ULTRATHIN CONDUCTING POLYMER TRANSDUCERS:

FABRICATION, CHARACTERIZATION, AND MODELING

by

Ngoc Tan Nguyen

M. Sc., Inje University, 2014

B. Eng. Danang University of Science and Technology, 2010

A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF

THE REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY
in
THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES
(Electrical &Computer Engineering)

THE UNIVERSITY OF BRITISH COLUMBIA

(Vancouver)

September 2018

© Ngoc Tan Nguyen, 2018

 Thèse de doctorat

Pour obtenir le grade de Docteur de

L’UNIVERSITÉ POLYTECHNIQUE HAUTS-DE-FRANCE

ET L’UNIVERSITÉ DE COLOMBIE BRITANNIQUE

Spécialité micro et nanotechnologies, acoustiques et télécommunications
Et Doctor of Philosophy in Electrical &Computer Engineering (PhD)

Présentée et soutenue par Ngoc Tan, NGUYEN.

Le 21/09/2018, à Villeneuve d’Ascq

Ecole doctorale :
Sciences Pour l’Ingénieur (SPI)

Equipe de recherche, Laboratoire :

Institut d’Electronique, de Micro-Electronique et de Nanotechnologie/Département d’Opto-Acousto-Electronique
(IEMN/DOAE)
Transducteurs ultra fin à base de polymères conducteurs :
fabrication, caractérisation et modélisation

Composition du jury

Président du jury
M. Edmond CRETU, Professeur de l’Université de Colombie Britannique, Vancouver

Rapporteurs
M. Alejandro A. FRANCO, Professeur des Universités, UPJV/ CNRS UMR 7314, Amiens.
M. Herbert SHEA, Professeur l’Université de EPFL, Neuchâtel.

Examinateurs
M. Mu CHIAO, Professeur de l’Université de Colombie Britannique, Vancouver
Mme. Ludivine FADEL, Professeur des Universités, IMS UMR 5218, Talence

Directeur de thèse
M. Éric CATTAN, Professeur des Universités, UPHF / IEMN, Valenciennes
M. Sébastien GRONDEL, Professeur des Universités, UPHF / IEMN, Valenciennes
M. John D.W. MADDEN, Professeur de l’Université de la Colombie Britannique, Vancouver

Membres invités
M. Cédric PLESSE, Maître de Conférence HDR, LPPI, Cergy Pontoise.
Abstract
Recently, ultrathin poly (3,4-ethylenedioxythiophene) (PEDOT) – based ionic actuators have
overcome some initial obstacles to increase the potential for applications in microfabricated
devices. While microfabrication processing of trilayer actuators that involve no manual
handling has been demonstrated, their mechanical performances remain limited for practical
applications.
The goal of this thesis is to optimize the transducers in thin films fabrication by micro
technologies, fully characterize the electrochemomechanical properties of the resulting
trilayers, and develop a model to simulate their bidirectional electromechanical ability
(actuation and sensing).
At first, ultrathin PEDOT-based trilayer actuators are fabricated via the vapor phase
polymerization of 3,4-ethylenedioxythiophene combining with the layer by layer synthesis
process. This constitutes the first full characterization of ionic PEDOT-based microactuators
operating in air of such a small thickness (17 µm) having bending deformation and output
force generation of 1% and 12 µN respectively.
Secondly, electrical, electrochemical and mechanical properties of the resulting
microactuators have been thoroughly studied. Non-linear characterization was extended to
volumetric capacitance dependence on voltage window. Damping coefficient was
characterized for the first time.
Thirdly, a nonlinear multi-physics model was proposed as a method of simulating actuator
and sensor responses in trilayers, represented using a Bond Graph formalism, and was able
to implement all of the characterized parameters. The concordance between the simulations
and the measurements confirmed the accuracy of the model in predicting the non-linear
dynamic behavior of the actuators. In addition, the information extracted from the model also
provided an insight into the critical parameters of the actuators and how they affect the
actuator efficiency, as well as the energy distribution.
Finally, a nouveau bidirectional electromechanical linear model was introduced to simulate
the sensing ability of the trilayer transducer and was confirmed via experimental results in
both frequency and time domains of a sinusoidal input displacement. The resulting actuators
and the proposed models are promising for designing, optimizing, and controlling of the
future soft microsystem devices where the use of polymer actuators should be essential.
Keywords: Electroactive polymer, conducting polymer transducers, interpenetrating networks, non-
linear characterization, non-linear dynamic modeling, Bond Graph formalism.

iii
Résumé
Récemment, les actionneurs ioniques ultra-minces à base de poly (3,4-
éthylènedioxythiophène) (PEDOT) ont surmonté certains obstacles initiaux pour augmenter
le potentiel d'applications dans les dispositifs microfabriqués. Bien que la microfabrication
d’actionneurs à trois couches, n’impliquant aucune manipulation manuelle, ait été
démontrée, leurs performances mécaniques restent limitées pour des applications pratiques.
Le but de cette thèse est d'optimiser les transducteurs dans la phase de fabrication des
couches minces en utilisant des micro technologies, de caractériser complètement les
propriétés électrochimiques des transducteurs ainsi obtenus, et de développer un modèle
pour simuler leurs capacités électromécaniques bidirectionnelles (actionnement et
détection).
Tout d'abord, les actionneurs à trois couches ultra-minces à base de PEDOT sont fabriqués
par polymérisation en phase vapeur de 3,4-éthylènedioxythiophène en réalisant un procédé
de synthèse couche par couche. Le travail présenté constitue la première caractérisation
complète de microactionneurs ioniques à base de PEDOT fonctionnant dans l’air d’une si
faible épaisseur (17 µm) présentant une déformation en flexion et une génération de force
de 1% et 12 µN respectivement. En effet, les propriétés électriques, électrochimiques et
mécaniques des microactionneurs ont été minutieusement étudiées. La caractérisation non
linéaire a été étendue à la dépendance de la capacité volumétrique sur une fenêtre de
tension. Le coefficient d'amortissement a été caractérisé pour la première fois. Par ailleurs,
un modèle multi-physique non linéaire a été proposé comme méthode de simulation des
réponses en mode actionneur et capteur dans des couches multiples, représenté à l'aide d'un
formalisme Bond Graph, et a été capable de mettre en œuvre tous les paramètres
caractérisés. La concordance entre les simulations et les mesures a confirmé l'exactitude du
modèle pour prédire le comportement dynamique non linéaire des actionneurs. En outre, les
informations extraites du modèle ont également permis de mieux comprendre les
paramètres critiques des actionneurs et leur incidence sur l'efficacité de l'actionneur et sur la
distribution de l'énergie.
Enfin, un nouveau modèle linéaire électromécanique bidirectionnel a été introduit pour
simuler la capacité de détection du transducteur à trois couches et a été confirmé par des
résultats expérimentaux dans les domaines fréquentiel et temporel d'un déplacement
d'entrée sinusoïdal. Les actionneurs résultants et les modèles proposés sont prometteurs
pour la conception, l'optimisation et le contrôle des futurs dispositifs de microsystèmes
souples dans lesquels l'utilisation d'actionneurs en polymère devrait être essentielle.
Keywords: Polymère électroactif, transducteurs polymères conducteurs, réseaux interpénétrés,
caractérisation non linéaire, modélisation dynamique non linéaire, formalisme Bond Graph.

iv
Résumé Substantiel
Introduction
Les polymères conducteurs (CP) sont des polymères organiques conducteurs électricique.
Depuis la découverte de conducteurs organiques polymères en 1977 (Chiang et al. 1977;
Shirakawa et al. 1977), les connaissances dans ce domaine se sont considérablement
développées. En particulier, le poly (3,4-éthylènedioxythiophène) (PEDOT) joue un rôle
important dans les applications antistatiques, électriques et électroniques (Kirchmeyer et
Reuter, 2005). L'idée d'utiliser des polymères conducteurs comme actionneurs pour la
conversion directe d'énergie électrique en énergie mécanique, basée sur le processus
d'oxydo-réduction entre les PC et les ions mobiles dans l'électrolyte, a été initialement
proposée par Baughman (Baughman, 1996; Baughman, et al., 1991). De remarquables
progrès ont été réalisés sur les actionneurs à base de PC par Pei et Inganäs, Smela et Inganäs,
Otero et, récemment, par Vidal et al. (Maziz et al. 2014; Otero et al. 1992; Otero 2013; Pei et
Inganäs 1993; Smela 2003; Vidal et al. 2006; Vidal et al. 2004). Ces actionneurs offrent des
avantages par rapport aux autres types de muscles artificiels en proposant une combinaison
de tension de fonctionnement faible (typiquement 1-2 V) et de biocompatibilité (Ateh et al.
2006; Cellot et al. 2015; Richardson-Burns, al. 2007), d’une plage de déformation importante
allant de 1 à 10%, de force bloquante et de travail spécifique élevé. Ils peuvent être envisagés
dans diverses applications, notamment les cathéters (Della Santa et coll. 1996; Jager 2002),
l'affichage dynamique en braille (Runyan et Carpi 2011; Spinks et coll. 2003), les valves (Jager
2002) et la reconnexion des vaisseaux sanguins. (Mirfakhrai et al. 2007).
Récemment, le problème de la faible vitesse d’actionnement a été abordé par Maziz et al.
(Maziz et al. 2016b; Maziz et al. 2014). Ils ont mis en évidence une vitesse de fonctionnement
en kHz (à la résonance) des actionneurs à base de PEDOT, l'épaisseur totale de l'actionneur
étant de plusieurs micromètres. Le Laboratoire de Physicochimie des Polymères et des
Interfaces de l’Université de Cergy-Pontoise, l’Institut d’électronique de microélectronique et
de nanotechnologie de l’Université Polytechnique des hauts de France et le laboratoire de
mécatronique moléculaire de l’Université de Colombie Britannique ont une longue et
fructueuse collaboration (depuis plus de 10 ans) sur les matériaux et dispositifs à base de
polymères conducteurs (CP), ainsi que sur l'optimisation et la caractérisation, la modélisation
et les applications de ces matériaux. Actuellement, l'un des principaux objectifs du groupe est
de développer des microsystèmes à partir de transducteur polymère à base de PC. Comme
cela a été évoqué, les transducteurs à base de CP sont aussi des candidats potentiels pour des
applications à l'échelle microscopique (épaisseur de l'actionneur <20 µm) nécessitant une
faible tension de fonctionnement, des contraintes élevées, un fonctionnement rapide à l'air
libre et, en particulier, une compatibilité avec les techniques de microfabrication en salle
blanche. Cependant, pour amener ces matériaux vers de nouveau procédé de fabrication
comme proposé par Maziz et al. (Maziz, et al. 2016a) vers une production à grande échelle,

v
l'optimisation des techniques de fabrication, ainsi que la caractérisation complète et la
modélisation de ces micro structures sont des étapes essentielles.

Le but de cette thèse est de fabriquer, caractériser les propriétés électro-chimico-mécaniques

et de modéliser les transducteurs tricouches ultra-minces à base de PEDOT. La thèse est
divisée en six chapitres. Un bref contenu de chaque chapitre est donné ci-dessous:
Chapitre 2:
Ce chapitre contient une brève introduction sur les mécanismes de fonctionnement des
actionneurs à base de CP. Le processus de sélection des matériaux utilisés pour fabriquer la
tri couche à base de CP dans cette étude sont décrits. Par la suite, les différentes méthodes
de synthèse utilisées pour réaliser un polymère conducteur PEDOT sont brièvement passées
en revue et la polymérisation en phase vapeur est sélectionnée pour le procédé de fabrication
de l’électrode PEDOT, car elle permet de changer plus facilement l’épaisseur de l’électrode.
Le processus de fabrication en trois couches est ensuite décrit et la structure obtenue est
analysée avec un profilomètre, des techniques MEB et EDX afin de déterminer la rugosité et
la structure exactes du tri couche. Les résultats de rugosité de surface montrent une
asymétrie entre les deux électrodes de PEDOT, suggérant une différence de propriétés
électriques des deux électrodes.
Chapitre 3:
Ce chapitre présente la caractérisation électro-chimico-mécaniques des tri couches obtenus
comprenant les informations sur les méthodes de caractérisation et les résultats obtenus.
Tout d'abord, les propriétés électrochimiques de la tri couche, telles que la conductivité
électronique en fonction de l'état d'oxydation du PEDOT, la conductivité ionique et la capacité
volumétrique des électrodes de PEDOT en fonction de la vitesse de balayage et de la tension,
et la conductivité ionique du réseau semi-interpénétré NBR-PEO, sont déterminés. On mesure
aussi le degré de déformation en flexion à partir duquel est calculé le couplage entre les
domaines électrochimique et mécanique - un rapport déformation-charge empirique. Enfin,
les propriétés mécaniques sont déterminées: le module de Young des électrodes PEDOT en
fonction de l’état d’oxydation et de la structure en trois couches, le rapport d’amortissement,
la densité en trois couches et les forces en fonction de la tension et de la fréquence sont
déterminés.
Chapitre 4:
Le chapitre 4 est consacré à la modélisation des actionneurs à tri couches. Le formalisme Bond
Graph est introduit et la raison de son utilisation expliquée. Un modèle électro-chimio-
mécanique simple basé sur la théorie de poutres dynamiques d'Euler-Bernoulli est décrit.
Ensuite, un modèle électro-chimio-mécanique dynamique non linéaire basé sur la méthode
des éléments finis rigides dynamiques est présenté. Les propriétés des matériaux mesurés
(chapitre 3) sont utilisées dans le modèle et les simulations sont comparées aux déviations
d'actionneur observées de manière expérimentale au fil du temps. Le modèle proposé permet

vi
d’extraire des informations sur les flux d’énergie à l’intérieur de la tri couche et d’évaluer les
facteurs clés affectant les performances de l’actionneur (comme par exemple la déformation
en sortie).
Chapitre 5:
Un modèle simple, comprenant à la fois la détection et l'actionnement, est proposé,
démontrant l'avantage de la formulation Bond Graph dans les modes de fonctionnement
électrique vers mécanique et mécanique vers électrique. Le modèle initial est basé sur
l'utilisation aussi d’un rapport déformation sur charge comme coefficient de couplage
empirique.
Chapitre 6:
Dans ce dernier chapitre, les conclusions de cette recherche sont résumées et les perspectives
de travail à court et long terme sont développées.

Résultats
Tout d'abord, est décrit la fabrication d'une structure en trois couches d'un semi-NBR / PEO
pris en sandwich entre deux électrodes PEDOT, réalisée par empilement par la technique dite
couche par couche. La procédure de fabrication est résumée dans la figure 1.

Fig.1 Procédé de fabrication d'un microactionneur montrant le processus multicouche suivi d'un micro-
usinage au laser
En introduisant le PEO dans les trois couches, la conductivité ionique et l'adhésion mécanique
entre ces couches augmentent. Cette méthode de fabrication nous a permis d’obtenir une
épaisseur fine et contrôlable. Par rapport à de précédents travaux, des modifications notables
ont été apportées : a) l’optimisation des électrodes de PEDOT entraînant l’augmentation du
temps de polymérisation en phase vapeur de 30 à 50 minutes, b) l’utilisation d’une plaque
chauffante sous atmosphère contrôlée (Sawatec HP-200) permettant un contrôle plus précis
de la température et du vide pendant le polymérisation en phase vapeur, c) tout le processus

vii
de fabrication a été effectué dans une salle blanche, ce qui garantit une reproductibilité
accrue. Des lots d'échantillons tri couches ont été fabriqués et les mesures en déformation
ont montré que les actionneurs fonctionnaient correctement. Le pourcentage de PEO dans la
couche PEDOT est passé également à 10% par rapport aux 20% utilisés dans les travaux de A.
Maziz (Maziz 2014). Ce pourcentage assure un équilibre entre une conductivité électronique
élevée (environ 200 S / cm) et une densité de charge volumétrique élevée (entre 2,3 x 10 7 C
/ m3 et 1,0 x 108 C / m3) de la couche PEDOT. De plus, nous avons mis en évidence une légère
asymétrie de rugosité de surface entre les couches PEDOT supérieure et inférieure (figure 2)
qui a des conséquences sur les propriétés électriques des électrodes PEDOT et enfin sur les
propriétés mécaniques (déplacement, force…) de l'actionneur.

Fig.2 a) Image MEB de la coupe transversale de la tri couche non gonflé de liquide ionique, rugosité de
la surface par profilomètrie de la surface supérieure de l'électrode supérieure, c) de la surface
supérieure du polymère support d’électrolyte (NBR/POE) sur un bicouche PEDOT/SPE, d) de la surface
supérieure de l'électrode inférieure, e) de la surface inférieure de l'électrode inférieure
Le procédé de polymérisation en phase vapeur de cette thèse utilise la solution oxydante
conventionnelle constituée de 55% de Fe(TOs)3 dans du butanol. Cependant, la littérature a
montré que ce pourcentage élevé produit une solution oxydante fortement acide et réactive,
entraînant une polymérisation difficilement contrôlable et des défauts de structure du film
déposé (Shi, et al. 2017). Des recherches sur l'optimisation du pourcentage de solution
oxydante, ou l'ajout d'inhibiteurs de bases tels que la pyridine pour contrôler la vitesse de
réaction, sont suggérées pour améliorer encore les propriétés électrochimiques de l'électrode
en PEDOT. Ceci est envisagé dans le futur.

viii
Deuxièmement, le transducteur à base de polymère conducteur résultant a été caractérisé.
Ses propriétés électrochimiques, notamment la conductivité ionique des électrodes de
PEDOT et de la couche polymère support d’électrolyte ainsi que la conductivité électronique
du PEDOT en fonction de l’état d’oxydation et la capacité volumétrique du PEDOT avec sa
dépendance en tension et avec la vitesse de balayage ont toutes été étudiées. De plus, les
propriétés mécaniques telles que le module de Young du PEDOT en fonction de l’état
d’oxydation, le module de Young de la couche polymère support d’électrolyte, le taux
d’amortissement ont également été mesurés. Enfin, le couplage entre l'électrochimique et la
mécanique représentant via un rapport empirique déformation/charge a été déterminé.
Le tableau ci-dessous résume les propriétés électriques, électrochimiques et mécaniques
critiques des actionneurs tri couches. Ces valeurs obtenues permettent d'évaluer le modèle
dynamique non linéaire établi et seront discutées dans le chapitre suivant.
En outre, ces paramètres obtenus nous fournissent un aperçu de la structure physique du
système et du mécanisme de travail de l'actionneur. Plus important encore, les facteurs clés
ayant une incidence sur le comportement d’actionnement sont tous déterminés, ce qui nous
fournit les meilleures indications pour améliorer les performances de l’actionneur.
Table: Résumé des paramètres mesurés de l'actionneur ultramince tri couches
Conductivité électronique du Electrode supérieur de PEDOT:
PEDOT (S/cm) -2.62U3 + 2.76U2 + 9.89U + 7.49
Electrode inférieur de PEDOT:
-16.6U3 + 1.1U2 + 51.7U + 59.1
Conductivité ionique des Electrode supérieur de PEDOT: 0.025
électrodes de PEDOT (S/m) Electrode inférieur de PEDOT: 0.030
Capacité volumétrique des 1.1 × 106 𝑈 6 − 1.9 × 107 𝑈 5 + 1.1 × 108 𝑈 4
électrodes de PEDOT (F/m3) − 2.8 108 𝑈 3 + 3.4 × 108 𝑈 2 − 1.7
× 108 𝑈 + 6.9 × 107
Conductivité ionique du 0.038
NBR/PEO (S/m)
Rapport déformation/charge 𝛼 = 3.6 ± 0.5 × 10−10
(m3/C)
Module Young du NBR/PEO 329 ± 50
(kPa)
Module Young du tri couche −0.09U3 − 0.08U2 − 0.26U + 9.54
(MPa)
Module Young du PEDOT −0.23U3−0.19U2− 0.63U + 22.70
(MPa)
Amortissement 0.39
3
Densité du matériau (kg/m ) 2000

ix
 Différence de déformation 0.56
maximum (%)
1
Déformation linéaire (%)

Troisièmement, un modèle dynamique non linéaire et un modèle linéaire dynamique plus

simple ont été proposés pour simuler l'actionnement de ces matériaux tri couches ultra-
minces. Une comparaison entre ces deux modèles a montré que le modèle linéaire est plus
simple à construire et que la caractérisation des paramètres requis est simple aussi.
Cependant, il est plus approprié pour prédire de petits déplacements et l’on a des pertes de
précision pour des déplacements et des tensions importants. Le modèle Bond Graph par mot
pour les modèles linéaire et non linéaire est présenté dans la figure 3.

