durch Zylinderaktuator–Experimente bestätigt.
Die Resultate der vorliegenden Arbeit liefern grundlegende Beiträge betreffend
Modellierung und Charakterisierung von Aktuatoren aus dielektrischen Elastomeren.
Für die Auslegung und die Optimierung von Aktuator-Anwendungen können die
vorliegenden Ergebnisse mittels passender numerischer Modelle verwendet werden.
Die nächste Herausforderung stellt die Untersuchung von Schadensmechanismen dar,
welche wichtig für die Kommerzialisierung von Aktuator-Anwendungen ist.

viii
1 Introduction

1.1 General remarks

Electroactive polymers (EAP) are functional materials that can be used as actuators in
active structures, in particular when large deformations are required. EAP actuators
transform electrical energy directly into mechanical work and produce large strains, in
the order of 10% to 30%.
There are two classes of so called “electroactive polymers”: electric EAPs [Kim 2000,
Pelrine 1998, Pelrine 2000a, Su 1999, Zhang 1998 and Zhang 2002] and ionic EAPs
[Baughman 1999, Calvert 1998, Gandhi 1995 and Shahinpoor 2001]. The mode of
action of electric EAPs is based on Coulomb forces generated by electric fields. In
ionic EAPs, the mobility or diffusion of ions is exploited to produce large elongations.
The rapid development in EAP technology that has occurred in the last ten years is the
result of collaboration between scientists from various disciplines such as mechanics,
physics, materials science, chemistry and electrical engineering.
One class, the dielectric elastomers (DE), belongs to the electric EAPs and has
outstanding properties. The combination of large elongation, high energy density,
good efficiency and high speed of response [Bar Cohen 2001 and Pelrine 2000a] is
unique to dielectric EAPs and is essential to certain applications such as mobile
(micro-) robots.
Actuators made of dielectric elastomers consist basically of a capacitor with a thin
passive elastomer film sandwiched between two compliant electrodes [Pelrine 1998],
see Figure 1.1. According to [Pelrine 1998], the electrostatic pressure pel acting on the
insulating elastomer film can be calculated for a given applied voltage, U, and film
thickness, d:
2
⎛U ⎞
pel = ε 0 ⋅ ε r ⋅ ⎜ ⎟ (1.1)
⎝d⎠

where ε0 is the free-space permittivity (8.85•10-12 As/Vm) and εr is the relative

dielectric constant of the elastomer. Due to the mechanical compression, the elastomer
film contracts in the thickness direction and expands in the film's planar directions.
Ideally, the electrodes have to behave as very compliant layers in order to present a
minimum of resistance to the deformation. This in-plane expansion is exploited to

1
generate motion or forces. Dielectric elastomers are pre-stretched, typically up to five
times their in-plane dimensions, in order to improve their performance.

Electrodes

Elastomer U=0

Voltage off

pel

d U>0

Voltage on

Figure 1.1. Working principle of a dielectric elastomer actuator: activation with a high
voltage leads to vertical contraction and in-plane expansion

Dielectric elastomers have many potential applications as actuators. Examples of such

actuators include artificial muscles [Kovacs 2006, Pei 2003 and Pelrine 2002], mini-
and micro-robots [Pelrine 2001a], color displays [Aschwanden 2006], electroacoustic
transducers [Heydt 1998], in aircraft [Michel 2006], diaphragm actuators for pumps
[Ashley 2003 and Pelrine 2001a] and generators [Pelrine 2001b].
It has to be pointed out that the denotation “electroactive polymer” is somewhat
misleading in the case of dielectric elastomers. In fact the polymer itself behaves as a
passive layer. In this sense, it would be more accurate to denote a dielectric elastomer
actuator as an “electroactive system”.

1.2 Design of dielectric elastomer actuators

For the design and optimization of actuator applications, models and simulation
techniques are required in order to determine stress and strain states as functions of the
applied voltage. There are various aspects which have to be considered in the
modeling and simulation of dielectric elastomer actuators:

2
(i) Constitutive model of the elastomer. The constitutive model is the link between
the stress and strain states of the elastomeric film. The preferred elastomer for DE
actuators is VHB 4910 (3M), see section 2.3. This material undergoes large
strains, shows a dissipative behavior and is incompressible [Pelrine 1998, Pelrine
2000a and Kofod 2001a]. In order to model such a material, a constitutive model
is required which describes the large strains and time-dependence of the
mechanical response with an appropriate mathematical formulation, as well as the
corresponding material parameters. More information about the constitutive
model is given in section 1.3.
(ii) Electromechanical coupling. The relationship between the applied voltage and
the stress and strain states in the actuator has to be described with suitable
equations.
(iii) Numerical models. Numerical models are required for the design and
optimization of ‘real’ actuators. The constitutive model of the elastomer and the
electromechanical coupling equations have to be implemented into a numerical
model for simulation of the actuator behavior.
(iv) Experimental validation. The formulation and the verification of the constitutive
model, the electromechanical coupling and the numerical model require
experimental data. Experiments have to be representative of relevant actuator
configurations.
A general simplification for modeling dielectric elastomers proposed by Bhattacharya
in [Bar-Cohen 2001] is that the passive mechanical response of the elastomer can be
decoupled from the electromechanical problem. Electrostatic forces arise only at the
interface between the electrodes and the elastomer. The elastomer behaves as a passive
layer and no direct interaction occur between electrical field and its mechanical
properties. This fact was confirmed experimentally by [Kofod 2003], who
demonstrated that electrostrictive effects are negligibly small for VHB 4910. Thus, the
constitutive model of the elastomer includes only purely mechanical equations while
the electromechanical coupling is considered through kinetic boundary conditions
representing the electrostatic forces.
The generally accepted approach to electromechanical coupling in dielectric elastomer
actuators is described by eq. (1.1). This equation was proposed by [Pelrine 1998] and
verified experimentally for silicone actuators by [Kofod 2005]. No comprehensive
analysis of this expression has been carried out so far.
The determination of suitable constitutive equations capable of describing the large
strain response (with prestrain) and time-dependence over a large deformation range

3
represents the main challenge in modeling dielectric elastomer actuators, as described
in the next section.

1.3 Constitutive equations

Equations that specify the stress components of the elastomer in terms of strain
components [Holzapfel 2000] are necessary for the constitutive model. The framework
of nonlinear continuum mechanics is used for this purpose. A few comments on
constitutive equations in nonlinear continuum mechanics are given here.
For the description of materials which undergo large strains, the theory of
hyperelasticity is used [Holzapfel 2000] to describe the non-dissipative mechanical
response. The existence of a Helmholtz free-energy function is postulated for so-called
hyperelastic material. When this function is solely dependent on the deformation
gradient, the Helmholtz free-energy function is referred to as a strain energy potential,
see section 2.1.
A large number of constitutive equations have been proposed for hyperelastic
materials, e.g. see [Holzapfel 2000 and Truesdell 1992]. Most of the equations follow
the so-called phenomenological approach, describing the macroscopic nature of the
observed mechanical behaviour of materials. By contrast, the statistical approach (see
for example [Boyce 2000] for a review) represents an attempt to take the
microstructure of the materials into consideration.
Generally the phenomenological approach, but inevitably also the statistical approach,
is concerned with fitting mathematical equations to experimental data. The fitting
parameters correspond to the parameters of the strain energy potential. The
constitutive model must be able to describe the three-dimensional stress / strain
behavior. This means that the model has to correctly predict the material response for
various types of experimental configurations. Thus, different types of experimental
data are required for the formulation and verification of the constitutive model.
In addition to the large strain elastic response, time effects have to be included in the
model formulation. This is relevant especially for materials with a pronounced
dissipative behavior such as VHB 4910.
Time-dependent response can be described within the framework of the theory of large
deformations by so called quasilinear viscoelastic models [Fung 1993]. These models
describe the time dependence through hereditary integrals and therefore imply that no
microstructural changes occur in the material due to the strain history.

4
In order to describe microstructural changes, nonlinear viscoelastic or viscoplastic
models can be used. For example, Rubin [Rubin 2002] proposed a model with internal
variables that allow for the description of history-dependence for large deformations in
materials with dissipative behavior.

1.4 Literature review on modeling of DE-actuators

In the present section, a review is provided on previous work dealing with modeling
and simulation of dielectric elastomer actuators. The purely hyperelastic models
without viscoelastic effects were considered with the following works:
Kofod [Kofod 2001a] performed uniaxial tensile tests and verified the range of validity
of various hyperelastic material models for the one-dimensional stress state for the
elastomer VHB 4910. The best results were obtained with a four-parameter Ogden
[Ogden 1972] model, which allows the fitting of the stress-strain curves for
elongations up to four times the initial length.
An analytical model for cylindrical silicone actuators was derived by [Carpi 2004]
using linear elastic theories. The Young’s modulus was obtained by fitting a uniaxial
tensile test. The comparison between experiments and modeling of one single actuator
showed good agreement.
Kofod [Kofod 2005] proposed an analytical model for a stripe actuator made on a
laminate of two silicone films. The Ogden strain energy function was used for the
constitutive model of the elastomer. The material parameters were obtained by fitting
uniaxial data. Two different kinds of experiments were performed: (i) constant strain
experiments where the force was measured and (ii) constant load experiments where
the strain was measured. A comparison of experimental data and modeling resulted in
an experimental verification of eq. (1.1) for silicone for the constant strain experiment.
The discrepancy between the model and experimental data varied between 15 and 37%
for the strain values for the constant load experiments.
In [Goulbourne 2005], an analytical model is derived for the investigation of the
behavior of an inflatable dielectric elastomer membrane which has a potential
application as a cardiac pump. The material parameters of the mechanical model (an
Ogden strain energy form is used here) are obtained by fitting a uniaxial tensile test.
The model is used to study the behavior of the inflatable membrane by changing a
number of design variables. A comparison between the model and experimental data
of the elastomer membrane is missing.

5
Lochmatter [Lochmatter 2006a] investigated the performance of a planar DE actuator
with a hyperelastic film model. The material parameters were obtained by fitting
uniaxial tensile test data. A quasi-static activation cycle of the planar actuator was
simulated. The theoretical efficiency (10%) of the planar actuator was calculated.
Models that also take into account the time-dependence have been presented, as
summarized below:
The dynamic behavior of a VHB 4910 stripe actuator loaded with a constant weight
was investigated by [Sommer-Larsen 2002]. The mechanical model considered
viscoelasticity and large deformations by using the Rivlin-Sawyer equation [Rivlin
1971]. The model parameters were fitted to a uniaxial tensile and a uniaxial relaxation
test. The dynamic strain response of the stripe actuator was calculated, but not verified
experimentally.
In [Mockensturm 2006], an analytical model is derived for the dynamic behavior of a
spherical inflatable membrane fabricated from VHB 4910. A Mooney-Rivlin [Mooney
1940] strain energy form was used for the constitutive model. Time effects are
included. The model is used for the design of pumps. There is no comparison between
the model and experimental data.
Yang [Yang 2005] proposed an analytical model to model an annular membrane made
of dielectric elastomers. Finite strains (Mooney-Rivlin strain energy form) and
viscoelasticity [Christensen 1981] were applied for the mechanical model. The
material parameters were fitted to uniaxial creep tests and the strain behavior of the
annular membrane was predicted for a stretch ratio up to three.
In [Plante 2006a and Plante 2006b] a detailed study on modeling was carried out by
considering so-called circular actuators and diamond actuators. A modified Bergström-
Boyce model [Bergström 1998], a constitutive model which describes viscoelasticity
and large strains, was applied for the diamond actuators. The hyperelastic part of the
model was adapted: The eight-chain model was replaced by an Ogden strain energy
form.
An analytical model which combines the constitutive model and electromechanical
coupling was proposed. The model was used to predict the work cycles of diamond
actuators. Diamond actuators at a single prestrain level were created and work cycles
with five different velocities were driven for experimental verification. The material
parameters of the adapted Bergström-Boyce model (ten in total) were found by fitting
the model to experimental data for a single velocity. The actuator behavior was then
predicted for the other four velocities and compared to the experimental data. The
agreement between the model and experiments was reasonable. By considering a

6
power balance, the analytical model also was used to calculate the current
consumption. There are two points which restrict the model proposed by Plante:
(i) The material parameters are only fitted for a single prestrain configuration.
Validity for different prestrains is not verified.
(ii) The material parameters were fitted to a diamond actuator which has a
deformation state that is to a certain extent equi-biaxial. It was not evaluated
whether the model with the material parameter fitted by the diamond actuator is
able to describe different deformation states, e.g. the uniaxial behavior. Uniaxial
tests were carried out experimentally but not used for model verification.
Failure criteria and an analytical model were used to investigate failure mechanism for
the circular actuators. The constitutive behavior of the analytical model was given by
the Ogden strain energy form. Viscoelasticity was not considered in the constitutive
equations but indirectly, by fitting tensile tests with the Ogden strain energy form at
‘low’ and ‘high’ stretch rates of the tensile test. The two sets of material parameters (at
low and high stretch rates) were used to predict the failure of circular actuators which
were activated slowly and rapidly. For rapidly activated actuators, the model
overestimated the experimental critical strain up to 57%. For slowly activated
actuators, the prediction of the model was reasonable, but only experimental data for
three different prestrains were used for verification. In order to draw serious
conclusions about the failure mechanism, more detailed information is required. A
quantitative relationship between the circular actuator and the diamond actuator is
missing.
In all cases the electromechanical coupling was considered either with the full
electrostatic pressure as proposed by Pelrine (eq. (1.1)) or else by half of it
[Goulbourne 2005, Mockensturm 2006 and Yang 2005]. Even though Pelrine's
equation is generally established for the mechanical coupling, some researchers
[Goulbourne 2005, Mockensturm 2006 and Yang 2005] use half of the electrostatic
pressure proposed by Pelrine, corresponding to the behavior of a plate capacitor. A
detailed theoretical and experimental investigation of the electromechanical coupling
is thus required.
The majority of the models in literature describes large strains and neglects
viscoelasticity. The material parameters of all cited references except [Plante 2006a]
are evaluated by means of uniaxial tensile tests. This is not sufficient for the
characterization of the three-dimensional material behavior [Boyce 2000]. Therefore,
most of the models are successful in describing the uniaxial behavior but are either
restricted to one activation configuration or fail to make reasonable predictions for the

7
actuator behavior. This is due to the missing characterization of the multiaxial
behavior for material parameter evaluation.
Plante's model [Plante 2006a] is calibrated for a biaxial state and is able to predict
actuation cycles at different velocities for diamond actuators. The model is, however,
restricted to one single prestrain (5x2.2) and is not verified for different deformation
states.
From a scientific point of view, the ‘three-dimensionality’ of a model in continuum
mechanics is important. Otherwise the model is reduced to a sterile curve fitting
exercise. A phenomenological approach in continuum mechanics is always restricted
to specific deformation and time ranges, and material parameters are determined by
fitting experimental data. Therefore, an essential step consists in the verification
whether the model and experimental data agree for all relevant deformation states.
The application of all cited models is in practice of limited usefulness; most models do
not describe the actuator behavior reasonably or have not been verified experimentally.
All models are restricted to a small deformation range in their predictive capabilities.
For general actuator design and optimization, a ‘global’ constitutive model is required
which is able to describe (i) the actuator behavior over a large strain and time range,
and (ii) various deformation configurations (uniaxial as well as multiaxial).
Dielectric elastomer actuators are usually pre-stretched by a stretch ratio of between 3
and 5. Systematic investigations considering this deformation range are required, but
corresponding experimental studies are absent from literature.

1.5 Objectives of the present work

The main objectives of the present thesis concern (i) constitutive model formulation,
(ii) electromechanical coupling, (iii) numerical implementation of the models, (iv)
experimental characterization and (v) an overall verification with different types of
actuators.
(i) The first aim is to define a three-dimensional constitutive model that includes
large strains and viscoelasticity and which is able to predict the states of stress
and strain of arbitrary actuator configurations as a function of the applied voltage,
over a large deformation and time range. The material parameters of the model
have to be determined by optimization techniques. The model should be able to
rationalize the observations from different experiments: actuator tests, uniaxial
tensile and relaxation tests, aspiration tests [Nava 2007] and compression tests.
Circular actuators are considered as a basic model system. Different strain energy

8
 potential formulations have to be investigated for large-strain elasticity. In
literature, the Mooney-Rivlin or the Ogden form (see section 1.4) are used for
modeling dielectric elastomer actuators. Both strain energy forms are able to well
describe the uniaxial response. Along with the Ogden and the Mooney-Rivlin
form, the Yeoh [Yeoh 1990] and the Arruda-Boyce [Arruda 1993] form are also
evaluated here. The Yeoh form is a third degree polynomial of the first invariant
of the right Green deformation tensor (see section 2.1). The Arruda-Boyce strain
energy form has two material parameters which can be related to physical
properties of the elastomer.
(ii) A fundamental investigation of the electromechanical coupling has to be
carried out. The aim is to verify Pelrine's equation theoretically as well as
experimentally. The difference in electrostatic force distributions in a dielectric
elastomer and a plate capacitor is investigated. For the theoretical understanding
of electromechanical coupling, numerical analysis is performed to determine the
charges and electrostatic forces. The experimental verification of the
electromechanical coupling considers measurement of the dielectric constant for
different pre-stretches as well as experiments with spring rolls.
(iii) Numerical models must include the implementation of both the constitutive
model and the electromechanical coupling. Commercial finite element programs
are used for the numerical calculations. Various constitutive models are
implemented. The challenge lies in the algorithm development for
electromechanical coupling. Various methods have been investigated. In one
approach, the measured displacement of the active zone is imposed as kinetic
boundary condition (by prescribing the membrane thickness variation) and the
‘required’ activation voltage represents the result of the calculation. In a second
approach, the activation voltage is the input while the time history of the radial
strain is the output of the calculation. This is achieved by an iterative algorithm
included into a user program. The advantage of this direct approach is that the
voltage is a direct input parameter. The disadvantage is that it is time-consuming
and might present convergence problems (especially when large strains occur).
(iv) A systematic experimental characterization of the material VHB 4910 has to
be performed. Experiments under different prestrain and voltage conditions are
carried out for the circular actuators. Thus a large deformation range is included.
Several experiments are carried out for each condition. This is important since
statistical aspects have to be considered (the electrode is applied by hand).
Uniaxial tensile and relaxation tests, aspiration tests and compression tests are
performed to describe the passive mechanical behavior in different deformation

9
 states. The experiments serve for the formulation and verification of the
constitutive model.
(v) Spring rolls are used for a further verification of the whole model system. By
blocking the strain of a spring roll in an experiment, the electromechanical
coupling is studied by measuring the force. This leads to an evaluation of both the
mathematical form of eq. (1.1) and the dielectric constant of the elastomer.

10
2 Theoretical Background and Material Details
This chapter provides an introduction to the theoretical framework of nonlinear
continuum mechanics, to the optimization methods applied in this work and to the
utilized materials.

2.1 Nonlinear continuum mechanics

2.1.1 Finite strains

The following part is adapted from [Holzapfel 2000]. In continuum mechanics, two
different configurations are considered for the description of the deformation of a body
B, viz. the reference and the current configuration. The position of a particle, Y, in the
body, B, is given by the position vector, x, in the current configuration and by the
position vector, X, in the reference configuration. In this section, vectors are singly
underlined and tensors doubly underlined.
Reference configuration Current configuration

dX dx
Y
X
x
x3

x2
x1

Figure 2.1. Deformation of a solid body

The reference configuration is a single-constant configuration and is independent of

time. The current configuration describes the position of each particle Y at time t. The
mapping relation between the current and the reference configuration is given by a
function, f, which specifies how the particle Y moves through space as time
progresses.

x = f ( X ,t) (2.1)

11
The deformation of the body is described by the deformation gradient, F, which gives
the incremental deformation of a material line element, dX (in the reference
configuration), to the material line element dx (in the current configuration).

∂x
F= (2.2)
∂X

The relationship between a material line element, dX, of the reference configuration
and the line element, dx, of the current configuration is given by

d x = Fd X (2.3)

The magnitude, |ds|, of the material line element, dx, in the current configuration may
be calculated using eq. (2.3) such that
T
( d s ) 2 = d x ⋅ d x = d X ⋅ F F d X = d X ⋅ Cd X (2.4)

where C=F TF is called the right Green deformation tensor. It is also convenient to
define the left Green deformation tensor, B=FFT. With the help of this tensor, the
magnitude, |dS|, of the material line element, dX, in the reference configuration is
given:
−1
( d S )2 = d X ⋅ d X = d x ⋅ B d x (2.5)

The stretch ratio, λ, of a material line element is defined in terms of the ratio of the
lengths |ds| and |dS|,

ds
λ= (2.6)
dS

There are various tensors which define the strain state. One of them is the Lagrangian
strain tensor E.

1
E = (C − I ) (2.7)
2

where I is the unit tensor. Strain states can also be characterized by scalar terms, the
so-called invariants. There are three invariants for the right Green deformation tensor
and the left Green deformation tensor:

I1 = C ⋅ I = B ⋅ I (2.8)

I2 =
1
2
[ ] [1
(C ⋅ I ) 2 − C ⋅ I = ( B ⋅ I ) 2 − B ⋅ I
2
] (2.9)

12
 I 3 = det(C ) = det( B) = J 2 (2.10)

where J corresponds to the ratio between a volume element in the current

configuration, dv, and the same in the reference configuration, dV:

dv
J = det( F ) = (2.11)
dV

The determinant of F is a pure measure of dilatation. For an incompressible material, J

is equal to 1.
A further strain measure which is used later is the rate of deformation tensor, D. It is
related to the velocity gradient, L, the derivative of the velocity field, v(x,t), in the
current configuration:

∂v
L= = D +W (2.12)
∂x

where D is the symmetrical part of the velocity gradient, L:

1 T T
D = sym( L) = ( L + L ) = D (2.13)
2

W is the spin tensor, the skew symmetric part of L:

1 T T
W = ( L − L ) = −W (2.14)
2

2.1.2 Stresses

An elastic material for which a strain energy function exists is called a Green elastic or
hyperelastic material. The strain energy function, W, represents the elastic energy
stored in the material. The stress state depends only on the deformation state of the
material itself (no time-dependent effects occur) and is given by

∂W ∂W
P=F = 2F (2.15)
∂E ∂C

where P is the first Piola-Kirchhoff (or nominal) stress tensor. The first Piola-
Kirchhoff stress tensor characterizes the stress state in the reference configuration. By
multiplying the first Piola-Kirchhoff stress tensor by a unit vector, N, in the reference
configuration, the first Piola-Kirchhoff (or nominal) traction vector, p, in the reference
configuration is obtained. For the stress states in the present configuration, the Cauchy
stress tensor, T, is used:

13
 ∂W T
T = 2F F (2.16)
∂C

Multiplying the Cauchy stress tensor with a unit vector, n, in the current configuration
gives the Cauchy (or true) traction vector, t, in the current configuration. Numerous
definitions exist for stress tensors. The Kirchhoff stress tensor, τ, is introduced for later
reference:

τ = JT (2.17)

where J is the volume ratio. The Kirchhoff stress tensor is work conjugate [Holzapfel
2000] to the rate of deformation tensor, D. Work conjugate means that the double
contraction of a stress tensor and the associated rate of deformation tensor describes
the real physical power during a dynamic process [Holzapfel 2000], i.e. the rate of
internal mechanical work per unit reference volume, denoted by wint

wint = τ ⋅ D (2.18)

2.1.3 Strain energy functions and nearly incompressible materials

As mentioned previously, the stress tensor in hyperelastic materials is described using

a strain energy function. Strain energy potentials are given either as a function of the
invariants or of stretch ratios. An example is the so-called Yeoh form [Yeoh 1990]:
3
W = ∑ Ci 0 ( I1 − 3) i (2.19)
i =1

For finite element simulations, mixed formulations which use both displacement and
pressure degrees of freedom have been shown to be very efficient and accurate in the
solution of problems for nearly or totally incompressible materials. The material laws
have to be modified in order to treat the pressure as a separate variable. To separate the
hydrostatic pressure from the stress tensor, the following strain energy function is used
[Sussmann 1987]:

W = W ( I r1 , I r 2 ) + Wh ( I r 3 ) (2.20)

where Ir1, Ir2 and Ir3 are the reduced invariants and Wh represents a hydrostatic work
term. For the Yeoh form (eq. 2.19), the strain energy is adapted as follows:
3
K0
W = ∑ Ci 0 ( I r1 − 3) i + ( J − 1) 2 (2.21)
i =1 2

14
where Ci0 and K0 (bulk modulus) are material parameters. The reduced invariants are
[e.g. Sussmann 1987]:
−1 / 3
I r1 = I 3 I1 (2.22)
−2 / 3
I r 2 = I3 I2 (2.23)

I r3 = I3 (2.24)

2.1.4 Quasilinear viscoelasticity

In order to introduce time-dependent effects, the concept of quasilinear viscoelasticity

[Fung 1993] is used. For infinitesimal strains, linear viscoelasticity is given by the
basic hereditary formulation:
t t
P (t ) = ∫ 2G (t − t ' )e&dt '+ I ∫ K (t − t ' )ϕ&dt ' (2.25)
0 0

where e is the mechanical deviatoric strain tensor, φ is the mechanical volumetric

strain and P is the nominal stress tensor. The time-dependent functions, G(t) and K(t),
are the shear and bulk relaxation moduli. In the present work the time dependence of
the volumetric behavior is neglected, i.e. K(t)=K0, while G(t) is expressed in terms of
an exponential series known as the Prony series:

⎛ M
⎞
G (t ) = G0 ⎜1 − ∑ g i e −t /τ i ⎟ (2.26)
⎝ i =1 ⎠

where G0 is the instantaneous shear modulus, τi are the relaxation times that
characterize the decay time and gi are the corresponding weighting factors that define
the amount of decay of the stress response. Using integration by parts, eq. (2.25) can
be rewritten as
t
P (t ) = 2G0 e(t ) + ∫ 2G& e(t − t ' )dt '+ K 0ϕ (t ) I (2.27)
0

A suitable generalization for the finite strains of the hereditary integral formulation is
obtained as follows [ABAQUS 2004]:

⎡ t −1 ⎛ G& (t ' ) D ⎞ ⎤
τ (t ) = τ 0 (t ) + sym ⎢ ∫ Ft (t − t ' )⎜⎜ τ 0 (t − t ' ) ⎟⎟ Ft (t − t ' )dt '⎥ (2.28)
⎢⎣ 0 ⎝ G0 ⎠ ⎥⎦

15
where τ0D is the deviatoric part of the instantaneous Kirchhoff stress, τ0, and Ft(t-t’) is
the deformation gradient of the state t-t’ relative to the state at time t. A transformation
is performed on the stress relating the state at time t-t’ to the state at time t.

2.2 Optimization procedure

To determine the material parameters of the constitutive model, the ‘inverse problem’
has to be solved. In other words the material parameters are adapted iteratively by an
optimization procedure until experimental data and the corresponding finite element
calculations agree to a specified extent.
For example, for a circular strain test, the material parameters (x1, x2, …, xn) have to
be chosen in such a way that the experimental and calculated nominal radial strain
history (the finite element output) agree. The calculated strain history, sc(t), is given by
discrete strain values, sc,k (k=1..n), at corresponding time values, tk. In order to obtain
the n material parameters (x1, x2, …, xn), a minimum has to be found for the error
function, fe. The function fe is the sum of the squares of the differences between the
calculated strains, sc,k, and the experimental strains, se,k:

f e ( x1 , x2 ,..., xn ) = ∑ ( sc ,k ( x1 , x2 ,..., xn ) − se, k ) 2 (2.29)

k

where se,k are discrete points of the experimental strain history, se(t). The minimum of
a function with n independent variables is searched for in a mathematical way. An
iterative solution technique is used to find the minimum of the function, fe.
There are two classes of optimization algorithms, viz. direct and indirect methods. The
direct methods require only the function values. The indirect methods also require
derivatives of the error function. In the present work, the function fe is evaluated with
the output of the finite element simulation and hence only the function value is known
and not its derivatives. Therefore, a direct method is required. A review of direct
methods is given by [Lewis 2000]. An effective and computationally compact direct
method [Lewis 2000] is the simplex algorithm of Nelder and Mead [Nelder 1965]
which is used in the present work. A simplex is a polytope of N+1 vertices in N
dimensions, e.g. for N=2 the simplex is a triangle, for N=3 the simplex is a
tetrahedron.
The Nelder-Mead method can be understood in principle for the simple example of
two material parameters, x1 and x2, with N=2. The method is a pattern search that
compares function values at the three vertices of the triangle. The diameter of the
simplex is calculated after each step. The aim of the algorithm is to minimize the

16
diameter of the simplex. The so-called ‘worst’ vertex which ‘causes a large diameter’
is rejected and replaced with a new vertex. A new triangle is formed and the search is
continued. The process generates a sequence of triangles for which the function values
at the vertices get smaller and smaller. The size of the triangles is reduced until the
diameter of the simplex is less than the specified tolerance. Further details on the
Nelder-Mead method are given in [Lagarias 1998 and Nelder 1965].
Like all general purpose multidimensional optimization algorithms, Nelder-Mead
occasionally gets stuck in a rut. The standard approach to handle this is to restart the
algorithm with a new simplex starting at the current best value.