Fig.3 Actionneur tri couches modèle Bond Graph par mots.

Le modèle non linéaire proposé a montré des avancées par rapport aux travaux existants sur
les actionneurs tri couches fonctionnant dans l’air en raison du fait qu'il permet de prédire
avec précision un actionnement dynamique non linéaire en utilisant des propriétés
électrochimiques et mécaniques non linéaires des matériaux: conductivités électriques et
ioniques du PEDOT; valeurs de capacité volumétrique; et la variation du module de Young. De
plus la prise en compte de la masse et l’amortissement associés à la poutre est aussi une
nouveauté. De plus, toutes les caractéristiques physiques utilisées dans le modèle ont été
mesurées. La figure 4 représente le modèle dynamique global non linéaire de l'actionneur à
base de polymère conducteur.

x
Fig.4 Modèle global de l'actionneur à base de polymère conducteur: 1. Source de tension, 2. Modèle
électrochimique, 3. Couplage entre partie électrochimique et partie mécanique, 4. Modèle mécanique,
5. Conditions aux limites: contrainte initiale, 6. Déplacement de sortie.
La concordance entre la simulation et les résultats expérimentaux dans les domaines
temporel et fréquentiel (illustrés dans la figure 5) confirme la capacité de prédire la flexion de
l'actionneur tri couches.

xi
Fig.5 Comparaison entre les simulations et les expériences sur a) La déformation en fonction de
l’amplitude de la tension appliquée à une fréquence constante de 0,1 Hz, b) La déformation en fonction
de la fréquence de la tension appliquée à une valeur constante de ± 1,1 V c) L'erreur relative des
simulations comparée aux données mesurées.
De plus, l'approche représentée en langage Bond Graph, offre un moyen simple d'accéder aux
informations énergétiques (figure 6), d'évaluer les paramètres critiques affectant les
performances de l'actionneur et de mieux comprendre les phénomènes qui se produisent
dans l'actionneur.

Fig.6 Couplage électrique / mécanique instantané.

xii
De plus, le modèle peut également être utilisé pour optimiser la conception du transducteur
avant sa fabrication. Le logiciel de simulation permet de faire varier les paramètres d’entrée,
tels que les dimensions et les propriétés du matériau (conductivités, module de Young,…) sur
des plages définies, afin de définir celle qui produira la réponse souhaitée. Ces valeurs
serviront de guide au concepteur pour ajuster les propriétés du transducteur. L'outil peut
également être utilisé pour déterminer les propriétés des matériaux qui limitent le plus la
réponse afin de piloter le développement des matériaux ou de permettre l'évaluation de
matériaux émergents. La figure ci-dessous donne une évaluation de la sensibilité à la
différence de déformation, au module de Young, à la conductivité électrique PEDOT, à la
conductivité ionique du polymère support d’électrolyte et à la longueur de l'actionneur. Les
résultats sont montrés pour une basse fréquence (Fig. 7a) et une haute fréquence (Fig. 7b).

Fig.7 Facteurs clés affectant le déplacement de l’actionneur sous une tension carrée (Vpp = 2,4 V), a) à
basse fréquence f = 0,1 Hz, et b) à haute fréquence f = 16 Hz. Pour chaque paramètre il y a trois colonnes
qui, de gauche à droite représentent la valeur réelle (colonne bleue), l’effet lorsque la valeur est
augmentée de 20% (colonne orange) et l’effet lorsque la valeur est réduite de 20% (colonne grise).

xiii
Par ailleurs, une relation simple entre une déformation manuelle et la tension de sortie aux
bornes du transducteur a été proposé dans un modèle linéaire dans le but de créer un
modèle unique permettant de fonctionner à la fois en mode d'actionnement et en mode de
détection. Le rapport déformation/charge a été recaractérisé et la concordance entre la
simulation et la mesure expérimentale de la tension de sortie a confirmé la capacité de
prédire la fonction de détection de ces matériaux tri couche, malgré une petite erreur. Le
modèle montre sa capacité à prédire assez bien la modification de tension électrique sous
un déplacement pas à pas.

Conclusion
Les développements dans les transducteurs à base de polymères conducteurs opérant dans
l'air ont permis de les rendre, entre autres, plus efficaces en termes de force générée, de
déplacement et de fréquence de fonctionnement. Récemment, ils ont été fabriqués avec des
épaisseurs acceptables pour une intégration potentielle dans des microsystèmes. Le but
ultime est de disposer d’un matériau dont les propriétés sont bien connues et qui puisse être
modélisé afin d’être prédictif dans la fabrication d’un microsystème souple comprenant des
transducteurs souples. C'est pourquoi cette thèse s'est concentrée sur la caractérisation et la
modélisation d’un certain type de transducteurs à base de polymères conducteurs. À notre
connaissance, il s'agit du premier effort visant à caractériser et à modéliser de manière
complète un tel transducteur tri couche très mince, dont l'épaisseur totale n'est que de 17
µm - après avoir été gonflé dans un liquide ionique.

La première étape de ce travail consistait à améliorer un processus de microfabrication

compatible avec les salles blanches récemment mis au point pour les couches conductrices
en polymère. Bien que le rendement de ces matériaux reste faible avec ce type de tri couche,
les caractéristiques physiques pertinentes d’un transducteur performant ont été entièrement
caractérisées, notamment les facteurs affectant le transport ionique et électronique, le
couplage charge / déformation, les rigidités des matériaux et les amortissements. Lors de
l'actionnement, une différence de déformation d'environ 1%, une force de blocage de 12 µN
et une tension de détection de 0,7 mV ont été obtenues à partir d'un transducteur de
longueur x largeur = 6 mm x 1 mm. Cette déflexion et cette tension de détection
correspondent à la réponse attendue et qui a été aussi validée par un modèle non linéaire
mis en place dans le cadre de cette thèse.

Le polymère conducteur PEDOT a été utilisé comme matériau actif en constituant les deux
électrodes du transducteur. Étant donné que la polymérisation du 3,4-
éthylènedioxythiophène (EDOT) est un processus complexe et sensible et que le PEDOT est
un matériau présentant des propriétés électro-chimio-mécanique non linéaires, les processus
de fabrication, de caractérisation et de modélisation ont été confrontés à divers problèmes.
La résolution de ces problèmes a conduit aux principales contributions scientifiques et
techniques décrites ci-dessous:

xiv
- L'amélioration du procédé de fabrication des transducteurs. Dans la configuration à trois
couches, un semi-NBR / PEO est pris en sandwich entre deux électrodes PEDOT / PEO,
produites par empilement couche par couche. Cette technique de fabrication permet
d’ajuster l’épaisseur de chaque couche à notre guise. En outre, a été étudié l'asymétrie de
rugosité de surface entre les couches PEDOT d’électrodes supérieure et inférieure, ce qui a
des conséquences sur les propriétés électriques des électrodes de PEDOT et, enfin, sur les
propriétés mécaniques (déplacement, force…) de l'actionneur.

- La caractérisation complète des propriétés électro-chimio-mécanique critiques des

transducteurs tri couches a été effectuée. Les propriétés électrochimiques incluant la
conductivité ionique des électrodes PEDOT et de la couche de polymère support d’électrolyte,
la conductivité électronique du PEDOT en fonction de son état d’oxydation et la dépendance
de la capacité volumétrique du PEDOT sur la fenêtre de potentielle ont toutes été étudiées.
En outre, les propriétés mécaniques telles que le module de Young du PEDOT en fonction de
l’état d’oxydation, le module de Young de la couche polymère support d’électrolyte et le taux
d’amortissement ont aussi été mesurés. Enfin, le couplage électrochimique et mécanique a
été évalué via un rapport empirique déformation-charge, la déformation linéaire a été
déterminée et la force de blocage mesurée en fonction du potentiel appliqué.

- Le développement d'un modèle dynamique non linéaire - et d'une version beaucoup plus
simple avec un modèle linéaire dynamique - capable de simuler l'actionnement de tri couches
ultra-minces. Une comparaison entre ces deux modèles a montré que le modèle linéaire est
simple à construire et qu'il est plus facile d'obtenir tous les paramètres requis. Il est plus
approprié de prévoir dans ce cas de simuler de petits déplacements Il fournit une prévision
moins précise que le modèle non linéaire pour des tensions appliquées plus importants
produisant de grands déplacements. Le modèle non linéaire proposé montre des avancées
intéressantes par rapport aux travaux existants sur les actionneurs tri couches en raison de sa
capacité à prédire avec précision un actionnement dynamique non linéaire prenant en
compte des propriétés électrochimiques et mécaniques non linéaires du matériau:
conductivités électriques et ioniques du PEDOT; valeurs de capacité volumétrique; et la
variation du module de Young. De plus, le modèle utilisé prend en compte aussi la masse et
l’amortissement associés à la micro poutre de transducteur. Toutes les caractéristiques
physiques utilisées dans le modèle ont été mesurées. La concordance entre la simulation et
les résultats expérimentaux dans les domaines temporel et fréquentiel confirme la capacité
de prédire la flexion de l'actionneur tri couches. En outre, l'approche est représentée dans un
langage Bond Graph, offrant un moyen simple d'accéder aux informations énergétiques,
d'évaluer les paramètres critiques affectant les performances de l'actionneur et de mieux
comprendre les phénomènes qui se produisent dans l'actionneur ionique.

- Le développement d'un modèle linéaire unique fonctionnant à la fois en mode

d'actionnement et en mode de détection. Une relation simple entre la déformation

xv
provoquée mécaniquement et la tension de sortie a été utilisé dans le modèle simple linéaire.
Le rapport déformation-charge a été re caractérisé en mode de détection et la concordance
entre la simulation et la valeur expérimental de la tension de sortie a confirmé la capacité de
prédire la fonction de détection tri couche, malgré une erreur faible. Une autre avancée de
ce modèle est la possibilité de prédir le taux de décroissance de la tension de détection sous
un déplacement pas à pas.

References

Ateh, D. D., H. A. Navsaria, and P. Vadgama

2006 Polypyrrole-based conducting polymers and interactions with biological
tissues. J R Soc Interface 3(11):741-52.
Baughman, R. H.
1996 Conducting polymer artificial muscles. Synthetic Metals 78(3):339-353.
Baughman, R. H., et al.
1991 Micro Electromechanical Actuators Based on Conducting Polymers. In
Molecular Electronics: Materials and Methods. P.I. Lazarev, ed. Pp. 267-289.
Dordrecht: Springer Netherlands.
Cellot, G., et al.
2015 PEDOT:PSS Interfaces Support the Development of Neuronal Synaptic
Networks with Reduced Neuroglia Response In vitro. Front Neurosci 9:521.
Chiang, C. K., et al.
1977 Electrical Conductivity in Doped Polyacetylene. Physical Review Letters
39(17):1098-1101.
Della Santa, A., A. Mazzoldi, and D. De Rossi
1996 Steerable microcatheters actuated by embedded conducting polymer
structures. Journal of Intelligent Material Systems and Structures 7(3):292-300.
Jager, E.W., Immerstrand, C., Peterson, K.H. et al.
2002 The Cell Clinic: Closable Microvials for Single Cell Studies. Biomedical
Microdevices 4(3):10.
Kirchmeyer, Stephan, and Knud Reuter
2005 Scientific importance, properties and growing applications of poly(3,4-
ethylenedioxythiophene). Journal of Materials Chemistry 15(21):2077-2088.
Maziz, A., et al.
2016a Top-down Approach for the Direct Synthesis, Patterning, and Operation of
Artificial Micromuscles on Flexible Substrates. ACS Appl Mater Interfaces 8(3):1559-
64.
Maziz, Ali
2014 Microactionneurs à base de polymères conducteurs électroniques : Vers
l’intégration aux microsystèmes par de nouveaux procédés d’élaboration,
Laboratoire de Physico-chimie des Polymères et des Interfaces, Université de Cergy
Pontoise.
Maziz, Ali, et al.
2016b Top-down Approach for the Direct Synthesis, Patterning, and Operation of
Artificial Micromuscles on Flexible Substrates. ACS Applied Materials & Interfaces
8(3):1559-1564.

xvi
Maziz, Ali, et al.
2014 Demonstrating kHz Frequency Actuation for Conducting Polymer
Microactuators. Advanced Functional Materials 24(30):4851-4859.
Mirfakhrai, Tissaphern, John D. W. Madden, and Ray H. Baughman
2007 Polymer artificial muscles. Materials Today 10(4):30-38.
Otero, T. F., et al.
1992 Electrochemomechanical properties from a bilayer: polypyrrole / non-
conducting and flexible material — artificial muscle. Journal of Electroanalytical
Chemistry 341(1):369-375.
Otero, Toribio F.
2013 Biomimetic Conducting Polymers: Synthesis, Materials, Properties, Functions,
and Devices. Polymer Reviews 53(3):311-351.
Pei, Qibing, and Olle Inganäs
1993 Electrochemical applications of the bending beam method; a novel way to
study ion transport in electroactive polymers. Solid State Ionics 60(1):161-166.
Richardson-Burns, S. M., et al.
2007 Polymerization of the conducting polymer poly(3,4-ethylenedioxythiophene)
(PEDOT) around living neural cells. Biomaterials 28(8):1539-52.
Runyan, N. H., and F. Carpi
2011 Seeking the 'holy Braille' display: might electromechanically active polymers
be the solution? Expert Rev Med Devices 8(5):529-32.
Shi, Wei, et al.
2017 Micron-thick highly conductive PEDOT films synthesized via self-inhibited
polymerization: roles of anions. Npg Asia Materials 9:e405.
Shirakawa, Hideki, et al.
1977 Synthesis of electrically conducting organic polymers: halogen derivatives of
polyacetylene, (CH). Journal of the Chemical Society, Chemical Communications
(16):578-580.
Smela, E.
2003 Conjugated polymer actuators for biomedical applications. Advanced
Materials 15(6):481-494.
Spinks, G. M., et al.
2003 Ionic Liquids and Polypyrrole Helix Tubes: Bringing the Electronic Braille
Screen Closer to Reality. Proceedings of SPIE - The International Society for Optical
Engineering, 2003. Vol. 5051, pp. 372-380.
Vidal, Frédéric, et al.
2006 Conducting IPN actuators: From polymer chemistry to actuator with linear
actuation. Synthetic Metals 156(21–24):1299-1304.
Vidal, Frédéric, et al.
2004 Long-life air working conducting semi-IPN/ionic liquid based actuator.
Synthetic Metals 142(1):287-291.

xvii
Lay Summary
This work constitutes the most thorough characterization and modeling to date of ionic
conducting polymer-based actuators, applied to PEDOT microactuators operating in air. Non-
linear characterization was extended to volumetric capacitance dependence on voltage
window. Damping coefficient was characterized for the first time. These and other measured
properties were included in a nonlinear multi-physics model that is demonstrated as an
effective method for simulating actuation in trilayers. In addition, a new bidirectional
electromechanical linear model was introduced to simulate both the actuation and sensing
ability of the trilayer transducer.

xviii
Preface

This dissertation is formatted in accordance with the regulations of the University of

Polytechnique Haut-de-France and submitted in partial fulfillment of the requirements for a
PhD degree awarded jointly by the University of Polytechnique Haut-de-France and the
University of British Columbia. Versions of this dissertation will exist in the institutional
repositories of both institutions.

All aspects of the material appearing in this thesis have been originally written by the author
unless otherwise stated.

This work has been done in the IEMN-DOAE laboratory and in Molecular Mechatronics Lab
under the supervision of Prof. Eric Cattan, Prof. Sébastien Grondel, and Prof. John D. W.
Madden.

A version of chapter 2 has been published. [T.N. Nguyen], K. Rohtlaid, C. Plesse, G.T.M.
Nguyen, C. Soyer, S. Grondel, E. Cattan, J.D.W. Madden, F. Vidal, Ultrathin electrochemically
driven conducting polymer actuators: fabrication and electrochemomechanical
characterization, Electrochimica Acta, 265(2018) 670-80. All of the fabrication and
characterization have been performed with the supervision of Prof. Eric Cattan, Prof. John D.
W. Madden, and Prof. Frédéric Vidal. Dr. Cédric Plesse, Dr. Giao T.M. Nguyen, Dr. Caroline
Soyer, Prof. Sébastien Grondel helped to reviewed the results and revise the manuscript. I
conducted the PEDOT-based trilayer fabrication process and the trilayer characterization
including geometries and surface roughness, electrochemical and mechanical properties. The
sections on “PEDOT electrode fabrication” and “Optimization of electrochemical properties
of PEDOT electrodes” was written by K. Rohtlaid.
A version of chapter 3 and 4 has been submitted. [T.N. Nguyen], Y. Dobashi, C. Soyer, C. Plesse,
G.T.M. Nguyen, F. Vidal, E. Cattan, S. Grondel, J.D.W. Madden, Non-linear dynamic modeling
of ultrathin conducting polymer actuators including inertial effects, Smart Materials and
Structures, May 2018. All the experiments and simulations were conducted by the author
under the supervision of Prof. Eric Cattan, Prof. John D. W. Madden, and Prof. Sébastien
Grondel. Yuta Dobashi helped to setup the experiments.
The section 4.3 of chapter 4 was presented SPIE conference on electroactive polymer
actuators and devices, 2017 (N.T. Nguyen, C. Plesse, F. Vidal, C. Soyer, S. Grondel, J.D.W.
Madden, E. Cattan, Microfabricated PEDOT trilayer actuators: synthesis, characterization, and
modeling, SPIE Smart Structures and Materials + Nondestructive Evaluation and Health
Monitoring, SPIE2017, p. 13).