2.3 Material details

2.3.1 Elastomers

Actuator technologies are often compared by different actuator output parameters

[Bar-Cohen 2001]: maximum strain, maximum pressure, efficiency and speed of
response.
For dielectric elastomer actuators, SRI International [Kornbluh 1999, Kornbluh 2000
and Pelrine 2000a] evaluated different elastomers by circular strain tests and
theoretical considerations regarding the actuator parameters. The conclusion [Pelrine
2000a] was that acrylic and silicone actuators are the most promising materials for the
dielectric elastomer technology.
In particular, the acrylic elastomer VHB 4910, commercially available from the
company 3M (in the form of a membrane with 1 mm thickness, 23 mm width and
user-defined length), has shown the largest activation strain (over 200%) [Kornbluh
2000 and Pelrine 2000a], a high elastic energy density (3.4 J/cm3) [Kornbluh 2000]
and a high electrical breakdown strength (218 MV/m) [Kofod 2003]. The material is
widely used in the research community and has also been used in the present work.
The chemical composition is not declared by the manufacturer.
VHB 4910 is usually prestrained in two directions, see Figure 2.2. The pre-stretch ratio
(or prestrain) in the x-direction is defined as λx = x1/X and the pre-stretch ratio (or
prestrain) in the y-direction is defined as λy = y1/Y. For an equi-biaxial deformation
state, λx= λy, the pre-stretch ratio is described by λp with λp= λx= λy.

17
 Reference configuration Current configuration

Z z1 x1
X y1
Y z
y
x

Figure 2.2. Illustration of the prestrain

The reasons for pre-stretching the elastomeric film are:

(i) The electrical breakdown strength increases from 18 to 218 MV/m by equi-
biaxially pre-stretching the film from no prestrain to λp=6 [Kofod 2003]
(ii) The thickness of the film decreases. A lower voltage has to be applied to obtain
the same electrostatic pressure (see eq. (1.1))
(iii) The prestrain avoids so-called pull-in failures [Plante 2006a]

2.3.2 Electrodes

Various electrode materials have been investigated in [Carpi 2003 and Kofod 2001a].
So far, several different types of electrodes are being used in the research community.
No theoretical study has been performed, nor has one electrode material shown any
major advantages. In general, the electrode should be conductive and compliant.
Gold electrodes [Zhang 2004] were used in the first part of the present work. Thin
layers (thickness: 250 nm) of beaten gold (from Brandenberger AG, Thalwil,
Switzerland) were applied (article 1). Beaten gold has the advantage that so-called
‘self healing’ effects occur. At the position of the electrode where an electrical
breakdown occurs, the thin gold layer evaporates and the actuator does not fail.
In the second part of this work (article 3), the gold electrodes were replaced with
silicone/graphite electrodes. The reason for this was that gold electrodes constrain the
elongation of the actuator mechanically, in contrast to the silicone/graphite electrodes,
see section 4.2. The silicone/graphite electrode is a mixture of 11 g graphite powder
(TIMREX SP30) and 10 ml silicone oil (DC 200/100 cs).

18
3 Modeling Dielectric Elastomer Actuators

3.1 Article 1

This section contains a reprint of the article:

M. Wissler and E. Mazza, Modeling and simulation of dielectric elastomer actuators,
Smart Materials and Structures, 2005, vol. 14, pp. 1396-1402

Background motivation: The motivation for this article was to create a model for the
simulation of a circular actuator including large strains and viscoelasticity. The model
was compared to experimental data.

Summary of methods: Uniaxial tensile and relaxation tests were carried out. A finite
element model for a uniaxial tensile test was created and the material parameters were
extracted by an optimization procedure which led to a good agreement between
experimental data and simulation. The ‘quasilinear visco-hyperelastic model’ was used
as a constitutive model for the elastomer as a general model formulation. For the
hyperelastic part, the strain energy form of Yeoh and for the viscoelastic part the
Prony series were applied. The concept of ‘quasilinear viscoelasticity’ was evaluated
by uniaxial relaxation tests.
A circular actuator with prestrain 4 was created and experimentally activated by a
constant voltage (3.5 kV). A corresponding finite element model was created. For the
simulation of the activation, the inverse problem was solved. The thickness of the
actuator was prescribed corresponding to the experimental strain history. The output of
the simulation was the out-of-plane stress which corresponded to the electrostatic
pressure in eq. (1.1). With this equation, the voltage history was calculated and
compared to the experimental voltage history.

Summary of results: The uniaxial relaxation tests show that the concept of
‘quasilinear viscoelasticity’ is able to describe the mechanical response of the material
VHB 4910. A set of parameters was determined by fitting uniaxial tensile test data.
Experimental results from the circular actuator test agree to a great extent with the
corresponding simulations.

19
Main conclusions, link to the next article: The ‘three-dimensionality’ of the model is
warranted for both the uniaxial and the multi-axial (actuator) behavior. Experimental
data and simulations agree to a great extent. Different strain energy forms will be
investigated in article 2 in order to evaluate their predictive capabilities for the same
data sets. Further analytical models are required for the optimization of the actuators,
due also to the duration of the numerical calculations, see article 2.

20
 Modeling and simulation of dielectric elastomer
actuators
Michael Wissler1 and Edoardo Mazza2
1
Swiss Federal Laboratories for Materials Testing and Research (EMPA), 8600
Dübendorf, Switzerland
2
Swiss Federal Institute of Technology (ETH), Dept. of Mechanical Engineering,
8092 Zürich, Switzerland

Abstract
Dielectric elastomers are used as base material for so called electroactive polymer
(EAP) actuators. A procedure and a specific constitutive model (for the acrylic
elastomer VHB 4910) are presented in this work for finite element modeling and
simulation of dielectric elastomer actuators of general shape and set-up. The Yeoh
strain energy potential and the Prony series are used for describing the large strain time
dependent mechanical response of the dielectric elastomer. Material parameters were
determined from uniaxial experiments (relaxation tests and tensile tests). Thereby the
inverse problem was solved using iterative finite element calculations. A pre-strained
circular actuator was built and activated with a predefined voltage. A three
dimensional finite element model of the circular actuator was created and the
electromechanical activation process simulated. Simulation and actual measurements
agree to a great extent, thus leading to a validation of both, the constitutive model and
the actuator simulation procedure proposed in this work.

1 Introduction

Electroactive polymers (EAP) are functional materials that can be used as actuators in
adaptive structures, in particular when large deformations are required. EAP transform
electric energy directly into mechanical work and produce large strains, in the order of
10% to 30%. Modeling of actuators made of dielectric elastomers is investigated in
this paper. Dielectric elastomer actuators are a category of EAP which were shown to
provide excellent overall performance [1], combining large elongation, high energy
density, good efficiency and high speed of response. Examples of applications [2, 3] of
dielectric elastomers as actuators include mobile mini- and microrobots, micropumps
and microvalves, micro air vehicles, disk drives, prosthetic devices and flat panel
loudspeakers.

21
Actuators made of dielectric elastomers consist basically of a compliant capacitor,
with a thin passive elastomer film sandwiched between two compliant electrodes [2,
4]. The electrostatic pressure acting on the insulating elastomer film can be calculated
for a given applied voltage and film thickness [2]. Due to the mechanical compression,
the elastomer film contracts in the thickness direction and expands in the film plane
directions. This in-plane expansion is exploited to generate motion or forces. In order
to improve their performance dielectric elastomers are pre-stretched, typically up to
five times of their in-plane dimensions.
Models are required for the design and optimization of EAP actuators. So far only few
works were presented in literature on modeling and simulation of dielectric elastomers
[1, 5-9]. In most cases functioning of EAP actuators is predicted by using analytical
expressions which directly relate voltage with displacement or force exerted. These
analytical models assume a linear or non-linear elastic mechanical response of the
elastomer and do neither account for its three dimensional viscoelastic behavior nor for
the anisotropy induced by pre-stretching of EAP. Analytical models describe the
behavior of simple actuator set-up (such as planar actuators with simple kinematic
boundary conditions) and thus their application in practice for general actuator design
and optimization is of limited use.
The finite element (FE) method is used here for simulating the behavior of dielectric
elastomer actuators. In a FE analysis there is no limitation in actuator geometry and
set-up. The main difficulties in the FE simulation of EAP actuators lay in (i) modeling
the activation through an applied voltage and (ii) the definition of the elastomer
constitutive model. Dielectric elastomers behave as insulating layers; do not change
their shape as a consequence of the electric field but only as a consequence of the
electrostatic mechanical pressure exerted by the electrodes in the thickness direction.
Since the electric field between the electrodes is inversely proportional to the film
thickness, the activation pressure depends on the deformation of the actuator, thus
leading to electromechanical coupling.
The problem can be treated however as fully uncoupled according to the procedure
applied in the present work: (i) the contraction in thickness direction and thus the
current thickness is imposed as a kinematic boundary condition, (ii) the required
compressive forces are calculated, (iii) from the mechanical pressure in thickness
direction and the current film thickness, the corresponding voltage is determined.
Motion of the actuator is considered as an input whereas the required voltage to realize
it is the output of the calculation. The results of the FE calculation depend on the
electrical properties and on the (passive) mechanical properties of the elastomer film.

22
The determination of the relative dielectric constant can be accomplished through
standard methods [10]. The main challenge for the realization of a FE model of an
EAP actuator consists therefore in the definition of its constitutive model. The acrylic
elastomer used in the present work behaves as a nonlinear viscoelastic material.
The experiments performed for determining the constitutive equations of the elastomer
are described in section 2 of this paper: these are large strain uniaxial relaxation tests,
large strain uniaxial tensile tests and experiments with a biaxially pre-strained circular
actuator. The nonlinear viscoelastic model used for the description of the mechanical
behavior of the dielectric elastomer is presented in section 3. In section 4 the FE
models for the analysis of the experiments are outlined along with the procedure for
determining the material parameters from curve fitting and optimization algorithms.
Section 5 presents the results of the uniaxial tests with the corresponding material
model. The FE simulation of the circular actuator described in section 6 provides a
validation of both, the constitutive model and the FE procedure applied for the analysis
of EAP actuators.

2 Experiments

The dielectric elastomer used in the present work is the commercially available acrylic
elastomer VHB 4910 (3M), which has been shown to produce large strains and high
elastic energy density [11].

2.1 Relaxation tests

Uniaxial relaxation tests were performed at room temperature (23° C) by stretching the
samples to a predefined elongation and measuring the force decrease over a time
period of approximately 0.5 hour. For each sample the elongation was quickly (0.2-0.4
s.) applied using a hand-operated set-up. The nominal or engineering strain s1 (the
change of length divided by the original length) was defined for a gauge length
corresponding to the whole length of the sample between the clamping points.
Fourteen experiments have been performed: the nominal strain was 20, 50, 100, 200,
300, 400 and 500%, and two tests were carried out for each elongation. The resulting
force history has been measured with a force transducer (HBM type S2) which was
connected to a PC LabView system for data acquisition. The undeformed sample
geometry was set to a ratio between length and width of 10:1, in order to ensure
uniaxial stress conditions. The undeformed thickness of the samples was 1 mm.

23
2.2 Tensile tests

Uniaxial tensile tests have been performed for two samples with a Zwick (Z010)
machine at room temperature (23 °C). The time history of the nominal strain (for both
samples) is shown in figure 1: the sample was first loaded up to 300% nominal strain
linearly over a time period of 900 s, the strain was hold fix for 900 s, and the sample
was than unloaded with a negative strain rate of -0.33 %/s. The sample geometry
corresponded to the one of the relaxation tests. The strain in these tests was not
determined from the relative displacement of the clamping points but monitored with
an elongation sensor (Zwick multisens) in the middle of the sample, over an
undeformed gauge length of 50 mm.

300

250
Nominal strain [%]

200

150

100

50

0 500 1000 1500 2000 2500 3000

Time [s]

Figure 1. Control profile of the strain controlled tensile test.

2.3 Circular strain test

A multi-axial electromechanical experiment has been performed on a biaxially

prestrained circular actuator (in a so called circular strain test [4, 12]) at room
temperature (23° C). The elastomer film was pre-strained radially up to a nominal
radial strain of 308% (according to the procedure described in [13]) and fixed on a
circular frame (radius R = 75 mm), see figure 2. The thickness of the film was
measured by a thickness sensor (Mitutoyo ID-F125/150) before (1 mm) and after (0.06
mm) prestraining.
After prestraining, a circular area (radius r0 = 22 mm) at the center of the film was
coated with beaten gold (Brandenberger AG, Thalwil, Switzerland) used as electrode
on the upper and lower side. The electrodes were connected to a high voltage supplier

24
through a thin metal wire. A constant voltage of 3.5 kV was applied for 3 min. The
nominal radial strain sr = r1/r0 - 1 of the coated area was measured with a video-
extensometer (Ovex ME-46) connected to a PC LabView system.

Coated area

r0 r1

circular frame

Voltage off Voltage on

Figure 2. Arrangement of the circular actuator for electromechanical measurements (r0

= 22 mm and R = 75 mm). The maximum nominal radial strain sr under activation was
14%.

3 Constitutive model

In order to describe the mechanical behavior of the dielectric elastomer a three

dimensional constitutive model is required and the material parameters have to be
determined from the experiments. In this work large strain elastic response is modeled
using the strain energy potential of Yeoh [14] and the so called Prony series are used
to describe the time dependence of the mechanical response (quasi-linear
viscoelasticity, [15]). The material is considered to be incompressible, an assumption
justified by previous measurements in uniaxial tension in which lateral contraction was
monitored.

3.1 Strain energy potential

In continuum mechanics large strain elastic (i.e. hyperelastic) materials are

characterized through a strain energy potential W, which represents the strain energy
of the material as a function of the deformation. The Yeoh-form [14] of the strain
energy potential (a particular type of the so called reduced polynomial form) is
described by three material parameters C10, C20 and C30, according to the following
equation:

W = C10 ( I1 − 3) + C 20 ( I1 − 3) 2 + C30 ( I1 − 3) 3 (1)

25
I1 in eq. (1) is the first invariant of the left Cauchy-Green deformation tensor and
depends on the principal stretch ratios λi (i = 1, 2 and 3) as

I1 = λ1 + λ2 + λ3
2 2 2
(2)

The stretch ratio λi (i = 1, 2 and 3) is defined as the current length divided by the
original length in the principal directions of the deformation. For a totally
incompressible material, the nominal or engineering stress σi (the current force per
unit area of the undeformed configuration) is given by the derivative of the strain
energy potential with respect to the stretch ratio λi:

∂W 1
σi = − p (3)
∂λi λi

In eq. (3) p is the hydrostatic pressure which depends on the kinetic boundary
conditions. For the uniaxial tensile test (loading in direction 1), the nominal stresses in
the directions 2 and 3 are zero (σ2 = σ3 = 0). With this boundary condition, the
nominal stress for the Yeoh form can be calculated as

1 2 2
σ 1 = 2 ⋅ (λ1 − ) ⋅ [C10 + 2 ⋅ C 20 ⋅ (λ1 +
2
− 3) + 3 ⋅ C30 ⋅ (λ1 +
2
− 3) 2 ] (4)
λ1 2
λ1 λ1

In the last expression, the relation

λ2 = λ3 = 1 / λ1 (5)

has been used which corresponds to an isochoric deformation (λ1λ2λ3=1). Eq. (4) is
used to determine the hyperelastic material parameters C10, C20 and C30 from uniaxial
test data.

3.2 Viscoelasticity

Time dependence of the mechanical response in a so called quasi-linear viscoelastic

model [15] is described by using time dependent coefficients in the strain energy
potential. The time function for these parameters is described here with the so called
Prony series [15]. The hyperelastic material parameters CijR (in this case C10 = C10R,
C20 = C20R and C30 = C30R) are defined as
N
t
Cij R (t ) = Cij 0 ⋅ [1 − ∑ g k ⋅ (1 − exp(− ))] = Cij 0 ⋅ f (t ) (6)
k =1 tk

where Cij0 describe the instantaneous elastic response and gk and tk characterize the
relaxation behavior. The relaxation function f(t) characterizes the relaxation behavior,

26
with f(t=0)=1. The material parameters gk and tk are determined from uniaxial
relaxation tests. The relaxation curves for the measured nominal stress can be fitted
through the parameters σ∞, σi and τi in the following equation
N
t
σ (t ) = σ ∞ + ∑ σ i ⋅ exp(− ) (7)
i =1 τi

where σ∞ is the long term nominal stress, σi and τi are parameters which depend on the
relaxation behavior. The instantaneous nominal stress σ0 is
N
σ 0 = σ (t = 0) = σ ∞ + ∑ σ i (8)
i =1

The assumption of quasi-linear viscoelasticity is that the stress relaxation function is

independent of the magnitude of the deformation. This can be verified experimentally
through the function s∞(t):
N
t
σ ∞ + ∑ σ i ⋅ exp(− )
σ (t ) i =1 τi
s∞ (t ) = = (9)
σ∞ σ∞

A material is quasi-linear viscoelastic when the function s∞(t) is independent of the

imposed elongation in the relaxation test. The parameters of the Prony series are
evaluated by combining eq. (6) and eq. (9):
N N
t t
σ ∞ + ∑ σ i ⋅ exp(− ) 1 − ∑ g k ⋅(1 − exp(− ))
σ (t ) i =1 τi k =1 tk
s∞ (t ) = = = (10)
σ∞ σ∞ N
1 − ∑ gk
k =1

From eq. (10) the following relation is obtained, where gk and tk can be obtained by
coefficient comparison from the parameters σ∞, σi and τi:
N
t
N
σi t
∑g k ⋅ exp(− )
tk
∑σ ⋅ exp(−
τi
)= k =1
N
(11)
i =1 ∞
1 − ∑ gk
k =1

The stress relaxation function is normalized here with respect to the long term stress
σ∞ (and not the instantaneous stress σ0) due to the experimental conditions. In fact the
initial strain is applied over a short ramp of variable duration so that determination of
the initial stress value is uncertain. The influence of this uncertainty vanishes due to
the fading memory of the material.

27
4 Numerical calculations

4.1 Finite Element models

The general purpose FE program ABAQUS [16] has been used for the present
calculations. Two FE-models have been created. The model used for simulation of the
tensile test consists of triangular plane stress elements. The 3D model for the analysis
of the circular actuator experiment is shown in figure 3 and consists of 140 solid
hybrid elements (8-node linear brick).
For the tensile test simulation the time history of uniaxial displacement corresponded
to the elongation control profile of figure 1. True (Cauchy) and nominal stresses were
determined as a function of time for each parameter set of the quasi-linear viscoelastic
model and compared with the experimental results.
The circular actuator simulation included three steps: (i) prestrain in radial direction by
308%, by imposing the displacement of the nodes at the model boundary, (ii)
relaxation for 1 hour, and (iii) actuation by imposing a time history of thickness
variation in the activated region corresponding to the measured radial displacement
from the experiments. The required voltage was determined from the calculated
stresses in thickness direction in the activated region. Thereby was assumed that the
compressive forces arise from the electrostatic pressure [2]. Electrostrictive effects are
neglected, which is a common simplification for dielectric elastomers [10]. The
electrostatic pressure pel corresponds to the negative Cauchy stress t3 in the direction of
the electric field and is defined by
2
⎛U ⎞
pel = ε 0 ⋅ ε r ⋅ ⎜ ⎟ = −t3 (12)
⎝d⎠

ε0 is the free-space permittivity (8.85·10-12 As/Vm), εr is the relative dielectric

constant (4.7 for VHB 4910 [10]), U is the electrical voltage applied to the electrodes
and d is the thickness of the polymer film. Eq. (12) has been implemented in the
ABAQUS user subroutine UVAR and used to calculate the electric voltage from the
current values of film thickness and stress t3 as a function of time.

28
 Activated area

3
2

Figure 3. FE-model for the circular actuator. The activated area is indicated.

4.2 Procedure for material parameter extraction

The software Matlab with the standard tool ‘cftool’ has been used for determining the
hyperelastic (eq. (4)) and the viscoelastic (eq. (7)) coefficients from the relaxation
tests. Matlab has been used also for the optimization procedure applied for the material
parameter identification in the tensile tests. The parameters were obtained as follows:
(i) Matlab calls a FE tensile test simulation with a specific material parameter set; (ii)
ABAQUS calculates the force versus time curve; (iii) the difference ferror between the
time integrals of the simulated and the experimental curve is than calculated; (iv) a
new material parameter set is determined by Matlab, according to a specific
optimization algorithm (Nelder-Mead simplex method). The iteration has been stopped
when a minimum of ferror had been found.
No material parameter optimization has been performed for the circular actuator
simulation, which has been used as a validation of the present modeling approach.

5 Uniaxial test results

5.1 Relaxation tests

The experimental results of the relaxation tests for nominal strains of 50, 200 and 500
% are shown in figure 4. A simple curve fitting procedure has been used for
determining the parameters σ∞, σi and τi, eq. (7), from the experimental data. Good
results were obtained with N=4 in eqs. (6) – (11) (four terms in the Prony series).

29
 0.25

Experiment: s1 = 500 %
0.20 Experiment: s1 = 200 %

Nominal stress [MPa]

Experiment: s1 = 50 %
0.15 Fit wtih eq. (7)

0.10

0.05

0.00
0 250 500 750 1000
Time [s]

Figure 4. Nominal stress in function of time for nominal strains (s1) of 50, 200 and
500%.

In figure 5 the relaxation function s∞(t) is plotted against time for all relaxation
experiments.

4
Relaxation function [-]

0
0 250 500 750 1000
Time [s]

Figure 5. The relaxation function s∞ from all relaxation tests.

The excellent agreement of the functions s∞(t) for different elongations confirms that
quasi-linear viscoelastic behavior can be assumed for this material. With the average
of the relaxation functions s∞(t), the material parameters gk and tk were determined
according to the procedure of section 3.2. The results are reported in Table 1, column
A.

30
From the measured stress relaxation functions the best fit parameters σ∞, σi, τi and
with eq. (8), the instantaneous nominal stresses σ0 are determined. The hyperelastic
behavior can be characterized from the instantaneous nominal stress in function of the
corresponding stretch ratio, figure 6.

0.8

0.6
Nominal stress [MPa]

0.4

Experiment
Fit with eq. (4)
0.2

0.0
1 2 3 4 5 6 7
Stretch ratio [-]

Figure 6. Evaluation of the hyperelastic material parameters for the Yeoh-form.

The parameters C10R, C20R and C30R were obtained by fitting the data of figure 6 with
the Yeoh hyperelastic model, eq. (4). The corresponding values are reported in Table
1, column A.

5.2 Tensile tests

In figure 7 the experimental force time history is compared with the results of FE
tensile test simulations. The dashed curve represents a simulation based on the
material parameters from the relaxation tests, Table 1, column A. Specific
optimization (for experiment 1) of the parameters C10R, C20R, C30R and gk, tk using the
procedure described in section 4 led to the results of the continuous curve in figure 7.
The corresponding material parameters are given in Table 1, column B.
The agreement of simulation and experiment obtained with the optimized material
parameters is good, although the stress decrease in the relaxation phase of the
simulation is too rapid. The results obtained with the material parameters from the
relaxation tests are less satisfactory. As described in section 3, there are some
uncertainties connected with the present relaxation tests: (i), the application of the
initial strain was hand-controlled; (ii), strain was measured from the relative

31
displacement of the clamping points (end effects). The errors arising from these
uncertainties might be responsible for the deviation between simulation and
experiment shown in figure 7. The optimized parameters listed in Table 1, column B
have been used for the circular actuator simulation, section 6.

0.6

0.4
Force [N]

Experiment 1
0.2
Experiment 2
Calculation A
0.0 Calculation B

-0.2
0 500 1000 1500 2000 2500
Time [s]

Figure 7. Force time history: Experimental data and simulations of the tensile tests.
Calculation A: using data from relaxation tests. Calculation B: optimized parameters.

Table 1. Material parameters from relaxation tests (column A) and tensile tests
(column B)
Parameter A B
C10 [MPa] 0.0803 0.0693
C20 [MPa] -7.65 x 10-4 -8.88 x 10-4
C30 [MPa] 9.84 x 10-6 16.7 x 10-6
g1 [-] 0.570 0.547
t1 [s] 0.311 0.135
g2 [-] 0.189 0.198
t2 [s] 3.35 0.382
g3 [-] 0.0860 0.110
t3 [s] 35.7 37.1
g4 [-] 0.0543 0.0384
t4 [s] 370 217

32
6 Results for the circular actuator

The experimental data of nominal radial strain reported in figure 8 were obtained with
an activation voltage of 3.5 kV. The curve in figure 8 has been used in order to define
the time history of the elastomer thickness imposed as kinematic boundary condition
in the FE calculation.

15.0

12.5
Nominal radial strain [%]

10.0

7.5

5.0

2.5

0.0

0 25 50 75 100 125 150 175

Time [s]

Figure 8. Measured nominal radial strain in the circular actuator activated by a

constant voltage of 3.5 kV.

Figure 9 shows the FE-results for the activation voltage at different calculation time
points. The time history of the calculated voltage is reported in figure 10. The
agreement with the prescribed experimental voltage of 3.5 kV is excellent. The initial
deviation of maximum 20% decreases with time and disappears after 80 s.
Improvements of the short term relaxation model in the tensile test analysis would
probably contribute reducing the voltage overestimation in the initial phase of the
simulation.

33
 Voltage (V)
5000
4000
3000
2000
1000
0

(a) (b) (c)

Figure 9. FE calculation of the applied voltage at time 1 s (a), 10 s (b) and 100 s (c).

4000

3000
Voltage [V]

2000
FE Calculation
1000
Experiment

0 25 50 75 100 125 150 175

Time [s]

Figure 10. Circular actuator: experimental and calculated activation voltage.

7 Discussion and conclusions

A three dimensional constitutive model of the dielectric elastomer VHB 4910 has been
derived from uniaxial tests and applied for the simulation of a biaxially pre-strained
circular actuator. The good agreement between FE calculation and measurements with
the circular actuator represents a validation of the constitutive model formulation
proposed in this work.
The model parameters were determined from uniaxial data whereas a biaxial pre-
straining and a three-axial stress state characterize the experiments with the circular

34
actuator. A number of strain energy formulations can be used for fitting the data from
the uniaxial experiments. A comparison of the constitutive behavior predicted using
the so called Mooney-Rivlin [17], Ogden [18] and Yeoh formulations for VHB 4910 is
discussed in [19]: this work highlights the importance of performing multi-axial
experiments in order to determine constitutive models of EAP materials. Other strain
energy functions for hyperelastic materials will be evaluated in future with multi-axial
large strain experiments. As well, the short term viscoelastic response has to be further
characterized in order to eliminate the deviations between simulation and experiments
observed in uniaxial (tensile test, figure 7) and in multi-axial (figure 10) experiments.
The results presented in section 6 confirm the validity of eq. (12) for the calculation of
the electrostatic forces in EAP actuators. This relation along with the constitutive
model represent the input for the procedure described in this paper which allow EAP
actuators to be simulated by FE. This procedure can be used for design and
optimization of EAP actuators of general shape.
The main disadvantage of the method presented here is the fact that the electrical
voltage is the output of the calculation (with the elastomer film deformation as the
input). Techniques for incorporating the electrical voltage as the input of the FE
analysis are under evaluation and will be used for future simulations.
Future work on simulation and modeling of EAP actuators will also consider: (i) the
influence of the compliant electrode layers (whose stiffness was neglected in the
present study); and (ii) the material behavior in cyclic loading conditions and the
microstructural damage mechanisms, which might lead to fatigue failure of the
actuators.