xix
Table of Contents
Abstract ..................................................................................................................................... iii
Résumé...................................................................................................................................... iv
Résumé Substantiel ................................................................................................................... v
Lay Summary ..........................................................................................................................xviii
Preface .................................................................................................................................... xix
Table of Contents ..................................................................................................................... xx
List of Figures .........................................................................................................................xxiii
List of Tables .......................................................................................................................... xxix
Abbreviations ..........................................................................................................................xxx
Acknowledgements................................................................................................................ xxxi
Dedication ............................................................................................................................. xxxii
Chapter 1: Introduction ................................................................................................... 1
1.1 Mammalian muscles .................................................................................................... 2
1.2 Artificial muscles .......................................................................................................... 2
1.3 Motivation and problem statement ............................................................................ 9
1.4 Thesis structure............................................................................................................ 9
Chapter 2: PEDOT-based trilayer fabrication process ....................................................... 16
2.1 Introduction ............................................................................................................... 17
2.2 The selection of materials for CP-based trilayer actuators ....................................... 22
2.2.1 Electrodes of the microactuators ....................................................................... 22
2.2.2 Solid polymer electrolyte layer ........................................................................... 22
2.2.3 The electrolyte .................................................................................................... 24
2.2.4 Microactuator fabrication technique ................................................................. 26
2.3 Materials .................................................................................................................... 26
2.4 PEDOT synthesis route ............................................................................................... 27
2.5 PEDOT-based trilayer fabrication process ................................................................. 30
2.5.1 Trilayer fabrication process ................................................................................ 30
2.6 PEDOT-based trilayer patterning ............................................................................... 33
2.6.1 Fabrication of samples for the characterization process ................................... 35
2.7 Analysis of the texture of the trilayer structure ........................................................ 36
2.8 Conclusion .................................................................................................................. 39
Chapter 3: Electrochemomechanical characterization of the trilayer structure ................ 45

xx
 3.1 Introduction ............................................................................................................... 46
3.2 Electro-chemical properties ....................................................................................... 48
3.2.1 Ionic conductivity of the SPE and PEDOT layers ................................................. 49
3.2.2 Electrical conductivity of the PEDOT electrodes ................................................ 52
3.2.3 Volumetric capacitance of the PEDOT electrodes .............................................. 55
3.2.4 Possible short circuit between two PEDOT layers .............................................. 58
3.3 Mechanical properties ............................................................................................... 59
3.3.1 Youngs moduli of the SPE layer and of the trilayer actuator ............................. 59
3.3.2 Damping ratio ..................................................................................................... 62
3.3.3 Blocking force characterization .......................................................................... 63
3.4 Empirical strain-to-charge ratio ................................................................................. 64
3.4.1 Strain to charge ratio .......................................................................................... 64
3.4.2 Linear strain ........................................................................................................ 67
3.5 Conclusion .................................................................................................................. 69
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior ......................................................................................................... 73
4.1 Motivation.................................................................................................................. 74
4.1.1 Objectives............................................................................................................ 74
4.1.2 Proposed methodology....................................................................................... 74
4.2 State of art ................................................................................................................. 74
4.2.1 A summary of Black box, white box, grey-box models for CP actuators ............ 75
4.2.1.1 Black-box model............................................................................................... 75
4.2.1.2 Grey-box model ............................................................................................... 75
4.2.1.3 White-box model ............................................................................................ 78
4.2.2 Why the choice of the Bond Graph language? ................................................... 80
4.3 Dynamic Bond Graph modeling ................................................................................. 84
4.3.1 Actuation description.......................................................................................... 84
4.3.2 Word Bond Graph model .................................................................................... 86
4.3.3 BG submodels ..................................................................................................... 87
4.3.3.1 Electrical model................................................................................................ 87
4.3.3.2 Electromechanical coupling ............................................................................. 92
4.3.3.3 Mechanical model............................................................................................ 93
4.3.4 BG global models ................................................................................................ 99
4.4 Simulation tests ....................................................................................................... 101

xxi
 4.4.1 Software implementation ................................................................................. 101
4.4.2 Comparison between the linear and nonlinear simulations ............................ 102
4.5 Comparison to experimental tests .......................................................................... 106
4.5.1 Time domain responses .................................................................................. 106
4.5.2 Frequency responses ....................................................................................... 110
4.6 Parameter sensitivity and power performance analysis ......................................... 113
4.7 Conclusion ................................................................................................................ 117
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators ..... 112
5.1 Introduction ............................................................................................................. 123
5.2 Theories on the mechanoelectrical effects ............................................................. 123
5.3 Sensing modeling ..................................................................................................... 129
5.3.1 Mechanoelectrical coupling ........................................................................... 129
5.3.2 Global model .................................................................................................. 131
5.4 Experimental setup .................................................................................................. 131
5.5 Results ...................................................................................................................... 134
5.5.1 Comparison between the model simulation and experimental results ......... 134
5.6 Discussion................................................................................................................. 139
5.6.1 Force in response to a step displacement ..................................................... 141
5.7 Conclusion ................................................................................................................ 142
Chapter 6: Conclusion and outlook ............................................................................... 145
Appendix...................................................................................................................... 149
A.2 Chapter 2: PEDOT-based trilayer fabrication process ................................................... 149
A.2.1 Optimization of electrochemical properties of PEDOT electrodes ........................ 149
A.2.2 Surface measurement method ............................................................................... 153
A.3 Chapter 3: Electrochemomechanical characterization of the trilayer structure ........... 154
A.3.1 Qualitative explanation the apparent capacitance of the PEDOT electrodes at
extreme low scan rate ..................................................................................................... 154
A.4 Chapter 4: Linear dynamic and non-linear dynamic model to predict PEDOT-based trilayer
actuation behavior ............................................................................................................... 155
A.4.1 The coupling matrix derivation method ................................................................. 155
A.5 Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators . 156
A.5.1 A possible qualitative sensing model ..................................................................... 156

xxii
List of Figures
Chapter 1: Introduction
Fig. 1. Dependence of blocking stress on actuation strain for various artificial muscles.
Abbreviations: PVDF - piezoelectric polymer [32], PZT - piezoelectric ceramic [33], TE 100K -
Thermal expansion actuators 100K [34], SMA – shape memory alloy [35], CP – conducting
polymer [36], IPMC – ionic polymer metal composite [37], PZT-PT – piezoelectric single
crystal [33], P(VDF-TrFE) – relaxor ferroelectric polymer [38], DEA – dielectric elastomer
[35], PAM – pneumatic artificial muscle [19], Gels – Gel actuators [39], NM – natural muscle
[1]. .............................................................................................................................................. 4
Fig. 2. Dependence of specific work on actuation strain for various artificial muscles. ........... 5
Fig. 3. Chemical structures of some typical conducting polymers. ........................................... 8

Chapter 2: PEDOT-based trilayer fabrication process

Fig. 1. Actuation mechanism of conducting polymer actuators: a) anion driven mechanism,
b) cation driven mechanism [6] (Reprinted with permission from Springer Nature). The ovals
represent positive electronic charge on the backbone. .......................................................... 17
Fig. 2. Cation driven actuation in a PEDOT-based actuator..................................................... 18
Fig. 3. Bending actuation mechanism of PEDOT-based trilayer actuators. The anode
contracts due to ion expulsion and the cathode expands due to cation incorporation, leading
to bending. ............................................................................................................................... 18
Fig. 4. A possible model taking into account two electrochemical doping process: some ions
(II), deeply trapped near the polymer chains are released at only very low potentials. Other
ions (I) are shallowly trapped, they form a double layer at a certain distance from the chain.
(Reproduced with permission from [10], The Electrochemical Society.) ................................ 19
Fig. 5. a) Schematic of a micropump driven by a PPy-based CP actuator [12] (Reprinted with
permission from John Wiley and Sons), b) A diagram pump: the mechanism of flap check
valves and a bottom view of the assembled micropump [13] (Reprinted with permission
from Elsevier), c) Micro autofocus lens actuator that uses bending conducting polymer
actuators [14] (Reprinted with permission from Elsevier), d) A prototype of a robotic fish
[15] (Reprinted with permission from IOP Publishing) ............................................................ 20
Fig. 6. a) A schematic drawing of a closable microvials [16] (Reprinted with permission from
Springer Nature), b) A sequence of pictures (left) showing the grabbing and lifting of a 100
µm glass bead and schematic drawings of the motion (right) [17] (Reprinted with permission
from The American Association for the Advancement of Science) ......................................... 21
Fig. 7. Schematic of a) full-IPN and b) semi-IPN where the pink, blue, and empty dot
represent for the crosslinking between two polymer chains. ................................................. 23

xxiii
Fig. 8. a) properties of ionic liquids incorporated in NBR/PEO matrix, and b) strain
differences calculated for PEDOT-NBR/PEO-PEDOT actuator swollen with different ionic
liquids [42] (Reprinted with permission from Elsevier). .......................................................... 25
Fig. 9. Molecular formula of a) the precursor PEGM, b) PEDGM, c) the linear chain NBR, d)
the initiator DCPD, and e) the ionic liquid EMITFSI [46] (Reprinted with permission from
American Chemical Society, 2008.) ......................................................................................... 27
Fig. 10. Oxidation of EDOT with iron (III) p-toluene sulfonate ................................................ 28
Fig. 11. Fabrication process of PEDOT electrodes [51] ............................................................ 28
Fig. 16. Relation between the rotational speed of the spin-coater for deposition of the
oxidant solution on a silicon wafer and the thickness of the PEDOT electrode, where the
mPEG content is 10% of oxidant solution................................................................................ 30
Fig. 13. Microactuator fabrication process showing the multilayer process followed by laser
micromachining ....................................................................................................................... 31
Fig. 14. Relationship between rotation speed of spin-coater for deposition of NBR/PEO and
its resulting thickness............................................................................................................... 32
Fig. 15. Trilayer structures during washing step in methanol, the rotational coating speeds of
PEDOT:NBR/PEO:PEDOT are a) 1000:2000:1000 b) 1500:2000:1500 c) 2500:2000:2500 d)
3000:2000:3000; the acceleration and coating duration are constant at 1000 rpm s-1 and 30
s, respectively. Les coloration is seen during washing for the fast spun, thin layers of oxidant
solution, indicating less weakly polymerized EDOT. ............................................................... 33
Fig. 16. a) The wafer after drying process, the white trace on the top right indicating the
area has been peeled off for characterization: EDX and SEM, b) the resulting trilayer
actuator after laser patterning process, the dimensions of the trilayer are length x width = 5
mm x 1 mm c) SEM images of the trilayer after swollen in EMITFSI ....................................... 35
Fig. 17. a) The EDX of the trilayer structure, b) The anticipated morphology of the trilayer
structure PEDOT:NBR/PEO:PEDOT .......................................................................................... 37
Fig. 18. a) SEM image of the trilayer cross section in dry state, profilometer surface
roughness of the b) top surface of the top electrode layer, c) top surface of the SPE of a
bilayer PEDOT/SPE, d) top surface of the bottom electrode layer, e) bottom surface of the
bottom electrode layer ............................................................................................................ 38

Chapter 3: Electrochemomechanical characterization of the trilayer structure

Fig. 1. Actuator behavior under a step voltage a) The applied step voltage and the response
current, b) Bending response of the actuator, c) A picture of the bending actuator (length x
width x thickness = 4 mm x 1 mm x 0.017 mm), the + and – indicate the polarity of the
applied voltage......................................................................................................................... 47
Fig. 2. a) EIS measurement using 4-point setup, b) NBR/PEO layer, bottom bilayer, and top
bilayer configuration, and c) a zoomed image of an area in the red rectangular A-A and its
equivalent circuit showing the current direction at low and high operating frequency. ....... 50

xxiv
Fig. 3. Impedance values obtained from EIS measurements of a pure EMImTFSI, a top bilayer
of NBR/PEO-PEDOT, a bottom bilayer of PEDOT-NBR/PEO, and a single NBR/PEO layer. ..... 51
Fig. 4. Three-electrode system for setting the oxidation state of the PEDOT and for cyclic
voltammetry analysis. .............................................................................................................. 53
Fig. 5. A built in-house four-line probe to measure the electronic conductivity of the PEDOT
electrodes, where the distance between two copper lines is d = 2 mm. ................................ 54
Fig. 6. PEDOT electrical conductivity as a function of oxidation state: a) on the bottom
PEDOT surface of the trilayer structure, b) on the top PEDOT surface of the trilayer structure
.................................................................................................................................................. 55
Fig. 7. a) Cyclic voltammograms of PEDOT electrodes obtained in neat EMImTFSI within a
potential window of -0.6 V to +0.7 V with different scan rates: b) Volumetric capacitance of
PEDOT electrodes as function of the scanning rate. ............................................................... 56
Fig. 8. Volumetric capacitance as a function of the voltage window from 0.5 Vpp to 3.3 Vpp.
.................................................................................................................................................. 57
Fig. 9. The current response to an input step voltage at low frequency (f = 0.005 Hz) .......... 59
Fig. 10. Young’s moduli of the NBR-PEO layer alone - and of a trilayer structure as a function
of oxidation state - were measured using a Bose Electroforce 3000 dynamic mechanical
analyzer .................................................................................................................................... 60
Fig. 11. Young’s modulus of the trilayer actuator and the PEDOT layer as a function of
applied voltage......................................................................................................................... 61
Fig. 12. Experimental setup to measure the damping ratio of the trilayer actuator and the
blocking force ........................................................................................................................... 62
Fig. 13. Beam vibrating as function of time. ............................................................................ 62
Fig. 14. Force generated as a function of the magnitude and the frequency of the applied
voltage from a trilayer actuator of length x width x thickness: 6 mm x 1 mm x 0.017 mm.... 63
Fig. 15. Actuator at the a) neutral state where the bending radius is equal r 0 = 12 mm, and
b) excited state where the bending radius is rt = 2.8 mm. The orange square on the
background of the pictures has the dimension of 1 x 1 mm. .................................................. 64
Figure 16. Current and charge response to a triangular input voltage. .................................. 65
Fig. 17. Charge density and strain of the trilayer actuator as a function of time under a
triangular wave voltage excitation .......................................................................................... 65
Fig. 18. Uniaxial deformation of the trilayer actuator under a step voltage ........................... 67

Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
Fig. 1. a) The equivalent electrical circuit of the diffusive elastic model, which includes: a
double layer capacitance C, electrolyte resistance R, and the impedance associated with
ionic charge transport dominated by diffusivity ZD, b) Example of an equivalent electrical
circuit including a finite diffusion element Cjk and Rijk, a resistance Rej, and capacitance at the
interface between the solution and the polymer ................................................................... 76

xxv
Fig. 2. a) a CP actuator in neutral state and its dimensions (shown in the magnified section in
the middle), and b) a voltage excitation induces a tips displacement w(L, t) and an angular
rotating angle 𝜃(𝑥, 𝑡) of the actuator, which is equivalent to an external force F(L, t) or a
moment M(x, t), respectively. The Young’s moduli of the PEDOT electrodes and the
NBR/PEO layer are EP and ES, respectively. ............................................................................. 85
Fig. 3. Trilayer actuator Words BG model................................................................................ 86
Fig.4. a) 2D diffusion transmission line across the thickness and the length of a trilayer
actuator, and simplified electrochemical model of the trilayer actuator for b) nonlinear
model, and c) linear model ...................................................................................................... 90
Fig. 5. BG representation of a) the nonlinear electrochemical model, b) the linear
electrochemical model. ........................................................................................................... 91
Fig. 6. Coupling between electrical sub-system and mechanical sub-system. ........................ 92
Fig. 7. Rigid Finite Element Model (RFEM) of a trilayer actuator, in which it is divided into
(n+1) rigid elements: from RFE0 to RFEn, the elements each have a mass and are linked by a
spring-damper element SDE1 to SDEn, shown in their initial positions. The integrated
coordinator includes X0 axis along the length of the actuator, X1 axis perpendicular to the X0
and X2 axis perpendicular to the X0X1 surface. ........................................................................ 94
Fig. 8. Trilayer actuator in a bent state. The bending angle between the element (i-1) and the
element i is θi........................................................................................................................... 95
Fig. 9. Mechanical BG model.................................................................................................... 97
Fig. 10. BG representation of the mechanical model of the trilayer actuator, the block 1 (red
color) indicates the point force F(L, t) obtaining from the electromechanical coupling matrix
applying to the free end of the beam, the block 2 (green color) presents the boudary
condition where one end of the beam is fixed, the block 3 (pink color) is a BG presentation
of the equation (25), and finally the block 4 (orange color) introduces the method to extract
the tips displacement of the beam w(L, t). .............................................................................. 99
Fig. 11. CP actuator global model: 1. Voltage source, 2. Electrochemical model, 3. Coupling
between electrochemical and mechanical part, 4. Mechanical model, 5. Boundary
conditions: initial stress, 6. Displacement output. ................................................................ 100
Fig. 12. Linear BG global model of CP actuators, where a source of voltage is employed as
the input and the tips displacement of the beam is the output of the model...................... 101
Figure 13. Interface of 20-sim simulation software .............................................................. 102
Fig. 14. Comparison between the output current of the linear model and nonlinear (6 RFEs)
responses to a step voltage of 2 Vpp at f = 0.1 Hz .................................................................. 104
Fig. 15. Comparison between the linear model and the nonlinear model in bending
displacement of the actuator in response to the step input voltage with various magnitudes
................................................................................................................................................ 105

xxvi
Fig. 16. Trilayer actuator under voltage excitation: a) Top PEDOT electrode voltage: -1.1 V,
Bottom PEDOT electrode: 1.1 V, b) neutral state, c) Top PEDOT electrode: 1.1 V, Bottom
PEDOT electrode: -1.1 V......................................................................................................... 107
Fig. 17. a) Triangular wave and rectangular wave applied voltage, b) Measured and
simulation current response .................................................................................................. 108
Fig. 18. Modeled voltage drops along the length of the beam due to the existence of
electrical resistance of PEDOT electrodes. The black curve is the voltage seen at the fixed
end of the cantilever where electrical contact is made. The dashed green line is the
predicted voltage as seen at 0.9 of the length, and the others are in between. .................. 109
Fig. 19. Comparison between the simulations and the experiments on the strain as a
function of the magnitude of applied voltage at a constant frequency of 0.1 Hz ................ 110
Fig. 20. a) Comparison between the simulations and the experiments on the strain as a
function of the frequency of applied voltage at a constant magnitude of ±1.1 V, b) The
relative error of the simulations compared to the measured data. ..................................... 111
Fig. 21. Comparison between the experiment and the simulation in terms of the dynamic
mechanical behavior of the beam subjected to a voltage excitation of 2.2 Vpp at 0.1 Hz. .. 112
Fig. 22. Key factors affecting the actuator displacement under a square wave voltage (Vpp =
2.4 V), a) at low frequency f = 0.1 Hz, and b) at high frequency f = 20 Hz. For each parameter,
three columns from left to right represent for its real value (blue column), its value
increased by 20% (orange column), and its value reduced by 20% (grey column). .............. 113
Fig. 23. Evolution of energy over half a period of applied voltage: a) In the electrochemical
part, b) In the mechanical part. ............................................................................................. 115
Fig. 24. Instantaneous electrical to mechanical coupling. ..................................................... 116

Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators
Fig. 1. a) Schematic of CP free-standing film sensing mechanism before and after subjected
to a linear force, and b) Schematic of a bending trilayer sensor [4] ..................................... 123
Fig. 2. Effect of counter-ion and redox state on the voltage output of polypyrrole mechanical
sensor [2] (Reprinted with permission from John Wiley and Sons) ...................................... 125
Fig. 3. Schematic diagram illustrating the net ion transport expected in both the
“Deformation Induced Ion Flux” model and based on elastic energy considerations [2]
(Reprinted with permission from John Wiley and Sons). In the first case compression of the
polymer leads to an increase in ion concentration within the polymer, and a net flux of ions
out. In the second, it is the compressive stress that squeezes ions out. .............................. 126
Fig. 4. Experimental scheme and induced voltage mechanism of a thin ionic conducting
polymer film under an applied stress [10] ............................................................................. 127
Fig. 5. Schematic experiments and comparison of output voltage in two different trilayer
configurations: Gold-NBR/PEO-Gold, and PEDOT-NBR/PEO-PEDOT with the dimension of L =
15mm, b = 5mm, ht = 0.17 mm at various EMITFSI-PC concentrations, under a 2% applied

xxvii
strain [11] (Retrieved from https://tel.archives-ouvertes.fr/tel-01413677. By permission
from author.). In this case a positive voltage suggests anion motion. .................................. 128
Fig. 6. Global Word BG model, where (1) and (3) represent the input voltage and the tip
displacement, respectively, as the trilayer structure in actuation mode; the (4) and (5) are
the input displacement and the voltage output, respectively, when the trilayer is operated as
a sensor, (2) is the global BG model ....................................................................................... 131
Fig. 7. a) Experimental setup to measure the voltage response to a sinusoidal displacement,
and b) An output voltage response to an input displacement of the magnitude of 2 mm and
frequency of 3 Hz on a trilayer of 7 mm x 1 mm x 0.017 mm. .............................................. 133
Fig. 8. Experimental setup to measure the decaying of blocking force of the actuator in
response to a step displacement ........................................................................................... 134
Fig. 9. a) Schematic of a trilayer transducer under a displacement excitation and the
produced voltage polarity b) Voltage magnitude and the strain-to-charge ratio as a function
of frequency when the magnitude of the sinusoidal displacement is constant at 2mm, c)
Comparison between the simulation and the experiment of the magnitude of output voltage
................................................................................................................................................ 136
Fig. 10. a) Voltage magnitude and the strain-to-charge ratio as a function of magnitude
when the frequency of displacement is constant at 3 Hz, b) Comparison between simulation
and experiment of the magnitude of output voltage ............................................................ 138
Fig. 11. Comparison between the simulation and measured voltage responding to an applied
step displacement .................................................................................................................. 139
Fig. 12. a) Experiment of applying step displacement, and b) The decay of force response of
the beam as a function of time. ............................................................................................. 142

xxviii
List of Tables
Chapter 2: PEDOT-based trilayer fabrication process
Table 1. Summary some important properties of three different PEO-based IPNs ................ 24
Table 2. Summary the sample configuration for different characterizations ......................... 35

Chapter 3: Electrochemomechanical characterization of the trilayer structure

Table 1. Thickness in swollen state and ionic conductivity of the NBR/PEO alone, top bilayer,
and bottom bilayer .................................................................................................................. 52
Table 2. Summary strain-to-charge ratio for different materials and actuator configuration66
Table 3 Summary of measured parameters of the ultrathin trilayer actuator ....................... 68

Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
Table 1. Summarize on various model approaches of CP-based actuators ............................ 81
Table 2. Notations for continuous, lumped, and distributed modeling approaches .............. 85
Table 3 Summary of derived parameters for the linear model ............................................. 101
Table 4. Pros and cons of the linear model in comparison to the nonlinear model ............. 103

xxix
Abbreviations

BG Bond Graph
CP conducting polymer
CV cyclic voltammetry
EAP electroactive polymer
EDOT 3,4-ethylenedioxythiophene
EDX energy-dispersive X-ray spectroscopy
EIS electrochemical impedance spectroscopy
EMITFSI 1-ethyl-3-methylimidazolium bis(trifluoromethanesulfonyl)imide
IPMC ionic polymer-metal composite
IPN interpenerating network
LbL Layer by Layer
NBR nitrile butadiene rubber
PEDOT poly(3,4-ethylenedioxythiophene)
PEDOT:PSS poly(3,4-ethylenedioxythiophene): poly(styrene sulfonate)
PEGDM poly(ethylene glycol) dimethacrylate
PEGM poly(ethylene glycol) methyl ether methacrylate
PEO polyethylene oxide
PPy polypyrrole
PVDF polyvinylidene difluoride
RFE rigid finite element
RFEM rigid finite element method
SDE spring damper element
SEM scanning electron microscopy
sIPN semi interpenerating network
SPE solid polymer electrolyte
VPP vapor phase polymerization

xxx
Acknowledgements
Financial support for my study was generously provided by the European Union’s Horizon
2020 research and innovation programme under the Marie Skłodowska-Curie grant
agreement No 641822 (http://micact.eu), National Research Agency (ANR) under program
PIA EQUIPEX LEAF (ANR-11-EQPX-0025), RENATECH and MICRO-TIP projects.

This study could not be finished without collaboration and support of my supervisors, my
colleagues, and my family.

First of all, I gratefully acknowledge my supervisors: Prof. Eric Cattan, Prof. Sébastien Grondel,
and Prof. John D.W. Madden for their enduring patience and thoughtful advices, for granting
me the freedom to pursue my ideas, and for giving me the opportunity to interact with
experts of electroactive polymer community.

I would like to acknowledge the LPPI team: Prof. Frédéric Vidal, Dr. Cédric Plesse, Dr. Giao
Nguyen, and Kätlin Rohtlaid for giving me the opportunity to visit their lab, to discuss the
challenges arising in transducers’ fabrication process, and to collaborate with me on some
parts of the thesis.

Thank you to my past and present colleagues in MAMINA group in IEMN, Dr. Caroline Soyer,
Dr. Damien Faux, Mohamed Bentefrit, Lauréline Seurre, and in Molecular Mechatronics
Laboratory in UBC, Dr. Ali Mahmoudzadeh, Yuta Dobashi, Saquib Sarwar, Claire Preston, and
Saeedeh Ebrahimi for their great friendship and help.

Finally, thanks to my loving wife Minh Quan and my daughter Minh Anh, who always had faith
in me through the most difficult time of my life. I owe special gratitude to my Mother and my
Brother for their unconditional love and support, encouragement, patience, without which
this work would not have been possible. You have been supported me in one way or another
along the way.
Thank you all.

xxxi
Dedication

I dedicate this thesis

To the memory of my late Father.

xxxii
Chapter 1: Introduction

Chapter 1: Introduction

Contents

1.1 Mammalian muscles ................................................................................................................ 2

1.2 Artificial muscles ...................................................................................................................... 2
1.3 Motivation and problem statement ........................................................................................ 9
1.4 Thesis structure ....................................................................................................................... 9

1
Chapter 1: Introduction

This chapter reviews the motivation for the work and outlines the structure of the thesis. At
first, a brief review on mammalian muscles is given. Then contractile materials, known as
artificial muscles, are described. Particular attention is paid to conducting polymer actuators
and sensors, which are the subjects of this thesis.

1.1 Mammalian muscles

Mammalian muscles are highly optimized systems as their structure are considered to be the
same for all animals. They are driven by a complex mechanism where a chemical reaction
drives reversible hydrogen bonding between proteins actin and myosin, sliding them past
each other to produce strain. Mammalian muscle performance has been reported reach a
peak stress of 150-300 kPa at a strain of about 25% and the maximum power output is 150-
225 W/kg. At low and intermediate operating frequencies, natural muscle has the average
power of 50 W/kg the energy density of 20-70 J/kg [1]. Because of their micro fiber structure,
natural muscles are easy to scale. However, the biggest challenge is the requirement of a
complex accompanying system to feed and keep them alive. This issue has driving the
researchers to keep searching possible materials and structures, which can replicate natural
muscles.

1.2 Artificial muscles

Actuating materials, or “artificial muscles” have long been of interest. A growing interest in
artificial muscles has been observed over two last decades. Aiming to develop a soft robotic
programmable motion has driven researchers around the world to experiment with a variety
of materials and designs. The designed robots were expected to mimic the bending and
flexing in a way that can interact more naturally with living organisms. These are materials or
devices having the ability to reversibly contract, expand, or rotate within one component
under an external stimulus [2], and promise compact, lightweight means to mechanically
power everything from robots, exoskeletons, and medical prosthetics to micro-fluidic circuits
and micro-tools. Different materials have been developed for artificial muscles including
shape memory alloys (SMAs) [3], electroactive polymers (EAPs) [4], and piezoelectric and
dielectric materials [5]. Like natural muscles, artificial muscles are characterized by a
reciprocating or oscillatory types of motion that may be evident as linear [6, 7], bending [8-
12], torsional [13] [14, 15] [3]or torsional induced linear actuation [16] and a combination of
these processes [5]. Based on the actuation mechanism, these artificial muscles can be
classified into four categories: pneumatic, thermal, ionic-based, and electric field driven
actuation - wherein the actuation is due to Coulombic forces. The relation between blocking
stresses and actuation strains for a variety of the artificial muscles can be found in the Fig. 1.

Pneumatic artificial muscle (PAM), so-called McKibben Artificial Muscle, Fluidic Muscle, or the
Biomimetic Actuator [17], is an actuator, which is shortened in the actuating length when it
is pressurized [18]. The common structure of a PAM is a long and hollow cylinder made of

2
Chapter 1: Introduction

synthetic or natural rubber, wrapped inside a braided mesh shell built at a predetermined
angle. The working mechanism of the PAM relies on the supply of compressed-air to the
interior of the cylinder, which contracts in length and expands radially. Upon the air being
released, the mesh plays a role as a spring that restores the cylinder to its original form. The
most significant advantage of PAM in control applications is that only pressure parameter
needs to be controlled [19]. In addition, PAMS are also lightweight, easy to fabricate, and
have load-length curves similar to human muscle [20, 21]. However, PAMs are not very
accurate, have a low force output, and require lot of additional equipment such as: a source
of pressurized air, a feedback sensor, and a solenoid for controlling flow. These deficiencies
motivate research on alternatives to PAMs going forward.

In the class of thermal actuation, shape memory alloys (SMAs) are the well-known actuators,
which belong to the class of shape memory materials. They are metallic alloys having the
ability to return to their original shape when subjected to a temperature or magnetic field
memorization process between two transformation phases [22-25]. Under some certain
conditions, superelasticity or visco-elasticity are also commonly observed [24]. The SMAs
have been used in various fields such as automotive, aerospace, robotics, and biomedical
applications [26-28]. They have been proven to provide unique characteristics such as bio-
compatibility, high stress, high work per volume and work density (for NiTi SMA type: 200
MPa, 10%, 100 J/cm3, 15 kJ/kg) [29]. However, they also have several disadvantages such as
low accuracy, limited strain, fatigue, stability, and low energy efficiency [30, 31]. Therefore,
intensive work is required to enhance the performance of SMAs.

Recently, new members of the thermal actuation family have been warmly welcomed. Lima
et al. [15] have introduced hybrid yarn actuators in which a guest material is incorporated
into the pore space between multi-walled carbon nanotube bundles. The actuators produce
a torsional rotation and contraction performance when an electrical, chemical, or photonic
excitation is induced on the host material, resulting in a dimension change of the guest and
the yarn. Later on, Haines et al. [16] demonstrated that inexpensive high-strength polymer
fibers used for fishing line and sewing thread can be used for tensile and torsional muscles.
These fishing line artificial muscles produce a large stroke (up to 49%), and high stress.
Considering they are of the same length and weight, these fishing line can lift a loads over 100
times larger compare to that of human.

3
Chapter 1: Introduction

Fig. 1. Dependence of blocking stress on actuation strain for various artificial muscles.
Abbreviations: PVDF - piezoelectric polymer [32], PZT - piezoelectric ceramic [33], TE 100K -
Thermal expansion actuators 100K [34], SMA – shape memory alloy [35], CP – conducting
polymer [36], IPMC – ionic polymer metal composite [37], PZT-PT – piezoelectric single
crystal [33], P(VDF-TrFE) – relaxor ferroelectric polymer [38], DEA – dielectric elastomer [35],
PAM – pneumatic artificial muscle [19], Gels – Gel actuators [39], NM – natural muscle [1].

Electric field actuation includes piezoelectric and dielectric elastomer actuation (DEAs).
Piezoelectric materials are known as materials that undergo a small reversible change in
dimensions when subjected to an electric field. An electric polarization is developed when
they are strained [33]. Piezoelectricity can be used in various applications such as sound
production and detection, microbalances, and ultrafine focusing of optical assemblies. In
addition, it is a basic component in various scientific devices, such scanning probe
microscopes [40], the ignition source for cigarette lighters, and the time reference source in
quartz watches [41]. The most common type of piezo ceramic is lead zirconate titanate known
as PZT. However, it is known as a brittle material causing limitations in the applications where
high strain is required.

For applications requiring deformable materials researchers have developed flexible material
which can withstand a large amount of strain [35]. DEAs are made of thin film of electrically
insulating elastomeric materials, which are coated on both sides with the electronically
conducting compliant electrodes. When a voltage is applied to the electrodes, the films

4
Chapter 1: Introduction

expand in area and compress in thickness. DEAs are a promising EAP technology that can be
applied in various technologies, such as actuators, strain sensors, and electrical generators
used to harvest mechanical energy [29, 42, 43]. However, the widespread adoption of DEAs
has been challenged by early breakdown and the high electric field requirements, and large
support frames to hold the DEA membranes [29]. The miniaturization to the micro scale of
the DEA fabrication process has recently been investigated. A thin film of 5 µm was fabricated
and stacked on the top of each other up to 50 layers to increase the total height of the system
to obtain usable displacement and reduce the controlling voltage down to 150 V [44-46],
which still remains high and be restricted for the human affinity applications.

Fig. 2. Dependence of specific work on actuation strain for various artificial muscles.

Gels, ionic polymer-metal composites (IPMCs), carbon nanotubes (CNTs), and conducting
polymers fall under the ionic-based classification. Gels are found in manufactured foods,
cosmetics, and in medical creams [47, 48]. They can be extremely soft, biocompatible and
biodegradable [49, 50]– all properties of interest for medical and in vivo applications.
Basically, ionic gels contain a crosslinked polymer in an electrolyte solution (it can be water-
based or organic solvent). As actuators, gels are exciting because of the very large dimensional
changes they can produce (more than 40% in strain) [32], albeit at blocking forces (maximum
pressure at 0.3 MPa). Depending on types of stimulus signal, gel actuators can be classified
into temperature-responsive gels [51], photoresponsive gels [52], chemical driven gels [53],

5
Chapter 1: Introduction

and electrically driven gels where an electric field can induce a flow of positively charged
surfactant binding to one side of the gel [54]. Gels also have a great potential as sensors. But
the actual uses of gels are somewhat restricted. Gel actuators have very low blocking force
due to their low modulus (normally < 4 MPa) [47]. In addition, delivery of chemical energy
requires pumping systems, which may add to the weight and complexity. The other
limitations of gel actuators are low response speed - which is due to diffusion rates - and
drying out of hydrogels when they are used outside of a wet environment. Recent work by
Spinks in which gels are used as the swelling element in McKibben actuators (PAMs that swell
with gel instead of air) show large strains (9%) and relatively high forces (2 N) when heated
or swollen in the hot water, but remain slow [55].

Carbon nanotubes (CNTs), which were discovered in 1991 by Iijima [56], have attracted a lot
of interest thanks to their intrinsic electrical and mechanical properties. An individual CNT
possesses a high Young’s modulus of approximately 1 TPa, maximum tensile strength of 100
GPa, and a conductivity of 106 – 107 S/m [57]. In addition, CNTs can be incorporated with
different guest materials to form composite materials and benefit from those. Although the
mechanical strength of an individual CNT is extremely high, the effective strength of CNT
bundles is pretty low because of weak shear interactions between adjacent shells and tubes
[58]. CNTs alone have been used in tensile and torsional actuators [14, 59, 60] where the
working mechanism is based on the formation of a double layer capacitance to balance the
charges injected by an applied potential – with the formation of this double layer also
appearing to push nanotubes apart [60, 61]. Due to their high stiffness, the obtained strain is
pretty small, ~ 0.07 % [59]. However, the high porosity and fast ion diffusion are the main
contribution to a high strain rate of 19 % s-1 [62]. CNTs remain expensive to mass produce [63,
64] which has limited their widespread applications in artificial muscles and other applications
to date.

IPMCs, which are structured of an ionic conducting polymer sandwiched by two metal
electrodes, are one of the promising electroactive polymers (EAP) for soft actuators and
sensors. Their actuation mechanism is based on mobile cations accompanied by water (or
solvent) flow. An electric field induces a change in ion concentration, which attracts water
molecules (or other solvent) to one side of the polymer, causing a deflection of the total
structure toward one of the metal electrodes – and away from the side that swells. IPMCs
offer a huge advantages that attract the use in various biomedical and human interface
applications such as, low driving voltage (1 – 3 V), quite high frequency response up to several
hundred hertz [65, 66], large displacement (maximum strain 3%), softness (maximum
pressure 30 MPa) [37], ready manufacturing into a wide range of shapes, operation in wet
conditions (water, solutions) as well as dry (open air) conditions [67-69]. IPMCs also exhibit
sensing properties - generating an electric current in response to bending moments imposed
on them. However, a big drawback of IPMCs is a slow relaxation due to the backward

6
Chapter 1: Introduction

movement of water molecules out of the cation-rich areas. They also require a constant
applied voltage and some power to maintain their operating position [70]. Recently, ionic
liquid was incorporated into the IPMCs instead of solvent, to reduce the relaxation and
increase the operating lifetime of the actuators - at the cost of very low response speeds [71-
74].

Conducting polymers (CPs) are organic polymers that conduct electricity. Since the discovery
of polymeric organic conductors in 1977 [75, 76], knowledge in this field has grown
dramatically. Based on the heteroatoms present in the structure, conducting polymers can be
classified into three types: i) no heteroatom such as poly(acetylene)s (PA), and poly(p-
phenylene vinylene) (PPV), ii) nitrogen containing such as poly(pyrrole)s (PPy), and
polyanilines (PANI), and iii) sulfur-containing such as poly(thiophene)s (PT), poly(3,4-
ethylenedioxythiophene) (PEDOT), and poly(p-phenylene sulfide) (PPS). Molecular structure
of some popular conducting polymers is described in Fig. 3. PEDOT holds an important role in
anti-static, electric and electronic applications [77]. The idea of using conducting polymers as
actuators for the direct conversion of electrical energy to mechanical energy, based on the
redox process between the CPs and the mobile ions in electrolyte, was initially proposed by
Baughman [8, 78]. Remarkable progress in conducting polymer actuators has been reported
by Pei and Inganäs, Smela and Inganäs, Otero, and, recently, by Vidal et al. [11, 12, 79-83].
These actuators have offered advantages over other types of artificial muscles by offering a
combination of low operating voltage (typically 1-2 V), biocompatibility [84-86], large strain
range from 1 up to 10 %, high blocking force and high specific work (Fig. 1 and Fig. 2). They
can be found in various potential applications including catheters [87, 88], dynamic Braille
display [89, 90], valves [88], and blood vessel reconnection [61].

CP actuators suffer from very low efficiency (less than 1%) and electrochemical coupling
under 1% [5], however the operating efficiency can be improved by recover the stored energy
under the form of electrochemical, mechanical elastic and potential energy. In addition, the
actuation response is normally slow since their working mechanism relies on ion migration.
Recently, the problem of low actuation speed has been addressed by Maziz et al. [83, 91].
They have demonstrated a kHz operating speed (at resonance) of PEDOT-based actuators,
where the total thickness of the actuator is several micrometers. Later on, they showed that
this actuator is compatible with a microfabrication process that includes: i) compatibility with
photolithography process and plasma etching [92, 93] ii) the layer-by-layer technique
combined with the vapor phase polymerization to fabricate thin actuators, ii) integration the
electrical contact onto the electrodes and patterns them into desired shape with high
precision [91, 92]. In the remainder of the thesis, this fabrication technique and the
subsequent developments are built upon.

Considering CP-based transducers and IPMCs, they both show quite similar advantages, as
reported previously, and can both be adapted to similar applications. In term of working

7
Chapter 1: Introduction

mechanism, they are based on ion movements due to the redox reactions in the first case,
and mobile cations accompanied by water (or solvent) flow in the second case. These result
in a contraction or an expansion of the CP electrode or an osmotic pressure gradient towards
the negative electrode of IMPC. However, while the modelization of IPMCs was well
established, the working mechanism of CP-based actuators is still unclear.

Fig. 3. Chemical structures of some typical conducting polymers.

The Laboratoire de Physicochimie des Polymères et des Interfaces at the University of Cergy-
Pontoise, the Institut d'électronique de microélectronique et de nanotechnologie at the
University of Valenciennes et du Hainaut Cambrésis, and the laboratory of Molecular
Mechatronics at the University of British Columbia have had a long (more than 10 years) and
fruitful collaboration on conducting polymer (CP) materials and devices, including
optimization and characterization, modeling, and applications of the materials. Currently, one
of the main goals of the group is to develop microsystems from CP materials. CP actuators, as
demonstrated previously, are potential candidates for microscale applications (thickness of
the actuator < 20 µm), where a low operating voltage, high strain, fast operation in open air,
and especially a compatibility with clean room techniques are required. However, to bring
this material and the new fabrication method proposed by Maziz et al. [36] from the lab into
large scale manufacturing and production, the optimization on the fabrication techniques, as
well the full characterization and modeling of this ultrathin structure are the essential steps.

8
Chapter 1: Introduction

1.3 Motivation and problem statement

The goal of this thesis is to fabricate, characterize electro-chemo-mechanical properties, and
to model the ultrathin PEDOT-based trilayer transducers.

For this purpose, a versatile fabrication technique, which offers high precision and reliability,
is required to produce a good trilayer. The layer-by-layer fabricating technique, which is
compatible with the photolithography process including plasma dry etching, is selected to
fabricate the trilayer structure. Characterization methods have been developed to respond
to measurements made on very thin films to understand their transducer mechanisms. The
characterization process of the trilayer transducers may also bring some difficulties, and since
these trilayers are thin and easy to damage under mechanical stress. In addition, properties
can be varied with oxidation state.

Ultimately, to fully exploit the thin trilayer structures, modeling is required to predict behavior
of micro-actuators and micro-sensors, to optimize their design before mass production and
finally to understand and control device operation. As demonstrated previously, the actuation
mechanism of the PEDOT-based actuators is a complex process, that involves electrical
stimulation, redox reactions, and induces mechanical strain. These can challenge the
modeling process, where a complex model including different domains is needed.

Within this thesis, these challenges on the layer-by-layer fabrication technique and the
characterization as well as the modeling process of the resulting trilayer transducers will be
confronted and the outcome will be drawn.

1.4 Thesis structure

The thesis is divided in six chapters. The brief content of each chapter is drawn as below:

Chapter 2:

Within this chapter, a brief introduction on working mechanism of CP-based actuators is

given. The process of selection of the materials used to fabricate the CP-based trilayer in this
study is described. Later on, different synthesis methods used to synthesize PEDOT
conducting polymer are briefly reviewed and the vapor phase polymerization is selected for
the PEDOT electrode fabrication process since it allows an easier means to change the
electrode’s thickness. The trilayer fabrication process is then described and the obtained
trilayer is analyzed with profilometer, SEM and EDX techniques to determine the exact surface
roughness and structure of the trilayer. The roughness surface results show an asymmetric
between the two PEDOT electrodes suggesting a difference in electrical properties of the two
electrodes.