8 References

[1] Y. Bar-Cohen, Electroactive Polymer (EAP) Actuators as Artificial Muscles -

Reality, Potential and Challenges (2001) Washington, SPIE Press
[2] R.E. Pelrine, R.D. Kornbluh, J.P. Joseph, Electrostriction of polymer dielectrics
with compliant electrodes as a means of actuation, Sens. Actuators A, 64 (1998)
77-85
[3] R. Pelrine, P. Sommer-Larsen, et al., Applications of dielectric elastomer
actuators, Proc. SPIE, 4329 (2001) 335-349
[4] R. Pelrine, R. Kornbluh, et al., High-speed electrically actuated elastomers with
strain greater than 100%, Science, 287 (2000) 836-839
[5] H. Kim, S. Oh, et al., Actuator model of electrostrictive polymers (EPs) for
microactuators, Proc. SPIE, 4329 (2001) 482-490

35
[6] F. Carpi, D. Rossi, Dielectric elastomer cylindrical actuators: electromechanical
modelling and experimental evaluation, Mat. Sci. Eng. C, 24 (2004) 555-562
[7] P. Sommer-Larsen, G. Kofod, et al., Performance of dielectric elastomer
actuators and materials, Proc. SPIE, 4695 (2002) 158-166
[8] N. Goulbourne, M. Frecker, et al., Modeling of a dielectric elastomer diaphragm
for a prosthetic blood pump, Proc. SPIE, 5051 (2003) 319-31
[9] G. Kofod, Dielectric elastomer actuators, PhD Thesis, (2001) Denmark, Riso-R-
1286(EN)
[10] G. Kofod, P. Sommer-Larsen, et al., Actuation response of polyacrylate dielectric
elastomers, J. Intel. Mat. Syst. Str., 14 (2003) 787-793
[11] R. Kornbluh, R. Pelrine, et al., Ultrahigh strain response of field-actuated
elastomeric polymers, Proc. SPIE, 3987 (2000) 51-64
[12] R. Pelrine, R. Kornbluh, G. Kofod, High-strain actuator materials based on
dielectric elastomers, Adv. Mater., 12 (2000), no. 16, 1223-1225
[13] R. Zhang, A. Kunz, et al., Dielectric elastomer actuators for a portable force
feedback device, Proceedings of the 4th International Conference, EuroHaptics
(2004) Munich, Germany, 300-307
[14] O.H. Yeoh, Characterization of elastic properties of carbon-black-filled rubber
vulcanizates Rubber Chem. Technol., 63 (1990) 792-805
[15] Y. C. Fung, Biomechanics. Mechanical properties of living tissues, 2nd edn.
(1993) New York, Springer-Verlag
[16] ABAQUS/Standard Manual, Version 6.4.1, Hibbitt, Karlsson, Sorensen, Inc.,
2003
[17] M. Mooney, A theory of large elastic deformation, J. Appl. Phys., 11 (1940) 582-
592
[18] R. W. Odgen, Large deformation isotropic elasticity – on the correlation of
theory and experiment for incompressible rubberlike solids, Proc. R. Soc. Lond.
A, 326 (1972) 565-584
[19] M. Wissler, E. Mazza, Modeling of a pre-strained circular actuator made of
dielectric elastomers, Sens. Actuators A, 120 (2005) 184-192

36
3.2 Article 2

This section contains a reprint of the article:

M. Wissler and E. Mazza, Modeling of a pre-strained circular actuator made of
dielectric elastomers, Sensors and Actuators A, 2005, vol. 120, pp. 184-192

Background motivation: An analytical model for the circular actuator is derived for
optimization processes and for calculating the onset of electromechanical instabilities.
In previous work, the Mooney-Rivlin or the Ogden strain energy forms are often used
to model dielectric elastomer actuators. In this article, these two strain energy forms
are investigated and compared to the Yeoh form which showed good predictive
capabilities in article 1.

Summary of methods: A hyperelastic analytical model (without time-dependent

effects) was derived for the circular actuator. Three strain energy functions (Yeoh,
Ogden and Mooney-Rivlin form) were applied. The material parameters of the
analytical model were obtained by fitting uniaxial data. A finite element model was
created for the verification of the analytical model which includes simplifying
assumptions. Time-dependent simulations were also performed (according to article 1)
and the three strain energy functions were compared.

Summary of results: All three strain energy forms are able to fit the uniaxial data. For
the actuator behavior, simulation and analytical results show large discrepancies
between the applied voltage and strain, even though the uniaxial behavior is identical.
Simulations with the Yeoh strain function agree to a great extent with the experiments
(see also article 1). By contrast, the predictions for the Ogden form lay around 30%
below the experimental values. The values of the Mooney-Rivlin form lay
approximately 200% above the experimental values. The analytical models agree to a
great extent to the hyperelastic finite element models. Electromechanical instability
was predicted by the analytical models.

Main conclusions, link to the next article: The results show that the selection of an
appropriate strain energy form is essential. The analytical model provides a
relationship between voltage, prestrain and active strain. The circular actuator is a
useful model system for investigating the main features of dielectric elastomer
actuators. The analytical model describes electromechanical instabilities and can be

37
used for the optimization of the actuators. It was shown that the active strain is
strongly dependent on the prestrain. The next steps are to broaden the experimental
range, to find appropriate material parameters and to establish direct simulation
techniques.

38
Modeling of a pre-strained circular actuator made of
dielectric elastomers
Michael Wissler1 and Edoardo Mazza2
1
Swiss Federal Laboratories for Materials Testing and Research (EMPA), 8600
Dübendorf, Switzerland
2
Swiss Federal Institute of Technology (ETH), Dept. of Mechanical Engineering,
8092 Zürich, Switzerland

Abstract
Dielectric elastomers are used for the realization of actuators with large deformations
and belong to the group of so called electroactive polymers (EAP). Models are
required for the design and optimization of EAP actuators. Thereby the constitutive
behavior of the elastomer is of crucial importance and typically uniaxial experiments
are performed in order to determine the mechanical properties of these materials. In
this paper a pre-strained circular actuator made of a dielectric elastomer is
investigated: constitutive models based on uniaxial data are verified by comparing
calculation results with experimental observations. An analytical model is derived for
the instantaneous response to an activation voltage in the pre-strained circular actuator
and a finite element model is used to simulate the time dependent behavior.
Hyperelastic models are used and three strain energy formulations (Yeoh, Ogden and
Mooney-Rivlin) are compared in their predictive capabilities. The results of the
calculations with the three strain energy forms differ significantly, although all forms
were successfully fitted to the same uniaxial data set. Predictions of the actuator
behavior with the Yeoh form agree to a great extent with the measurements. The
results of the present work show that the circular actuator set-up represents a valid
model system for the characterization and optimization of the electromechanical
behavior of dielectric elastomers.

Keywords: EAP; modeling; dielectric elastomer; simulation; actuator

39
1 Introduction

Dielectric elastomers are used in actuator material systems able to generate large
deformations by transforming electrical energy directly into mechanical work. They
belong to the group of so called electroactive polymers (EAP). Several applications [1,
2] are envisaged for these actuator materials, such as mobile mini- and microrobots,
micropumps and microvalves, micro air vehicles, disk drives, prosthetic devices and
flat panel loudspeakers.
Actuators made of dielectric elastomers consist of a compliant capacitor, with a thin
passive elastomer film sandwiched between two compliant electrodes [1, 3]. The
electrostatic pressure acting on the insulating elastomer film can be calculated for a
given applied voltage and film thickness [1]. Due to the mechanical compression, the
elastomer film contracts in the thickness direction and expands in the film plane
directions. This in-plane expansion is exploited to generate motion or forces in
dielectric elastomer actuators. In order to increase the efficiency of the conversion of
electrical energy into mechanical work dielectric elastomers are pre-stretched,
typically up to five times their in-plane dimensions. Voltages are in the order of up to
5000 V (corresponding to an electric field of 200 MV/m in case of 25 µm film
thickness) and deformations are in the order of 30% for pre-strained elastomer
actuators.
Modeling techniques are required for the design and optimization of EAP actuators.
An approach for modeling electromechanical coupling in dielectric EAP is proposed in
[4]: the problem is solved by decoupling the electrostatic problem from the passive
mechanical response of the material. The characterization of the constitutive behavior
of the elastomeric film represents in this context the essential task for modeling
purposes. The models have to describe time dependent mechanical response for large
deformations.
Kofod [5] performed uniaxial tensile tests and verified the range of validity of various
hyperelastic material models for the one dimensional stress state. Best results were
obtained with a four-parameter Ogden model [6], which allows fitting the stress-strain
curves for elongations up to 4 times the initial length. The influence of pre-straining in
dielectric elastomer actuators has been studied with a Mooney-Rivlin model in [7].
Viscoelasticity and large deformations in dielectric elastomers have been investigated
with uniaxial tests by Sommer-Larsen et al. [8] and the experimental data were
rationalized by using the Rivlin-Sawyer model [9]. An analytical model for cylindrical
silicone actuators has been derived in [10] by using linear elastic theories.

40
In-plane pre-straining and out-of-plane electrostatical activation induce a three-axial
stress and deformation state in dielectric elastomer actuators. Under such
circumstances hyperelastic-viscoelastic models derived from uniaxial tests are often
inadequate for predicting the mechanical response: large strain biaxial and three-axial
experiments have to be performed in order to determine the strain energy function that
characterizes the non-dissipative component of the mechanical behavior.
In the present work a biaxially pre-strained circular actuator is proposed as model
system for the characterization of the electromechanical behavior of a dielectric
elastomer. In this system, like in all dielectric elastomer actuators, the deformation of
the actuator for a given activation voltage depends on the three dimensional
mechanical behavior of the film. The instantaneous (time independent) response is
calculated analytically using hyperelastic models and a relationship between applied
voltage and in-plane deformation is derived. Simplifications of the analytical model
are verified through a finite element simulation of the circular actuator. Time
dependent material behavior is than included in the finite element analysis by using
hyperelastic-viscoelastic models determined from relaxation and tensile tests. The
numerical results are compared with experimental observations. Three strain energy
formulations are considered and compared in their predictive capabilities.

2 Hyperelastic material behavior

In the first part of this work analytical models are derived for a dielectric elastomer in
a pre-strained circular actuator. In the analytical calculations the constitutive behavior
is simplified by neglecting the time dependence of the mechanical response.
Additional assumptions are that (i) the material is incompressible, and (ii) the material
is isotropic in the reference undeformed configuration. Viscoelastic behavior is
considered in the numerical calculations presented in section 5.

2.1 Strain energy functions

The mechanical behavior of hyperelastic materials is characterized by the strain energy

function W. Many functional forms of the strain energy W exist [11]. In this work,
three forms are considered, the Yeoh [12], Ogden [6] and Mooney-Rivlin [13] form. In
the Yeoh form [12] the strain energy depends on I1, the first invariant of the so called
left Cauchy-Green deformation tensor:

W = C10 ( I1 − 3) + C 20 ( I1 − 3) 2 + C30 ( I1 − 3) 3 (1)

41
C10, C20 and C30 are material parameters. I1 can be calculated from the eigenvalues of
the deformation gradient tensor, i.e. the so called principal stretch ratios λi (i=1,2 and
3) as

I1 = λ1 + λ2 + λ3
2 2 2
(2)

The strain energy form of Mooney-Rivlin [13] is given by

W = C10 ( I1 − 3) + C01 ( I 2 − 3) (3)

C10 and C01 are material parameters. Here W depends on I1 and on I2, the second
invariant of the left Cauchy-Green deformation tensor, which is given by
−2 −2 −2
I 2 = λ1 + λ2 + λ3 (4)

The Ogden form [6] is expressed in function of the principal stretch ratios:
N
µi
W =∑
α α α
⋅ (λ1 + λ2 + λ3 − 3)
i i i
(5)
i =1 α i

µi and αi are material parameters and N is the number of terms. For an incompressible
material, the principal Cauchy stresses ti (force per unit area of the deformed
configuration) can be determined from the derivative of the strain energy potential
with respect to the stretch ratio λi (e.g. [6]):

∂W
t i = λi ⋅ −p (6)
∂λi

The hydrostatic pressure p in eq. (6) depends on the kinetic boundary conditions.

2.2 Uniaxial tension

For the uniform extension (here in direction 1) with simple (uniaxial) tension, lateral
contraction occurs so that the stresses in the directions 2 and 3 are zero (t2 = t3 = 0).
The isochoric deformation (λ1λ2λ3=1) gives the following relation between the
principal stretch ratios:

λ2 = λ3 = 1 / λ1 (7)

By using the kinetic boundary conditions and eq. (7) the Cauchy principal stress t1 can
be calculated for the three strain energy forms as follows:
Yeoh form:

42
 1 ⎡ 2 2 ⎤
t1 = 2 ⋅ (λ1 − ) ⋅ ⎢C10 + 2 ⋅ C 20 ⋅ (λ1 + − 3) + 3 ⋅ C30 ⋅ (λ1 + − 3) 2 ⎥
2 2 2
(8 a)
λ1 ⎣ λ1 λ1 ⎦

Ogden form:
N −α i
t1 = ∑ µ i ⋅ (λ1 i − λ1
α
2 ) (8 b)
i =1

Mooney-Rivlin form:

1 1
t1 = 2 ⋅ (λ1 − ) ⋅ C10 + 2 ⋅ (λ1 −
2
) ⋅ C01 (8 c)
λ1 λ12

2.3 Biaxial tension

For the in-plane biaxial extension with equal stretches in the directions 1 and 2 (λ1=λ2),
out-of-plane contraction occurs so that the stress in the direction 3 is zero (t3 = 0). The
isochoric deformation (λ1λ2λ3=1) gives the following relation between the principal
stretch ratios:

λ3 = 1/ λ12 = 1/ λ2 2 (9)

By using the kinetic boundary condition and eq. (6) the Cauchy principal stresses t1=t2
can be calculated for the three strain energy forms as follows:

∂W ∂W
t1 = t2 = λ1 − λ3 (10)
∂λ1 ∂λ3
∂W
λk ⋅ (k=1, 3) can be calculated as:
∂λk
Yeoh form:

∂W
λk ⋅ = 2 ⋅ λk 2 ⋅ ⎡⎣C10 + 2 ⋅ C20 ⋅ ( I1 − 3) + 3 ⋅ C30 ⋅ ( I1 − 3) 2 ⎤⎦ (11 a)
∂λk

Ogden form:

∂W N
= ∑ µ i ⋅ λk i
α
λk ⋅ (11 b)
∂λk i =1

Mooney-Rivlin form:

∂W 1
λk ⋅ = 2 ⋅ λ1 ⋅ C10 − 2 ⋅ 2 ⋅ C01
2
(11 c)
∂λk λ1

43
Cauchy principal stresses in directions 1 and 2 can be readily calculated from eq. (10)
and (11).

3 Pre-strained circular actuator

3.1 Actuator set-up

A multi-axial electromechanical experiment has been performed on a biaxially pre-

strained circular actuator (in a so called circular strain test [3, 14]) at room temperature
(23° C). The elastomer film was pre-strained radially up to a nominal radial strain of
308% (according to the procedure described in [15]) and fixed on a circular frame
(radius R = 75 mm), see Figure 1. The thickness of the film was measured by a
thickness sensor (Mitutoyo ID-F125/150) before (1 mm) and after (0.06 mm) pre-
straining.

Coated area 3
2
2 1
R 1
Infinitesimal element
r0 r1 in the activated area

Activated area
Passive area
Voltage off Voltage on Circular frame

Figure. 1. Arrangement of the circular actuator for electromechanical measurements

with pictures of the device (left: voltage off; right: voltage on).

After pre-straining, a circular area (radius r0 = 22 mm) at the center of the film was
coated with beaten gold (Brandenberger AG, Thalwil, Switzerland) used as electrode

44
on the upper and lower side. The electrodes were connected to a high voltage supplier
through a thin metal wire. A constant voltage of 3.5 kV was applied for 3 min. The
nominal radial strain sr = r1/r0 - 1 of the coated area was measured with a video-
extensometer (Ovex ME-46) connected to a PC LabView system.

3.2 Analytical model

The analytical model is derived for the elastomer film in the activated area (Figure 1),
which is subjected to deformations in two steps: (I) pre-straining: homogeneous
equibiaxial deformation in the film plane 1-2 with contraction in direction 3; (II)
activation: compression in the out-of-plane direction through electrostatical forces
with out-of-plane contraction and in-plane elongation. Three configurations are
considered for the analytical model, see Figure 2: the undeformed or reference
configuration 0 corresponds to the initial unstressed state of the elastomeric film; the
pre-strained configuration I; the activated configuration II.

3
0 2
1

I t2I

t1I
II t3II
t2II

t1II

Figure. 2. Three configurations considered in the analytical model: 0 (undeformed

configuration), I (pre-strained configuration) and II (activated configuration)

The basic assumptions for the present model are: (i) the material is incompressible, (ii)
the material is isotropic in the reference configuration, (iii) the material is elastic (no
time dependent response) and (iv) the electromechanical coupling is due solely to the
electrostatic forces (no electrostriction). In the following, the index I refers to the pre-
strained configuration and the index II to the activated configuration, Figure 2.

3.2.1 Pre-strained configuration

The deformation gradient of the in-plane isochoric equibiaxial deformation

corresponds to the deformation described in section 2.3. Equations (10) and (11) give

45
the Cauchy stresses in the principal directions 1, 2 and 3 for each strain energy
function.

3.2.2 Activated configuration

The activation of the dielectric elastomer leads to a further reduction of the film
thickness. The principal stretch ratio λ3II in direction 3 is expressed as function of the
pre-strain stretch ratio λ1I and the parameter k which characterizes the change of the
thickness due to electrostatic forces:

1
λ3II = λ3I − k = −k (12)
(λ1I ) 2

Since only compressive forces are applied, the range of k is given by:

1
0<k < (13)
(λ1I ) 2

Incompressibility yields

λ3II ⋅ (λ1II ) 2 = 1 (14)

so that the principal in-plane stretch ratios λ1II and λ2II can be written in function of the
pre-strain stretch ratio λ1I and k

1 1
λ1II = λ2II = = (15)
λ3II 1
−k
(λ1I ) 2

In the following, the pre-strain stretch ratio is replaced by λp = λ1I = λ2I and the index
‘p’ refers to the pre-strain phase. In the experiment the in-plane nominal strain s1 is
measured and is defined as:

λ1II 1
s1 = ( − 1) ⋅100% = ( − 1) ⋅100% (16)
λ1I
1− k ⋅ λp
2

The definition of the kinetic boundary conditions for the activated configuration
requires a simplification. In fact, due to the in-plane elongation of the activated region
of the actuator, the in-plane stresses decrease (since the film is clamped at its outer
radius, in the non-activated region). Consideration of the stress relaxation would
require the non-activated region to be included in the mechanical analysis. However, if
the radial elongation due to activation is small with respect to the pre-strain radial
elongation, the reduction of the in-plane stress components can be neglected. The

46
errors arising from this approximation will be evaluated from a comparison with finite
element calculations in section 4.
With eq. (6) the Cauchy principal stresses t1II and t2II in directions 1 and 2 are:

∂W
t1II = λ1II − p II = t1I = t p (17 a)
∂λ1II

∂W
t 2II = λ2II − p II = t 2I = t p (17 b)
∂λ2
II

where tp corresponds to the pre-stress which is equal in direction 1 and 2. From eq.
(17a) the hydrostatic pressure pII can be evaluated:

∂W ∂W ∂W ∂W
p II = λ1II − t p = λ1II II − λ1I I + λ3I I (18)
∂λ1
II
∂λ1 ∂λ1 ∂λ3

The Cauchy principal stress in direction 3, t3II becomes

∂W ∂W ∂W ∂W ∂W ∂W ∂W
t 3II = λ3II − p II = λ3II II − λ1II II + t p = λ3II II − λ1II II + λ1I I − λ3I I (19)
∂λ3
II
∂λ3 ∂λ1 ∂λ3 ∂λ1 ∂λ1 ∂λ3

3.2.3 Electromechanical coupling

The Cauchy-stress t3II corresponds to the (negative) electrostatic pressure pel acting
across the electrodes [1],
2
⎛U ⎞
pel = ε r ⋅ ε 0 ⋅ ⎜ II ⎟ = −t3II (20)
⎝d ⎠

Thereby, εo is the free-space permittivity (8.85·10-12 As/Vm), εr is the elastomer

specific relative permittivity, U is the electrical voltage applied to the electrodes and
dII is the thickness of the polymer film in the activated configuration. The thickness dII
is related to the original thickness d0:

1
d II = λ3II ⋅ d 0 = ( − k) ⋅ d0 (21)
λp2

The combination of eq. (19), (20) and (21) yields an expression for the voltage U in
function of the kinematic quantities and derivatives of the strain energy function W:

47
 2
d 1 II
U = − 0 ⋅ ( 2 − k ) 2 ⋅ t3
2

ε r ⋅ε0 λp
(22)
d0
2
1 ⎛ ∂W ∂W ∂W ∂W ⎞
=− ⋅ ( 2 − k ) 2 ⋅ ⎜⎜ λ3II II − λ1II II + λ1I I − λ3I I ⎟⎟
ε r ⋅ε0 λp ⎝ ∂λ3 ∂λ1 ∂λ1 ∂λ3 ⎠

The expressions for the different strain energy potentials (eq. (11)) are inserted in eq.
(22). Eq. (16) is used to express the out-of-plane elongation parameter k in function of
the measured in-plane strain s1. In this way a relation between the applied voltage U,
the pre-strain of the elastomeric film λp and the ‘activated’ strain s1 is obtained.

4 Calculation results for VHB 4910

4.1 Material parameters from uniaxial tests

The dielectric elastomer used in the present work is an acrylic polymer, VHB 4910
(3M). Uniaxial tests were performed in order to characterize the mechanical behavior
of this material [16]. Uniaxial relaxation experiments were performed at room
temperature by stretching the samples to a predefined elongation and measuring the
force decrease over a time period of approximately 0.5 hour. Fourteen experiments
have been performed, the nominal strain being between 20% and 500%. Uniaxial
tensile tests have been performed for two samples with a Zwick (Z010) machine at
room temperature. The time history of the elongation consisted of three steps: the
sample was first loaded up to 300% nominal strain linearly over a time period of 900 s,
the strain was hold fix for 900 s, and the sample was than unloaded with a negative
strain rate of -0.33 %/s. Testpieces had undeformed dimensions of 150 x 10 x 1 mm3
(length x width x thickness). A finite element model was created and the material
modeled as quasi-linear viscoelastic using different strain energy formulations.
Material parameters were obtained from the experimental data using an optimization
algorithm with iterative finite element simulations. The corresponding results for the
instantaneous uniaxial response of the Yeoh, Ogden and Mooney-Rivlin strain energy
forms are shown in Figure 3 and, in terms of material parameters, in Table 1. For all
strain energy formulations good fitting of the experimental data up to a stretch ratio of
5 can be obtained.

48
 3

Yeoh
Ogden

Cauchy stress t [MPa]

2 Mooney-Rivlin

1 2 3 4 5
Stretch ratio for uniaxial extension λ [-]

Figure 3. Uniaxial instantaneous hyperelastic response: Yeoh, Ogden and Mooney-

Rivlin model from uniaxial experiments. The Cauchy stress t = t1 is obtained from eq.
8 a, b and c, with λ = λ1.

Table 1. Material parameters for the Yeoh, Ogden and the Mooney-Rivlin form
Yeoh
C10 [MPa] 0.0693
C20 [MPa] -8.88 x 10-4
C30 [MPa] 16.7 x 10-6
Ogden
α1 [-] 1.76
µ1 [MPa] 0.156
Mooney-Rivlin
C10 [MPa] 0.0450
C01 [MPa] 0.0412

4.2 Results for the circular actuator

The principal Cauchy stresses for the pre-strained and activated configurations were
calculated using the material parameters of Table 1. The in-plane Cauchy stress tp
(principal stress in direction 1 and 2 in the pre-strained configuration) is plotted in
Figure 4 for all three forms in function of the in-plane stretch ratio λp.

49
 10.0
Yeoh
Ogden
7.5 Mooney-Rivlin

Cauchy stress t [MPa]

5.0

2.5

0.0

1 2 3 4 5
Stretch ratio for the biaxial extension λ [-]

Figure 4. Pre-strained configuration: in-plane Cauchy stress t = tp vs. pre-strain stretch

ratio λ = λp for the three strain energy forms calculated from eq. (10) and (11).

The results for the Ogden and the Yeoh model are similar, whereas the pre-stress tp of
the Mooney-Rivlin model is about one order of magnitude higher.
The results of the activated configuration for a pre-strain with λp = 4.08 are presented
in Figure 5. The pre-strain with λp = 4.08 corresponds to the measured film thickness
change from 1 mm (undeformed configuration) to 0.06 mm (pre-strained
configuration) in the experiment. The nominal strain in the radial direction (s1, eq.
(16)) is plotted against activation voltage, using the results of section 3. For VHB
4910, a relative permittivity εr = 4.7 has been used, [17]. The analytical results are
compared with the strain-voltage curves obtained with a FE calculation of the whole
circular actuator (see section 5). In order to enable a direct comparison with the
analytical results, for this particular numerical analysis time dependent material
properties have been neglected.

50
 20
Yeoh: anal.
Yeoh: FE

Nominal radial strain [%]

15 Ogden: anal.
Ogden: FE
MR: anal.
10 MR: FE

0 2500 5000 7500 10000

Voltage [V]

Figure 5. Activation: instantaneous nominal radial strain vs. voltage. Comparison of

analytical (anal.) and numerical (FE) results for the different strain energy forms.
Yeoh, Ogden and Mooney-Rivlin (MR).

5 Finite Element modeling of the circular actuator

For the calculations of the circular planar actuators the general purpose FE program
ABAQUS [18] has been used. The FE model of the circular actuator is shown in
Figure 6 and consists of 140 solid hybrid elements (8-node linear brick).

Coated area

3
2

Figure 6. FE-model for the circular actuator. The coated area is indicated.

Time dependent material behavior is described by a quasi-linear viscoelastic model,

[19]. The time functions for the parameters of the strain energy functions were
determined from uniaxial relaxation and tensile tests, [16].

51
The circular actuator simulation included three steps: (i) pre-strain (λp = 4.08) in radial
direction, by imposing the displacement of the nodes at the model boundary, (ii)
relaxation over a time period of one hour, and (iii) activation by imposing a thickness
reduction in the activated region. The time history of thickness reduction corresponds
to the radial strain history measured in the experiment (assuming incompressibility).
The measured deformation represents therefore the input of the finite element analysis,
whereas the voltage needed to realize this deformation history is the output of the
calculation, and allows comparison with the experimental values and validation of the
model. The required voltage is determined from the calculated stresses in thickness
direction in the activated region: the electrostatic pressure pel corresponds to the
negative Cauchy stress t3 in the direction of the electric field, eq. (20); with eq. (20)
the voltage can be calculated from the current values of film thickness and stress t3.
Further details on the simulation of EAP actuators by FE are given in [16].
The measured time dependent in-plane elongation history of the actuator that has been
used as the input for the FE simulation is shown in Figure 7. The result of the
calculation is the time history of the required activation voltage. A constant voltage of
3500 V has been applied in the experiment. The results of the FE-modeling for the
three strain energy potentials are presented in Figure 8.
15.0

12.5
Nominal radial strain [%]

10.0

Uapplied = 3500 V
7.5

5.0

2.5

0.0

0 25 50 75 100 125 150 175

Time [s]

Figure 7. Nominal radial strain (s1, eq. (16)) history measured in the circular actuator
experiment.

52
 12500

10000

Yeoh
7500
Voltage [V]
Ogden
Mooney-Rivlin
5000 Experiment

2500

0
0 50 100 150 200
Time [s]

Figure 8. FE-calculation: time history of the voltage required to obtain the observed
radial elongation history for the Yeoh-, Ogden- and Mooney-Rivlin-form. Constant
voltage (3500 V) was applied in the experiment.

6 Discussion

Fitting of the hyperelastic parameters for the instantaneous response to the uniaxial
experimental data yields good results for all three strain energy formulations. This is in
contrast with the results presented in [5] for the same dielectric elastomer, for which
only the Ogden form allowed the stress-stretch data to be reproduced over a large
range of elongation.
As expected, major differences in the biaxial (pre-straining) and three-axial
(activation) response are obtained from the different strain energy forms. In particular,
the results for Ogden and Yeoh are similar for uniaxial and for equibiaxial loading, but
differ considerably for the three-axial deformation of the activated phase. This
highlights the importance of performing multi-axial tests for characterizing the
mechanical behavior of EAP materials. Experiments with the pre-strained circular
actuator can serve for this purpose.
The circular actuator can be studied with analytical models. Figure 5 compares
analytical and numerical calculations of the instantaneous behavior in the activation
phase. The discrepancies between numerical and analytical calculations are due to the
simplified kinetic boundary conditions of the analytical model. The curves of Figure 5
show that neglecting the reduction of in-plane stresses due to the radial elongation of

53
the film does not affect the results to a significant extent for nominal radial strains in
the order of up to 10%.
The calculation of the time-dependent behavior of the circular actuator was performed
here by FE, due to the difficulties in evaluating the hereditary integrals of the
viscoelastic model in the analytical framework. This FE simulation allowed comparing
the strain energy functions in their predictive capabilities. The results of Figure 8
demonstrate that the Yeoh form performs best for the present application. Some
mismatching is visible for the initial part of the activation phase.
The analytical model allows studying the instantaneous response of the circular
actuator. This is useful for design and optimization of EAP actuators. For instance, the
occurrence of “electromechanical instability” can be investigated with this model.
Instability occurs due to the fact that the electrostatic pressure increases when the film
thickness decreases. If the increase in out-of-plane stiffness of the film does not
compensate for the increased electrostatic forces, thickness will decrease in an
unstable manner (mechanical collapse). This phenomenon is evident from the curves
shown in Figure 9: the nominal radial strain is plotted against the required voltage for
the three strain energy forms (λp = 4.08). Instability occurs when the slope of the curve
is vertical: increase in lateral expansion (i.e. decrease in thickness) occurs without
increase of applied voltage. Instability occurs here for approximately 5000 V and
45000 V for the Ogden and Mooney-Rivlin models, whereas no instability is found for
the Yeoh model (for the present level of pre-stretching). It must be underlined that for
EAP actuators instability is not expected to represent the failure mechanism when the
so called electrical breakdown [17] occurs at a lower voltage. Indeed voltage of 45000
V over a film thickness < 0.1 mm cannot realistically be expected to occur without
prior electrical breakdown.