9
Chapter 1: Introduction

Chapter 3:

This chapter presents the electrochemomechanical characterization of the resulting trilayers,

including characterization techniques and the obtained results. At first, the electro-chemical
properties of the trilayer, such as the electronic conductivity as the function of oxidation state
of PEDOT, the ionic conductivity, and the volumetric capacitance of PEDOT electrodes as a
function of the scanning rate and the voltage window, and the ionic conductivity of NBR-PEO
semi interpenetrating network, are determined. The degree of bending is measured, from
which the coupling between the electro-chemical and the mechanical domains - an empirical
strain-to-charge ratio - is computed. Finally, the mechanical properties: Young’s modulus of
the PEDOT electrodes as a function of oxidation state, and of the trilayer structure, the
damping ratio, the trilayer density and the forces as a function of voltage and frequency are
determined.

Chapter 4:

Chapter 4 is devoted to the modeling of the trilayer actuators. The Bond Graph formalism is
introduced, and the reason for its use explained. A simple electro-chemo-mechanical model
based on Euler-Bernoulli dynamic beam theory is described. Later, a non-linear dynamic
electro-chemo-mechanical model based on the dynamic rigid finite element method is
introduced. The measured material properties (Chapter 3) are used in the model, and the
simulations are compared with experimentally observed actuator deflections through time.
The proposed model enables to extract the information on the energy flows inside the trilayer
and evaluate the key factors affecting the actuator’s performance (strain output).

Chapter 5:

A simple model including both sensing and actuation is proposed, demonstrating the
advantage of the Bond Graph formulation in running in both forward (electrical to
mechanical) and backward (mechanical to electrical) modes. The initial model is based on
using the strain to charge ratio as an empirical coupling coefficient.

Chapter 6:

In this last chapter, the findings of this investigation are summarized and the most conclusions
are drawn. Furthermore, the prospects for ongoing work are developed.

10
Chapter 1: Introduction

[1] G.H.R. Pollack, Frederick B.; Blyakhman, Felix A.; Yakovenko, Olga V.; Dunaway, Dwayne L.,
Natural Muscle as a Biological System, in: Y. Bar-Cohen (Ed.) Electroactive Polymer (EAP) Actuators
as Artificial Muscles: Reality, Potential, and Challenges, SPIE, USA, 2004.
[2] S.M. Mirvakili, Niobium Nanowire Yarns and Their Application as Artificial Muscle: University of
British Columbia; 2013.
[3] K.P. Jamie, H. Elliot, J.W. Robert, A novel low-profile shape memory alloy torsional actuator,
Smart Materials and Structures, 19(2010) 125014.
[4] Y. Bar-Cohen, Electroactive Polymers as Artificial Muscles: A Review, Journal of Spacecraft and
Rockets, 39(2002) 822-7.
[5] J.D.W. Madden, N.A. Vandesteeg, P.A. Anquetil, P.G.A. Madden, A. Takshi, R.Z. Pytel, S.R.
Lafontaine, P.A. Wieringa, I.W. Hunter, Artificial muscle technology: Physical principles and naval
prospects, IEEE Journal of Oceanic Engineering, 29(2004) 706-28.
[6] A.D. Santa, D.D. Rossi, A. Mazzoldi, Characterization and modelling of a conducting polymer
muscle-like linear actuator, Smart Materials and Structures, 6(1997) 23.
[7] A. Fannir, Actuators based on conducting polymers with a linear deformation in open air and
compatible with a space environment: Université de Cergy Pontoise; 2017.
[8] R.H. Baughman, L.W. Shacklette, R.L. Elsenbaumer, E.J. Plichta, C. Becht, Micro Electromechanical
Actuators Based on Conducting Polymers, in: P.I. Lazarev (Ed.) Molecular Electronics: Materials and
Methods, Springer Netherlands, Dordrecht, 1991, pp. 267-89.
[9] J.D. Madden, P.G. Madden, I.W. Hunter, Polypyrrole actuators: modeling and performance, 2001,
pp. 72-83.
[10] P.G.A. Madden, Development and Modeling of Conducting Polymer Actuators and the
Fabrication of a Conducting Polymer Based Feedback Loop: MASSACHUSETTS INSTITUTE OF
TECHNOLOGY; 2003.
[11] E. Smela, Conjugated polymer actuators for biomedical applications, Advanced Materials,
15(2003) 481-94.
[12] T.F. Otero, E. Angulo, J. Rodríguez, C. Santamaría, Electrochemomechanical properties from a
bilayer: polypyrrole / non-conducting and flexible material — artificial muscle, Journal of
Electroanalytical Chemistry, 341(1992) 369-75.
[13] K.J. Gabriel, W.S.N. Trimmer, J.A. Walker, A micro rotary actuator using shape memory alloys,
Sensors and Actuators, 15(1988) 95-102.
[14] J. Foroughi, G.M. Spinks, G.G. Wallace, J. Oh, M.E. Kozlov, S. Fang, T. Mirfakhrai, J.D.W. Madden,
M.K. Shin, S.J. Kim, R.H. Baughman, Torsional Carbon Nanotube Artificial Muscles, Science,
334(2011) 494.
[15] M.D. Lima, N. Li, M. Jung de Andrade, S. Fang, J. Oh, G.M. Spinks, M.E. Kozlov, C.S. Haines, D.
Suh, J. Foroughi, S.J. Kim, Y. Chen, T. Ware, M.K. Shin, L.D. Machado, A.F. Fonseca, J.D.W. Madden,
W.E. Voit, D.S. Galvão, R.H. Baughman, Electrically, Chemically, and Photonically Powered Torsional
and Tensile Actuation of Hybrid Carbon Nanotube Yarn Muscles, Science, 338(2012) 928.
[16] C.S. Haines, M.D. Lima, N. Li, G.M. Spinks, J. Foroughi, J.D. Madden, S.H. Kim, S. Fang, M. Jung de
Andrade, F. Goktepe, O. Goktepe, S.M. Mirvakili, S. Naficy, X. Lepro, J. Oh, M.E. Kozlov, S.J. Kim, X.
Xu, B.J. Swedlove, G.G. Wallace, R.H. Baughman, Artificial muscles from fishing line and sewing
thread, Science, 343(2014) 868-72.
[17] E.T. Roche, R. Wohlfarth, J.T.B. Overvelde, N.V. Vasilyev, F.A. Pigula, D.J. Mooney, K. Bertoldi,
C.J. Walsh, A Bioinspired Soft Actuated Material, Advanced Materials, 26(2014) 1200-6.
[18] G. Andrikopoulos, G. Nikolakopoulos, S. Manesis, A Survey on applications of Pneumatic
Artificial Muscles, 2011 19th Mediterranean Conference on Control & Automation (MED)2011, pp.
1439-46.
[19] C. Ching-Ping, B. Hannaford, Measurement and modeling of McKibben pneumatic artificial
muscles, IEEE Transactions on Robotics and Automation, 12(1996) 90-102.
[20] S.C. Obiajulu, E.T. Roche, F.A. Pigula, C.J. Walsh, Soft Pneumatic Artificial Muscles With Low
Threshold Pressures for a Cardiac Compression Device, (2013) V06AT7A009.

11
Chapter 1: Introduction

[21] M. Doumit, A. Fahim, M. Munro, Analytical Modeling and Experimental Validation of the
Braided Pneumatic Muscle, IEEE Transactions on Robotics, 25(2009) 1282-91.
[22] P.S. Lobo, J. Almeida, L. Guerreiro, Shape Memory Alloys Behaviour: A Review, Procedia
Engineering, 114(2015) 776-83.
[23] J. Mohd Jani, M. Leary, A. Subic, M.A. Gibson, A review of shape memory alloy research,
applications and opportunities, Materials & Design (1980-2015), 56(2014) 1078-113.
[24] W.M. Huang, Z. Ding, C.C. Wang, J. Wei, Y. Zhao, H. Purnawali, Shape memory materials,
Materials Today, 13(2010) 54-61.
[25] A. Nespoli, S. Besseghini, S. Pittaccio, E. Villa, S. Viscuso, The high potential of shape memory
alloys in developing miniature mechanical devices: A review on shape memory alloy mini-actuators,
Sensors and Actuators A: Physical, 158(2010) 149-60.
[26] J. Van Humbeeck, Non-medical applications of shape memory alloys, Materials Science and
Engineering: A, 273-275(1999) 134-48.
[27] L. Sun, W.M. Huang, Z. Ding, Y. Zhao, C.C. Wang, H. Purnawali, C. Tang, Stimulus-responsive
shape memory materials: A review, Materials & Design, 33(2012) 577-640.
[28] M. Sreekumar, T. Nagarajan, M. Singaperumal, M. Zoppi, R. Molfino, Critical review of current
trends in shape memory alloy actuators for intelligent robots, Industrial Robot, 34(2007) 285-94.
[29] P. Brochu, Q. Pei, Advances in Dielectric Elastomers for Actuators and Artificial Muscles,
Macromolecular Rapid Communications, 31(2010) 10-36.
[30] K. Marjaana, L. Tomi, Long-term behaviour of binary Ti–49.7Ni (at.%) SMA actuators—the
fatigue lives and evolution of strains on thermal cycling, Smart Materials and Structures, 19(2010)
115019.
[31] T.W. Duerig, D. Stöckel, A. Keeley, Actuator and Work Production Devices, Engineering Aspects
of Shape Memory Alloys, Butterworth-Heinemann1990, pp. 181-94.
[32] R. Pelrine, R. Kornbluh, J. Joseph, R. Heydt, Q. Pei, S. Chiba, High-field deformation of
elastomeric dielectrics for actuators, Materials Science and Engineering: C, 11(2000) 89-100.
[33] A.J. Moulson, J.M. Herbert, Piezoelectric Ceramics, Electroceramics, John Wiley & Sons,
Ltd2003, pp. 339-410.
[34] J.E. Huber, N.A. Fleck, M.F. Ashby, The selection of mechanical actuators based on performance
indices, Proceedings of the Royal Society of London Series A: Mathematical, Physical and Engineering
Sciences, 453(1997) 2185.
[35] R.D. Kornbluh, R. Pelrine, Q. Pei, S. Oh, J. Joseph, Ultrahigh strain response of field-actuated
elastomeric polymers, SPIE's 7th Annual International Symposium on Smart Structures and
Materials, SPIE2000, p. 14.
[36] A. Maziz, C. Plesse, C. Soyer, E. Cattan, F. Vidal, Top-down Approach for the Direct Synthesis,
Patterning, and Operation of Artificial Micromuscles on Flexible Substrates, ACS Appl Mater
Interfaces, 8(2016) 1559-64.
[37] S. Mohsen, J.K. Kwang, Ionic polymer–metal composites: IV. Industrial and medical applications,
Smart Materials and Structures, 14(2005) 197.
[38] F. Xia, Z.Y. Cheng, H.S. Xu, H.F. Li, Q.M. Zhang, G.J. Kavarnos, R.Y. Ting, G. Abdul‐Sadek, K.D.
Belfield, High Electromechanical Responses in a Poly(vinylidene fluoride–trifluoroethylene–
chlorofluoroethylene) Terpolymer, Advanced Materials, 14(2002) 1574-7.
[39] C.W. Peak, J.J. Wilker, G. Schmidt, A review on tough and sticky hydrogels, Colloid and Polymer
Science, 291(2013) 2031-47.
[40] J.A. Christman, R.R. Woolcott, A.I. Kingon, R.J. Nemanich, Piezoelectric measurements with
atomic force microscopy, Applied Physics Letters, 73(1998) 3851-3.
[41] E.R. Keeler, Electronic watch counting circuit, in: US (Ed.), Timex Corp US, 1968.
[42] C. Federico, A. Iain, B. Siegfried, F. Gabriele, G. Giuseppe, G. Massimiliano, G. Christian, J.-M.
Claire, K. William, K. Guggi, K. Matthias, K. Roy, L. Benny, M. Marc, M. Silvain, N. Stephan, O.B.
Benjamin, P. Qibing, P. Ron, R. Björn, R. Samuel, S. Herbert, Standards for dielectric elastomer
transducers, Smart Materials and Structures, 24(2015) 105025.

12
Chapter 1: Introduction

[43] S. Rosset, H.R. Shea, Small, fast, and tough: Shrinking down integrated elastomer transducers,
Applied Physics Reviews, 3(2016) 031105.
[44] P. Lotz, M. Matysek, H.F. Schlaak, Fabrication and Application of Miniaturized Dielectric
Elastomer Stack Actuators, IEEE/ASME Transactions on Mechatronics, 16(2011) 58-66.
[45] D. Gatti, H. Haus, M. Matysek, B. Frohnapfel, C. Tropea, H.F. Schlaak, The dielectric breakdown
limit of silicone dielectric elastomer actuators, Applied Physics Letters, 104(2014) 052905.
[46] S.-A. Susana, K. Florian, M. Holger, F.-Z. Florentine, F.S. Helmut, A novel application of dielectric
stack actuators: a pumping micromixer, Smart Materials and Structures, 27(2018) 074008.
[47] P. Calvert, Hydrogels for Soft Machines, Advanced Materials, 21(2009) 743-56.
[48] P. Calvert, Gel Sensors and Actuators, MRS Bulletin, 33(2011) 207-12.
[49] L. Taylor Danielle, M. in het Panhuis, Self‐Healing Hydrogels, Advanced Materials, 28(2016)
9060-93.
[50] C.J. Ferris, K.G. Gilmore, G.G. Wallace, M. in het Panhuis, Biofabrication: an overview of the
approaches used for printing of living cells, Applied Microbiology and Biotechnology, 97(2013) 4243-
58.
[51] L.-W. Xia, R. Xie, X.-J. Ju, W. Wang, Q. Chen, L.-Y. Chu, Nano-structured smart hydrogels with
rapid response and high elasticity, 4(2013) 2226.
[52] F. Carpi, Electromechanically active polymers : a concise reference, Cham: Springer; 2016.
[53] Q. Zhao, J.W.C. Dunlop, X. Qiu, F. Huang, Z. Zhang, J. Heyda, J. Dzubiella, M. Antonietti, J. Yuan,
An instant multi-responsive porous polymer actuator driven by solvent molecule sorption, Nature
Communications, 5(2014) 4293.
[54] Y. Osada, H. Okuzaki, H. Hori, A polymer gel with electrically driven motility, Nature, 355(1992)
242.
[55] D. Sangian, S. Naficy, G.M. Spinks, Thermally activated paraffin-filled McKibben muscles, Journal
of Intelligent Material Systems and Structures, 27(2016) 2508-16.
[56] S. Iijima, Helical microtubules of graphitic carbon, Nature, 354(1991) 56.
[57] A. Eatemadi, H. Daraee, H. Karimkhanloo, M. Kouhi, N. Zarghami, A. Akbarzadeh, M. Abasi, Y.
Hanifehpour, S.W. Joo, Carbon nanotubes: properties, synthesis, purification, and medical
applications, Nanoscale Research Letters, 9(2014) 393-.
[58] T. Filleter, R. Bernal, S. Li, H.D. Espinosa, Ultrahigh Strength and Stiffness in Cross-Linked
Hierarchical Carbon Nanotube Bundles, Advanced Materials, 23(2011) 2855-60.
[59] R.H. Baughman, C. Cui, A.A. Zakhidov, Z. Iqbal, J.N. Barisci, G.M. Spinks, G.G. Wallace, A.
Mazzoldi, D. De Rossi, A.G. Rinzler, O. Jaschinski, S. Roth, M. Kertesz, Carbon Nanotube Actuators,
Science, 284(1999) 1340.
[60] T. Mirfakhrai, O. Jiyoung, K. Mikhail, F. Eddie Chi Wah, Z. Mei, F. Shaoli, H.B. Ray, D.W.M. John,
Electrochemical actuation of carbon nanotube yarns, Smart Materials and Structures, 16(2007) S243.
[61] T. Mirfakhrai, J.D.W. Madden, R.H. Baughman, Polymer artificial muscles, Materials Today,
10(2007) 30-8.
[62] J.D.W. Madden, J.N. Barisci, P.A. Anquetil, G.M. Spinks, G.G. Wallace, R.H. Baughman, I.W.
Hunter, Fast Carbon Nanotube Charging and Actuation, Advanced Materials, 18(2006) 870-3.
[63] J. Lee, T. Aida, "Bucky gels" for tailoring electroactive materials and devices: the composites of
carbon materials with ionic liquids, Chemical Communications, 47(2011) 6757-62.
[64] M. Tunckol, J. Durand, P. Serp, Carbon nanomaterial–ionic liquid hybrids, Carbon, 50(2012)
4303-34.
[65] J.D. Carrico, M. Fleming, M.A. Tsugawa, K.K. Leang, Chapter 11 Precision Feedback and
Feedforward Control of Ionic Polymer Metal Composite Actuators, Ionic Polymer Metal Composites
(IPMCs): Smart Multi-Functional Materials and Artificial Muscles, Volume 1, The Royal Society of
Chemistry2016, pp. 354-85.
[66] F. Carpi, Electromechanically active polymers, 1st ed., Place of publication not identified:
Springer; 2016.

13
Chapter 1: Introduction

[67] K.J. Kim, S. Tadokoro, Electroactive polymers for robotic applications : artificial muscles and
sensors, London: Springer; 2007.
[68] F. Carpi, E. Smela, Biomedical applications of electroactive polymer actuators, Chichester, West
Sussex: John Wiley & Sons; 2009.
[69] C. Jo, D. Pugal, I.-K. Oh, K.J. Kim, K. Asaka, Recent advances in ionic polymer–metal composite
actuators and their modeling and applications, Progress in Polymer Science, 38(2013) 1037-66.
[70] S. Nemat-Nasser, Micromechanics of actuation of ionic polymer-metal composites, Journal of
Applied Physics, 92(2002) 2899-915.
[71] M. Safari, L. Naji, R.T. Baker, F.A. Taromi, The enhancement effect of lithium ions on actuation
performance of ionic liquid-based IPMC soft actuators, Polymer, 76(2015) 140-9.
[72] J.-W. Lee, S.M. Hong, C.M. Koo, High-performance polymer ionomer–ionic liquid membrane
IPMC actuator, Research on Chemical Intermediates, 40(2014) 41-8.
[73] B.J. Akle, M.D. Bennett, D.J. Leo, High-strain ionomeric–ionic liquid electroactive actuators,
Sensors and Actuators A: Physical, 126(2006) 173-81.
[74] M.D. Bennett, D.J. Leo, Ionic liquids as stable solvents for ionic polymer transducers, Sensors
and Actuators A: Physical, 115(2004) 79-90.
[75] H. Shirakawa, E.J. Louis, A.G. MacDiarmid, C.K. Chiang, A.J. Heeger, Synthesis of electrically
conducting organic polymers: halogen derivatives of polyacetylene, (CH), Journal of the Chemical
Society, Chemical Communications, (1977) 578-80.
[76] C.K. Chiang, C.R. Fincher, Y.W. Park, A.J. Heeger, H. Shirakawa, E.J. Louis, S.C. Gau, A.G.
MacDiarmid, Electrical Conductivity in Doped Polyacetylene, Physical Review Letters, 39(1977) 1098-
101.
[77] S. Kirchmeyer, K. Reuter, Scientific importance, properties and growing applications of poly(3,4-
ethylenedioxythiophene), Journal of Materials Chemistry, 15(2005) 2077-88.
[78] R.H. Baughman, Conducting polymer artificial muscles, Synthetic Metals, 78(1996) 339-53.
[79] Q. Pei, O. Inganäs, Electrochemical applications of the bending beam method; a novel way to
study ion transport in electroactive polymers, Solid State Ionics, 60(1993) 161-6.
[80] T.F. Otero, Biomimetic Conducting Polymers: Synthesis, Materials, Properties, Functions, and
Devices, Polymer Reviews, 53(2013) 311-51.
[81] F. Vidal, C. Plesse, D. Teyssié, C. Chevrot, Long-life air working conducting semi-IPN/ionic liquid
based actuator, Synthetic Metals, 142(2004) 287-91.
[82] F. Vidal, C. Plesse, G. Palaprat, A. Kheddar, J. Citerin, D. Teyssié, C. Chevrot, Conducting IPN
actuators: From polymer chemistry to actuator with linear actuation, Synthetic Metals, 156(2006)
1299-304.
[83] A. Maziz, C. Plesse, C. Soyer, C. Chevrot, D. Teyssié, E. Cattan, F. Vidal, Demonstrating kHz
Frequency Actuation for Conducting Polymer Microactuators, Advanced Functional Materials,
24(2014) 4851-9.
[84] G. Cellot, P. Lagonegro, G. Tarabella, D. Scaini, F. Fabbri, S. Iannotta, M. Prato, G. Salviati, L.
Ballerini, PEDOT:PSS Interfaces Support the Development of Neuronal Synaptic Networks with
Reduced Neuroglia Response In vitro, Front Neurosci, 9(2015) 521.
[85] S.M. Richardson-Burns, J.L. Hendricks, B. Foster, L.K. Povlich, D.H. Kim, D.C. Martin,
Polymerization of the conducting polymer poly(3,4-ethylenedioxythiophene) (PEDOT) around living
neural cells, Biomaterials, 28(2007) 1539-52.
[86] D.D. Ateh, H.A. Navsaria, P. Vadgama, Polypyrrole-based conducting polymers and interactions
with biological tissues, J R Soc Interface, 3(2006) 741-52.
[87] A. Della Santa, A. Mazzoldi, D. De Rossi, Steerable microcatheters actuated by embedded
conducting polymer structures, Journal of Intelligent Material Systems and Structures, 7(1996) 292-
300.
[88] E.W. Jager, Immerstrand, C., Peterson, K.H. et al. , The Cell Clinic: Closable Microvials for Single
Cell Studies, Biomedical Microdevices, 4(2002) 10.