54
 300 Yeoh
Ogden

Nominal radial strain [%]

Mooney-Rivlin
200

100

0 20000 40000 60000

Voltage [V]

Figure 9. Electromechanical instability occurs when the strain vs. voltage curve has an
infinite slope.

Instability and more in general the relationship between in-plane strain and applied
voltage strongly depends on the pre-strain level. This relation can be studied with the
analytical model. Calculations were performed for the Yeoh form. Nominal radial
strain curves vs. applied voltages were calculated for different values of the pre-strain
stretch ratio λp. The results are shown in Figure 10.

40
rain [%]

30
Nominal radial st

20

10 10000
7500
]
g e [V

5000
01
2 2500
3
Volta

4
Pre-st 5 0
retch λ
p [- ]

Fig 10. Dependence of the instantaneous strain-voltage curves on pre-strain for the
Yeoh strain energy potential.

For a given voltage, the lateral elongation increase with increasing pre-stretch level up
to about λp = 2.5, and than decreases. These curves along with the information of

55
maximum allowable voltage for each pre-strain level (given by the breakdown
strength) can be used for selecting appropriate pre-straining of EAP actuators.

7 Conclusions

The results of the analytical and numerical calculations show that the selection of an
appropriate strain energy form is essential to a reliable characterization of the behavior
of dielectric elastomer actuators. In particular, it was demonstrated that no conclusions
can be drawn from uniaxial test data alone.
The model system proposed in this paper for the evaluation of electromechanical
models of EAP actuators offers several advantages:
(i) Important characteristics of the actuator can be studied with analytical
calculations. Qualitative analyses can be easily performed on the influence of
specific design parameters. Use of (time independent) long term moduli (in
place of the instantaneous moduli applied in the present work) would yield
quantitative results on the long term response of the actuator to a given voltage.
(ii) The multi-axial stress and deformation state of the pre-strained circular actuator
is representative for the working conditions of EAP actuators. Models that are
validated on a circular actuator arrangement are expected to usefully describe
the behavior of other arrangements.
(iii) The circular actuator can be easily realized and used for laboratory tests. In-
plane elongation, undeformed and pre-strained thickness, and voltage are the
data required for model fitting and can be easily determined in the experiment.
The Yeoh strain energy form with parameters determined from uniaxial tests led to an
excellent prediction of the observed electromechanical behavior. The agreement
between experiment and numerical calculations confirms (i) the constitutive model
(including the time dependent behavior) and (ii) the modeling approach adopted for
electromechanical coupling.
Further experiments with the circular actuator set-up will be performed in order to
confirm the validity of the present constitutive model. In particular, the influence of
the pre-strain level will be investigated. Future work on the characterization of
dielectric EAP actuators includes: (i) realization of multi-axial relaxation and creep
tests; (ii) verification of the viscoelastic model with cyclic experiments, in order to
study the evolution of the hysteretic behavior; and (iii) investigation of the role of the
electrodes compliance [20].

56
8 References
[1] R.E. Pelrine, R.D. Kornbluh, J.P. Joseph, Electrostriction of polymer dielectrics
with compliant electrodes as a means of actuation, Sens. Actuators, A 64 (1998)
77-85
[2] R. Pelrine, P. Sommer-Larsen, R. Kornbluh, R. Heydt, G. Kofod, Q. Pei, P.
Gravesen, Applications of dielectric elastomer actuators, Proc. SPIE, 4329
(2001) 335-349
[3] R. Pelrine, R. Kornbluh, Q. Pei, J. Joseph, High-speed electrically actuated
elastomers with strain greater than 100%, Science, 287 (2000) 836-839
[4] Y. Bar-Cohen, Electroactive Polymer (EAP) Actuators as Artificial Muscles -
Reality, Potential and Challenges, vol. PM98, SPIE Press, Washington, 2001, p.
309
[5] G. Kofod, Dielectric elastomer actuators, PhD Thesis, Riso-R-1286(EN),
Denmark, 2001, p. 13
[6] R.W. Odgen, Large deformation isotropic elasticity – on the correlation of theory
and experiment for incompressible rubberlike solids, Proc. R. Soc. Lond., A 326
(1972) 565-584
[7] H. Kim, S. Oh, K. Hwang, H. Choi, J.W. Jeon, J.D. Nam, Actuator model of
electrostrictive polymers (EPs) for microactuators, Proc. SPIE, 4329 (2001) 482-
490
[8] P. Sommer-Larsen, G. Kofod, M. Shridhar, M. Benslimane, P. Gravesen,
Performance of dielectric elastomer actuators and materials, Proc. SPIE, 4695
(2002) 158-166
[9] R.S. Rivlin, K.N. Sawyers, Nonlinear continuum mechanics of viscoelastic
fluids, Ann. Rev. Fluid Mech, 3 (1971) 117-146
[10] F. Carpi, D. De Rossi, Dielectric elastomer cylindrical actuators:
electromechanical modeling and experimental evaluation, Mat. Sci. Eng., C 24
(2004) 555-562
[11] O.H. Yeoh, Some forms of the strain energy function for rubber Rubber, Chem.
Technol., 66 (1993) 754-771
[12] O.H. Yeoh, Characterization of elastic properties of carbon-black-filled rubber
vulcanizates, Rubber Chem. Technol., 63 (1990) 792-805
[13] M. Mooney, A theory of large elastic deformation, J. Appl. Phys., 11 (1940) 582-
592
[14] R. Pelrine, R. Kornbluh, G. Kofod, High-strain actuator materials based on
dielectric elastomers, Adv. Mater., 12 (16) (2000) 1223-1225
[15] R. Zhang, A. Kunz, G. Kovacs, S. Michel, A. Mazzone, Dielectric elastomer
actuators for a portable force feedback device, The 4th European Conference on
EuroHaptics 2004, Munich Germany, June 2004, 300-307
[16] M. Wissler, E. Mazza, Modeling and simulation of dielectric elastomer actuators,
submitted to Smart Mater. Struct.

57
[17] G. Kofod, P. Sommer-Larsen, R. Kornbluh, R. Pelrine, Actuation response of
polyacrylate dielectric elastomers, J. Intel. Mat. Syst. Str., 14 (2003) 787-793
[18] ABAQUS/Standard Manual, 2003. Version 6.4.1, Hibbitt, Karlsson & Sorensen,
Inc.
[19] Y.C. Fung, Biomechanics. Mechanical properties of living tissues, 2nd edn.,
Springer-Verlag, New York, 1993, p. 277
[20] F. Carpi, P. Chiarelli, A. Mazzoldi and D. De Rossi, Electromechanical
characterisation of dielectric elastomer planar actuators: comparative evaluation
of different electrode materials and different counterloads, Sens. Actuators, A
107 (2003) 85-95

58
3.3 Article 3

This section contains a reprint of the article:

M. Wissler and E. Mazza, Mechanical behavior of an acrylic elastomer used in
dielectric elastomer actuators, Sensors and Actuators A, 2007, vol. 134, pp. 494-504

Background motivation: In article 1 and 2 the circular actuator was identified as a

valuable modeling system, including a constitutive model and an electromechanical
coupling. The simulation at prestrain λp = 4 and voltage U = 3.5 kV with the Yeoh
strain energy form showed a good agreement between simulation and experiment. In
order to characterize a larger strain range, more experiments are required at different
prestrains and voltages. Several experiments have to be carried out for each condition
in order to verify the experimental data statistically. Furthermore, an actuator
simulation is required where the voltage and not the strain history is the input. In
article 2 it was shown that the selection of an appropriate strain energy form is
essential. Alternative strain energy forms are also investigated here.

Summary of methods: A large experimental study investigating circular actuators

was performed at three different prestrains and four different voltage levels. A total
number of 40 actuators was tested. Each experiment was carried out 3-4 times for
statistical aspects. A novel approach is proposed for finite element analysis. The
material parameters were calibrated for the three different prestrains at 2 kV. Different
strain energy functions were investigated, viz. the Yeoh form, the Ogden form (with 6
parameters) and the Arruda-Boyce form. Uniaxial experiments and corresponding
finite element simulations were carried out for verification. Further, the mechanical
model was evaluated by considering a cyclic actuation.

Summary of results: The comparison between experiments and simulations for the
circular actuators shows a reasonable agreement for all three strain energy forms. By
contrast, a discrepancy between experimental data and simulations is evidenced for the
uniaxial behavior. The novel finite element approach successfully simulates the
actuator behavior in nearly all cases, including those with large prestrains and large
active strains. The quasilinear visco-hyperelastic model fails to describe the cyclic
actuator behavior.

59
Main conclusions, link to the next article: A large effort was made for the
experimental characterization of circular actuators. These experiments are also
important for the whole research community for model verification purposes. The
‘global’ actuator behavior is described well by the constitutive model and the applied
simulation techniques. The reason for the discrepancy between experimental data and
simulations for the uniaxial behavior lies in the electromechanical coupling which is
investigated in article 4.

60
Mechanical behavior of an acrylic elastomer used in
dielectric elastomer actuators
Michael Wissler1,2 and Edoardo Mazza2,1
1
Swiss Federal Laboratories for Materials Testing and Research (EMPA), 8600
Dübendorf, Switzerland
2
Swiss Federal Institute of Technology (ETH), Dept. of Mechanical Engineering,
8092 Zürich, Switzerland

Abstract
The paper reports on extensive experimental work for the characterization of a
dielectric elastomer used as base material for electroactive polymer (EAP) actuators.
The mechanical behavior of the acrylic elastomer VHB 4910 is characterized using
large strain experiments (uniaxial and equibiaxial deformation) under force and
displacement controlled loading conditions. Next to tensile and relaxation tests,
experiments were conducted also using so called circular actuators. Over 40 actuators
were produced (with different in-plane prestrain levels) and activated with voltages
between 2000 and 3500 V. The experimental data are useful for determining
constitutive model parameters as well as for validating models and simulation
procedures for electromechanical coupling in EAP actuators. A novel approach is
proposed for finite element analysis of dielectric elastomer actuator, which has been
used in the present work for the evaluation of the experimental observations from
circular actuators. Material parameters of different visco-hyperelastic models have
been determined from a subset of the experimental data and the predictive capabilities
of the models evaluated through comparisons with the remaining data. The prediction
of the circular actuator behavior was satisfactory so that the proposed models might be
useful for actuator design and optimization purposes. Limitations of the proposed
constitutive model formulation are presented.

Keywords: Modeling, EAP, dielectric elastomer, simulation, actuator

61
1 Introduction

Dielectric elastomers (DE) are materials that are used as actuators in adaptive
structures, in particular when large deformations are required. DE actuators belong to
the category of so called “electro active polymer actuators”: they transform electric
energy directly into mechanical work and produce large strains, in the order of 10% to
30%. They were first proposed in 1998 for use as actuators [1]. Important progresses
in this emerging field were achieved in the selection of suitable elastomers and
actuator configurations (e.g. [2], [3] ), in the development of technologies for the
realization of the actuators (e.g. [4]), in the evaluation of potential applications and the
characterization of simple actuators (e.g. [5]). The present work focuses on modeling,
which represents a fundamental task for the development of optimized actuators, for
simulating and predicting the efficiency, functionality and reliability of DE actuators.
Account of previous work on DE modeling is provided in our recent papers [6], [7].
Dielectric elastomer actuators were shown to provide good overall performance [5],
combining large elongation, high energy density, good efficiency and high speed of
response. In particular an acrylic elastomer is investigated here, VHB4910 (3M),
which is widely used for the realization of DE actuators.
The actuators consist basically of a compliant capacitor, with a thin passive elastomer
film sandwiched between two compliant electrodes. The electrostatic pressure acting
on the insulating elastomer film can be calculated for a given applied voltage and film
thickness [1]. Due to the mechanical compression, the elastomer film contracts in the
thickness direction and expands in the film plane directions. This in-plane expansion is
exploited to generate motion or forces. In order to improve their performance
dielectric elastomers are pre-stretched, typically up to five times of their in-plane
dimensions.
Characterization of the constitutive behavior of the elastomer is essential for modeling
purposes. Fitting of large strain constitutive models using the results of conventional
uniaxial tests is not sufficient for this purpose. In fact, in DE actuators the material is
subjected to cyclic loading and unloading in a multiaxial stress state with finite
deformations. Suitable material laws must be developed in order to describe the
multiaxial large deformation behavior of the material and to reproduce the time and
history dependence of its mechanical response.
A systematic and comprehensive experimental characterization has been carried out at
the DE laboratory of EMPA, with large strain uniaxial displacement controlled tests
and multi-axial experiments using a circular actuator [6]. A total number of 40 circular
actuators were built and tested in order to characterize the electro-mechanical response

62
at several pre-strain levels and with different activation voltages. One of the main
objectives of the present paper is to share the experimental data with the DE research
community so that other researchers might use these data for model development and
validation purposes. The experimental details are presented in section 2.
Section 3 describes our evaluation of the experimental data: a quasi-linear viscoelastic
constitutive model (with three different strain energy formulations) is applied and the
corresponding material parameters are determined from the experimental data. This
section describes also the calculation procedure applied for the solution of the inverse
problem. In particular a novel approach is proposed for simulation of the active DE
behavior by finite element (FE) calculations.
The proposed constitutive models are discussed and their predictive capabilities are
evaluated in section 4. The results of a multi-cycle experiment with the circular
actuator (up to 75 activation-deactivation cycles) are shown in order to demonstrate
the limitations of a quasi-linear viscoelastic constitutive model formulation.

2 Experiments

2.1 Material details

VHB 4910 is an acrylic elastomer produced by 3M. The chemical composition is not
declared by the manufacturer. The material is commercially available in form of a
membrane with 1 mm thickness, 23 mm width and arbitrary length. Testing was
performed at room temperature (23° C).

2.2 Uniaxial displacement controlled experiments

Uniaxial tensile tests were performed with a Zwick (Z010) machine (tensile testing
equipment) and relaxation tests with a hand-operated setup under displacement control
conditions (see Figure 1 for the control profiles). For the tensile test the probe
dimensions in the undeformed state were: 1 mm thickness, 10 mm width and 150 mm
length in axial direction. The strain was monitored with an elongation sensor (Zwick
multisens) in the middle of the sample, over an undeformed gauge length of 50 mm.
For the relaxation test the undeformed sample geometry was set to a ratio between
length and width of 10:1, in order to ensure uniaxial stress conditions, with thickness
of 1mm. Several experiments (figure 1) were carried out: (i) relaxation tests (Rel50,
Rel200 and Rel500), (ii) tensile tests at fixed deformation rate (Ten500), (iii) tensile
tests with a deformation history comprising a loading ramp, hold time, and unloading

63
(Ten300). Experimental results (force history for each test) are reported in figure 2.
The nominal stress, the force divided by the initial cross section, is plotted against the
time. Each curve corresponds to the average of two experiments, whereby the scatter
between experiments with identical control profiles was very low. Analysis of the
relaxation curves led to the conclusion that this material might be modeled as
quasilinear viscoelastic [6].

Rel500
500 Ten500

400
Nominal strain [%]

300

200 Rel200 Ten300

100
Rel50
0

0 1000 2000 3000

Time [s]

Figure 1. Control profile for the uniaxial tests.

0.125

0.100
Ten500
Nominal stress [MPa]

Rel500
0.075

0.050

Rel200 Ten300
0.025
Rel50
0.000

0 1000 2000 3000

Time [s]

Figure 2. Nominal stress vs. time for the uniaxial experiments

64
2.3 Experiments with circular actuators

2.3.1 Description of the tests

Multi-axial electromechanical experiments were performed using biaxially prestrained

circular actuators (in a so called circular strain test [8]) at room temperature (23° C),
see figure 3. The actuators consist of a VHB4910 membrane radially pre-strained and
fixed on a circular frame with radius R= 75 mm. A circular area (radius r0 = 7.5 mm)
at the center of the membrane is coated with a mixture of graphite powder (TIMREX
SP30, 11 g) and silicone oil (DC 200 / 100 cs, 10 ml) for the electrodes.
Radial pre-straining was realized by a special device developed at the DE laboratory of
EMPA for biaxial stretching of elastomeric films. With this machine the pre-stretch
ratio λp (for the circular actuator: the deformed radial length divided by the
undeformed radial length of the membrane) can be arbitrarily prescribed. Three
families of actuators were realized with prestretch ratios λp = 3, 4 and 5. After radial
stretching, the film was fixed uniformly along a circular frame. The circular area at the
center of the film was coated with the graphite / silicone electrode on the upper and
lower side (electrode thickness: approximately 40 µm). The electrodes were connected
to a high voltage supplier through a thin metal wire.

Active zone (Coated area) Passive zone

Circular frame

r0 r1
M

Voltage off Voltage on

Figure 3. Arrangement of the circular actuator for electromechanical measurements (r0

= 7.5 mm and R = 75 mm).

The experiment consisted of one single activation cycle in which a constant voltage of
2, 2.5, 3 or 3.5 kV was applied for 900 s. The nominal radial strain sr = r1/r0 - 1 of the
coated area was measured with a video-extensometer (Ovex ME-46) connected to a
PC LabView system for data acquisition and analysis.

65
A total number of about 40 actuators were used for the present experimental work,
with identical conditions of pre-strain and voltage applied to 3 or 4 actuators, in order
to evaluate the repeatability of the test results.

2.4 Experimental results

Measurement results are presented in Figure 4, 5 and 6. For each test, the time history
of nominal radial strain sr is reported. For each actuator group with same testing
conditions an average strain history curve is calculated which is then used for material
modeling, section 3.
25
2000 V
2500 V
20 3000 V
3500 V
Radial strain sr [%]

15

10

0
0 200 400 600 800 1000
Time [s]

Figure 4. Circular actuator tests at λp = 3. The filled symbols represent the average

2000 V
80 2500 V
3000 V
3500 V
60
Radial strain sr [%]

40

20

0
0 200 400 600 800 1000
Time [s]

Figure 5. Circular actuator tests at λp = 4. The filled symbols represent the average

66
 50 2000 V
2500 V
40 3000 V

Radial strain sr [%]

30

20

10

0
0 200 400 600 800 1000
Time [s]

Figure 6. Circular actuator tests at λp = 5. The filled symbols represent the average

Generally, larger scatter is observed for higher activation voltage, with the exception
of the three tests at λp = 3, 3500 V. This might be attributed to the in-homogeneity of
the electrodes, which is expected to increase at larger deformations. At λp = 5, 3500V
the actuators consistently experienced an electric breakdown (no data points
available). The largest radial strain values (up to 70%) were measured for λp = 4,
3500V.

70

60

50 Prestrain 3
40
Prestrain 4
Strain [%]

Prestrain 5
30

20

10

0 1000 2000 3000 4000

Voltage [V]

Figure 7. Time vs. strain at the time 900 s

In Figure 7 the average strains measured at the end of the experiment (900 seconds)
are plotted as a function of activation voltage. The difference between strains at λp = 3
and λp = 4 is considerably larger than the strain difference between λp = 4 and λp = 5.

67
This is the consequence of two concurrent effects: The radial deformation depends on
(i) the applied electrostatic pressure in thickness direction, and (ii) the stiffness of the
membrane. The electrostatic pressure is assumed here to follow eq. (1) (see [1]):
2
⎛U ⎞
pel = ε r ⋅ ε 0 ⋅ ⎜ ⎟ (1)
⎝d⎠

d: membrane thickness, U: voltage, εr (4.7 for VHB 4910, [9]): the dielectric constant
and ε0: the vacuum permittivity (8.5x10-12 As/Vm). Larger pre-strain leads to a
reduction in d, and therefore to a larger electrostatic pressure, for the same activation
voltage. On the other hand the material becomes stiffer (larger tangent modulus) at
higher values of λp so that lower radial strains are obtained for the same electrostatic
pressure.

3 Material modeling

The demands on a constitutive model describing the mechanical response of the

elastomer in the experiments of section 2 are challenging. Non-linear three
dimensional constitutive models have to be formulated and material parameters
determined in order to describe the time dependent mechanical response over a wide
deformation range: uniaxial stretch up to λ = 6, biaxial deformation (from a
kinematical point of view, a circular DE actuator is akin to a biaxial test) at different
pre-strain and voltage levels.
These challenges correspond to two classical problems of material modeling: (i) the
model must be capable to describe the large strain response in uniaxial as well as
multiaxial loading conditions; (ii) the time dependent component of the model must be
capable to describe the deformation history in force controlled experiments as well as
the stress history under displacement controlled loading conditions.
Further challenges are related to the electromechanical coupling in DE: (i) prediction
of the mechanical loading as function of the applied voltage; (ii) influence of the
electrodes on the electromechanical performance; (iii) simulation of the active
behavior by numerical methods (finite elements).

3.1 Constitutive models

The quasilinear viscoelastic model proposed by Fung [10] and applied already for DE
modeling in [6] is used here. It comprises hyperelastic equations (for the large strain
elastic response) and a viscoelastic dissipative component (describing the time

68
dependence of the mechanical response). For the hyperelastic part three different
model formulations are investigated: the strain energy potentials of Yeoh [11], Ogden
[12] and Arruda-Boyce [13]. Incompressible material behavior is assumed for all three
formulations.
The Yeoh form describes the strain energy function W as:

W = C10 ( I1 − 3) + C 20 ( I1 − 3) 2 + C30 ( I1 − 3) 3 (2)

C10, C20 and C30 are material parameters. I1 can be calculated from the eigenvalues of
the deformation gradient tensor, i.e. the so called principal stretch ratios λi (i=1,2 and
3) as

I1 = λ1 + λ2 + λ3
2 2 2
(3)

The Ogden model is characterized through the following strain energy potential with
material parameters µi and αi:
n
µi α
W =∑
α α
(λ1 + λ2 + λ3 )
i i i
(4)
i =1 α i

A strain energy function with three terms (n=3) is used here.

The strain energy form of Arruda-Boyce is based on statistical mechanics and includes
only two material parameters (A and N), which can be related to properties of the
elastomer microstructure:

⎡1 1 2 11 3
W = A ⋅ ⎢ ( I1 − 3) + ( I1 − 9) + 2
( I1 − 27)
⎣2 20 N 1050 N
(5)
19 4 519 5 ⎤
+ 3
( I1 − 81) + ( I1 − 243)⎥
7000 N 673750 N 4 ⎦

The time dependence of the mechanical response is described by assuming time

dependent coefficients in the strain energy functions. The corresponding time
functions are defined through the relaxation function g(t):
K
t
g (t ) = 1 − ∑ g k ⋅ (1 − exp(− )) (6)
k =1 tk

where gk and tk characterize the relaxation behavior. Here K is chosen as 4. For a

specific material the quasi-linear viscoelastic model is appropriate when the relaxation
function g(t) is independent of the applied strain. The uniaxial relaxation tests (section
2) were used to confirm the validity of this assumption for VHB 4910 [6]. The stress -
deformation behavior for arbitrary loading histories is calculated through so called

69
hereditary integrals. Further details on the quasi-linear viscoelastic model are
described in [6].

3.2 Finite element calculations and constitutive model parameters

Material parameters (the parameters of the strain energy functions and of the
relaxation function) were determined by fitting the experimental results of the circular
actuator experiments with activation voltage of 2 kV, at λp = 3, 4 and 5. The resulting
constitutive models were then applied for predicting the response of the other circular
actuator tests (with higher voltages) and uniaxial tests.
In [7] an analytical model is presented for analyzing the circular actuator test in case of
purely elastic material behavior. The problem cannot be solved analytically when the
time dependence of the material response is considered. For this reason, a finite
element simulation of the experiment has been used for the determination of the
material parameters from the circular actuator tests.
The software ABAQUS 6.5.1 [14] has been used for the calculations. The
axisymmetric model consists of 4-node elements (CAX4H: 4-node bilinear
axisymmetric quadrilateral, hybrid, constant pressure). Subsequent mesh refinement
was introduced (with higher mesh density at the extremity of the active zone) in order
to avoid artifacts due to discretization errors. The model considers a symmetry plane in
the horizontal direction (Figure 8). The axis of rotational symmetry is perpendicular to
the membrane plane and passes through the centre of the coated area, see Figure 3.
The electromechanical pressure pel acts on the active zone and depends on the current
membrane thickness and the activation voltage, according to eq. 1.

pel Nodes constrained in radial direction

Rotation axis
Plane of vertical symmetry

Figure 8. Finite element model of a circular dielectric elastomer actuator

70
In [6] a finite element calculation of the circular actuator behavior was proposed in
which the measured radial displacement of the active zone is imposed as kinematic
boundary condition (by prescribing the membrane thickness variation); in those
simulations the corresponding “required” activation voltage represent the result of the
calculation, to be compared with the experimental activation voltage. Here a novel
approach is proposed in which the activation voltage is the input and the time history
of the radial strain the output of the calculation. Two approaches were implemented to
this end:
a) the straight forward approach consists of applying the electromechanical pressure pel
as a ‘surface pressure’. Due to thickness reduction during the activation process, the
amplitude of the pressure (for a given activation voltage) changes according to eq. 1.
The corresponding evolution of the amplitude of pel is defined by a Fortran code (in a
so called “user subroutine”) which calculates the current pressure as a function of the
current thickness d and the activation voltage U (which is constant for each
experiment). The main disadvantage of this approach is that a large displacement
analysis with kinetic (force) boundary conditions requires considerably larger number
of iterations (and in certain cases does not converge) as compared with a calculation
with kinematic (displacement) boundary conditions.

b) this approach aims at realizing the same loading history by applying kinematic
boundary conditions: the vertical displacement uv of the nodes at the upper border of
the active zone is prescribed in order to achieve (at the upper face of the coated zone) a
stress component in vertical direction equal to current value of pel (depending on the
activation voltage and the current membrane thickness). The vertical displacement uv
is controlled using a Fortran code (embedded in a user subroutine): the velocity change
∆v for the current calculation increment is defined as

U pres − U calc
∆v = 0.2 ⋅ ⋅v (7)
abs(U pres − U calc )

where v: velocity in the previous increment, Upres is the prescribed activation voltage
(e.g. 2 kV), and Ucalc is the calculated activation voltage of the previous increment,
based on the values of vertical stress and thickness according to eq. 1. In this way, the
vertical displacement uv is continuously adjusted in order to minimize the discrepancy
between the prescribed activation voltage and the current voltage level.
Procedure (b) provides considerable advantages in terms of calculation times and
convergence. However, this control procedure leads to an unstable behavior when
dealing with the sudden (stepwise) change of the activation voltage at the beginning of

71
the test: for this reason procedure (a) was applied in the initial phase (20 sec.) and
procedure (b) for the remaining duration of the active phase (880 sec.).
In Figure 9 an example is shown of the performance of the proposed procedure (Yeoh
model, λp=4) The applied voltage is compared with the constant prescribed activation
voltage during the active phase. The discrepancy is larger for higher voltage level, but
overall the agreement is satisfactory.

4000

3000
Voltage [V]

2000

1000

0
0 200 400 600 800 1000
Time [s]

Figure 9. Comparison of realized (symbols) and experimentally prescribed (horizontal

line) activation voltages for the finite element calculation of the circular actuator
(example with λp=4).

The simulation of one experiment consists of four steps: (i) radial pre-strain; (ii) hold
time of 1 h (with stress relaxation, corresponding to the experiments); (iii) application
of the activation voltage step according to procedure (a); (iv) control of the constant
voltage level according to procedure (b).
The material parameters for the quasilinear visco-hyperelastic behavior are determined
using an optimization procedure (described schematically in Figure 10) which is
programmed to run automatically in Matlab [15]. Starting with initial material
parameters as a vector X0 Matlab writes an ABAQUS input file with these material
parameters. The components of X0 are the material parameters of the strain energy
function (Cij for Yeoh, µi, αi for Ogden and A, N for Arruda-Boyce) and the relaxation
function parameters (tk and gk). Matlab calls ABAQUS to run the simulation with the
material parameters X0. When the simulation is completed the ABAQUS results
(calculated strain history) are extracted and compared with the experimental strain
history. Next, the error function f (eq. 8) is calculated.