14
Chapter 1: Introduction

[89] G.M. Spinks, G.G. Wallace, J. Ding, D. Zhou, B. Xi, J. Gillespie, Ionic Liquids and Polypyrrole Helix
Tubes: Bringing the Electronic Braille Screen Closer to Reality, Proceedings of SPIE - The
International Society for Optical Engineering2003, pp. 372-80.
[90] N.H. Runyan, F. Carpi, Seeking the 'holy Braille' display: might electromechanically active
polymers be the solution?, Expert Rev Med Devices, 8(2011) 529-32.
[91] A. Maziz, C. Plesse, C. Soyer, E. Cattan, F. Vidal, Top-down Approach for the Direct Synthesis,
Patterning, and Operation of Artificial Micromuscles on Flexible Substrates, ACS Applied Materials &
Interfaces, 8(2016) 1559-64.
[92] A. Khaldi, A. Maziz, C. Plesse, C. Soyer, D. Teyssié, F. Vidal, E. Cattan, Patterning innovative
conducting interpenetrating polymer network by dry etching, IEEE/ASME International Conference
on Advanced Intelligent Mechatronics, AIM2014, pp. 1424-9.
[93] A. Khaldi, C. Plesse, C. Soyer, E. Cattan, F. Vidal, C. Legrand, D. Teyssié, Conducting
interpenetrating polymer network sized to fabricate microactuators, Applied Physics Letters,
98(2011) 164101.

15
Chapter 2: PEDOT-based trilayer fabrication process

Chapter 2: PEDOT-based trilayer fabrication

process

Contents

2.1 Introduction ........................................................................................................................... 17

2.2 The selection of materials for CP-based trilayer actuators ................................................... 22
2.2.1 Electrodes of the microactuators ................................................................................. 22
2.2.2 Solid polymer electrolyte layer ..................................................................................... 22
2.2.3 The electrolyte .............................................................................................................. 24
2.2.4 Microactuator fabrication technique ............................................................................ 26
2.3 Materials ................................................................................................................................ 26
2.4 PEDOT synthesis route........................................................................................................... 27
2.5 PEDOT-based trilayer fabrication process ............................................................................. 30
2.5.1 Trilayer fabrication process........................................................................................... 30
2.6 PEDOT-based trilayer patterning ........................................................................................... 33
2.6.1 Fabrication of samples for the characterization process .............................................. 35
2.7 Analysis of the texture of the trilayer structure .................................................................... 36
2.8 Conclusion.............................................................................................................................. 39

16
Chapter 2: PEDOT-based trilayer fabrication process

2.1 Introduction
The actuation mechanism of conducting polymer (CP) actuators is based on their reversible
redox chemistry. The actuation is obtained when the CP is oxidized or reduced
electrochemically [1]: ions and solvent molecules, coming from a surrounding electrolyte, are
inserted or expelled from the CP in order to ensure the overall electroneutrality, which results
in a variation of the CP volume, or in short, the CP actuator responds to the consumed charge.
The degree of oxidation of the CPs, or in other words, the density of the polarons that carry
electronic charge in conducting polymers, can be controlled by the applied voltage. These ion
movements are principally responsible for the volume changes and, thus, the actuation
response that is referred to as electro-chemo-mechanical deformation.

Fig. 1 shows two models of operating mechanism of actuation in CPs [2, 3]. By electrochemical
reduction, the CPs experience two paths that depend on the size of doping anions. When
small anion like ClO4- is involved, the anion moves out from the bulk of CPs into electrolyte
during electrochemical reduction. In this path, the CPs is shrunk in reduced state and the
actuation is named anion driven, as shown by Fig. 1a. Otherwise, when a large negative cluster
like 1-Ethyl-3-methylimidazolium (EMI+) bis-trifluoromethan sulfonylimide (TSFI-) is used, it is
twisted and immobilized in polymer network [4]. In this cation-driven path, the CPs are
reduced by insertion of cations (Li+, or EMI+ for example) from the electrolyte, and then
swollen by the additional cations [5], as described in Fig. 1b.

Fig. 1. Actuation mechanism of conducting polymer actuators: a) anion driven mechanism, b) cation
driven mechanism [6] (Reprinted with permission from Springer Nature). The ovals represent positive
electronic charge on the backbone.

The cation driven reaction of PEDOT, one of the most widely studied conjugated polymers, is
described in Fig. 2. Providing an ionically conducting medium is available, PEDOT can be used

17
Chapter 2: PEDOT-based trilayer fabrication process

as the active component in CP actuators. To operate in air, CP actuators are usually built using
a solid polymer electrolyte (SPE) film sandwiched between two PEDOT electrodes
(PEDOT//SPE//PEDOT) [7-9], as demonstrated in Fig. 3. The SPE layer plays a role as an ion
reservoir as well as a supporting structure deciding the mechanical strength of actuator.
When two CP electrodes are subjected to an applied voltage (typically smaller than ±2 V), one
CP layer will be reduced leading to an increase in volume whereas the other layer will be
oxidized and shrunk in volume. Finally, the bending movement is induced.

Fig. 2. Cation driven actuation in a PEDOT-based actuator.

The expansion and speed of the actuation of CP actuators are affected by the
electrochemomechanical properties of the polymer, the diameter and the speed of the ions
moving into and out of the polymer, as well as the SPE’s ability to facilitate ion movement.
These key factors will be discussed in detail in chapters 4.

Fig. 3. Bending actuation mechanism of PEDOT-based trilayer actuators. The anode contracts due to
ion expulsion and the cathode expands due to cation incorporation, leading to bending.

18
Chapter 2: PEDOT-based trilayer fabrication process

Fig. 4. A possible model taking into account two electrochemical doping process: some ions (II),
deeply trapped near the polymer chains are released at only very low potentials. Other ions (I) are
shallowly trapped, they form a double layer at a certain distance from the chain. (Reproduced with
permission from [10], The Electrochemical Society.)

In addition to the redox process happening in the CP electrodes as mentioned above, another
possible contribution to the bending of the actuator comes from the forming of double layer
between the electrolyte and the CP’s nanostructure [10, 11]. Fig. 4 shows a two-step doping
process of polypyrrole conducting polymer thin film in lithium perchlorate in propylene
carbonate (1M). At low potential in redox process, the ionic double capacitance layer is
formed at the surface of the polymer chains arising from the accumulation near this chain of
the weakly trapped ions. When the applied potential reaches the oxidation potential of the
CP, a certain amount of ions approaches the polymer chains which are switched in their
oxidation state [10]. However, the identification of the percentage of this contribution into
the actuation mechanism remains open.

There were efforts to make use of CP actuators in different applications. Naka et al. have
developed a micropump driven by a PPy-based conducting polymer soft actuators [12]. The
micropump can transport fluids in one direction without backflow by means of two soft
actuators – consisting of a pair of conducting polymer driven bending structures - that open
and close (Fig. 5a). At a bigger scale, Fang et al. [13] introduced a diaphragm pump based on
circular CP actuators. The flow rate achieved with the diaphragm pump is 1260 µL/min, which
is achieved when the operating frequency is about 0.5 Hz.

19
Chapter 2: PEDOT-based trilayer fabrication process

Fig. 5. a) Schematic of a micropump driven by a PPy-based CP actuator [12] (Reprinted with

permission from John Wiley and Sons), b) A diagram pump: the mechanism of flap check valves and a
bottom view of the assembled micropump [13] (Reprinted with permission from Elsevier), c) Micro
autofocus lens actuator that uses bending conducting polymer actuators [14] (Reprinted with
permission from Elsevier), d) A prototype of a robotic fish [15] (Reprinted with permission from IOP
Publishing)

An autonomously powered and controlled robotic fish is shown in Fig. 5d. This robot has an
active flexural joint tail fin, which can be activated through polypyrrole (PPy) based actuator
developed by McGovern et al. [15]. The authors demonstrated an ability to control wirelessly
in real time the robotic fish, and a directional control. The highest speed of the fish was
approximately 33 mm/s corresponding to a flapping frequency of 0.6 – 0.8 Hz.

Fig. 5c shows micro autofocus lens actuators using a trilayer actuator structuring of a
polyvinylidene difluoride (PVDF) membranes sandwiched between two layers of a poly
ethylenedioxythiophene/poly (styrene sulfonic acid) (PEDOT/PSS) [14]. The actuator was
fabricated by casting for easily producing bending conducting polymer actuators with high
mechanical strength.

Jager implemented several designs aimed at medical applications. A closable microvial for
single cell studies (Fig. 6a) [16] was fabricated from two polypyrrole/Au microactuators. It
might also act as a microrobot for handling micrometer-size objects (including cells) in

20
Chapter 2: PEDOT-based trilayer fabrication process

aqueous solution [17]. In an effort to individually control the microactuators and avoid the
short circuiting due to electrical connections, a microactuator fabrication method was
developed, where CP actuators were patterned using microfabrication technology including
photolithography [18].

Fig. 6. a) A schematic drawing of a closable microvials [16] (Reprinted with permission from Springer
Nature), b) A sequence of pictures (left) showing the grabbing and lifting of a 100 µm glass bead and
schematic drawings of the motion (right) [17] (Reprinted with permission from The American
Association for the Advancement of Science)

Wolff and Beiski proposed and patented a controllable drug delivery system where a
polypyrrole actuator was used in different microvalve configurations, that can be inserted
into the oral cavity of an adult human, for example as part of a dental prosthesis [19]. The
two biggest advantages of this drug delivery system are a novel route for drug delivery via the
buccal mucosa, and a controllable and miniaturized drug delivery system that is the size of
two teeth.

21
Chapter 2: PEDOT-based trilayer fabrication process

In conclusion, conducting polymer- based actuators have received significant attention

because they offer lot of advantages over other electroactive polymers such as dielectric
elastomers and piezoelectric polymers. CP-based actuators have been known for their highly
reversible, fast switching redox processes [20]. Also, they are lightweight, biocompatible,
produce high stress (~ 10 MPa) [21], high power/weight ratio [22], significant strain (up to
1%), require low operating voltages (typically < 2 Vpeak-peak (Vpp)) [23] in solutions or in
open-air condition [24], can be electronically controlled with high frequency response ( 50 Hz
or more) and are potentially suitable for microscale applications [24, 25].

In this chapter, we will introduce the selection process of the materials from the type of
conducting polymer, the solid polymer electrolyte, and the electrolyte, which are suitable for
the microactuator fabrication process. And then, we propose the materials to be used and
describe the Vapour Phase Polymerization (VPP) synthesis of PEDOT electrodes to provide
highly electroactive electrodes, suitable for effective microactuators. Monolayers of PEDOT,
bilayers of NBR-PEO/PEDOT have been also prepared for specific measurements. Trilayer
microbeams of the actuators will be prepared using the Layer-by-Layer (LbL) approach and
patterned by femtosecond laser. In the final step, the geometric properties of the resulting
devices are then characterized.

2.2 The selection of materials for CP-based trilayer actuators

2.2.1 Electrodes of the microactuators

Polypyrrole (PPy) and PEDOT are the most popular materials to prepare CP-based actuators.
PPy-based actuators have been known for their high actuation speeds, large strains, and
acceptability for various applications. However, their notable disadvantages are high rigidity,
low ionic conductivity, and suffering a considerable damage in over-oxidation. Whereas,
PEDOT-based actuators have been described as thermally and chemically stable actuators.
PEDOT itself shows high electrochemical stability and good conductivity [8, 26-28]. In
comparison to PPy, PEDOT is typically a higher conductivity material, which can reduce the
voltage drop along the length of the actuator and increase the actuation efficiency in general.
In addition, PEDOT and its based devices can be adapted to various fabrication processes, for
example spin coating process. These are the reasons why PEDOT is chosen as the CP material
used for the electrodes of the actuators in this study.

2.2.2 Solid polymer electrolyte layer

The performance of CP-based actuators does not depend only on the CP performance but
also on the SPE layer. The electrolyte medium providing the ions necessary to the redox
process has a great effect on electromechanical response of the actuators. Low ionic
conductivity of the SPE layer will lead to a slower response and higher energy loss, which is
further investigated in section 4.3.2. And a stiffer SPE reduces the magnitude of the actuation.

22
Chapter 2: PEDOT-based trilayer fabrication process

Since the CP materials are often brittle, the mechanical toughness and stretchability of the
actuators is mainly dependent on the SPE layer.

Therefore, a full interpenetrating polymer network (full-IPN) or a semi interpenetrating

polymer network (sIPN) is employed to meet the requirement of both ionic conductivity and
mechanical properties.

IPNs were discovered in the 1960s. Several group of researchers [29, 30] found out that if two
immiscible polymers were formed into an IPN, the glass-transition temperatures of these
polymers were shifted towards each other, resulting from an increase in the miscibility of
these polymers caused by the presence of cross-links. Based on the chemical bonding and
rearrangement pattern, IPN can be classified into: sIPN and full-IPN in terms of structure, or
sequential IPN and simultaneous IPN regarding their synthesis pathway.

In a full-IPN, shown in Fig. 7a, two polymers are present as cross-linked networks, however
there is negligible bonding between these polymers. These can be prepared sequentially or
simultaneously.

Fig. 7. Schematic of a) full-IPN and b) semi-IPN where the pink, blue, and empty dot represent for the
crosslinking between two polymer chains.

In a sIPN, only one cross-linked polymer is present as a host network (Fig. 7b). In comparison
to a full IPN, this can enhance the miscibility of the linear polymer in the hosted matrix.
However, the mobility of the linear polymer increases the ability of the phase segregation
when two components are not thermodynamically compatible. Applications of IPNs related
to this work are as solid polymer electrolytes for fuel cell and as soft mechanical actuators
[28, 31, 32].

Different partner materials used to fabricate an “ideal” IPN for actuator applications have
been investigated. Poly(ethylene oxide) (PEO) is normally the first candidate in an IPN since it
is well known for its high ionic conductivity property, which can reach 0.1 S/cm when
incorporating the ionic liquid EMITFSI at ambient temperature [33]. The synthesis of this PEO
network is via free radical copolymerization of poly(ethylene glycol) methyl ether

23
Chapter 2: PEDOT-based trilayer fabrication process

methacrylate and poly(ethylene glycol) dimethacrylate. However, PEO films show poor
mechanical properties and weak dimensional stability [34]. Adding a second networkto PEO
[35], namely hydroxytelechelic polybutadiene (HTPB) [36], or polytetrahydrofurane (PTHF)
[37], or nitrile butadiene rubber (NBR) [32], was proposed to improve the mechanical
properties.

Table 1 describes the most important parameters of HTPB/PEO, PTHF/PEO, and NBR/PEO
IPNs. As can be seen, among the three types of IPNs, NBR/PEO IPN shows quite good ionic
conductivity, high strain at break, small Young’s moduli, and can operate very high frequency
(in kHz range). In addition, the NBR/PEO IPN had shown its compatibility with the reactive
ionic etching process, which is an indispensable step in photolithography for patterning the
microbeam. These properties make NBR/PEO become a promising candidate for the
lithography process and microactuator system.

Table 1. Summary some important properties of three different PEO-based IPNs

HTPB/PEO PTHF/PEO NBR/PEO

Ionic conductivity (S/m) 0.25 [38] 0.04 [37] 0.1 [33]
(incorporated with EMITFSI)
Elongation at break in dry state N/A 110 [37] 150 [33]
(%)
Young’s moduli (MPa) 1.2 [38] 6 [37] 0.8 [33]
Maximum operating frequency 18 [37] 125 [37] 900 [39]
(Hz)
Durability (cycles) 3.5 x 106 [28] N/A N/A

2.2.3 The electrolyte

A variety of types of electrolytes have been used to incorporate in microactuators. Solutions

of lithium perchlorate (LiClO4) and lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) in
propylene carbonate (PC) are the powerful tools for microactuators since they display a high
ionic conductivity and can be tuned by changing their concentration in PC. However, one of
their notable drawbacks is volatility, which quickly reduces the actuator performance after a
short amount of time of continuous operation. Encapsulation of the actuators is necessary for
a long-term functioning.

Ionic liquids can get over these problems since they have a large window of electrochemical
stability, high thermal stability, and non-volatility, so thereby non-flammability [40, 41].
However, due to a high viscosity, the conductivity of ionic liquids in general is not as high as
solvent based electrolyte. 1-ethyl-3-methylimidazolium bis(tri- fluoromethylsulfonyl)imide
(EMITFSI), 1-allyl-3-methylimidazolium bis(trifluoromethylsulfonyl) imide (AMITFSI), 1-butyl-
3- methylimidazolium bis(trifluoromethylsulfonyl)imide (BMITFSI), 1-methyl-1-propylpiper-
idinium bis(trifluoromethylsulfonyl)imide (MPPPTFSI), 1-ethyl-3-methylimidazolium bis(fluor-

24
Chapter 2: PEDOT-based trilayer fabrication process

osulfonyl)imide (EMIFSI), 1- butyl-1-methylpyrrolidinium bis(fluorosulfonyl)imide (Pyr14FSI),

1-butylpyridinium tetra-fluoroborate (BuPyBF4), and 1-ethyl-3-methylimidazolium
tetrafluoroborate (EMIBF4) are the most commonly used to incorporate into SPE layers in
microactuators. Fig. 8 shows the properties of variety of ionic liquids and the strain difference
their produce when they are incorporated in PEDOT-NBR/PEO-PEDOT actuators. The EMITFSI
displays a medium ionic conductivity, quite high maximum swelling, and highest strain
difference.

These are the reason why EMITFSI is chosen as the electrolyte to incorporate to the
microactuators in our study.

Fig. 8. a) properties of ionic liquids incorporated in NBR/PEO matrix, and b) strain differences calculated
for PEDOT-NBR/PEO-PEDOT actuator swollen with different ionic liquids [42] (Reprinted with
permission from Elsevier).

25
Chapter 2: PEDOT-based trilayer fabrication process

2.2.4 Microactuator fabrication technique

For microscale actuators, the thickness must be less than 20 µm to be compatible with, in
particular, etching processes, which is the reason why the layer-by-layer (LbL) approach has
been developed recently [43]. The bending microactuators were fabricated by sequentially
stacking layers using a layer-by-layer polymerization of conducting polymer electrodes and a
SPE. Each layer was deposited by spin-coating to accurately control the thicknesses. PEDOT
can be fabricated in-situ via electropolymerization [44] or via vapor phase polymerization
(VPP) [45] (see 2.4). In this work the first and last layers of the actuators were obtained using
VPP of 3,4-ethylenedioxythiophene (EDOT). The intermediate SPE layer was synthesized as a
semi-IPN combining a poly(ethylene oxide) (PEO) network for ionic conductivity and nitrile
butadiene rubber (NBR) for mechanical strength [9, 27] (described in 2.5).