72
 f ( X ) = ∑ f i = ∑ ( sc ( X ) − se ) i
2 2
(8)

f depends on the material parameters (vector X) and is defined as the sum of the
squares of the differences fi between the calculated strains sc and the experimental
strains se. The optimization function fminsearch (in Matlab) is based on the Nelder-
Mead-Simplex algorithm and changes the material parameters (vector X) in order to
find a minimum of f. The whole procedure is repeated until the minimum of f is found.

X0 X

f=fminsearch(X) no Minimum of f ? yes

f = ∑ fi
2
Run ABAQUS

Extract data ssc

[%]

ffi
Strain [%]

i
Strain

Interpolation
ss e
e

Time [s]
Time [s]
Figure 10. Optimization procedure for evaluation of the material parameters

The results of the combined optimization for the circular actuator tests at 2 kV, with λp
= 3, 4 and 5 are given in Figure 11.

73
 5

Radial strain [%]

3

2 Yeoh
Experiment
Arruda-Boyce
1
Ogden

0 200 400 600 800 1000

Time [s]

(i)

10.0

7.5
Radial strain [%]

5.0
Yeoh
Experiment
2.5 Arruda-Boyce
Ogden

0.0
0 200 400 600 800 1000
Time [s]

(ii)

74
 10.0

7.5

Radial strain [%]

5.0 Arruda-Boyce
Ogden
Experiment
2.5
Yeoh

0.0
0 200 400 600 800 1000
Time [s]

(iii)

Figure 11. Optimization of the three models (Yeoh, Ogden and Arruda-Boyce) for 2
kV: (i) λp = 3; (ii) λp =4; (iii) λp = 5.

The material parameters obtained through the optimization procedure are reported in
Tables 1 and 2.

Table 1. Optimized hyperelastic material parameters

C10 C20 C30
Yeoh [MPa] [MPa] [MPa]
0.0827 -0.000747 5.86e-6
µ1 [MPa] α1 [-] µ2 [MPa] α2 [-] µ3 [MPa] α3 [-]
Ogden
0.00858 1.293 0.0843 2.3252 -0.0233 2.561
Arruda- A [MPa] N [-]
Boyce 0.0686 124.88

Table 2. Optimized parameters of the relaxation function

g1 t1 g2 t2 g3 t3 g4 t4
[-] [s] [-] [s] [-] [s] [-] [s]
Yeoh 0.478 0.153 0.205 0.464 0.0727 32.021 0.0492 215.85
Ogden 0.534 0.357 0.134 3.254 0.0495 46.11 0.0290 476.75
Arruda-
0.452 0.341 0.144 2.326 0.0746 33.07 0.0215 313.8
Boyce

75
3.3 Comparison of simulations and experimental results

The radial strain histories calculated with the parameters of Tables 1 and 2 for the
whole experimental range of circular actuator tests are reported in Figures 12 to 14 for
the different constitutive model formulations.

25
2000 V
2500 V
20 3000 V
3500 V
Radial strain [%]

15

10

0 200 400 600 800 1000

Time [s]

(i)

80 2000 V
2500 V
3000 V
60 3500 V
Radial strain [%]

40

20

0 200 400 600 800 1000

Time [s]

(ii)

76
 40

30

Radial strain [%] 20

10

2000 V
2500 V
0
3000 V

0 200 400 600 800 1000

Time [s]

(iii)

Figure 12. Comparison between experimental data and simulation for the Yeoh model
for prestrain λp = 3 (i), λp = 4 (ii) and λp = 5 (iii). The filled symbols represent the
experimental data and the open symbols the corresponding simulation.

2000 V
2500 V
20
3000 V
3500 V
15
Radial strain [%]

10

0 200 400 600 800 1000

Time [s]

(i)

77
 80
2000 V
2500 V
3000 V
60 3500 V

Radial strain [%]

40

20

0
0 200 400 600 800 1000
Time [s]

(ii)

2000 V
2500 V
40 3000 V
Radial strain [%]

20

0 200 400 600 800 1000

Time [s]

(iii)

Figure 13. Comparison between experimental data and simulation for the Ogden
model for prestrain λp = 3 (i), λp = 4 (ii) and λp = 5 (iii). The filled symbols represent
the experimental data and the open symbols the corresponding simulation.

78
 2000 V
25 2500 V
3000 V
20
3500 V

Radial strain [%] 15

10

0 200 400 600 800 1000

Time [s]

(i)

80

60
2000 V
2500 V
Radial strain [%]

3000 V
40
3500 V

20

0 200 400 600 800 1000

Time [s]

(ii)

79
 50
2000 V
2500 V
40 3000 V

Radial strain [%]

30

20

10

0 200 400 600 800 1000

Time [s]

(iii)

Figure 14. Comparison between experimental data and simulation for the Arruda-
Boyce model for prestrain λp = 3 (i), λp = 4 (ii) and λp = 5 (iii). The filled symbols
represent the experimental data and the open symbols the corresponding simulation.

The corresponding predicted uniaxial behavior for the different models are compared
with the experimental data in Figures 15 to 17. The uniaxial test simulation was
performed according to the procedure described in [6].

0.2
Rel50
Rel200
Rel500
Nominal stress [MPa]

Ten300
Ten500
0.1

0.0

0 1000 2000 3000

Time [s]

Figure 15. Uniaxial behavior: experimental data (filled symbols) compared to the
Yeoh model simulation (open symbols).

80
 0.2
Rel50
Rel200
Rel500

Nominal stress [MPa]

Ten300
Ten500
0.1

0.0

0 1000 2000 3000

Time [s]

Figure 16. Uniaxial behavior: experimental data (filled symbols) compared to the
Ogden model simulation (open symbols).

0.2
Rel50
Rel200
Rel500
Nominal stress [MPa]

Ten300
Ten500
0.1

0.0

0 1000 2000 3000

Time [s]

Figure 17. Uniaxial behavior: experimental data (filled symbols) compared to the
Arruda-Boyce model simulation (open symbols).

4 Discussion

The predictive capabilities of the three different models used in the present work can
be evaluated from the results of section 3.3. All three models show reasonable
agreement with respect to the actuator behavior, whereas the correspondence between
experiment and simulation for the uniaxial test is not satisfactory. Obviously, an

81
optimization procedure that would include all the uniaxial tests could lead to improved
correspondence for these experiments (but probably to worse predictive capabilities in
the biaxial tests).
A comparison of the radial strain predicted by the different models at the end of the
active phase (900 sec) and the corresponding scatter band of the experimental values is
reported in Figures 18.

25

20 Yeoh
upper border
Arruda - Boyce
15 lower border
Strain [%]

Ogden
10

0
2000 2500 3000 3500
Voltage [V]

(i)

60
upper border
lower border
Ogden
Strain [%]

40 Yeoh
Arruda - Boyce

20

0
2000 2500 3000 3500
Voltage [V]

(ii)

82
 upper border
lower border
40
Yeoh
Arruda - Boyce
Ogden
Strain [%]
20

0
2000 2500 3000
Voltage [V]

(iii)

Figure 18. Comparison between simulation and experimental values (with scatter
band) for the strain at 900 s for λp= 3 (i), λp= 4 (ii) and λp= 5 (iii)

In case of λp=4 and 5, at higher activation voltages the discrepancy between simulation
and experimental data is larger. This might be considered as a limitation of the
proposed constitutive model. At this deformation level, however, the scatter in the test
data is quite large and in certain cases electric breakdown occurred. It is possible that
in-homogeneities in the electrodes or in the thickness distribution influence the
actuator behavior to such extent that the electromechanical coupling might not be
correctly described by eq. 1.
Mechanical instability (one possible failure mode of DE actuators, see [7]) is predicted
in certain calculations (Arruda-Boyce-model at λp=4, 3.5 kV, Ogden model at λp=5, 3.
kV). The Arruda-Boyce-model at prestrain λp=5, 3 kV converges, but the radial strain
reaches a value of 120 %.
It is noteworthy that the Arruda-Boyce model shows good predictive capabilities
despite the lower number of material parameters available as degree of freedom for the
optimization procedure.

4.1.1 Cyclic behavior

An additional experiment with a circular actuator has been performed in order to

evaluate the predictive capabilities of the proposed quasi-linear viscoelastic model in
case of cyclic behavior. The actuator was activated with a voltage of 3 kV over a time

83
period of 240 seconds and then deactivated for 60 seconds. The cycle was repeated 75
times.
The quasi-linear viscoelastic model leads to excellent predictions when fitted to one
specific pre-strain and voltage level, as demonstrated by the comparison between
simulation (here a Yeoh model) and measured radial strain history for the first cycle of
this experiment, Figure 19.

25

20

15
Strain [%]

10 Experiment
FE simulation
5

0 100 200 300

Time [s]

Figure 19. Calculated and measured radial strain history of the first cycle (Yeoh
model)

Simulation of the radial strain for the whole cyclic history of the experiments shows an
increasing discrepancy with respect to the measurements, Figure 20.

40

30
Strain [%]

20

10

Experiment
0 FE-calculation

0 10000 20000
Time [s]

Figure 20. Cyclic behavior: experiment and simulation of 75 cycles

84
The progressive increase in the maximum radial strain observed in the experiments
cannot be reproduced by the simulation with the quasi-linear viscoelastic model. This
increase might be attributed to changes in the mechanical response of the elastomer
due to the loading history, which could be described with a viscoplastic constitutive
model with internal variables (and corresponding evolution equations). Further
experiments (with different voltage range, frequencies and pre-strain level) will help
understanding whether this observation is related to material degradation or to
temperature effects influencing the material response and the electromechanical
behavior.

5 Conclusions

The mechanical behavior of the acrylic elastomer VHB 4910 used in DE actuators has
been investigated. Next to force and displacement controlled uniaxial data,
experimental observations from tests with 40 circular actuators are reported. The
activation behavior for different prestrain levels and at different activation voltages is
presented. Multiple tests at same prestrain and voltage levels allow evaluating the
scatter of the electromechanical response obtained from virtually identical actuators.
The experimental data presented in section 2 represent a unique collection that can be
used for material and actuator modeling purposes.
The numerical procedure adopted for simulating the electromechanical behavior of the
circular actuator allowed determining material model parameters from the
experimental data through an optimization algorithm. The novel approach proposed
for electromechanical coupling might be generally useful for finite element
simulations of DE actuators.
Three different constitutive model formulations were used and compared in their
predictive capabilities. It is noteworthy that the Arruda-Boyce model, having only 2
parameters for describing the three dimensional hyperelastic response, leads to
comparable matching as the Yeoh and Ogden models. Model parameters were
determined from a subset of circular actuator tests and then applied for simulating the
whole range of experiments performed. Since the prediction of the circular actuator
behavior was satisfactory, the proposed models might provide useful results for
actuator design and optimization purposes.
On the other hand, large discrepancies were obtained between the uniaxial data and the
corresponding simulations for all three model formulations. This mismatch represents
a limitation of the proposed models, which is related to the problem of describing the
time dependent material response over a wide range of uniaxial and multiaxial

85
deformation states and with force or displacement controlled experiments. Further
limitations of the proposed quasi-linear viscoelastic constitutive models have been
demonstrated when applied for predicting the multicycle actuator behavior.

Acknowledgements
Financial support from the Swiss National Science Foundation (Project 200021-
107661/1) is gratefully acknowledged.

6 References
[1] R.E. Pelrine, R.D. Kornbluh, J.P. Joseph, Electrostriction of polymer dielectrics
with compliant electrodes as a means of actuation, Sens. Actuators, A 64 (1998)
77-85
[2] R. Kornbluh, R. Pelrine, Q. Pei, S. Oh, J. Joseph, Ultrahigh strain response of
field-actuated elastomeric polymers, Proc. SPIE, 3987 (2000) 51-64
[3] R. Pelrine, P. Sommer-Larsen, R. Kornbluh, R. Heydt, G. Kofod, Q. Pei, P.
Gravesen, Applications of dielectric elastomer actuators, Proc. SPIE, 4329
(2001) 335-349
[4] X. Zhang, C. Löwe, M. Wissler, B. Jähne, G. Kovacs, Dielectric elastomers in
actuator technologies, Advanced Materials Engineering, 7, No.5, (2005) 361-367
[5] Y. Bar-Cohen, Electroactive Polymer (EAP) Actuators as Artificial Muscles -
Reality, Potential and Challenges, vol. PM98, SPIE Press, Washington, 2001
[6] M. Wissler, E. Mazza, Modeling and simulation of dielectric elastomer actuators,
Smart Materials and Structures, 14 (2005) 1396 - 1402
[7] M. Wissler, E. Mazza, Modeling of a prestrained circular actuator made of
dielectric elastomers, Sens. Actuators, A 120 (2005), 184-192
[8] R. Pelrine, R. Kornbluh, Q. Pei, J. Joseph, High-speed electrically actuated
elastomers with strain greater than 100%, Science, 287 (2000) 836-839
[9] G. Kofod, P. Sommer-Larsen, R. Kornbluh, R. Pelrine, Actuation response of
polyacrylate dielectric elastomers, J. Intel. Mat. Syst. Str., 14 (2003) 787-793
[10] Y.C. Fung, Biomechanics. Mechanical properties of living tissues, 2nd edn.,
Springer-Verlag, New York, 1993, p. 277
[11] O.H. Yeoh, Characterization of elastic properties of carbon-black-filled rubber
vulcanizates, Rubber Chem. Technol., 63 (1990) 792-805
[12] R.W. Odgen, Large deformation isotropic elasticity – on the correlation of theory
and experiment for incompressible rubberlike solids, Proc. R. Soc. Lond., A 326
(1972) 565-584
[13] E.M. Arruda, M.C. Boyce, A three-dimensional constitutive model for the large
stretch behavior of rubber elastic materials, J. Mech. Phys. Solids, 41, No.2
(1993) 389 - 412

86
[14] ABAQUS/Standard Manual, 2003. Version 6.4.1, Hibbitt, Karlsson & Sorensen,
Inc.
[15] Matlab 7.0, 2004, The MathWorks, Inc.

87
3.4 Article 4

This section contains a reprint of the article:

M. Wissler and E. Mazza, Electromechanical coupling in dielectric elastomer
actuators, submitted to Sensors and Actuators A

Background motivation: In article 3, the agreement between experiments and

simulations was shown to be reasonable for the actuator behavior but not for the
uniaxial behavior. This might be the consequence of the electromechanical coupling
not being adequately described by Pelrine’s equation (eq. (1.1)). Therefore, the
electromechanical coupling was evaluated both analytically and numerically. Further,
the dielectric constant was measured at different prestrains and spring roll experiments
were carried out for an overall verification.

Summary of methods: Energy considerations were performed analytically and

numerically for a circular actuator. For the analytical investigation, an energy balance
was used for determining the electrostatic pressure. An electrostatic analysis which
also took into account the electrodes was performed with COMSOL Multiphysics. The
capacity and the thickness of the film were measured at different prestrain levels and
the dielectric constant was determined. Six spring rolls were fabricated. The force
difference was measured in an experiment where the elongation of the spring roll was
constrained.

Summary of results: Results show that analytical and numerical considerations are
consistent. The latter provide a new physical interpretation of the electrostatic forces
acting on the DE film, with contributions from in-plane and out-of-plane stresses.
Pelrine's equation is confirmed by superposition of a hydrostatic stress state. The value
of the dielectric constant decreases with increasing pre-stretch ratio, from 4.7 for the
un-stretched film, down to 2.6 for an equi-biaxial prestrain with λp=5. Spring roll
measurements confirm a value of 3.2 at prestrain λp=4.

Main conclusions, link to the next section: The equation proposed by Pelrine was
verified theoretically and experimentally. The dielectric constant is about 3.2 and not
4.7 for a relevant prestrain of 4. With the modified dielectric constant, uniaxial
simulations agree to a great extent to the corresponding experiments without loss of
predictability of the actuator behavior (see article 3).

88
 Electromechanical coupling in dielectric elastomer
actuators
Michael Wissler1,2 and Edoardo Mazza2,1
1
Swiss Federal Laboratories for Materials Testing and Research (EMPA), 8600
Dübendorf, Switzerland
2
Swiss Federal Institute of Technology (ETH), Dept. of Mechanical Engineering,
8092 Zürich, Switzerland

Abstract
In this paper the electromechanical coupling in dielectric elastomer actuators is
investigated. An equation proposed by Pelrine et al. (Sensors and Actuators, A 64,
1998) is commonly used for the calculation of the electrostatic forces in dielectric
elastomer systems. This equation is analyzed here with (i) energy consideration and
(ii) numerical calculations of charge and force distribution. A new physical
interpretation of the electrostatic forces acting on the dielectric elastomer film is
proposed, with contributions from in-plane and out-of-plane stresses. Representation
of this force distribution using Pelrine’s equation is valid for an incompressible
material, such as the acrylic elastomer VHB 4910. Experiments are performed for the
measurement of the dielectric constant εr of the acrylic elastomer VHB 4910 for
different film deformations. The values of εr are shown to decrease with increasing
pre-stretch ratio λp, from 4.7 for the un-stretched film, down to 2.6 for equi-biaxial
deformation with λp=5. This result is important in that it corrects the constant value of
4.7 largely applied in literature for pre-stretched dielectric elastomer actuator
modeling. With the results of this work the predictive capabilities of a model
describing the three-dimensional passive and active actuator behavior are remarkably
improved.

Keywords: Electromechanical coupling, dielectric elastomer, dielectric constant,

actuator, modeling.

89
1 Introduction

Actuators made of dielectric elastomers (DE) consist of a compliant capacitor, with a

thin soft elastomer film sandwiched between two compliant electrodes [1, 2].
Application of a voltage leads to electrostatic forces that deform the elastomer
membrane, which contracts in the thickness direction and expands in the film plane.
This in-plane expansion is exploited to generate motion or forces. Dielectric
elastomers actuators transform electrical energy directly into mechanical work and are
able to generate in-plane deformations of up to 30% with respect to the un-activated
configuration. They belong to the group of so called electroactive polymers (EAP).
Modeling and simulation of DE systems are key steps in the design and optimization
of DE actuators. Challenges thereby are represented by the description of (i) the
passive mechanical behavior of the elastomers (with large strains, time and history
dependence of the mechanical response) and (ii) the mechanical forces generated by
the electric field, i.e. what is often referred to as “electromechanical coupling”.
Significant work has been performed for the mechanical characterization of the
elastomers [3-8], whereas only few papers exist on electromechanical coupling.
Electromechanical coupling is commonly described using a model proposed by Pelrine
et al. [1]. In this model electrostatic forces are assumed to act in the direction
perpendicular to the insulating elastomer film, and their magnitude is characterized by
pel, the so called “electrostatic pressure”. pel is calculated for a given applied voltage U
and film thickness z as

U2
pel = ε 0 ⋅ ε r ⋅ 2
= ε 0 ⋅ε r ⋅ E 2 (1)
z

Thereby, ε0 is the free-space permittivity (8.854·10-12 As/Vm), εr is the elastomer

dielectric constant and E is the electric field.
Eq. 1, which will be referred to as “Pelrine’s equation” in this paper, has been derived
for free boundary conditions and is generally accepted as representing the
electromechanical coupling in dielectric elastomer systems. Kofod et al. [4] validated
the equation for silicone experimentally. To our knowledge, no comprehensive
analysis of this expression has been presented so far.
The present investigations on electromechanical coupling are mainly motivated by
inconsistent results obtained from uniaxial mechanical tests and DE actuator
experiments recently presented in [8]. We reported there on an extensive experimental
characterization of so called “circular actuators” made of VHB 4910 (an acrylic
elastomer produced by 3M) with tests over a wide pre-strain and voltage range. Visco-

90
hyperelastic constitutive equations were used to describe the response of the elastomer.
The parameters of the mechanical model were determined from the results of circular
actuators experiments with radial pre-stretch ratios λp = 3, 4 and 5 (the pre-stretch ratio
λp is defined as the ratio between radius of the circular actuator after and before
stretching) at a constant voltage of 2 kV. Figure 1a shows the comparison between
simulation and experimental results for the model using the Arruda-Boyce strain
energy formulation [9]. Electromechanical coupling was thereby described by eq. (1)
and a value of 4.7 was used for the dielectric constant of VHB 4910, according to [10,
11]. Despite the good predictive capabilities of the model for the circular actuator test,
the constitutive equations failed to describe the passive uniaxial response, with
significant discrepancy between data points from tensile and relaxation experiments
and the corresponding simulations, Figure 1b.

0.2
Nominal stress [MPa] Rel50
10 Rel200
Rel500
Strain [%]

Ten300
0.1 Ten500
5

prestrain λp = 3
prestrain λp = 4
0 prestrain λp = 5 0.0
0 200 400 600 800 1000 0 1000 2000 3000
Time [s] Time [s]

(a) (b)
Figure 1. Experiment vs. simulation for (a) circular actuator tests at λp = 3, 4 and 5,
and (b) uniaxial experiments with three relaxation tests (Rel50, Rel200 and Rel500)
and two tensile tests (Ten300 and Ten500) [8]. The filled symbols represent the
experimental data and the open symbols the corresponding simulations.

Describing the mechanical behavior over a wide range of stress and deformation states
represents a common challenge for three-dimensional constitutive equations. We
therefore concentrated first on the mechanical model formulation when searching for
the causes of this inconsistency. The corresponding exercises using different strain
energy formulations (with the strain energy functions of Yeoh [12] and Ogden [13] in
addition to that of Arruda-Boyce) failed to overcome the inconsistency. We then
focused our attention on the electromechanical coupling model, thereby questioning (i)
the general validity of eq. (1), and more specifically (ii) the adopted value of 4.7 for
the dielectric constant εr. The results of these investigations are reported in this paper.

91
In section 2 we analyze Pelrine’s equation for the case of a circular actuator. In a first
step an energy balance is formulated which includes the energy from the power source,
the energy of the electrical field and the mechanical energy. In a second step, the
charge and the forces distributions on a circular actuator are calculated numerically. In
section 3 the dielectric constant is determined for VHB 4910 with two different
approaches: (i) the dielectric constant is measured using a capacitor set-up, at different
pre-strain levels (λp = 1, 3, 4 and 5) and frequencies (100 Hz and 10 kHz); (ii) the
electromechanical forces and thus the corresponding dielectric constant are measured
in experiments with a spring roll actuator. This test is similar to an experiment
(‘blocking force experiment’) proposed by Kofod et al. [4, 11] for measuring the
dielectric constant. The results of these investigation and their implications for DE
actuator models (in particular concerning the inconsistency illustrated in Figure 1) are
discussed in Section 4.

2 Electromechanical coupling

A circular actuator consisting of a dielectrica (elastomer) and two electrodes (Figure 2)

is selected as electromechanical system. The following assumptions hold for this
analysis: (i) electrostrictive effects are negligible [11], (ii) the electrodes are assumed
to be ideally compliant (they do not constrain the elastomer mechanically) and (iii) the
elastomer is incompressible.

2.1 Energy considerations

A closed system including the circular actuator is considered (Figure 2). Application
of a voltage U leads to a charge Q on the electrodes. The time rate of change of Q
correspond to the current I. Charge Q on the electrodes and voltage U are linked. Most
energy considerations in literature are formulated by either a constant voltage or a
constant charge [1]. This corresponds to a usual approach to calculate forces of plate
capacitors. In the present analysis charge and voltage are described as variables
dependent on the deformation of the elastomer film. For the present configuration the
deformation consists in equi-biaxial in-plane extension and the corresponding out-of-
plane contraction. For an incompressible material (volume is constant) this
deformation can be described by the film thickness z. U and Q are therefore handled as
functions of z.

92
 z

Dielectrica
Electrodes U
System‘s
I z boundary
U r

Active zone Passive zone

Figure 2. Sketch of the considered electromechanical system. The film thickness z is

indicated.

Three energy forms are considered for the energy balance: the energy from the
electrical power source Wext, the energy of the electric field Wel and the mechanical
energy Wm. All three energies are dependent on the thickness z. The energy balance
for the closed electromechanical system for an incremental deformation dz is given by
eq. (2).

dWext ( z ) dWel ( z ) dWm ( z )

= + (2)
dz dz dz

In the following Wext, Wel and Wm are determined. The time derivative of the energy
from the power source is related to current I and voltage U, as

dWext dQ
=U ⋅ I =U ⋅ (3)
dt dt

where t is the time. Considering z(t) and applying the chain rule leads to

dWext dQ( z )
= U ( z) ⋅ (4)
dz dz

Q and U are linked as Q(z) =C(z) ·U(z) with the capacitance C defined here as

ε 0 ⋅ε r ⋅ A ε 0 ⋅ ε r ⋅V0
C= = (5)
z z2

93
where A is the area coated by the electrodes and V0 is the elastomer volume (A·z) of
the active zone, which remains constant during deformation. dQ/dz can thus be
expressed as:

dQ( z ) d ⎛ U ( z ) ⎞ ε ⋅ ε ⋅V ⎛ dU U⎞
= ε 0 ⋅ ε r ⋅ V0 ⋅ ⎜ 2 ⎟ = 0 r2 0 ⋅ ⎜ − 2⋅ ⎟ (6)
dz dz ⎝ z ⎠ z ⎝ dz z⎠

dWext/dz is given by the combination of eqs. (4) and (6):

dWext ε 0 ⋅ ε r ⋅ V0 ⋅ U ⎛ dU U⎞
= 2
⋅⎜ − 2⋅ ⎟ (7)
dz z ⎝ dz z⎠

The rate of change of the energy of the power source depends on the rate of change of
the voltage. The energy Wel stored by the electric field in a capacitor is

1
Wel = C ⋅ U 2 (8)
2

by inserting C from eq. (5) the derivative dWel/dz results:

dWel 1 d ⎛ U ( z)2 ⎞ ε ⋅ ε ⋅V ⋅U ⎛ dU U ⎞
= ε 0 ⋅ ε r ⋅ V0 ⋅ ⎜⎜ 2 ⎟⎟ = 0 r 2 0 ⋅⎜ − ⎟ (9)
dz 2 dz ⎝ z ⎠ z ⎝ dz z⎠

Note that also dWel/dz depends on dU/dz. Equations (9), (7) and (2) lead to the
following result for the derivative of the mechanical energy dWm/dz:

dWm ε 0 ⋅ ε r ⋅ V0 ⋅ U 2
=− (10)
dz z3

The basic assumption of Pelrine’s model is that electrostatic forces are homogeneously
distributed over the coated area A= V0/z and act in z direction. The mechanical energy
represents the work of the external (electrostatic) forces on the film. With the
assumption that dWm/dz is the total external force, the corresponding pressure pz is
obtained as:
2
dWm 1 ⎛U ⎞
pz = − ⋅ = ε0 ⋅εr ⋅⎜ ⎟ (11)
dz A ⎝z⎠

This result confirms Pelrine’s equation, when comparing pz with pel from eq. (1).
Pelrine et al. refer to the contribution of lateral and out-of-plane effects, included in eq.
(3) in [1]. By the definition of the pressure pz (eq. (5) in [1] and eq. (11) in this work)
the presence of forces acting only in z-direction is assumed. The influence of radial
(“lateral”) electrostatic forces [1, 14] can be determined with a direct calculation of
charge and electrostatic force distributions. To this end, for the configuration shown in

94
Figure 2, a corresponding numerical calculation has been carried out and is described
in the next section.

2.2 Charge and force distributions

The simulation program COMSOL Multiphysics [15] has been used for the present
calculations. An axisymmetric model of a circular actuator (element type: ‘Lagrange-
quadratic’) including a dielectrica and two electrodes has been created, Figures 3 and
4. The electrodes are assumed to have straight edges. The following geometrical
parameters are used, Figure 3: the thickness of the dielectrica is zd = 60µm, this
corresponds to a VHB 4910 actuator pre-stretched by a factor of approximately λp=4.
The radius of the dielectrica is rd = 14 mm. The radius of the electrode is rel = 7 mm
and the thickness of the electrode is zel = 20µm. Microscopic investigations showed
that zel = 20µm is a reasonable value for a hand-applied graphite powder / silicone oil
electrode.
External surface area
z*
Lateral surface area
fe (r)
fl (z)
zel

fi (r)

rel
Internal surface area

z
zd

U r

rd

Figure 3. Sketch of the COMSOL Mulitphysics model. For the electrode, the
coordinates r, z and z* are introduced for later reference

The dielectric constant of the dielectrica was chosen as εr = 4.7. For the bordering
empty space the dielectric constant of vacuum εr = 1 was used. As boundary condition
a potential difference of U = 3 kV has been applied between the electrodes. Since the
electrodes are conductive, all the charges are on the surface area of the electrodes: the
so called internal surface area, lateral surface area and external surface area, indicated
in Figure 3. The mesh is shown in Figure 4. Singularities at corners have been closely
investigated at the upper electrode with a strong mesh refinement at the corresponding
locations.