For the LbL process, while the proof of concept has been demonstrated previously, the
resulting performances remained relatively poor (0.13% strain and 0.75 µN of generated
force) [43] mainly due to the low electroactivity of PEDOT electrodes obtained by VPP. The
process can also be difficult, given the fine dimensions required, and yields are low. Often it
takes many attempts to achieve one or two successful samples. Practical use of such
microactuators and integration into innovative microelectromechanical systems require
improved performance in mechanical response, and ideally also in yield. In this thesis we
show good electrical and mechanical response, and the process has been improved. Further
work will still be required to increase yield – perhaps through some optimization of the
chemistry.

2.3 Materials
Poly(ethylene glycol) methyl ether methacrylate (PEGM, Mn = 500 g mol-1), poly(ethylene
glycol) dimethacrylate (PEGDM, Mn = 750 g mol-1), cyclohexanone (>99.8%) and 3,4-
ethylenedioxythiophene (EDOT, distilled under reduced pressure) were obtained from Sigma
Aldrich. Iron(III) p-toluene sulfonate Clevios™ CB 55 V2 (55 wt% in butanol, Fe(OTs)3 in BuOH)
was purchased from HERAEUS. 1-butanol (99%), initiator dicyclohexyl peroxydicarbonate
(DCPD), 1-ethyl-3-methylimidazolium bis(trifluoromethanesulfonyl)imide (EMITFSI 99.9%)
and nitrile butadiene rubber (NBR), were used as supplied from Alfa Aesar, Groupe Arnaud,
Solvionic and LANXESS, respectively. The molecular structure of these materials are described
in Fig. 9.

26
Chapter 2: PEDOT-based trilayer fabrication process

Fig. 9. Molecular formula of a) the precursor PEGM, b) PEDGM, c) the linear chain NBR, d) the
initiator DCPD, and e) the ionic liquid EMITFSI [46] (Reprinted with permission from American
Chemical Society, 2008.)

2.4 PEDOT synthesis route

The PEDOT polymerization occurs through a three step mechanism (Fig. 10). The reaction is
initiated by an Fe(OTs)3, which generates a radical cation Fe3+. Radical termination of two
activated EDOT monomers provides carbon–carbon coupling. Typically, the anion (OTs)- acts
as a dopant to the polymer, stabilizing the positive charge on the sulfur. π-conjugation refers
to the alternating single and double bonds of the EDOT oligomers that typically result from
this synthesis process. This π-conjugation delocalizes the electrons and decreases the
oxidation potential. Consequently, existing PEDOT chains develop faster than new ones are
created since oligomers are more easy to oxidize than the monomers. The neutral PEDOT

27
Chapter 2: PEDOT-based trilayer fabrication process

chains are oxidized and a positive charge along the backbone with the ratio of every three or
four chain segments per charge is created. A “dopant” anion is electrostatically attracted to
the polymer and balances the charge [47]. The oxidation strength of the oxidant plays an
important role in PEDOT polymerization. At one hand, it must be high enough to drive the
reaction, but on the other hand it should be low enough to avoid unwanted bonding split in
the monomer leading to imperfection bonds, which can disrupt conjugation and reduce
conductivity [48, 49].

Fig. 10. Oxidation of EDOT with iron (III) p-toluene sulfonate

The fabrication process of the PEDOT electrodes was performed as described in the work of
Maziz [50] and is depicted in Fig. 11.

Fig. 11. Fabrication process of PEDOT electrodes [51]

28
Chapter 2: PEDOT-based trilayer fabrication process

The PEDOT electrode layers were obtained from EDOT VPP, first described by Winther-Jensen
[45]. The VPP of EDOT was carried out with a direct chemical oxidation using the commercial
solution of Iron(III) tosylate Fe(OTs)3 (55 wt% in butanol). In a departure from Winther-Jensen,
PEO precursors (mPEG) composed of 50 wt% of PEGM and 50 wt% of PEGDM were added as
a monomer and a crosslinker, respectively, to the Fe(OTs)3 solution. Additionally, DCPD (3
wt% vs mPEG) as the radical initiator for mPEG polymerization was introduced to the solution.
The latter was then stirred until dissolution and degassed. The oxidant solution was then spin-
coated with different rotational speeds from 1000 rpm to 3500 rpm (the acceleration and
duration were held constant at 1000 rpm s-1, and 30 s, respectively) onto a substrate and
exposed to EDOT vapor for a fixed time at a fixed temperature. EDOT VPP was carried out
under primary vacuum by heating liquid EDOT monomers that evaporate and polymerize on
the oxidant solution. The EDOT VPP was followed by the final heat treatment for mPEG
polymerization/crosslinking (3 h at 50°C) to obtain PEDOT/PEO composite electrodes. The
resulting electrode can be either i) washed in methanol to remove unreacted oxidant solution
and weakly bonded PEDOT chains, or ii) left as it is for subsequent trilayer fabrication, which
is described in the next section. This PEDOT electrode was swollen in ionic liquid (EMITFSI) for
further characterization. It is worth noting that introducing the PEO into PEDOT electrodes
helps to increase the specific capacitance, the cycling stability, and flexibility of the electrodes
[52], which improves the strain and the life time of the actuator.

The final properties (thickness, electronic conductivity and volumetric charge density) of the
PEDOT electrodes were studied and optimized [51] as a function of: composition of oxidant
solution (content of mPEG), rotation speed of spin-coater for deposition of oxidant solution,
as well as EDOT VPP time and temperature. The optimization of the PEDOT electrodes
performed by K. Rohtlaid [51] in a cooperating project with our group is detailed in the
Appendix A.2.1. Briefly, it can be concluded that to obtain PEDOT electrodes with high
electronic conductivity and high electroactivity using VPP, it is necessary to add PEO
precursors to the oxidant solution while not increasing their concentration above 20%. The
rotation speed can be chosen according to the desired CP thickness. In addition, the
polymerization time and temperature must not exceed 90 min and 50°C, respectively, in order
to maintain both, high electronic conductivity and good electrochemical properties. The
optimal parameters resulted in good electronic conductivities (around 200 S/cm) and high
electroactivity (between 2.3 x 107 C/m3 and 1.0 x 108 C/m3). This conductivity is much lower
than the previous conductivity value reported by Maziz [50] (before the PEDOT electrode is
swollen in EMITFSI), which is equal to 805 S/cm - corresponding to an mPEG content of 30%,
a rotational speed of the spin coater of 1000 rpm, a duration of 30 s, and the VPP duration of
30 min, producing a PEDOT electrode thickness of 0.34 µm. This decrease in the electronic
conductivity is attributed to different factors. At first, in our case, a thicker PEDOT layer was
fabricated by the VPP process. This was performed in a Sawatec vacuum hot plate instead of
a bell jar, providing a higher density of EDOT vapor - resulting in a higher PEDOT density than

29
Chapter 2: PEDOT-based trilayer fabrication process

is achievable in the bell jar. Thicker PEDOT layers show a lower electronic conductivity than
thinner ones [53]. Secondly, the PEO content in oxidant solution was below 20 % in order to
maintain a high volumetric charge density - at the cost of low electronic conductivity.

To fabricate samples for further characterization, the synthesis parameters are selected as
follow: mPEG content: 10%; EDOT VPP time: 50 min; EDOT VPP temperature: 40 oC;
acceleration: 1000 rpm s-1, duration: 30 s to obtain a high electronic conductivity and
volumetric charge density. With those parameters, the relation between the rotational speed
of the spin coater and the PEDOT thickness is determined as in Fig. 16, where the oxidant
solution was deposited on the silicon wafer. Finally, the rotational speed is chosen as 2000
rpm and the expected thicknesses of the PEDOT electrodes are 2.2 µm providing
experimentally highest bending strain in the trilayer configuration.

Fig. 12. Relation between the rotational speed of the spin-coater for deposition of the oxidant
solution on a silicon wafer and the thickness of the PEDOT electrode, where the mPEG content is 10%
of oxidant solution

2.5 PEDOT-based trilayer fabrication process

2.5.1 Trilayer fabrication process

In this section, the fabrication process of the CP-based trilayer actuators will be described. As
mentioned earlier, this process was developed from the LbL method reported by Maziz [43].
However, some noticeable changes have been made, such as the optimization of the PEDOT
electrodes described in the previous section, resulting in the increase of VPP time from 30
min to 50 min, a different vacuum hotplate (Sawatec HP-200) allowing a more precise
temperature and vacuum control, replacing the vacuum bell during the VPP. The fabrication
process was performed in a clean room environment (Class 100), which, given the potential
effects of humidity and temperature on the process, should ensure a higher reproducibility.

30
Chapter 2: PEDOT-based trilayer fabrication process

The trilayer fabrication process is summarized as in Fig. 13. This process includes multiple
steps where a trilayer structure was first coated and synthesized on a silicon wafer, followed
by the heat treatment during three hours and femtosecond laser patterning to produce a
cantilever beam shape. Lastly, the patterned beams were swollen in ionic liquid for
approximately one week. The details of each step will be described below.

Fig. 13. Microactuator fabrication process showing the multilayer process followed by laser
micromachining

The trilayer actuator was synthesized using the LbL method. In this process, all the layers are
spin-coated on top of the previous layers and the thickness can be controlled extremely
accurately by adapting the speed of rotation. The oxidant solution was prepared by adding
PEO precursors (50 wt% PEGM and 50 wt% PEGDM) into a commercial Fe(OTs)3 solution. The
PEO precursors (mPEG) were added to each layer and polymerized at the end of the process
throughout the trilayer structure to improve adhesion between the layers using the same
ratios as described above. The solution was then stirred for 10 min and spin-coated onto a
two-inch silicon wafer. The EDOT VPP was carried out in the same manner as in the PEDOT
electrode fabrication section - during 50 min and at a 40°C.

During the next step, the SPE layer, based on a semi-IPN architecture according to a procedure
described in [43], was prepared: The NBR solution (20wt% NBR and 80wt% cyclohexanone)
was mixed with PEO precursors (75 wt% PEGM and 25 wt% PEGDM vs. NBR). A free radical
initiator DCPD (3 wt% vs PEO precursors) was then added to the solution, stirred for 15 min
and degassed. A homogeneous mixture was obtained and spin-coated on top of the first
PEDOT electrode layer. The relationship between the spin coating speed and the thickness of
the NBR/PEO was investigated and shown in Fig. 14, which is similar to the results reported

31
Chapter 2: PEDOT-based trilayer fabrication process

by Maziz [39]. A rotational speed of 2000 rpm, 1000 rpm s-1, 30 s was chosen to fabricate the
SPE layer in this step. The pre-polymerization was carried out under a continuous supply of
N2 at 50°C during 45 min to enable the formation of the PEO network and to improve the
adhesion between the first two layers.

Fig. 14. Relationship between rotation speed of spin-coater for deposition of NBR/PEO and its
resulting thickness

The second PEDOT electrode layer was synthesized in the same way as the first one. The
oxidant solution was prepared, spin coated on top of the PEDOT/SPE bilayer, and EDOT VPP
was carried out under the vacuum in the same way as for the first electrode. During this step,
DCPD (3 wt% vs PEO network) was also added to the oxidant solution to initiate the
polymerization of the PEO network.

After EDOT VPP of the second electrode layer, the trilayer was placed under a continuous flow
of N2 for a final heat treatment step (3 h at 50°C) [43] in order to polymerize the PEO
precursors in every layer. This step provides a final intralayer cross-linking and interlayer
cobonding [5]. It is worth mentioning that there is no DCPD in the first PEDOT layer as it would
result in the early polymerization of the PEO precursors thus reducing the final cross-linking
with the SPE layer during the final heat treatment. However, during the final heat treatment,
the initiator from the second and third layers is assumed to initiate the polymerization of the
PEO precursors in the first layer to improve adhesion between all three layers. Finally, the
actuator was washed in methanol to remove any unreacted oxidant solution and any weakly
polymerized PEDOT. Depending on the thickness of the oxidant solution, the washing solution
can turn different shades, shown in Fig. 15. As shown previously, a thick oxidant solution
produces a thick and dense PEDOT electrode layer. However, a lot of short dimers, trimers
and short chains of PEDOT exist in this case and they are removed during the washing process,
resulting a blue color of methanol. As the thickness of oxidant solution is reduced, the number

32
Chapter 2: PEDOT-based trilayer fabrication process

of short dimers, trimers and such are reduced significantly leading to a more transparency of
methanol solution.

Fig. 15. Trilayer structures during washing step in methanol, the rotational coating speeds of
PEDOT:NBR/PEO:PEDOT are a) 1000:2000:1000 b) 1500:2000:1500 c) 2500:2000:2500 d)
3000:2000:3000; the acceleration and coating duration are constant at 1000 rpm s-1 and 30 s,
respectively. Les coloration is seen during washing for the fast spun, thin layers of oxidant solution,
indicating less weakly polymerized EDOT.

In conclusion the following parameters for the fabrication of the trilayer actuators using LbL
synthesis were chosen in order to obtain a thick, highly conductive and electroactive
electrode: mPEG content of 10 wt%, oxidant solution applied with a coating speed,
acceleration, and duration of 1500 rpm, 1000 rpm s-1, and 30 s, respectively; and EDOT VPP
at 40°C for 50 min producing a 2.2 µm in thickness of PEDOT layer (on silicon). And a rotational
speed of 2000 rpm, 1000 rpm s-1, 30 s was chosen to fabricate the SPE layer producing a 7 µm
in thickness of the NBR/PEO layer.

2.6 PEDOT-based trilayer patterning

In the past, various techniques have been investigated to pattern the CP-based trilayer
membrane into the desired beam shape at the micro-/nanoscale. Electropolymerization of
the CP materials onto a patterned metal electrode such as gold or platinum was used to
fabricate bilayer or trilayer structures, however the poor adhesion between gold layer and CP
as well SPE layer substantial reduced the life-time of the actuator, and the gold layer also
increases the total Young’s modulus of the actuator, resulting in a strain decrease [54-56].

33
Chapter 2: PEDOT-based trilayer fabrication process

Reactive-ion etching (RIE) was used to remove unwanted area on a CP-based membrane to
create high precision actuators [57, 58], however, this process involves the photolithography
process to deposit a sacrificial layer (poly(vinyl alcohol)), and photoresist (Megaposit SPR 220
7.0 photoresist) as locally protective layer, and so is considered to be a long and costly
process. Recently, a new patterning method has been introduced, where the oxidant solution
is printed using a syringe-based printing or microcontact printing followed by laser ablation
to cleanly cut the final device structures including the printed patterns [59]. This method
allows a versatile deposition of CP layer and can be used in mass productive process.
However, the resolution of the printing devices limits the precision of the resulting actuators.
Ultrafast laser micromachining was employed to pattern PPy-based actuators [60-63], which
works well. Use of electrical discharge machining is effective, but can be slow [64].

Among these patterning methods, contactless material processing by laser ablation

micromachining is chosen in our study enabling a mask-less, top-down, and flexible machining
technique. The trilayer structures are micro-patterned into microbeams using laser cutting
(Oxford Lasers, a THG crystal ( Third Harmonic generation) give the wavelength at 343 nm and
300 femtosecond diode-pumped (DPSS) lasers at 1030 nm (in IR), pulse length: 400 fs, spot
size: 10 µm, galvanometric mirror resolution: 1 µm, maximum power 2 W with frequency rate
at 200 kHz, the sample stage up to 300 × 300 mm2 with linear accuracy +/- 0.5 µm,
repeatability +/- 0.2µm; patterning conditions: power: 10%, cutting speed: 10 mm/s, cut
passes: 30 passes) and a resulting beam is shown in Fig. 16b. Femtosecond laser patterning
offers a high precision, small heat-affected zone, high reproducibility, short patterning time
in comparison to other patterning methods, and can be applied before or after swelling of the
trilayer structure. In this study, beams are designed with various geometries, where the
length and the width are varied from 4 mm to 7 mm and 0.8 mm to 1 mm, respectively. Here,
the ratio between the length and the thickness is from 200 to 350, which is high enough to
assume this is a thin beam. In addition, all the characterizations later are manually handled,
the beam size then must be big enough to facilitate the experiments. The rectangular contact
pad (length x width = 3 mm x2 mm) is added at one end of the beam in order to allow a large
clamping area, which is used to fix the beam and create the electrical contact to supply the
voltage.

A trilayer coated wafer, after the drying process, is shown in Fig. 16a. And in Fig. 16b, a cut
out trilayer actuator is seen in top view. After cutting, microbeams were then swollen in ionic
liquid (EMITFSI) until they reached swelling saturation point (around one week [39, 43])
enabling the incorporation of the ions necessary for the redox process (Fig. 16c). SEM images
(similar to those in Fig. 16c) show that the beam does not increase its length and width after
the swelling.

34
Chapter 2: PEDOT-based trilayer fabrication process

Fig. 16. a) The wafer after drying process, the white trace on the top right indicating the area has
been peeled off for characterization: EDX and SEM, b) the resulting trilayer actuator after laser
patterning process, the dimensions of the trilayer are length x width = 5 mm x 1 mm c) SEM images
of the trilayer after swollen in EMITFSI

2.6.1 Fabrication of samples for the characterization process

In the previous section, the LbL fabrication process used to produce the trilayer actuators has
been introduced. To characterize the electrochemomechanical properties of this resulting
device, samples corresponding to different measurements have been fabricated, as
summarized in Table 2.

Table 2. Summary the sample configuration for different characterizations

Single layer Single layer Top PEDOT on Bottom PEDOT:NBR/P

PEDOT NBR/PEO NBR/PEO PEDOT: EO:PEDOT
NBR/PEO
Surface x x x
Roughness
Ionic x x x
conductivity
Electronic x x
conductivity

35
Chapter 2: PEDOT-based trilayer fabrication process

Volumetric x
capacitance
Possible x
short circuit
Strain-to- x
charge ratio
Young’s x x
modulus
Damping x
ratio
Beam x
density
Strain x
difference
Linear strain x
Blocking x
force

A single sIPN of NBR/PEO (50/50) was fabricated for the ionic conductivity, and Young’s
modulus measurement.

To investigate the electronic and ionic conductivities of the PEDOT electrodes and their
asymmetric properties, a top bilayer (NBR/PEO coated with PEDOT/PEO) and bottom bilayer
(PEDOT coated with NBR/PEO) were manufactured, where the thickness of NBR/PEO is
increased to more easily distinguish between the ionic conductivity of PEDOT and of NBR/PEO
alone

The volumetric capacitance, the possible short circuit between two PEDOT electrodes, the
Young’s moduli, the damping ratio, and the linear strain of the trilayer structuring of PEDOT-
PEO/NBR-PEDOT were all measured on a fabricated trilayer produced as described in the
previous section.

It is worth of noticing that during the fabrication of a single PEDOT layer, the NBR/PEO, and
the bilayer, the initiator PCDP was added to each layer to polymerize the PEO. However, in
the trilayer configuration, the initiator was added into the membrane and the second PEDOT
layer only. This may lead to some differences in their properties.

2.7 Analysis of the texture of the trilayer structure

A trilayer sample was analyzed on EDX to confirm the three-layer structure, shown in Fig. 17a.
The green color represent sulfur, which is only present in PEDOT, and the red the carbon,
dominating in the NBR region

The sIPN morphology of the trilayer is described in Fig. 17b, where the PEO is covalently
crosslinked through all the trilayer improving the adhesion between the electrodes and the

36
Chapter 2: PEDOT-based trilayer fabrication process

SPE layer. The PEDOT chains are synthesized by VPP on the top and the bottom electrodes
and do not either crosslink to each other or inter-crosslink with PEO. The linear NBR chains
are only available in the middle layer.

Fig. 17. a) The EDX of the trilayer structure, b) The anticipated morphology of the trilayer structure
PEDOT:NBR/PEO:PEDOT

Table 2. Thicknesses of each layer before and after swelling in EMITFSI.