95
 Rotation axis
(a)

Electrode
Dielectrica
Vacuum

(b) Vacuum

Electrodes Dielectrica

Vacuum

Figure 4. (a) Sketch of the COMSOL Multiphysics model and (b) meshing of the part
indicated in (a)

All electric field variables (electric potential, electric field, surface charge density ρs)
are calculated by solving Poisson’s equation numerically. On the rotational axis a
suitable symmetry boundary condition is applied. Further information on the
numerical calculation can be found in [15].
The surface charge density of the internal surface area ρi(r) and the surface charge
density of the lateral surface area ρl(z*) are shown in Figure 5. The surface charge
density of the external surface area is neglected because there is no charge on this
surface except at the right corner, with a singularity which does not influence the
electromechanical behavior.
0.0050 1
Surface charge density [C/m ]

Surface charge density [C/m ]

2

0.1

0.0025 0.01

1E-3

0.0000 1E-4
0.0000 0.0025 0.0050 0.0075 0 10 20
Radial coordinate r [m] Coordinate z* [µm]

(a) (b)
Figure 5. Charge distributions on the electrode for (a) the internal surface area and (b)
the lateral surface area

96
The charges on the electrodes cause electrostatic forces. These forces are parallel to
the electric field. Both forces and electric field vectors are perpendicular to the
electrode’s surface area. The calculated electric field distribution is qualitatively
represented in Figure 6.

Figure 6. Electric field distribution: arrows length is proportional to field intensity.

The electrostatic forces are described by Maxwell’s stress tensor tMij [16]. The stress
vectors at the electrode surface are parallel to the electric field vectors and their
magnitude tM is calculated from the electric field magnitude E or alternatively from the
surface charge density ρs, see [3].

1 ρs2
tM = ε 0 ⋅ ε r ⋅ E 2 = (12)
2 2 ⋅ε0 ⋅εr

In Figure 7 the electrostatic stress distribution for the internal surface area tM,i and the
lateral surface area tM,l are shown. For the calculation of tM,l the dielectric constant of
vacuum (εr=1) was used because the lateral surface area borders the empty space.

97
 100000
Maxwell stress tM,i [N/m ] 10
10

Maxwell stress tM,l [N/m ]

9
2

2
10
75000 8
10
7
10
50000
6
10
5
25000 10
4
10

0 10
3

0.0000 0.0025 0.0050 0.0075 0 10 20

Radial coordinate r [m] Coordinate z* [µm]

(a) (b)
Figure 7. Electrostatic stress distribution for (a) the internal surface area and (b) the
lateral surface area

The stress for the internal surface area tM,i is constant with a singularity at the edge of
the electrode. Singularities occur also at the extremities of the lateral surface.
Subsequent mesh refinement have shown that (i) the singularity for tM,i has no
influence on the electromechanical behavior, and (ii) the value of the total radial force
Fr from integration of tM,l over the thickness converges to a finite value.
The mechanical pressure pz transmitted from the electrode to the dielectrica in
(negative) z-direction (see Figure 8) is constant over the coated area and has a value of
pz = 52018 N/m2.
The total force Fr caused by the Maxwell stress tM,l of the lateral surface area (see
Figure 3) is given by
z el

Fr = ∫ 2 ⋅ π ⋅ rel ⋅ tM ,l ( z*) ⋅ dz * (13)

0

Fr converges after mesh refinement to a value of 0.06784 N. As a simplification, tM,l

can be replaced by a statically equivalent constant stress distribution in radial direction
acting over the thickness of the elastomer film, see Figure 8. The corresponding
“lateral pressure” value (pr, Figure 8) is calculated taking into account the contribution
of both electrodes (factor 2 in eq. 14):

Fr Fr
pr = 2 ⋅ = (14)
2 ⋅ π ⋅ rel ⋅ z d π ⋅ rel ⋅ z d

The calculated value of pr for the applied mesh refinement is 51415 N/m2.

98
The electromechanical pressure pel calculated with eq. 1 is, for the present case, pel
=104036 N/m2. This value is double as much as the calculated value of pz and, with a
slight approximation due to numerical errors, pr.

1 1 U2
p z = pr = ⋅ pel = ⋅ ε r ⋅ ε 0 ⋅ 2 (15)
2 2 zd

z
pz

zd r pr

Active zone Passive zone

Figure 8. Out-of-plane pressure pz and lateral stress pr acting on the dielectrica in a

circular actuator, axisymmetric view.

This result shows that (i) the electrostatic forces act in out-of-plane and in-plane
direction, and (ii) the out-of-plane component is half the value predicted with Pelrine’s
equation.
For an incompressible material, however, Pelrine’s model can be shown to provide a
suitable description of electromechanical coupling. In fact, a superimposed hydrostatic
stress state does not affect, by definition, the deformation of the incompressible
elastomer. For the present case a hydrostatic stress with negative sign and magnitude
of 0.5pel is superimposed to the vertical and radial components pz and pr. The resulting
kinetic boundary condition is a homogeneous out-of-plane compression with pel, i.e.
the loading conditions predicted with Pelrine’s model.

3 Dielectric constant

The analysis presented in section 2 has confirmed that eq. 1 can be used for modeling
electromechanical coupling in DE systems. The value of the dielectric constant for
VHB 4910 is investigated in this section. Several researchers, [3, 5, 8, 17, 18], adopted
a deformation independent value of about 4.7, originally proposed by Kofod [10].
Different techniques were used here to measure εr for un-deformed and pre-stretched
elastomer films.

99
3.1 Capacitor set-up

A capacitor set-up has been used to measure the dielectric constant of VHB 4910 at
room temperature (23 °C). εr has been determined for different in-plane pre-stretch
ratios (λp = 1, 3, 4 and 5) at the frequencies 100 Hz and 10 kHz. The dielectric
constant can be obtained from measurements of capacitance and geometry of a
capacitor:

C⋅z
εr = (16)
A⋅ε0

The elastomeric film is pre-deformed by a pre-stretching machine [8, 19] and glued on
a circular frame (radius 15 mm), Figure 9. For each pre-stretch ratio five samples have
been created for multiple measurements. Each sample is inserted between two gold
plungers with a radius of 12.5 mm (Figure 9). The capacitance is measured by a LCR-
meter (an instrument for measuring the inductance (L), capacitance (C) and resistance
(R), 4263B from Agilent) connected to two gold tubes (Figure 9).

Circular Upper Lower

frame plunger plunger

(a) (b)

Figure 9. Device for measuring the dielectric constant: (a) disassembled and (b)
assembled device under the mechanical dial gage for thickness measurement.

The thickness is measured by a mechanical dial gage (Compac, Switzerland,

measurement range: 5 mm, resolution: 0.001 mm). The displacement of the upper
capacitor plate is measured with and without the elastomer sample. The thickness
corresponds to the displacement difference. The results of the dielectric constant are
presented in Table 1 and Figure 10.

100
Table 1. Dielectric constant (mean value ± standard deviation) for VHB 4910 at different
frequencies and pre-strain levels
εr [-] εr [-]
Pre-strain λp [-]
at 100 Hz at 10 kHz
1 4.68 ± 0.029 4.30 ± 0.025
3 3.71 ± 0.088 3.41 ± 0.098
4 3.34 ± 0.152 3.08 ± 0.184
5 2.62 ± 0.378 2.40 ± 0.410

5 5
Dielectric constant [-]

Dielectric constant [-]

4 4

3 3

2 2
Upper border Upper border
Value (frequency: 100 Hz) Value (frequency: 10 kHz)
1 1
Lower border Lower border
0 0
1 2 3 4 5 1 2 3 4 5
Pre-stretch ratio [-] Pre-stretch ratio [-]

(a) (b)
Figure 10. Dielectric constant in dependence of the pre-stretch ratio (with scatter
band) for (a) 100 Hz and (b) 10 kHz

3.2 “Spring roll” set-up

Dielectric constant measurements have been carried out using DE actuators in a

cylindrical configuration, called “spring roll”. Six spring rolls were manufactured and
tested. The VHB 4910 membrane was coated with a mixture of graphite powder
(Superior Graphite, ABG1005, 20 g) and silicone oil (Dow Corning, DC 200/50 cs, 45
g) as electrodes. A brief account on the actuator configuration is given below and
detailed information can be found in [19].
The elastomer film is pre-strained biaxially (using the pre-stretching device, [8, 19])
with the pre-stretch ratios λx = 4 and λy = 4, see Figure 11a. The pre-strained film is
fixed to a rigid frame and one electrode is applied on one side (Figure 11). The
geometrical parameters after pre-straining are x1 = 143 mm, y1 = 445 mm and z1 =
62.5 µm. The coated (active) zone is in the middle of the actuator with xa = 110 mm.
The ratio r of the coated area divided by the whole area is r = xa/x1 = 77 %. A second
layer consisting of the pre-stretched elastomer film and a second electrode is glued

101
over the first layer at the uncoated side. Then, both layers are wrapped around an
elastic core, Figure 11.
x
z
Fx Fx
z1 y

Passive zone

pel
x1 xa Active zone
pel

y1
Fx Fx

(a) (b)

Figure 11. (a) One layer (coated, pre-strained elastomer film) and (b) complete spring
roll actuator

The spring roll is hold at the extremities in x-direction by the force Fx. The experiment
is performed at room temperature (23 °C). A voltage U is applied with a high voltage
amplifier (Trek, Model 5/80) linearly increasing from 0 to 3 kV in 2 minutes. The
force Fx required to keep the length of the actuator constant is continuously measured
with a force transducer (HBM type S2). The experiment is controlled with a PC
LabView system. The measured force difference ∆F is defined as

∆F (U ) = Fx (U ) − Fx (U = 0) (17)

The actuators length remains unchanged despite the application of the voltage (and
thus application of electrostatic forces). This means that the electrostatic forces
together with the variation of axial stresses correspond to a hydrostatic stress state. As
a consequence, measurement of the axial force (or the corresponding stress
component) provides a direct measurement of the electrostatic stress:

∆Fi
pel = = ε0 ⋅εr ⋅ E2 (18)
2 ⋅ x1 ⋅ y1

This equation allows determining the dielectric constant from the measurement of the
force difference of an ‘ideal’ spring roll ∆Fi where the whole film is coated, r=100%.
The present spring rolls have r=77%. The presence of passive parts (the remaining

102
23%) causes a deformation of the film with elongation in the coated and contraction in
the uncoated area, Figure 12. As a consequence, the change in axial force is somewhat
lower with respect to the prediction according to eq. (18). For describing the change in
axial force the ratio k is introduced which is defined as k=∆F /∆Fi. The ratio k between
force reduction with r=77% and r=100% has been evaluated analytically and
numerically (finite element). The result of this analysis is that a constant value
corresponding to the geometrical ratio (i.e. k= r=0.77) represents a valid
approximation for the present case. By using this result, eq. (18) can be rewritten as:

∆Fi ∆F ∆F
pel = = = = ε0 ⋅εr ⋅ E2 (19)
2 ⋅ x1 ⋅ y1 2 ⋅ x1 ⋅ y1 ⋅ k 2 ⋅ x1 ⋅ y1 ⋅ r

Rotation axis

Passive parts
pel
Active part

Fx(U = 0) Fx(U > 0)

U=0 U>0

Figure 12. Sketch to illustrate the influence of the passive parts on the spring roll
behavior.

εr is evaluated by using eq. (19). The electromechanical pressure pel=∆F/(2x1y1r) as a

function of the electric field E is presented in Figure 13. By fitting the experimental
curve of pel with eq. (1), a dielectric constant of 3.24 is obtained. This result agrees to a
great extent with the values reported in Table 1 for λp=4.

103
 75000 Upper border

Electromechanical pressure pel [Pa]

Experimental value
Fitting with eq. 1 (εr = 3.24)
50000 Lower border

25000

0.0 2.0x10
7 7
4.0x10
Electric field E [V/m]

Figure 13. Electromechanical pressure (with scatter band) evaluated by spring roll
experiments compared to a fit with eq. (1). The fitting parameter εr is 3.24.

4 Discussion

The analysis presented in Section 2 provides a validation of the Pelrine’s equation for
modeling electromechanical coupling in a circular actuator and a new interpretation of
the electromechanical forces distribution. Energy and force considerations are
consistent.
The energy balance considered here leads to the same result as in Pelrine et al.
although the derivation is different. In fact, we considered the dependence of Q and U
on the deformation. As a consequence the derivatives dWext/dz and dWel/dz depends
on the voltage derivative dU/dz.
Electrostatic charge and force distribution calculations showed that half of the
electromechanical pressure acts as out-of-plane compression and the other half as in-
plane tension. The out-of-plane force corresponds to the one calculated for a standard
parallel plate capacitor. Due to the incompressibility of the material, electromechanical
coupling can be modeled with an out-of-plane pressure with magnitude pel according
to eq. (1). This approach is advantageous in actuator modeling since only out-of plane
forces have to be considered.
Note, the validity of Pelrine’s equation for modeling dielectric elastomer actuators is
not an implicitness. Consider for instance a spheric configuration (‘ballon actuator’,
Figure 14) with the radius of the elastomer rs and the thickness of the elastomer zs,

104
zs<<rs. The elastomer is prestrained and is coated at the inner and the outer surface.
The inner pressure of a gas (e.g. helium) maintains the prestrain. The capacity is
C=ε0εrAs/zs where As is the surface area of the sphere. The charge is Q=C·U and the
uniform surface charge density at the electrode is ρs=Q/As=CU/As=ε0εrU/zs. Eq. (12)
for the Maxwell stress gives tm=0.5ε0εr(U/zs)2. This is the half of the pressure that one
would obtain by direct application of Pelrine’s equation.

Figure 14. Ballon actuator, taken from [20]

The results of the dielectric constant measurements show a remarkable dependence on

the pre-strain and a weak dependence on the frequency. In the undeformed state (pre-
strain λp = 1) the value of the dielectric constant is around 4.7 which is the commonly
used value [10, 11]. An increase of the pre-strain causes a significant decrease of the
dielectric constant. For a relevant actuator pre-strain range (around λp= 4) the
dielectric constant is close to 3.
The present results are in contrast to the findings in [10, 11] where it is shown
experimentally that the dielectric constant of VHB 4910 has a negligible dependence
on pre-strain and frequency. A reason for the different outcome of the present tests
might be in the material composition, which might have changed recently.
Interestingly, in a 3M data sheet [21] of November 2005 the dielectric constant of
VHB 4910 is given as 3.21 for 1 kHz and 2.68 at 1 MHz.
The measurements with the capacitor set-up are confirmed by the experiment with the
spring roll actuator. An evaluation of the dielectric constant for a biaxially pre-strained
actuator (λp=4) gives a value of 3.24, in agreement with the results of section 3.1
where the dielectric constant at a pre-strain of λp=4 was measured as 3.34 for 100 Hz
and 3.08 for 10 kHz.
The results of the spring roll experiments confirm the validity of Pelrine’s equation for
DE actuator modeling. This is consistent with the experimental validation of
electromechanical coupling by Kofod et al. [4].

105
The stretching of the elastomer might cause an anisotropy of the dielectric constant.
The fact that the dielectric constant decreases in thickness direction by stretching
might cause a change of the dielectric constant in planar direction. Since the molecules
align in planar direction by stretching, it is expected that the polarizability increases in
planar direction and decreases in thickness direction. As a consequence an increase of
the dielectric constant in planar direction is expected. With our set-up it is not possible
to measure the dielectric constant in planar direction. In order to evaluate the influence
of an anisotropic dielectric constant with respect to the electromechanical coupling, the
numerical analysis has been repeated. Four calculations with COMSOL Multiphysics
were carried out using the dielectric constant in a tensor form. Thereby, the dielectric
constant in thickness direction εr,z was set to εr,z = 4.7 in each calculation. The
dielectric constant in radial direction εr,r was varied with values of 4.7 (isotropic case),
6, 8 and 10. Results shows, that the surface charge distribution of the internal surface
area and therefore the mechanical pressure pz remains unchanged. In contrast, the
surface charge distribution at the lateral surface area changes, which has an influence
on the radial pressure pr. In Figure 15, the ratio pr/pz is plotted against the dielectric
constant in radial direction, εr,r.

1.0

0.8

0.6
Ratio pr/pz [-]

0.4

0.2

0.0
4 5 6 7 8 9 10
Dielectric constant in radial direction εr,r [-]

Figure 15. The dielectric constant in radial direction εr,r is plotted against the ratio
pr/pz

These results show that contributions of out-of-plane and radial pressure are equal only
for the case of isotropic dielectric behavior. With εr,r = 10 (more than double as εr,z =
4.7), the radial pressure pr is 10% less as compared to the isotropic case. It can

106
however be stated that the anisotropy of the dielectric constant has a small influence
on the overall electromechanical behavior.
The results of section 2 and 3 showed that the Pelrine’s equation can be used for
circular actuator modeling but the dielectric constant has to be changed with respect to
the value used in [8]. As a consequence, the results of the circular actuator experiments
have been revisited and the constitutive model parameter identification corrected.

The circular actuators had pre-stretch ratios of λp= 3, 4 and 5. As a simplification, εr is

not considered as pre-strain dependent for the present analysis, and a constant value of
εr = 3.24 has been selected, according to the value obtained for the spring roll actuator,
see Figure 12.
The new constitutive model parameters have been determined for the Arruda-Boyce
strain energy form. In particular, the initial shear modulus (parameter A) is corrected
with respect to [8], with A = 0.0473 MPa instead of 0.0686 MPa. Figure 16
summarizes the new results for simulations of circulator tests and uniaxial response.
The circular test fit is identical to the one reported in Figure 1, and a significant
improvement is obtained for the prediction of the uniaxial behavior.

0.2

Rel50
Nominal stress [MPa]

10
Rel200
Rel500
Strain [%]

0.1
Ten300
5 Ten500

prestrain λp = 3
prestrain λp = 4
0 prestrain λp = 5 0.0
0 200 400 600 800 1000 0 1000 2000 3000
Time [s] Time [s]

(a) (b)
Figure 16. Experiment vs. simulation for the same testing data points shown in Figure
1: (a) circular actuator tests; and (b) uniaxial experiments. The filled symbols represent
the experimental data and the open symbols the corresponding simulations, based on εr
= 3.24.

107
5 Conclusions

Electromechanical coupling in DE actuator has been investigated. Validation of

Pelrine’s equation for modeling electromechanical coupling in circular actuators has
been provided through (i) energy consideration and (ii) numerical calculation of
charge and force distribution. The latter provided a new physical interpretation of the
electrostatic forces acting on the DE film, with contributions from in-plane and out-of-
plane stresses. Representation of this force distribution using Pelrine’s equation is
valid for an incompressible material, such as VHB 4910.
The value of the dielectric constant has been measured for different pre-stretched VHB
4910 membranes. The values are shown to decrease with increasing pre-stretch ratio,
from 4.7 for the un-stretched film, down to 2.6 for a equi-biaxial deformation with
λp=5. This result is important in that it corrects the constant value of about 4.7
originally proposed in [10] and, since then, largely applied for pre-stretched DE
actuator modeling, see e.g. [3, 5, 8, 17, 18]. The present values were obtained from a
capacitor set-up and confirmed by the analysis of spring roll and circular actuator
experiments.
Correction of the dielectric constant leads to a change in the VHB 4910 constitutive
model parameters with respect to the values proposed in [8]. The new model is capable
of describing the active and passive material behavior over a large range of loading
conditions and deformation histories.

Acknowledgment
Financial support from the Swiss National Science Foundation (Project 200021-
107661/1) is gratefully acknowledged.

6 References

[1] R.E. Pelrine, R.D. Kornbluh, J.P. Joseph, Electrostriction of polymer dielectrics
with compliant electrodes as a means of actuation, Sens. Actuators, A 64 (1998)
77-85
[2] R. Pelrine, R. Kornbluh, Q. Pei, J. Joseph, High-speed electrically actuated
elastomers with strain greater than 100%, Science, 287 (2000) 836-839
[3] N. Goulbourne, E. Mockensturm, M. Frecker, A nonlinear model for dielectric
elastomer membranes, Journal of Applied Mechanics, 72 (2005) 899 – 906
[4] G. Kofod, P. Sommer-Larsen, Silicone dielectric elastomer actuators: Finite-
elasticity model of actuation, Sens. Actuators, A 122 (2005), 273-283

108
[5] J. Plante, S. Dubowsky, Large-Scale Failure Modes of Dielectric Elastomer
Actuators, International Journal of Solids and Structures, in press (available
online)
[6] M. Wissler, E. Mazza, Modeling and simulation of dielectric elastomer actuators,
Smart Materials and Structures, 14 (2005) 1396 - 1402
[7] M. Wissler, E. Mazza, Modeling of a prestrained circular actuator made of
dielectric elastomers, Sens. Actuators, A 120 (2005), 184-192
[8] M. Wissler, E. Mazza, Mechanical behavior of an acrylic elastomer used in
dielectric elastomer actuators, Sens. Actuators, in press (available online)
[9] E.M. Arruda, M.C. Boyce, A three-dimensional constitutive model for the large
stretch behavior of rubber elastic materials, J. Mech. Phys. Solids, 41, No.2
(1993) 389 – 412
[10] G. Kofod, R. Kornbluh, R. Pelrine, P. Sommer-Larsen, Actuation response of
polyacrylate dielectric elastomers, Proc. SPIE 4329 (2001) 141 – 147
[11] G. Kofod, P. Sommer-Larsen, R. Kornbluh, R. Pelrine, Actuation response of
polyacrylate dielectric elastomers, J. Intel. Mat. Syst. Str., 14 (2003) 787-793
[12] O.H. Yeoh, Characterization of elastic properties of carbon-black-filled rubber
vulcanizates, Rubber Chem. Technol., 63 (1990) 792-805
[13] R.W. Odgen, Large deformation isotropic elasticity – on the correlation of theory
and experiment for incompressible rubberlike solids, Proc. R. Soc. Lond., A 326
(1972) 565-584
[14] R. McMeeking, C. Landis, Electrostatic forces and stored energy for deformable
dielectric materials, Journal of Applied Mechanics, 72 (2005) 581 – 590
[15] COMSOL Multiphysics, User’s Guide, 2005, version 3.2, COMSOL, Inc.
[16] J. C. Maxwell, A treatise on electricity and magnetism, Dover, Oxford, 1954
[17] P. Lochmatter, S. Michel, G. Kovacs, Electromechanical model for static and
dynamic activation of elementary dielectric elastomer actuators, Proc. SPIE 6168
(2006) 61680F-1 - 61680F-13
[18] E. Yang, M. Frecker, E. Mockensturm, Viscoelastic model of dielectric elastomer
membranes, Proc. SPIE 5759 (2005) 82 – 92
[19] G. Kovacs, P. Lochmatter, Arm wrestling robot driven by dielectric elastomer
actuators, Proc. SPIE, vol. 6168 (2006) pp. 616807-1 - 616807-12
[20] P. Lochmatter, Develompent of a shell-like electroactive polymer (EAP) actuator,
ETH dissertation, in preparation
[21] [Link]

109
110
4 Further Verification of the Actuator Model

4.1 Introduction

The results of section 3 are verified further and completed in this chapter. Four main
points are considered:
(i) In article 3 the agreement between experiments and simulations is good for the
circular strain test data, but the model fails to correctly describe the uniaxial
tensile tests. The reason for this is the dielectric constant which is about 3.2 and
not 4.7, see article 4. Identical to the Arruda-Boyce strain energy form in article
4, the material parameters of the Yeoh and the Ogden strain energy form are
adapted considering the correct value of the dielectric constant of VHB 4910.
This results in a modification of the predicted uniaxial behavior which is
presented in section 4.2.
(ii) In order to evaluate the predictive capability of the constitutive model, aspiration
and compression tests are performed. The aspiration test method is based on the
pipette aspiration technique [Aoki 1997] and is commonly used for testing the
biaxial behavior of biological soft tissues. The tests have been performed for
prestrained VHB-films with a radial pre-stretch ratio of λp = 3, 4 and 5.
Furthermore, compression tests are performed on un-stretched elastomer films.
Finite element models are created for both tests, the aspiration test in section 4.3
and the compression test in section 4.4. The simulations are compared to the
experiments in order to verify the mechanical model. The material parameters
listed in section 4.2 are used here.
(iii) Models in continuum mechanics are phenomenological. The model validity is
usually limited to a certain deformation range. An effective way of illustrating the
deformation range of a certain test for an incompressible material is to plot the
first invariant against the second invariant of the left Cauchy-Green deformation
tensor [Bird 1977 and Treloar 1975]. This results in a ‘map’ of the deformation
range which is presented in section 4.5. Each experiment represents a curve on
the map.
(iv) In article 4 spring rolls are used for the determination of the dielectric constant.
The evaluation of the dielectric constant for VHB 4910 yielded a value of 3.2
which is in agreement with LCR-meter experiments (article 4). One important
assumption was that the ratio k (section 3.2, article 4) is equal to the ratio r

111
 (section 3.2, article 4). This relationship is important for the design of spring rolls
and is analyzed in section 4.6.

4.2 Correction of material parameters

In order to consider the fact that the dielectric constant is about 3.2 and not 4.7 (article
4), the constitutive material parameters for the Yeoh and the Ogden strain energy
forms presented in article 3 are adapted as for the Arruda-Boyce strain energy form
(article 4).
The optimization procedure presented in article 3 has been carried out again for the
Yeoh and the Ogden strain energy form with a dielectric constant of 3.2 instead of 4.7.
The actuator behavior did not vary by using the new material parameters. By contrast,
the simulations of the uniaxial tensile and relaxation tests change. The new material
parameters for the Ogden, the Yeoh and the Arruda Boyce strain energy forms are
presented in Table 4.1 (hyperelastic parameters). The appropriate viscoelastic
parameters are listed in article 3, Table 2.

Table 4.1. New hyperelastic material parameters

C10 C20 C30
Yeoh [MPa] [MPa] [MPa]
0.05374 -4.86e-04 3.809e-06
µ1 [MPa] α1 [-] µ2 [MPa] α2 [-] µ3 [MPa] α3 [-]
Ogden
0.00592 1.293 0.0582 2.325 -0.0161 2.561
Arruda- A [MPa] N [-]
Boyce 0.0473 124.88

The comparison between experiments and simulations for the uniaxial tensile and
relaxation tests is illustrated in Figure 4.1 for all three strain energy forms.

112
 0.2 0.2
Rel50
Nominal stress [MPa]

Nominal stress [MPa]

Rel200 Rel50
Rel500 Rel200
Ten300 Rel500
0.1 Ten500 0.1 Ten300
Ten500

0.0
0.0
0 1000 2000 3000 0 1000 2000 3000
Time [s] Time [s]

(i) (ii)
0.2
Nominal stress [MPa]

Rel50
Rel200
Rel500
Ten300
0.1
Ten500

0.0
0 1000 2000 3000
Time [s]

(iii)
Figure 4.1. Uniaxial behavior for the adapted material parameters: experimental data
(filled symbols) compared to the simulations (open symbols) for (i) the Yeoh, (ii) the
Ogden and (iii) the Arruda-Boyce strain energy forms

The results of the uniaxial behavior show that this modification leads to an
improvement in the predictive capabilities of the constitutive model. Both the actuator
and the uniaxial behavior are described reasonably by all three strain energy forms.
In article 1, the Yeoh model provided good predictive capability. The dielectric
constant was 4.7 and the material parameters were different from Table 4.1. These
facts are in disagreement with the results of article 3, article 4 and this section. The
reason for this are that (i) the gold electrodes (article 1) cause a mechanical resistance
in the activation and (ii) the material parameter determination was verified solely on
one single prestrain and voltage level.
(i) The actuation strain (article 1, Figure 8) at 180 s is 14% at prestrain 4.08. By
contrast, actuators with silicone/graphite electrodes at prestrain 4 (article 3,
Figure 5) show a strain value of about 50% after 180 s. The silicone/graphite

113
 electrode is more compliant than the gold electrode. Therefore, graphite
powder/silicone electrodes were used in article 3. For investigating whether the
silicone/graphite electrode constrains the active strain, experiments have been
performed with two different silicone/graphite mixtures but identical conditions
[Wyrsch 2006]. Mixture 1 contained 11 g graphite powder (TIMREX SP30) and
10 ml silicone oil (DC 200/100 cs) and mixture 2 contained 13.5 g graphite
powder (TIMREX SP30) and 10 ml silicone oil (DC 200/100 cs). There was
almost no difference in the strain response between the two actuator types. Thus,
the used graphite/silicone electrode does not constrain the active strain.
(ii) A total number of 11 parameters (three for the Yeoh strain energy form and eight
for the Prony series) were determined in article 1 on the basis of uniaxial tests.
The modeling system had too many free parameters, which led to imprecise
results in the actuator behavior. By contrast, the parameter determination in
article 3 was more restricted. Three multiaxial tests (at various prestrain levels)
were involved in the fitting procedure. The detailed considerations of article 3,
article 4 and this chapter resulted in an improved model and a better
understanding.

4.3 Aspiration tests

4.3.1 Introduction

The aspiration device [Vuskovic 2001] is based on the pipette aspiration technique
[Aoki 1997] and is commonly used for testing the biaxial behavior of biological soft
tissues. In order to characterize the biaxial passive mechanical properties, the
aspiration test was used for the VHB 4910 membrane under different prestrain levels.
A finite element model was created and the constitutive model (with the material
parameters presented in section 4.2) was applied. The comparison between
experiments and simulations yields a further verification of the mechanical model.

4.3.2 Working principle of the aspiration device

The aspiration test device was originally developed by Vuskovic [Vuskovic 2001] and
further improved by Nava [Nava 2004 and Nava 2007]. The working principle of the
aspiration device [Nava 2007] is illustrated in Figures 4.2 and 4.3. The device consists
of a tube (outer radius: 13 mm, inner radius: 5 mm) in which the internal pressure can
be controlled according to a desired pressure law.

114
The experiment is performed by (i) gently pushing a tube against the tissue to ensure a
good initial contact and (ii) creating a vacuum (variable over time) inside the tube so
that the tissue is sucked through the aspiration area, Figure 4.2.
For an isotropic and homogenous film, a complete description of the deformation field
can be obtained by monitoring the side-view profile of the tissue during the vacuum
change. An optical fibre, which is connected to an external source of light, provides
the necessary illumination in the inner part of the tube.
The images of the side-view (see Figure 4.2, b) are reflected by a mirror and are
captured at a frequency of 25 Hz by a digital camera mounted on the upper part of the
device. The grabbed images are analyzed off-line in order to extract the profiles of the
deformed tissue. The present image acquisition and analysis technique allows the
tissue's vertical displacement, d, (Figure 4.2, b) to be measured with an accuracy of
0.05 mm. A standard personal computer (running NI LabView Version 6.1) controls
the pressure inside the device by means of a pump, an air reservoir and two valves.
Time histories of measured pressure and deformation profiles represent the input data
used to evaluate the mechanical properties and to determine the constitutive model.
Further information is given in [Nava 2007].

(a) (b)

Figure 4.2. (a) Sketch of the aspiration device and (b) image of the side view. The
light half-bubble is the part of the tissue that is aspirated inside the device. The point P
and the distance d are indicated for reference. The pictures are from [Nava 2007].

4.3.3 Description of the experiments

The samples (VHB 4910) for the aspiration tests were bi-axially pre-strained (see also
article 3). Radial pre-straining was realized by the pre-straining machine [Kovacs
2006] developed at EMPA's DE laboratory for biaxial stretching of elastomeric films.
Three types of samples were realized with prestretch ratios λp = 3, 4 and 5. After radial
stretching, the film was fixed uniformly along a circular frame.

115
 (a) (b)

Figure 4.3. (a) Aspiration test device with a prestrained VHB 4910 membrane and (b)
close-up of the device.

Two different aspiration cycles with a pressure of 40 or 55 mbar were applied for 100
s, Figure 4.4. Each experiment consisted of one single aspiration cycle. The
displacement d (Figure 4.2 b) was measured.

60
Pressure [mbar]

40

20 p = 40 mbar
p = 55 mbar

0 50 100
Time [s]

Figure 4.4. Imposed pressure history for the aspiration test

A total number of twelve samples was used for the experimental work. Two
experiments were performed for each prestrain and pressure history.

116
4.3.4 Experimental results

Measurement results are presented in Figure 4.5. The displacement, d, is plotted

against time. Each curve corresponds to the average of two experiments, whereby the
scatter between experiments was low (the relative variability was less than 7%).

3
Displacement [mm]

2
Prestrain 3, 40 mbar
Prestrain 3, 55 mbar
1
Prestrain 4, 40 mbar
Prestrain 4, 55 mbar
Prestrain 5, 40 mbar
0
Prestrain 5, 55 mbar

0 50 100
Time [s]

Figure 4.5. Experimental results for the aspiration test device

4.3.5 Finite element calculations

The software ABAQUS 6.5.1 [ABAQUS 2004] has been used for the present
calculations of the aspiration test. The axially symmetrical model (Figure 4.6) consists
of 4-node elements (CAX4H: 4-node bilinear axisymmetric quadrilateral, hybrid,
constant pressure). The homogeneous pressure, pas, corresponds to the imposed
pressure history in Figure 4.4 and acts on the aspiration area as a ‘surface pressure’.
The simulation of one experiment consists of three steps:
(i) pre-strain in the radial direction, by imposing the displacement of the nodes at the
model boundary (Figure 4.6)
(ii) hold time of 1 h. In the time span of 1h, the material is supposed to relax
completely (see also relaxation tests, article 1). This corresponds to the
experimental conditions of the aspiration test (at least 1 h elapsed between
prestretching and aspiration of the elastomer)
(iii) aspiration by imposing the pressure history (Figure 4.4) to the aspiration area.
Since VHB 4910 is very sticky, the area with contact to the tube is assumed to be

117
 fixed. Therefore, the nodes of this area were constrained in radial and vertical
direction.

Nodes constrained in radial and vertical direction

pas Nodes constrained in radial direction

d

Rotation axis

Figure 4.6. Finite element model of the aspiration test

The quasilinear visco-hyperelastic model with the material parameters presented in

Table 4.1 and Table 2 in article 3 was applied for the simulations.

4.3.6 Results

Figures 4.7, 4.8 and 4.9 show the comparison between experiments and simulations for
the Yeoh, the Ogden and the Arruda-Boyce models respectively.

4 4
Displacement [mm]

Displacement [mm]

3 3

2 2

p = 40 mbar, exp p = 40 mbar, exp

1
p = 55 mbar, exp 1 p = 55 mbar, exp
p = 40 mbar, FE p = 40 mbar, FE
0 p = 55 mbar, FE 0 p = 55 mbar, FE

0 50 100 0 50 100
Time [s] Time [s]

(i) (ii)

118
 4
Displacement [mm]
3

p = 40 mbar, exp
1 p = 55 mbar, exp
p = 40 mbar, FE
0 p = 55 mbar, FE

0 50 100
Time [s]

(iii)
Figure 4.7. Comparison between experimental data and simulation for the Yeoh
model for pre-strain λp = 3 (i), λp = 4 (ii) and λp = 5 (iii). The abbreviation ‘exp’ is
used for experimental data and ‘FE’ is used for finite element simulation.

4 4
Displacement [mm]

3 3
Displacement [mm]

2 2

p = 40 mbar, exp p = 40 mbar, exp

1 p = 55 mbar, exp 1 p = 55 mbar, exp
p = 40 mbar, FE p = 40 mbar, FE
0 p = 55 mbar, FE 0 p = 55 mbar, FE

0 50 100 0 50 100
Time [s] Time [s]

(i) (ii)
4
Displacement [mm]

p = 40 mbar, exp
1 p = 55 mbar, exp
p = 40 mbar, FE
0 p = 55 mbar, FE

0 50 100
Time [s]

(iii)
Figure 4.8. Comparison between experimental data and simulation for the Ogden
model for pre-strain λp = 3 (i), λp = 4 (ii) and λp = 5 (iii). The abbreviation ‘exp’ is
used for experimental data and ‘FE’ is used for finite element simulation.

119
 4 4
Displacement [mm]

Displacement [mm]
3 3

2 2

p = 40 mbar, exp p = 40 mbar, exp

1 p = 55 mbar, exp 1 p = 55 mbar, exp
p = 40 mbar, FE p = 40 mbar, FE
0 p = 55 mbar, FE 0 p = 55 mbar, FE

0 50 100 0 50 100
Time [s] Time [s]

(i) (ii)
4
Displacement [mm]

p = 40 mbar, exp
1 p = 55 mbar, exp
p = 40 mbar, FE
0 p = 55 mbar, FE

0 50 100
Time [s]

(iii)
Figure 4.9. Comparison between experimental data and simulation for the Arruda-
Boyce model for pre-strain λp = 3 (i), λp = 4 (ii) and λp = 5 (iii). The abbreviation ‘exp’
is used for experimental data and ‘FE’ is used for finite element simulation.

Experiments and simulations agree to a great extent, especially for the Yeoh strain
energy form. The aspiration tests characterize the biaxial behavior of the elastomer
whereas the circular strain tests also include the influence of the electromechanical
coupling.

4.4 Compression tests

4.4.1 Introduction

A compression test device is used for further characterization of the elastomer

behavior. A sample with thickness z is squeezed by two circular plungers with radius r.

120
The compression test also exerts a pressure on the elastomer as does the circular strain
test. In contrast to the circular strain test where no information about forces is
available, the average pressure is measured in the compression test.
These tests have various boundary conditions. The circular strain test corresponds to a
mainly biaxial deformation state. In the compression test, the VHB 4910 film adheres
to the plunger’s surface since the film is very sticky. This effect and the fact that the
film is nearly incompressible results in a ‘mixed mode’ deformation state (see section
4.5). The characterization of the compression behavior provides further verification of
the constitutive model.

4.4.2 Experimental details

Compression tests were performed with a Zwick (Z010) machine under displacement
control conditions at room temperature (23 °C), Figure 4.10. The radius, r, of the
plungers was 1.78 cm. In order to obtain a sample thickness which is of the same
magnitude as the radius r, five layers of un-stretched VHB 4910 were glued one on top
the other. The sample with an initial thickness of z0=5 mm was then fixed to a circular
frame with an inner radius of R = 75 mm.

Figure 4.10. Sketch of the two plungers of the compression test

The two concentric plungers with the radius r = 1.78 cm of the test device compressed
the sample in the middle of the circular frame. The distance z between the plungers
was provided by the Zwick (Z010) machine. The control profile of the displacement
d=z0-z is illustrated in Figure 4.11. Two identical samples were tested under identical
conditions. The force histories of both experiments are close, see Figure 4.12.

121
 0.0

-0.2

Displacement [mm]
-0.4

-0.6

-0.8

-1.0

-1.2

0 100 200 300 400

Time [s]

Figure 4.11. Control profile for the compression tests. The displacement is plotted
against time

4
Force [N]

0 100 200 300 400

Time [s]

Figure 4.12. Force vs. time for the two compression tests

4.4.3 Finite element simulations

The compression test simulations were performed with ABAQUS 6.5.1 [ABAQUS
2004]. The axially symmetrical model consists of 4-node elements (CAX4H: 4-node
bilinear axisymmetric quadrilateral, hybrid, constant pressure), Figure 4.13. Due to the
good adherence of the material
(i) the five layers were considered to be one layer with a thickness of z = 5 mm and

122
(ii) the two plungers were not directly included in the model. Instead, the nodes of
Border a and b are fixed in radial direction (Figure 4.13).
As further boundary condition the nodes of Border b were fixed in the vertical
direction and the vertical displacement, u2, of the nodes of Border a corresponds to the
displacement history in Figure 4.11.

Rotation axis

u2 Border a

2
r Border b
1

Figure 4.13. Finite element model of a circular dielectric elastomer actuator

The quasilinear visco-hyperelastic model with the parameters of Table 4.1 and Table 2
in article 3 were used. The output of the simulation was the distribution of Cauchy-
stresses s22 in the vertical direction at Border a as a function of the radius, r. In order to
obtain the total force in the vertical direction, the Cauchy stresses s22 along Border a
were multiplied by the corresponding area and summed up.

4.4.4 Results

Figure 4.14 shows the finite element model of the compression test and Figure 4.15
illustrates the comparison of the force history from experiment and simulation.

Figure 4.14. Finite element model of the compression test. The stress component s22
is given in MPa.

123
 6

Experiment
Yeoh
4 Ogden
Arruda-Boyce
Force [N]

0 100 200 300 400

Time [s]

Figure 4.15. Comparison between the experimental and numerical force history

The Yeoh strain energy form yields good agreement between simulation and
experiments. By contrast, the force values of the Ogden- and the Arruda-Boyce-strain
energy forms are about 40% too small. This finding corresponds with the results
shown in Figure 4.1: for small strains (up to 20%) the uniaxial behavior is predicted
well with the Yeoh model (Figure 4.1 (i)) whereas the Ogden and the Arruda-Boyce
strain energy forms (Figure 4.1 (ii) and (iii)) underpredict the force values.

4.5 Deformation state in the experiments

4.5.1 Introduction

The circular actuators made of VHB 4910 have a large deformation range. The
deformation corresponds to a biaxial deformation state with stretch ratios (including
the prestrain) of up to 6. In the uniaxial tensile tests, the stretch ratios are also as large
as 6. One possibility of analyzing the corresponding deformation states and ranges
[Bird 1977 and Treloar 1975] is to plot the first Invariant I1 as a function of the second
invariant I2 of the left Cauchy-Green deformation tensor. For an incompressible
material, the first and the second invariants are

I1 = λ1 + λ2 + λ3
2 2 2
(4.1)
−2 −2 −2
I 2 = λ1 + λ2 + λ3 (4.2)

124
where the stretch ratios λ1, λ2 and λ3 are the principal values of the deformation
gradient and incompressibility yields λ1λ2λ3=1. Three important deformation states are
uniaxial extension, equi-biaxial stretching and shear [Treloar 1975]. Uniaxial
extension is described by the stretch ratios as

1
λ 2 = λ3 = (4.3)
λ1

where λ1>1. The equi-biaxial behavior is given by

1 1
λ3 = = (4.4)
λ1 2
λ2 2

where λ1 and λ2 represent the in-plane stretch ratios and λ3 the out-of-plane stretch
ratio, λ1 ≥ 1. The shear behavior [Treloar 1975] is given by

λ3 = 1 / λ1 , λ2 = 1 (4.5)

where λ1 ≥ 1. A map of deformations is generated by plotting I1 vs. I2 for all three

deformation states [Bird 1977], Figure 4.16. All combinations of I1 and I2 within the
shaded zone are accessible [Bird 1977]. Each experiment represents a curve in this
plot. The uniaxial and the equi-biaxial behavior stand for the boundaries of the
allowable deformation range [Bird 1977].

Figure 4.16. Map of deformation range for the classical tests for hyperelastic
materials: uniaxial, equi-biaxial and shear behavior

The quality of a three-dimensional model can be evaluated based on its ability to

predict the mechanical response for different deformation states. To this end,

125
experimental data from different test configurations are necessary. In the following
sections, the deformation states for the experimental procedures applied in the present
work are discussed.

4.5.2 Uniaxial tensile tests

The stretch ratio ranges from 1 to 6 for the uniaxial tensile and relaxation tests. The
first invariant I1 vs. the second invariant I2 for the uniaxial behavior (λ1 = 1 to 6) is
illustrated in Figure 4.17.

40
Biaxial behavior

30
Shear behavior
Invariant I2 [-]

20

10
Uniaxial behavior

0
0 10 20 30 40
Invariant I1 [-]

Figure 4.17. Map of deformation range. The curve for the tensile test is indicated in
bold.

4.5.3 Circular strain test

The deformation of the active part of the circular actuators is mainly biaxial. The
validity of the mechanical model is estimated from Figure 12 in article 3 for the Yeoh
strain energy form. The smallest deformation is λ1= 3 corresponding to a pre-stretch
ratio λp=3 without activation. The largest stretch ratio where the model still provides
reasonable predictions is λ1= 5.75. This corresponds to 15% nominal strain for the pre-
stretch ratio 5. In Figure 4.18, I1 is plotted logarithmically vs. I2 for a biaxial behavior
from λ1= 3 to 5.75.

126
 Biaxial behavior
1000

100
Invariant I2 [-] Shear behavior

Uniaxial behavior
10

1
1 10 100 1000
Invariant I1 [-]

Figure 4.18. Map of deformation range (logarithmic scale). The curve for the circular
strain test is indicated in bold.

4.5.4 Aspiration test

The aspiration test corresponds to a great extent to a biaxial configuration. The upper
boundary for the stretch ratio λ1 was taken from the ABQAUS output file at point P
(Figure 4.2) for the Yeoh strain energy form. The largest value of λ1 is 6.09, achieved
at prestrain 5. The lowest value is λ1= 3 corresponding to a pre-stretch ratio λp=3,
identical to the circular strain test. The deformation range of the aspiration test lies
between λ1= 3 to 6.09, see Figure 4.19.
Biaxial behavior
1000

100
Invariant I2 [-]

Shear behavior

Uniaxial behavior
10

1
1 10 100 1000
Invariant I1 [-]

Figure 4.19. Map of deformation range (logarithmic scale). The curve for the
aspiration test is indicated in bold.

127
4.5.5 Compression test

For the compression test, the deformation behavior is strongly dependent on the
location of the material points. Therefore, four material points (volume elements with
the numbers 1690, 1699, 3601 and 3618) are considered in order to analyze the
deformation behavior, see Figure 4.20. In Figure 4.21 I1 is plotted against I2 for the
four elements. The dependencies between the invariants I1 and I2 were taken from the
output of the ABAQUS simulation.

3618
3601

1690 1699

Figure 4.20. Finite element model of the compression test. The elements with the
numbers 1690, 1699, 3601 and 3618 were used for the analysis of the deformation
behavior.

4.0 3.10

Element 1690 Element 1699

3.5 3.05
Invariant I2 [-]

Invariant I2 [-]

3.0 3.00

2.5 2.95
2.5 3.0 3.5 4.0 2.95 3.00 3.05 3.10
Invariant I1 [-] Invariant I1 [-]

(a) (b)

128
 4.5
3.01 Element 3601 Element 3618
4.0

Invariant I2 [-]
Invariant I2 [-]

3.5

3.00
3.0

2.5
3.00 3.01 2.5 3.0 3.5 4.0 4.5
Invariant I1 [-] Invariant I1 [-]

(c) (d)
Figure 4.21. Map of deformation range. The curves for the compression test are
indicated in bold for (a) element 1690, (b) element 1699, (c) element 3601 and (d)
element 3618.

Elements 1690 and 3601 exhibit a pure equi-biaxial deformation. The deformation of
element 3601 is very small compared to element 1690. Element 3618 undergoes a
deformation which is similar to the simple shear test. An interesting deformation
behavior is shown by element 1699. At small strains the deformation is similar to the
uniaxial tensile test and at higher strains the deformation corresponds rather to simple
shear.

4.5.6 Discussion

All the findings of section 4.5.2 to 4.5.5 for the uniaxial tests, the circular strain tests,
the aspiration tests and the compression tests are presented in one map, Figure 4.22.
The deformation range for which the simulations are in agreement with the
experiments is wide. From this point of view, the mechanical model (in particular the
Yeoh strain energy form) has excellent predictive capabilities. It is clear from the
corresponding curves that the circular strain and the aspiration test exhibit a similar
deformation behavior. A comparison between the circular strain and aspiration tests
therefore provides information about the validity of the model for electromechanical
coupling. The deformation range with reliable predictions is larger for the aspiration
test than for the circular strain test.
The deformation of the circular strain test and the aspiration experiment are mainly
equi-biaxial and thus comparable. However, simulation results are in better agreement
with experimental data for the aspiration test than for the circular strain test. The
agreement between experimental data and simulation is good for the circular strain test
up to 20-25% nominal strain and for the aspiration test up to 55% at λp=3, 40% at λp=4

129
and 22% at λp=5. This might lead to the conclusion that the disagreement between
simulation and experimental data in the circular strain test is due to Pelrine’s equation
failing to describe the electromechanical coupling at voltages which cause strains
greater than 20-25%. The electromechanical coupling at high voltages should be
further investigated in more detail.
The compression test covers a much smaller deformation range compared to the other
experiments but it has the advantage that different deformation states occur in one
single test.

Biaxial behavior
1000

100
Invariant I2 [-]

Shear behavior

Uniaxial behavior
10

1
1 10 100 1000
Invariant I1 [-]

Figure 4.22. Map of deformation range with all tests (bold lines): uniaxial tensile tests,
circular strain tests, aspiration tests and compression test.

4.6 Influence of the passive part of a spring roll activation

4.6.1 Introduction

In article 4 the value of the dielectric constant at a prestrain of 4 was determined as 3.2
by spring roll experiments and confirmed by LCR-measurements. As explained in the
article, the spring rolls had an active and a passive part.
By the evaluation of the dielectric constant with the spring rolls, it is assumed that the
geometrical ratio, r, corresponds to the ratio of force reduction, k. The ratio r is the
coated area divided by the whole area of the spring roll (r=xa/x1, Figure 11, article 4);
k is the axial force of a spring roll with ratio r divided by the axial force of a

130
geometrical identical spring roll without a passive part (r=1). As explained in the
article, r=k is an approximation.
In this section, the relationship between r and k is investigated. The results provide a
verification of the dielectric constant determination by spring roll experiments.
Furthermore, the relationship between r and k is important for the design of spring
rolls which are used as artificial muscles [Kovacs 2006] and for walking robots [Pei
2003]. Analytical and numerical calculations are performed for the evaluation of the
relationship between r and k. Viscoelastic effects are neglected. The analytical model
is more useful than the numerical one because it also contains the electromechanical
coupling. But with the analytical model it is not possible to consider the influence of
the intersection between the active and passive zones. Therefore the numerical model
serves as a verification of the analytical model.

4.6.2 Analytical model

Three planar configurations are considered for the analytical model, Figure 4.23. The
notation is slightly different from that in article 4. The configurations are:
(i) the undeformed configuration with the dimensions X, Xa, Xb, Y and Z,
(ii) the prestrained configuration (denoted with the index ‘I’) with the dimensions xaI,
xbI, xI, yI and zI and
(iii) the activated configuration (denoted with the index ‘II’) with the dimensions xaII,
xbII, xII, yII and zII.
The index ‘a’ refers to the active part and the index ‘b’ to the passive part.

131
 Z
Xb
(a) X
Xa
Y
zI

xbI

(b) xI pel
xaI

yI
zbII
xbII

(c) xII FxII

xaII
U>0
yII zaII

In-plane Thickness

Figure 4.23. Three configurations of deformation for the spring roll: (a) undeformed,
(b) prestrained and (c) activated configuration

A Cartesian and not a polar coordinate system is introduced for simplicity. For the
calculation results it is assumed that this approach is valid as long as the thickness of
one layer is much smaller than the radius of the spring roll.
The original thickness of the film, Z, is assumed to be 1 mm (this corresponds to the
film thickness of VHB 4910). The geometrical ratio r is

X a xaI
r= = I (4.6)
X x

The nominal actuation strain, sz, for the deformation of the actuator is introduced for
later reference

z aII
sz = I −1 (4.7)
z

[Link] Kinematics
For the pre-strained configuration the stretch ratios are

xI
λIx = = λx (4.8)
X

132
 yI
λIy = = λy (4.9)
Y

zI 1
λ =
I
= (4.10)
Z λx ⋅ λ y
z

where λx and λy correspond to the pre-stretch ratios of the spring roll. zI is given by

Z
zI = (4.11)
λx ⋅ λ y

The stretch-ratios in the y-direction in the active part λyaII and the passive part λybII do
not change during activation.

λIIya = λIIyb = λ y (4.12)

By transforming eq. (4.7), zaII becomes

z aII = ( s z + 1) ⋅ z I (4.13)

and thus the stretch ratio of the active area in the z-direction is

z aII s +1
λIIza = = z (4.14)
Z λx ⋅ λ y

Due to incompressibility (λxaII λyaII λzaII)=1, the stretch-ratio in the x-direction of the
active part is

xaII λ
λIIxa = = x (4.15)
X a sz +1

With eq. (4.15) the x-coordinate of the active part xaII is

λx
xaII = ⋅ Xa (4.16)
sz + 1

and xbII is the difference between xI (eq. (4.8)) and xaII (eq. (4.16))

xbII = x I − xaII (4.17)

A combination of eqs. (4.6), (4.16) and (4.17) yields the following expression for the
stretch ratio λxbII

xbII sz + 1 − r
λIIxb = = λx ⋅ (4.18)
Xb (1 − r ) ⋅ ( s z + 1)

133
where the relationship Xb=X-Xa has been used. Incompressibility yields for the stretch
ratio λzbII in the z-direction

1 1 ( s + 1) ⋅ (1 − r )
λIIzb = = ⋅ z (4.19)
λ ⋅ λ yb λ x ⋅ λ y
II
xb
II
sz + 1 − r

[Link] Kinetics
For an incompressible material, the principal Cauchy stresses, ti, are determined by the
derivative of the strain energy potential, W, with respect to the stretch ratio, λi, see
article 1.

∂W
t i = λi ⋅ − p = G ( λi ) − p (4.20)
∂λi

where i refers to a principal direction (x, y or z), p is the hydrostatic pressure and G is
defined here as:

∂W
G ( λi ) = λ i ⋅ (4.21)
∂λi

For the pre-strained configuration the stress in the z-direction has to be zero

t zI = G (λIz ) − p I = 0 (4.22)

The stress in the x-direction for the prestrained configuration is obtained by using eqs.
(4.20) and (4.22).

t xI = G (λ Ix ) − G (λIz ) (4.23)

The axial force FxI in the x-direction for the prestrained configuration is

FxI = t xI ⋅ y I ⋅ z I (4.24)

In the activated configuration, the stress tzaII in the z-direction for part a is

t zaII = G (λIIza ) − p aII = − pel (4.25)

where pel is the electrostatic pressure. Solving for paII and using eq. (4.20) for the x-
direction leads to

t xaII = G (λIIxa ) − G (λIIza ) − pel (4.26)

The axial force of the active and the passive part have to be equal.
FxII = t xaII ⋅ y II ⋅ z aII = t xbII ⋅ y II ⋅ z bII (4.27)

134
By evaluating the stress txbII in the y-direction for the passive part and using eqs. (4.26)
and (4.27), txbII is

sz + 1 − r
t xbII = ⋅ [G (λIIxa ) − G (λIIza ) − pel ] (4.28)
1− r

Using eq. (4.20), txbII can also be expressed as

t xbII = G (λIIxb ) − pbII (4.29)

The stress tzbII of the passive part in the z-direction is zero

t zbII = G (λIIzb ) − pbII = 0 (4.30)

Combining eqs. (4.29) and (4.30) gives for txbII

t xbII = G (λIIxb ) − G (λIIzb ) (4.31)

[Link] Electromechanical coupling

By equating eqs. (4.28) and (4.31), the following relationship is obtained

sz + 1 − r
t xbII = ⋅ [G (λIIxa ) − G (λIIza ) − pel ] = G (λIIxb ) − G (λIIzb ) (4.32)
1− r

The electromechanical pressure, pel, is given by Pelrine’s equation [Pelrine 1998]

2
⎛U ⎞
pel = ε r ⋅ ε 0 ⋅ ⎜⎜ II ⎟⎟ (4.33)
⎝ za ⎠

Solving for U and using eqs. (4.11), (4.13), (4.32) and (4.33) gives
1/ 2
Z ⋅ ( s z + 1) ⎛ 1− r ⎞
U= ⋅ ⎜⎜ G (λIIxa ) − G (λIIza ) − ⋅ (G (λIIxb ) − G (λIIzb )) ⎟⎟ (4.34)
λx ⋅ λ y ⋅ ε r ⋅ ε 0 ⎝ sz + 1 − r ⎠

This relationship represents a connection between the voltage, the geometrical

dimensions, the prestrain of the spring roll, the strain sz, the hyperelastic material
parameters and the dielectric constant.
The parameter k is determined by evaluating the force differences ∆F and ∆Feff. The
force difference, ∆Feff, corresponds to the difference of eqs. (4.24) and (4.27)

∆Feff = FxII − FxI (4.35)

The force difference ∆F of a spring roll without a passive part is

135
 2
⎛U ⎞
∆F = pel ⋅ y ⋅ z = ε r ⋅ ε 0 ⋅ ⎜ I ⎟ ⋅ y I ⋅ z I
I I
(4.36)
⎝z ⎠

and with eq. (4.11) yields:

Y
∆F = ε r ⋅ ε 0 ⋅ U 2 ⋅ ⋅ λx ⋅ λ y
2
(4.37)
Z

The ratio k is given by k=∆Feff/∆F. The whole system is described by the derived
equations. The equations are implemented into Matlab [Matlab 2004]. For the
comparison between analytical and numerical calculations, two different prestrain
levels, 1x1 (λx=λy=1, no prestrain) and 4x4 (λx=λy=4), are considered. The selected
parameters for the analytical model are listed in Table 4.2. The initial thickness at
prestrain 4x4 is the value of undeformed VHB 4910, 1mm. For prestrain 1x1 a value
of 50µm was chosen in order to ensure that the thickness is much smaller than the
radius of the elastic core of the spring roll which is an assumption of the analytical
model. The dielectric constant, εr, was 3.2 for both prestrains and the ratio r was 0.77.
The Yeoh strain energy form was used for W with the material parameters listed in
Table 4.1.

Table 4.2. Parameters for the analytical model

Prestrain 1x1 Prestrain 4x4
Dielectric constant εr [-] 3.2 3.2
Ratio r [-] 0.77 0.77
Initial thickness Z [mm] 0.05 1
C10 [MPa] 0.0537433 0.0537433
C20 [MPa] -4.857E-04 -4.857E-04
C30 [MPa] 3.809E-06 3.809E-06

4.6.3 Numerical considerations

Two axially symmetrical finite element models with hyperelastic material behavior
(the viscoelastic part was neglected) were created in ABAQUS in order to verify the
analytical calculations.
The first model was a spring roll with a prestrain 1x1 (λx=λy=1, no prestrain) and the
second model had a prestrain of 4x4 (λx=λy=4). For both models, the geometrical ratio,
r, was chosen as r=77%. The element type (4-node bilinear axially symmetrical
quadrilateral, reduced integration) was applied and the Yeoh model was used for the
material model with the parameters in Table 4.1.

136
A single layer was considered for both spring rolls in the finite element model (Figure
4.24). The spring roll with prestrain 4x4 was simulated in three steps, Figure 4.24:
(i) prestraining with a factor of 4 in the x-direction, the edge DC was displaced
vertically and the edge AB was fixed in the x-direction
(ii) prestraining in the z-direction (the edge AD was displaced) in such a way that the
stretch ratio in the z-direction and the r-direction was 4 after step (ii). The
geometrical parameters after step (ii) corresponded to the actual geometry of the
spring rolls in article 4, Figure 11: x1 = 143 mm, xa = 110 mm and z1 = 62.5 µm.
The radius of the spring roll, r0, was 3.5 mm
(iii) applying a displacement, ur, in the (negative) r-direction, see Figure 4.24

D C txbII
x

Rotation axis

r0
tzaII
ur
r
z1
zaII

A B
(i) (ii)

Figure 4.24. Finite element model of the spring roll (prestrain 4x4), (a) sketch and (b)
finite element model, the scale factor in the radial direction is 100.

Only step (iii) was performed for the spring roll with prestrain 1x1. The thickness z1
was chosen as 50 µm and the radius r0 as 5 mm. The same material parameters as for
the analytical model were applied for both simulations. The thickness, the prestrain (in
the radial and axial directions) and the ratio r were identical for the analytical and the
numerical models for each single prestrain, 1x1 and 4x4.

137
4.6.4 Results

A comparison between analytical and numerical calculations is shown in Figure 4.25.

The stresses tzaII of the active part and txbII of the passive part are plotted against the
thickness zaII.

Activation Not activated

Activation Not activated
0.00 0.2

Cauchy stress [MPa]

Cauchy stress [MPa]

II
txb analytical
II
0.1 txb numerical
II II
txb analytical tza analytical
-0.02
II II
txb numerical tza numerical
II 0.0
tza analytical
II
tza numerical
-0.04 -0.1
0.0495 0.0500 0.058 0.060 0.062
Thickness [mm] Thickness [mm]

(a) (b)
Figure 4.25. Analytical vs. numerical calculations for (a) a spring roll with prestrain
1x1 and (b) a spring roll with prestrain 4x4. The stresses are plotted against the
thickness.

The finite element and analytical results agree to a great extent and this provides a
verification of both the analytical and the numerical model. The distinctions between
analytical and numerical model are larger for prestrain 4x4 than 1x1. The
discrepancies between the analytical and numerical models might come from the
transition between the active and the passive part. The analytical model allows the
voltage, U, to be plotted against k for different ratios r. For prestrain 4x4, the voltage
U is calculated as function of k for the geometrical ratios r = 0.7, 0.8 and 0.9, Figure
4.26.

138
 1.0

r = 0.9

k [-] 0.8
r = 0.8

r = 0.7

0.6
0 1000 2000 3000
U [V]

Figure 4.26. Voltage U vs. ratio k for different geometrical ratios r = 0.7, 0.8 and 0.9.

The analytical and numerical models are in agreement for spring rolls with prestrain
1x1 and prestrain 4x4. The ratio k depends on the following parameters: the
geometrical ratio, r, the voltage, U, the dielectric constant, εr, the pre-stretch ratios, λx
and λy, the initial thickness, Z, and the material parameters, C10, C20 and C30. When the
voltage is zero, the ratio k corresponds to the ratio r. By increasing the voltage, k
increases only slightly.
From an experimental point of view, the ratio k should be considered as constant, as
confirmed by the fact that fitting the experimental electromechanical pressure with
Pelrine’s equation gives an excellent result. In this sense, the slight dependence of k on
the voltage is in agreement with the experiments.
The analytical model does not consider the electrodes. The electrode seems not to have
a strong influence. By assuming k=r, force measurements of spring rolls provide a
value for the dielectric constant of 3.24. This is in agreement with dielectric constant
measurements (article 4).
Future work should consider the influence of the electrodes by a spring roll activation
which can deform in the axial direction (no fixation).

139
5 Conclusions and Outlook

5.1 Main contributions of the present work

An important contribution of this thesis is represented by the mechanical model

(including the corresponding material parameters) valid over a wide deformation range
for dielectric elastomer actuators made of VHB 4910. A large number of experiments
including circular strain tests, uniaxial tests, aspiration tests and compression tests
were performed and the test data were shared with the scientific community for model
verification purposes. A finite element model was created and simulations were
performed for each test. Comparisons between experiments and simulations showed
that the constitutive model (in particular with the Yeoh strain energy form) is able to
describe the material behavior for several deformation and activation configurations.
Electromechanical coupling has been investigated and a new physical interpretation of
Pelrine’s equation was found. The dielectric constant has been determined directly
with a LCR-meter and with spring roll experiments. Specific conclusions were drawn
based on the results presented in each article that constitute the main part of the
present work.
Article 1 verifies the quasilinear visco-hyperelastic model, which includes large
deformation and viscoelasticity. The condition of validity for quasilinear
viscoelasticity (the stress relaxation function is independent of the magnitude of the
deformation) is verified based on relaxation tests. The strain energy potential of Yeoh
is used to describe large deformations. The material parameters were obtained by
fitting experimental data using an optimization algorithm. An inverse finite element
technique is introduced for the calculation of the circular actuator. Simulation results
have shown a good agreement between experiments and simulations for the circular
strain test.
Article 2 presents an analytical model for circular actuators. An equation that links the
voltage, the prestrain and the active strain was derived. The analytical model was
verified numerically. The analytical model is valid for hyperelastic materials (time-
dependent effects are neglected) and might be useful for the design of silicone
actuators with less pronounced viscoelastic behavior. The analytical model allows the
investigation of electromechanical instabilities which might lead to failure of the
actuators.

140
Article 2 demonstrates that the fitting of uniaxial data for the strain energy forms of
Yeoh, Ogden and Mooney-Rivlin leads to different simulation results in the
(multiaxial) VHB 4910 actuator behavior. This shows the importance of characterizing
the biaxial response of the elastomer. This step was omitted in most previous work
(see section 1.4) on dielectric elastomers, where material models are proposed based
on uniaxial test data only.
In article 3 an extensive experimental study on circular actuators is presented. A novel
finite element technique was used which allows direct simulation of actuator activation
by the applied voltage. Three different constitutive model formulations were
evaluated, viz. the strain energy forms of Yeoh, Ogden (six parameters instead of only
two as in article 2) and Arruda-Boyce. The corresponding material parameters were
determined with an enhanced optimization procedure.
The behavior of three actuators at prestrain 3, 4 and 5 was included in the optimization
procedure. With the optimized material parameters, the actuator behavior was
reasonably described by the model over a large strain range. By contrast, the uniaxial
tensile test simulation did not fit the experimental data. This discrepancy has been
further analyzed in article 4. Furthermore, limitations of the proposed quasilinear
visco-hyperelastic model were demonstrated when applied to the prediction of the
actuator behavior for a large number of activation cycles.
In article 4 the electromechanical coupling was investigated. Pelrine’s equation was
verified by an analytical approach. A numerical investigation provides the charge,
electric field and electrostatic force distribution for a circular actuator. It is found that
the electromechanical pressure acts in the thickness direction and with the same
magnitude in the lateral direction. By considering a superimposed hydrostatic stress
state the resulting out-of-plane pressure corresponds to the result obtained by Pelrine.
Measurements of the dielectric constant demonstrated that its value decreases with
increasing prestrain. The dielectric constant of VHB 4910 is about 3.2 and not 4.7 (as
proposed by Kofod and used by many researchers) in the relevant prestrain range. This
was also verified with the spring roll experiments.
With the findings of this paper the material parameters of the constitutive model were
adapted. A comparison between simulations (demonstrated with the Arruda-Boyce
strain energy form) and experiments showed a significant improvement for the
uniaxial behavior.
In section 5 the constitutive model parameters for the Yeoh and the Ogden strain
energy form were adjusted based on a new value of the dielectric constant. Results

141
have shown, that for all three strain energy formulations the uniaxial behavior is well
described.
Aspiration and compression tests are presented for the verification of the multiaxial
response predicted using the constitutive model. The results showed that the Yeoh
formulation yields excellent agreement between experiments and simulations. This is
not only a confirmation of the validity of the constitutive model but also, indirectly, of
the electromechanical coupling.
The deformation range of all performed experiments was studied by plotting the first
invariant vs. the second invariant of the left Cauchy Green deformation tensor. This
showed that the constitutive model is valid over a large deformation range.
The influence of the geometrical ratio between the active and passive parts of spring
rolls was investigated. This is important for (i) the design of spring rolls and (ii) the
interpretation of the measurements of the dielectric constant. Results showed that the
geometrical ratio, r, corresponds in a good approximation to the so-called force ratio,
k, introduced in article 4.

5.2 General design considerations and suggestions for future

work

The performance of dielectric elastomer actuators depends on the material properties

of the elastomer and their geometrical characteristics. Here, some considerations are
proposed on the material behavior and the geometrical design of the actuator with
respect to the findings of the present work.

5.2.1 Material behavior of the elastomer

The constitutive model presented in this work is valid for VHB 4910. For a different
material, but which is also incompressible and quasilinearly viscoelastic, the same
constitutive model formulation can be applied, provided the respective material
parameters are determined, e.g. with the procedures described in the present work. For
rubber-like materials (e.g. silicones) without a dissipative behavior, time dependence
does not have to be considered. Thus, the analytical model as proposed in article 2
might be used for the optimization of the actuators.
Models with internal state variables (e.g. [Rubin 2002]) that can be interpreted as
describing changes in the material microstructure, can be applied for elastomers with a
significant dissipative and history-dependent response.

142
For the characterization and modeling of a new DE material one of two cases may
occur:
(i) the material is given and an adequate constitutive model has to be determined and
(ii) the desired constitutive model is given (due to the required actuator performance)
and an adequate material has to be found or synthesized.
The formulation and the determination of the mechanical material parameters and the
dielectric constant are necessary to define a constitutive model for a given material. As
shown in the present work, the multiaxial material behavior has to be considered in the
fitting process of the material parameters.
The second case occurs when input voltages and the mechanical output
(stresses/strains) of the actuator are prescribed. A suitable elastomer is required which
can be, for example synthesized. For this purpose, it is useful to evaluate the link
between the microstructure of the elastomer and the constitutive model. For example,
the Arruda-Boyce strain energy form [Arruda 1993] is based on statistical mechanics
and the parameters are physically linked to the chain lengths and orientations involved
in the deformation of the three-dimensional network structure of the rubber [Holzapfel
2000]. By fitting the constitutive model parameters in such a way that the required
actuator output is achieved in the simulations, the material parameters (e.g. shear
modulus and number of chain segments) can be used as a basis for the synthesis of the
elastomer.

5.2.2 Geometrical design of the actuator

The results of the present work make fundamental contributions to the modeling of
dielectric actuators (not only their material). In particular, circular actuators have been
simulated. Suitable numerical models have to be developed for the design of actuators
for general applications. The design of the actuator can be optimized based on the
model presented here. One relevant design parameter in particular is the level of
prestretch (in each planar direction) to be applied. For spring rolls used as artificial
muscles [Kovacs 2006] or walking robots, axially symmetrical finite element models
can be created with axially symmetrical elements similar to the procedure described
for circular actuators. For applications with complex geometries such as shell-like
structures [Lochmatter 2006b] or blimps [Michel 2006], three-dimensional finite
element models with thin shell elements are required. Standard shell elements are not
able to transfer the electromechanical pressure in the lateral direction for hyperelastic
materials. New program codes have to be developed for the design of such structures.

143
The results of article 4 offer novel design possibilities for actuators. The finding that
high lateral forces arise at the lateral surface area might be useful for enhancing the
efficiency of the electromechanical coupling. The absorption of a larger charge density
would increase the electrostatic forces. This could be achieved by a smart design of the
electrodes: By applying a material with a high dielectric constant (e.g. ceramics) at the
lateral boundaries, the electrode could absorb a greater charge due to the high
polarization of the material with high dielectric constant.
In general, further investigation of the electrodes and the electromechanical coupling
are required to improve the predictive capabilities of the models, in particular for high
activation voltages. The physical mechanism of electrostatic force transmission from
the electrodes to the elastomer in the out-of-plane and in-plane directions has to be
understood in order to optimize the selection of electrode material.

5.2.3 Reliability

A further challenge is to describe the failure mechanisms by experiments and

simulations. In [Plante 2006a and Plante 2006b] three failure mechanisms are
described, viz. material strength failure, dielectric strength failure and pull-in
instability failure. The material strength is purely a mechanical failure caused by
stretching the elastomer over a certain limit. The dielectric strength is a purely
electrical failure mechanism and occurs when the electrical field is higher than the
electrical breakdown field. The pull-in instability occurs when the magnitude of the
electromechanical pressure increases at a greater rate than the mechanical out-of-plane
stress.
Plante performed experiments and calculations for circular actuators and for diamond
actuators relating failure. Failure maps which compare the experimental and the
calculated failure strain were created. For some cases, the experimental and the
calculated values agreed, for other cases the analytical prediction was 57% over the
experimental value. The failure models were only verified by a few pre-stretch ratios
(e.g. for the diamond actuator only one pre-stretch ratio was considered).
In general, reliability should be investigated more in detail. It is a relevant aspect for
dielectric elastomer technology, especially for the development of commercial
applications. For future work, the following points should be considered:
(i) The quantitative connection between failure mechanism of circular actuators and
‘real applications’ needs to be understood. The circular actuator is a good model
system as shown in the present work.

144
(ii) The influence of the electrode should be evaluated. The electrode composition
and shape might influence the failure behavior of the actuators.
(iii) The fatigue behavior of actuators should be investigated more in detail. Both
cyclic mechanical and actuator tests are required.
(iv) The influence of pre-stretching the elastomer with respect to failure behavior
needs to be understood. In general, time-dependent failure mechanisms have to
be identified.
(v) Different elastomers should be investigated, in order to evaluate their differences
and commonalities in failure mechanisms.

145
146
Bibliography

ABAQUS/Standard Manual, 2004. Version 6.5.1, Hibbitt, Karlsson & Sorensen, Inc.

Aoki T., T. Ohashi, et al., (1997). The pipette aspiration applied to the local stiffness
measurement of soft tissues, Ann. Biomed. Eng., vol. 25, pp. 581 - 587

Arruda E.M., M.C. Boyce (1993). A three-dimensional constitutive model for the large
stretch behavior of rubber elastic materials, J. Mech. Phys. Solids, vol. 41, No.2, pp.
389 - 412

Aschwanden M., A. Stemmer (2006). Polymeric, electrically tunable diffraction

grating based on artificial muscles, Optical letters, vol. 31, no. 17, pp. 2610-2612

Ashley S. (2003). Artificial muscles, Scientific American, vol. 289, no. 4, pp. 52-59

Bar-Cohen Y. (2001). Electroactive polymer (EAP) actuators as artificial muscles:

Reality, potential and challenges, SPIE Press, Bellingham, Washington

Baughman R.H., C. Cui, et al. (1999). Carbon nanotune actuators, Science, vol. 284,
pp. 1340-1344

Bergström J., M. Boyce (1998). Constitutive modeling of the large strain time-
dependant behavior of elastomers, Journal of Mechanics and Physics of Solids, vol.
46, No. 5, pp. 931-954

Bird R.B., R.C. Armstrong, O. Hassager (1977). Dynamics of polymeric liquids –

volume 1: Fluid mechanics, second edition, New York, Wiley

Boyce M.C., E.M. Arruda (2000). Constitutive models of rubber elasticity: A review,
Rubber Chemistry and Technology, vol. 73, no. 3, pp. 504-523

Calvert P., J. O’Kelly, C. Souvignier (1998), Solid freeform fabrication of organic-

inorganic hybrid materials, Mat. Sci. Eng. C, vol. 6, pp. 167-174

Carpi F., P. Chiarelli, et al. (2003). Electromechanical characterisation of dielectric

elastomer planar actuators: comparative evaluation of different electrode materials and
different counterloads, Sens. Actuators A, vol. 107, pp. 85-95

Carpi F., D. De Rossi (2004). Dielectric elastomer cylindrical actuators:

electromechanical modeling and experimental evaluation, Mat. Sci. Eng. C, vol. 24,
pp. 555-562

Christensen R.M. (1981). Theory of viscoelasticity: an introduction, Academic Press,

New York

COMSOL Multiphysics, 2005, User’s Guide, version 3.2, COMSOL, Inc.

147
Fung Y.C. (1993). Biomechanics. Mechanical properties of living tissues, 2nd edn.,
Springer-Verlag, New York, p. 277

Gandhi M.R., P. Murray, et al. (1995). Mechanisms of electromechanical actuation in

polypyrrole, Synthetic Metals, vol. 75, pp. 247-256

Goulbourne N., M. Frecker, et al. (2003). Modeling of a dielectric elastomer

diaphragm for a prosthetic blood pump, Proc. SPIE, vol. 5051, pp. 319-31

Goulbourne N., E. Mockensturm, M. Frecker (2005). A nonlinear model for dielectric

elastomer membranes, Journal of Applied Mechanics, vol. 72, pp. 899 – 906

Heydt R., R. Kornbluh, et al. (1998). Design and performance of an electrostrictive-

polymer-film acoustic actuator, Journal of Sound and Vibration, vol. 215, no. 2, pp.
297-311

Holzapfel A. (2000). Nonlinear solid mechanics – a continuum approach for

engineering, John Wiley & Sons, Chichester, UK

Kim J., J.Y. Kim, S.J. Choe (2000). Electro-active papers: its possibility as actuators,
Proc. SPIE, vol. 3987, pp. 203-209

Kim H., S. Oh, et al. (2001). Actuator model of electrostrictive polymers (EPs) for
microactuators, Proc. SPIE, vol. 4329, pp. 482-490

Kofod G. (2001a). Dielectric elastomer actuators, PhD. thesis, Riso-R-1286(EN)

Kofod G., R. Kornbluh, et al. (2001b). Actuation response of polyacrylate dielectric

elastomers, Proc. SPIE, vol. 4329, pp. 141 – 147

Kofod G., P. Sommer-Larsen, et al. (2003). Actuation response of polyacrylate

dielectric elastomers, J. Intel. Mat. Syst. Str., vol. 14, pp. 787-793

Kofod G., P. Sommer-Larsen (2005). Silicone dielectric elastomer actuators: Finite-

elasticity model of actuation, Sens. Actuators A, vol. 122, pp. 273-283

Kornbluh R., R. Pelrine, et al. (1999). High-field electrostriction of elastomeric

polymer dielectrics for actuation, Proc. SPIE, vol. 3669, pp. 149-161

Kornbluh R., R. Pelrine, et al. (2000). Ultrahigh strain response of field-actuated

elastomeric polymers, Proc. SPIE, vol. 3987, pp. 51-64

Kovacs G., P. Lochmatter (2006). Arm wrestling robot driven by dielectric elastomer
actuators, Proc. SPIE, vol. 6168, pp. 616807-1 - 616807-12

Lagarias J., J. Reeds, et al. (1998). Convergence properties of the Nelder-Mead

simplex method in low dimensions, Siam. J. Optim., vol. 9, no. 1, pp. 112-147

148
Lewis R., V. Torczon, M. Trosset (2000). Direct search methods: then and now,
Journal of Computational and Applied Mathematics, vol. 124, pp. 191-207

Lochmatter P., S. Michel, G. Kovacs (2006a). Electromechanical model for static and
dynamic activation of elementary dielectric elastomer actuators, Proc. SPIE, vol. 6168,
pp. 61680F-1 - 61680F-13

Lochmatter P., G. Kovacs, P. Ermanni (2006b). Design and characterization of shell-

like dielectric elastomer actuators, Proceedings of ICAST2006: 17th International
Conference on Adaptive Structures and Technologies, Oct. 16-19, Taipei, Taiwan

Matlab 7.0, 2004, The MathWorks, Inc.

Maxwell J. C. (1954). A treatise on electricity and magnetism, Dover, Oxford

McMeeking R., C. Landis (2005). Electrostatic forces and stored energy for
deformable dielectric materials, Journal of Applied Mechanics, vol. 72, pp. 581 – 590

Michel S., G. Kovacs, P. Lochmatter (2006). Propulsion unit for lighter-than-air

aircraft, PCT-Patent, No. WO 2006/108311 A1

Mockensturm E., N. Goulbourne (2006). Dynamic response of dielectric elastomers,

International Journal of Non-Linear Mechanics, vol. 41, pp. 388 – 395

Mooney M. (1940), A theory of large elastic deformation, J. Appl. Phys., vol. 11, pp.
582-592

Nava A., E. Mazza, et al. (2004). Evaluation of the mechanical properties of human
liver and kidney through aspiration experiments, Health Care, vol. 12, pp. 269 – 280

Nava A. (2007). In vivo characterization of the mechanical response of soft human

tissues, Ph.D. thesis, Institute for Mechanical Systems, ETH Zurich, Switzerland

Nelder J., R. Mead (1965). A simplex method for function minimization, Comput. J.,
vol. 7, pp. 308-313

Ogden R.W. (1972). Large deformation isotropic elasticity – on the correlation of

theory and experiment for incompressible rubberlike solids, Proc. R. Soc. Lond. A, vol.
326, pp. 565-584

Pei Q., R. Pelrine, et al. (2003). Electroelastomer rolls and their application for
biomimetic walking robots, Synthetic Metals, vol. 135- 136, pp. 129-131

Pelrine R., R. Kornbluh, J. Joseph (1998). Electrostriction of polymer dielectrics with

compliant electrodes as a means of actuation, Sens. Actuators A, vol. A 64, pp. 77-85

Pelrine R., R. Kornbluh, et al. (2000a). High-speed electrically actuated elastomers

with strain greater than 100%, Science, vol. 287, pp. 836-839

149
Pelrine R., R. Kornbluh, G. Kofod (2000b). High-strain actuator materials based on
dielectric elastomers, Adv. Mater., vol. 12, no. 16, pp. 1223-1225

Pelrine R., P. Sommer-Larsen, et al. (2001a). Applications of dielectric elastomer

actuators, Proc. SPIE, vol. 4329, pp. 335-349

Pelrine R., R. Kornbluh, et al. (2001b). Dielectric elastomers: Generator mode

fundamentals and applications”, Proc. SPIE, vol. 4329, pp. 148-156

Pelrine R., R. Kornbluh, et al. (2002). Dielectric elastomer artificial muscle actuators:
Toward biomimetic motion”, Proc. SPIE, vol. 4695, pp. 126-137

Plante, J.S. (2006). Dielectric elastomer actuators for binary robotics and
mechatronics, Ph.D. Thesis, Department of Mechanical Engineering, Massachusetts
Institute of Technology, Cambridge, MA.

Plante J., S. Dubowsky (2006b). Large-scale failure modes of dielectric elastomer

actuators, International Journal of Solids and Structures, vol. 43, pp. 7727–7751

Rivlin R.S., K.N. Sawyers (1971). Nonlinear continuum mechanics of viscoelastic

fluids, Ann. Rev. Fluid Mech., vol. 3, pp. 117-146

Rubin M.B., S.R. Bodner (2002). A three dimensional nonlinear model for dissipative
response of soft tissue, International Journal of Solids and Structures, vol. 39, pp.
5081-5099

Shahinpoor M., Kim K.J. (2001). Ionic polymer-metal composites: I. Fundamentals,

Smart Mater. Struct., vol. 10, pp. 819-833

Sommer-Larsen P., G. Kofod, et al. (2002). Performance of dielectric elastomer

actuators and materials, Proc. SPIE, vol. 4695, pp. 158-166

Su J., J.S. Harrison, et al. (1999). Electrostrictive graft elastomers and applications,
MRS Symposium Proceedings, Warrendale, vol. 600, pp. 131-136

Sussman T., K.J. Bathe (1987). A finite element formulation for nonlinear
incompressible elastic and inelastic analysis, Journal Computers & Structures, vol. 26,
pp. 357-409

Treloar L. (1975). The physics of rubber elasticity, third edition, Oxford, Clarendon
Press

Truesdell C., W. Noll (1992). The non-linear field theories of mechanics, 2nd edn.,
Springer-Verlag, Berlin

Vuskovic V. (2001). Device for in vivo measurement of mechanical properties of

internal human soft tissues, Ph.D. thesis, No. 14222, ETH Zurich, Switzerland,

150
Wissler M., E. Mazza (2005). Modeling and simulation of dielectric elastomer
actuators, Smart Materials and Structures, vol. 14, pp. 1396 - 1402

Wissler M., E. Mazza (2005). Modeling of a prestrained circular actuator made of

dielectric elastomers, Sens. Actuators A, vol. 120, pp. 184-192

Wyrsch J. (2006). Untersuchung des elektromechanischen Verhaltens von Aktoren aus

dielektrischen Elastomeren, Bachelerarbeit at Empa Dübendorf, Laboratory for
Materials and Engineering, Dubendorf, Switzerland

Yang E., M. Frecker, E. Mockensturm (2005). Viscoelastic model of dielectric

elastomer membranes, Proc. SPIE, vol. 5759, pp. 82 – 92

Yeoh O.H. (1990). Characterization of elastic properties of carbon-black-filled rubber

vulcanizates, Rubber Chem. Technol., vol. 63, pp. 792-805

Yeoh O.H. (1993). Some forms of the strain energy function for rubber, Rubber Chem.
Technol., vol. 66, pp. 754-771

Zhang Q.M., V. Bharti, X. Zhao (1998). Giant electrostriction and relaxor ferroelectric
behavior in electron-irradiated poly(vinylidene-fluoride-trifluorethylene) copolymer,
Science, vol. 280, pp. 2101-2104

Zhang Q. M., H. Li, et al. (2002). An all-organic composite actuator material with a
high dielectric constant, Nature, vol. 419, pp. 284-287

Zhang R., A. Kunz, et al. (2004). Dielectric elastomer actuators for a portable force
feedback device, The 4th European Conference on EuroHaptics 2004, Munich
Germany, June 2004, pp. 300-307

Zhang X., C. Löwe, et al. (2005). Dielectric elastomers in actuator technologies,

Advanced Materials Engineering, vol. 7, no. 5, pp. 361-367

151
152
Curriculum Vitae

Name: Michael Wissler

Date of birth: 25th October 1976
Citizen of: Sumiswald (BE), Switzerland

Education

1983 - 1985 Primary school, Oberrieden, Switzerland

1985 - 1991 Secondary school, Oberrieden, Switzerland

1991 - 1996 High School, MNG Rämibühl (Typus C), Zurich, Switzerland

1996 - 2002 Tertiary education at the ETH Zurich, Department of Materials.

Graduation as Dipl. Werkstoff.-Ing. ETH (Degree in Materials
Engineering)

2003 - 2007 Doctoral student at the Laboratory for Materials and Engineering,
Empa, Dübendorf, Switzerland and the Institute of Mechanical
Systems, ETH Zurich, Zurich, Switzerland

Professional experience

1996 Practical training for six weeks at Landis und Stäfa, Zug, Switzerland

1998 Practical training for three month at Copenhagen Pectin SA, Lille
Skensved, Denmark

2003 - 2007 Research assistant at the Laboratory for Materials and Engineering,
Empa, Dübendorf, Switzerland and the Institute of Mechanical
Systems, ETH Zurich, Zurich, Switzerland

153