Trilayer Thickness before swelling (µm) Thickness after swelling (µm)

Bottom PEDOT 2.1±0.3 3.5±0.5
electrode
SPE layer 7.0±0.3 10±0.5
Top PEDOT electrode 2.1±0.3 3.5±0.5

The thicknesses of each layer before and after immersion in EMITFSI are listed in Table 2. The
thickness measurements have been performed on a set of five different trilayer samples
prepared with the same fabrication process and the consistency of the thickness values in the
table demonstrate the reproducibility of the fabrication process.

37
Chapter 2: PEDOT-based trilayer fabrication process

Fig. 18. a) SEM image of the trilayer cross section in dry state, profilometer surface roughness of the
b) top surface of the top electrode layer, c) top surface of the SPE of a bilayer PEDOT/SPE, d) top
surface of the bottom electrode layer, e) bottom surface of the bottom electrode layer

It is important to note that after swelling in EMITFSI, the total thickness of trilayer actuator
can be easily verified by SEM (Fig. 18a) but the thickness change of each layer is impossible to
determine directly on the trilayer either by SEM or EDX because the ionic liquid covers the
cross section of the trilayer. Therefore, a single SPE layer was fabricated using the same
parameters in comparison with the SPE layer in trilayer structure. This single layer was then
swollen in EMITFSI during one week. It has been found that SPE layer increases by 43% its
thickness after swelling. From this increasing percentage and the total thickness of the
trilayer, the thicknesses of two conducting polymer electrode layers can be estimated. The
error is calculated based on five measurements made at different position along the sample.
The swollen percentage of SPE layer shown here is quite consistent with the results reported
by Maziz [39] for the same type of SPE.

Despite the fact of sharing the same fabrication parameters, differences between the bottom
and top PEDOT layers were observed with the roughness measurements conducted when the
layers were still dry. It is worth noticing that the profilometer measurements in Fig. 18c and
d were performed on the top surface of the SPE of a bilayer PEDOT/SPE and on the top surface
of a single bottom PEDOT layer, respectively. The roughness was measured here while the
layers were still on a silicon wafer and the average values obtained are 0.5 µm and around
0.1 µm respectively. The surface roughness of the top surface of the top electrode seen Fig.

38
Chapter 2: PEDOT-based trilayer fabrication process

18b is also measured as the actuator is still fixed on the silicon wafer and the average value
obtained (0.7 µm) is similar to that measured on PEDOT/SPE. The average roughness on the
bottom surface of the bottom electrode is measured after lift-off and once it has been flipped
on over onto a new silicon wafer. The value obtained in this case is very small, around 0.06
µm (Fig. 18e) (refer to Appendix A.2.2 for the surface roughness measurement method on
different samples). The average roughness of the top PEDOT electrode surface is then ten
times higher than for the bottom electrode.

The LbL fabrication method may explain this roughness difference. The SPE layer is spin-
coated on the bottom PEDOT layer and the pre-heat treatment (for evaporating the solvent
and crosslinking PEO precursors) may result in a slight contraction of this layer, producing a
slight and relatively regular folding on its surface. This folding depends on the thickness of the
SPE layer and the way in which the heat treatment is performed. The top PEDOT layer only
reproduces the shape of the SPE surface. The average roughness of the PEDOT surface in
contact with open-air (top PEDOT electrode) (Fig. 18d) is, quite logically, greater than the one
in contact with silicon substrate (Fig. 18e). The supposed contraction of the SPE layer during
the trilayer fabrication process is visible once the microbeams are lifted off and swollen in
EMITFSI. A slight curvature of the beam often appears which extends systematically the
bottom layer of PEDOT previously in contact with the silicon wafer, and contracts the top
layer of PEDOT. This indicates the presence of a weak stress inside the microbeam.

Another possible contribution to the roughness difference may come from the bottom-up
forming mechanism of PEDOT during the VPP of EDOT. Brooke et al. [65] and Evans et al. [66]
have shown that during the VPP of EDOT, PEDOT growth is via a bottom-up forming
mechanism, where the oxidant solution is transported up to the surface to form a new PEDOT
layer. This mechanism suggests that a denser PEDOT distribution is expected at the bottom
PEDOT surface in contact with silicon, while the top PEDOT surface is less dense.

2.8 Conclusion
In this chapter, I have described the fabrication of a trilayer structure of a semi-NBR/PEO
sandwiched between two PEDOT electrodes, produced by stacking LbL. By introducing the
PEO into all three layers, the ionic conductivity and the mechanical adhesion among these
layers increases. This fabricating method enabled us to obtain a thin, controllable thickness.
From previous work, noticeable changes have been made, such as the optimization of the
PEDOT electrodes resulting in the increase of VPP time from 30 min to 50 min, a different
vacuum hotplate (Sawatec HP-200) allowing a more precise temperature and vacuum control
replaced the vacuum bell during the VPP, and all the fabrication process was performed in
the clean room environment, which ensures a higher reproducibility -ten batches of trilayer
samples were fabricated and the bending experiment shown that the trilayer actuators in four
batches were working. The percentage of the PEO in the PEDOT layer is also changed to 10%

39
Chapter 2: PEDOT-based trilayer fabrication process

in comparison with the 20 % used in the work of A. Maziz [67]. This percentage ensures a
balance between high electronic conductivity (around 200 S/cm) and high volumetric charge
density (between 2.3 x 107 C/m3 and 1.0 x 108 C/m3) of the PEDOT layer. Moreover, we have
highlighted a slight asymmetry in surface roughness between the top and the bottom PEDOT
layers which might have consequences on the electrical properties of the PEDOT electrodes
and finally on the mechanical properties (displacement, force …) of the actuator.

The VPP process in this thesis has used the conventional oxidant solution consisting of 55%
Fe(TOs)3 in butanol. However, literature has shown that this high percentage produces a
highly acidic and reactive oxidant solution, resulting in uncontrollable polymerization and
structural defects in the deposited film [53]. Investigation on optimizing the percentage of
oxidant solution, or adding base inhibitors such as pyridine to control the reaction rate are
suggested to further improve the electrochemical properties of the PEDOT electrode. This is
left to future work.

It is quite possible to think in the future to have a wider choice of trilayers. Indeed, the work
of K. Rotlaid and F. Ribeiro go in this direction. In the case of K. Rotlhaid, the PEDOT electrodes
are replaced by PEDOT: PSS with incorporated PEO, which allows for thicker actuators (around
30μm) - thereby producing more force (this work has not been published yet). In the case of
F. Ribeiro, the possibility of incorporating the ionic liquid into the SPE at the time of
manufacture is a promising innovation (this work is still unpublished). In fact, the swelling
phase of our actuators modifies the total thickness by 160%, which potentially can have
consequences on the transfer of electrodes in a complete integration step – where the metal
electrodes in contacting with the PEDOT layers will be expanded leading to the facture of the
electrode after the actuator is swollen. The manufacturing characteristics of these future
materials will have to be taken into account to have the necessary elements for a good
interpretation of the modeling and their experimental mode of operation.

The electrical and mechanical properties of each layer and the three-layer structure must also
be evaluated experimentally in order to bring the parameters of interest to the model. This
characterization process will be described in the next chapter.

40
Chapter 2: PEDOT-based trilayer fabrication process

[1] R.H. Baughman, L.W. Shacklette, R.L. Elsenbaumer, E.J. Plichta, C. Becht, Micro Electromechanical
Actuators Based on Conducting Polymers, in: P.I. Lazarev (Ed.) Molecular Electronics: Materials and
Methods, Springer Netherlands, Dordrecht, 1991, pp. 267-89.
[2] T. Sendai, H. Suematsu, K. Kaneto, Anisotropic strain and memory effect in
electrochemomechanical strain of polypyrrole films under high tensile stresses, Japanese Journal of
Applied Physics, 48(2009) 0515061-4.
[3] K. Kaneto, H. Hashimoto, K. Tominaga, W. Takashima, Shape retention in polyaniline artificial
muscles, Japanese Journal of Applied Physics, 50(2011).
[4] D. Mecerreyes, Applications of Ionic Liquids in Polymer Science and Technology, Berlin, Heidelberg:
Springer; 2015.
[5] D. Mecerreyes, Applications of ionic liquids in polymer science and technology, Heidelberg ; New
York: Springer; 2015.
[6] K. Asaka, H. Okuzaki, SpringerLink (Online service), SpringerLINK ebooks - Chemistry and Materials
Science (2014), Soft Actuators Materials, Modeling, Applications, and Future Perspectives, S.l.:
Springer Japan; 2014.
[7] J. Citerin, A. Kheddar, M. Hafez, F. Vidal, C. Plesse, D. Teyssie, C. Chevrot, Characterization of a new
interpenetrated network conductive polymer (IPN-CP) as a potential actuator that works in air
conditions, Intelligent Robots and Systems, 2004 (IROS 2004) Proceedings 2004 IEEE/RSJ International
Conference on2004, pp. 913-8 vol.1.
[8] U.L. Zainudeen, M.A. Careem, S. Skaarup, PEDOT and PPy conducting polymer bilayer and trilayer
actuators, Sensors and Actuators, B: Chemical, 134(2008) 467-70.
[9] C. Plesse, F. Vidal, H. Randriamahazaka, D. Teyssié, C. Chevrot, Synthesis and characterization of
conducting interpenetrating polymer networks for new actuators, Polymer, 46(2005) 7771-8.
[10] J. Tanguy, N. Mermilliod, M. Hoclet, Capacitive Charge and Noncapacitive Charge in Conducting
Polymer Electrodes, Journal of The Electrochemical Society, 134(1987) 795-802.
[11] T. Yeu, K.M. Yin, J. Carbajal, R.E. White, Electrochemical Characterization of Electronically
Conductive Polypyrrole on Cyclic Voltammograms, Journal of The Electrochemical Society, 138(1991)
2869-77.
[12] Y. Naka, M. Fuchiwaki, K. Tanaka, A micropump driven by a polypyrrole-based conducting polymer
soft actuator, Polymer International, 59(2010) 352-6.
[13] Y. Fang, X. Tan, A novel diaphragm micropump actuated by conjugated polymer petals:
Fabrication, modeling, and experimental results, Sensors and Actuators A: Physical, 158(2010) 121-31.
[14] K. Ikushima, S. John, A. Ono, S. Nagamitsu, PEDOT/PSS bending actuators for autofocus micro lens
applications, Synthetic Metals, 160(2010) 1877-83.
[15] S. McGovern, G. Alici, V.T. Truong, G. Spinks, Finding NEMO (novel electromaterial muscle
oscillator): A polypyrrole powered robotic fish with real-time wireless speed and directional control,
Smart Materials and Structures, 18(2009).
[16] E.W. Jager, Immerstrand, C., Peterson, K.H. et al. , The Cell Clinic: Closable Microvials for Single
Cell Studies, Biomedical Microdevices, 4(2002) 10.
[17] W.H.J. Edwin, Ingan, xe, O. s, Lundstr, xf, I. m, Microrobots for Micrometer-Size Objects in
Aqueous Media: Potential Tools for Single-Cell Manipulation, Science, 288(2000) 2335-8.
[18] E.W.H. Jager, N. Masurkar, N.F. Nworah, B. Gaihre, G. Alici, G.M. Spinks, Patterning and electrical
interfacing of individually controllable conducting polymer microactuators, Sensors and Actuators B:
Chemical, 183(2013) 283-9.
[19] T. Göttsche, S. Haeberle, Integrated Oral Drug Delivery System with Valve Based on Polypyrrole,
Biomedical Applications of Electroactive Polymer Actuators, John Wiley & Sons, Ltd2009, pp. 301-16.
[20] C.A. Cutler, M. Bouguettaya, J.R. Reynolds, PEDOT polyelectrolyte based electrochromic films via
electrostatic adsorption, Advanced Materials, 14(2002) 684-8.
[21] J.D.W. Madden, N.A. Vandesteeg, P.A. Anquetil, P.G.A. Madden, A. Takshi, R.Z. Pytel, S.R.
Lafontaine, P.A. Wieringa, I.W. Hunter, Artificial muscle technology: Physical principles and naval
prospects, IEEE Journal of Oceanic Engineering, 29(2004) 706-28.

41
Chapter 2: PEDOT-based trilayer fabrication process

[22] A. Mazzoldi, A.D. Santa, D. De Rossi, Conducting Polymer Actuators: Properties and Modeling, in:
Y. Osada, D.E. De Rossi (Eds.), Polymer Sensors and Actuators, Springer Berlin Heidelberg, Berlin,
Heidelberg, 2000, pp. 207-44.
[23] E. Smela, Conjugated Polymer Actuators for Biomedical Applications, Advanced Materials,
15(2003) 481-94.
[24] A. Khaldi, C. Plesse, C. Soyer, E. Cattan, F. Vidal, C. Legrand, D. Teyssié, Conducting
interpenetrating polymer network sized to fabricate microactuators, Applied Physics Letters,
98(2011).
[25] A. Khaldi, A. Maziz, C. Plesse, C. Soyer, D. Teyssié, F. Vidal, E. Cattan, Patterning innovative
conducting interpenetrating polymer network by dry etching, IEEE/ASME International Conference
on Advanced Intelligent Mechatronics, AIM2014, pp. 1424-9.
[26] L. Groenendaal, F. Jonas, D. Freitag, H. Pielartzik, J.R. Reynolds, Poly(3,4‐ethylenedioxythiophene)
and Its Derivatives: Past, Present, and Future, Advanced Materials, 12(2000) 481-94.
[27] F. Vidal, C. Plesse, G. Palaprat, A. Kheddar, J. Citerin, D. Teyssié, C. Chevrot, Conducting IPN
actuators: From polymer chemistry to actuator with linear actuation, Synthetic Metals, 156(2006)
1299-304.
[28] F. Vidal, C. Plesse, D. Teyssié, C. Chevrot, Long-life air working conducting semi-IPN/ionic liquid
based actuator, Synthetic Metals, 142(2004) 287-91.
[29] S.C. Kim, D. Klempner, K.C. Frisch, H.L. Frisch, Polyurethane Interpenetrating Polymer Networks.
II. Density and Glass Transition Behavior of Polyurethane-Poly(methyl methacrylate) and
Polyurethane-Polystyrene IPN's, Macromolecules, 9(1976) 263-6.
[30] G.M. Yenwo, L.H. Sperling, J. Pulido, J.A. Manson, A. Conde, Castor oil based interpenetrating
polymer networks, IV. Mechanical behavior, Polymer Engineering & Science, 17(2004) 251-6.
[31] M.-J. Lee, Y.S. Choi, Y.S. Kang, J.-H. Choi, M.-S. Kang, All-solid-state proton conductive membranes
prepared by a semi-interpenetrating polymer network (semi-IPN), Journal of Materials Chemistry,
22(2012) 18522-7.
[32] L.J. Goujon, A. Khaldi, A. Maziz, C. Plesse, G.T.M. Nguyen, P.-H. Aubert, F. Vidal, C. Chevrot, D.
Teyssié, Flexible Solid Polymer Electrolytes Based on Nitrile Butadiene Rubber/Poly(ethylene oxide)
Interpenetrating Polymer Networks Containing Either LiTFSI or EMITFSI, Macromolecules, 44(2011)
9683-91.
[33] F. Nicolas, M. Ali, P. Cédric, T. Dominique, C. Claude, V. Frédéric, Robust solid polymer electrolyte
for conducting IPN actuators, Smart Materials and Structures, 22(2013) 104005.
[34] D. Mecerreyes, Springer-Verlag (Berlin), Applications of ionic liquids in polymer science and
technology, Heidelberg: Springer; 2015.
[35] P. Verge, F. Vidal, P.-H. Aubert, L. Beouch, F. Tran-Van, F. Goubard, D. Teyssié, C. Chevrot, Thermal
ageing of poly(ethylene oxide)/poly(3,4-ethylenedioxythiophene) semi-IPNs, European Polymer
Journal, 44(2008) 3864-70.
[36] C. Plesse, F. Vidal, C. Gauthier, J.-M. Pelletier, C. Chevrot, D. Teyssié, Poly(ethylene
oxide)/polybutadiene based IPNs synthesis and characterization, Polymer, 48(2007) 696-703.
[37] C. Plesse, A. Khaldi, Q. Wang, E. Cattan, D. Teyssié, C. Chevrot, F. Vidal, Polyethylene oxide–
polytetrahydrofurane–PEDOT conducting interpenetrating polymer networks for high speed
actuators, Smart Materials and Structures, 20(2011) 124002.
[38] F. Vidal, C. Plesse, G. Palaprat, J. Juger, C. Gauthier, J.-M. Pelletier, K. Masenelli-Varlot, C. Chevrot,
D. Teyssié, Influence of the poly(ethylene oxide)/polybutadiene IPN morphology on the ionic
conductivity of ionic liquid, European Polymer Journal, 49(2013) 2670-9.
[39] A. Maziz, C. Plesse, C. Soyer, C. Chevrot, D. Teyssié, E. Cattan, F. Vidal, Demonstrating kHz
Frequency Actuation for Conducting Polymer Microactuators, Advanced Functional Materials,
24(2014) 4851-9.
[40] Retraction notice to "Differential impact of diabetes and hypertension in the brain: adverse
effects in grey matter" [Neurobiol. Dis., 44 (2011) 161-173, http://dx.doi.org/10.1016/
j.nbd.2011.06.005], Neurobiol Dis, 68(2014) 229.

42
Chapter 2: PEDOT-based trilayer fabrication process

[41] M. Armand, F. Endres, D.R. MacFarlane, H. Ohno, B. Scrosati, Ionic-liquid materials for the
electrochemical challenges of the future, Nat Mater, 8(2009) 621-9.
[42] G. Petroffe, L. Beouch, S. Cantin, P.-H. Aubert, C. Plesse, J.-P. Dudon, F. Vidal, C. Chevrot,
Investigations of ionic liquids on the infrared electroreflective properties of poly(3,4-
ethylenedioxythiophene), Solar Energy Materials and Solar Cells, 177(2018) 23-31.
[43] A. Maziz, C. Plesse, C. Soyer, E. Cattan, F. Vidal, Top-down Approach for the Direct Synthesis,
Patterning, and Operation of Artificial Micromuscles on Flexible Substrates, ACS Appl Mater Interfaces,
8(2016) 1559-64.
[44] J.F. Hulvat, S.I. Stupp, Liquid-crystal templating of conducting polymers, Angew Chem Int Ed Engl,
42(2003) 778-81.
[45] B. Winther-Jensen, D.W. Breiby, K. West, Base inhibited oxidative polymerization of 3,4-
ethylenedioxythiophene with iron(III)tosylate, Synthetic Metals, 152(2005) 1-4.
[46] C. Largeot, C. Portet, J. Chmiola, P.L. Taberna, Y. Gogotsi, P. Simon, Relation between the ion size
and pore size for an electric double-layer capacitor, J Am Chem Soc, 130(2008) 2730-1.
[47] J.P. Lock, S.G. Im, K.K. Gleason, Oxidative Chemical Vapor Deposition of Electrically Conducting
Poly(3,4-ethylenedioxythiophene) Films, Macromolecules, 39(2006) 5326-9.
[48] S. Kirchmeyer, K. Reuter, Scientific importance, properties and growing applications of poly(3,4-
ethylenedioxythiophene), Journal of Materials Chemistry, 15(2005) 2077-88.
[49] K.H. Skorenko, A.C. Faucett, J. Liu, N.A. Ravvin, W.E. Bernier, J.M. Mativetsky, W.E. Jones, Vapor
phase polymerization and mechanical testing of highly electrically conductive poly(3,4-
ethylenedioxythiophene) for flexible devices, Synthetic Metals, 209(2015) 297-303.
[50] A. Maziz, C. Plesse, C. Soyer, E. Cattan, F. Vidal, Top-down Approach for the Direct Synthesis,
Patterning, and Operation of Artificial Micromuscles on Flexible Substrates, ACS Applied Materials &
Interfaces, 8(2016) 1559-64.
[51] T.N. Nguyen, K. Rohtlaid, C. Plesse, G.T.M. Nguyen, C. Soyer, S. Grondel, E. Cattan, J.D.W. Madden,
F. Vidal, Ultrathin electrochemically driven conducting polymer actuators: fabrication and
electrochemomechanical characterization, Electrochimica Acta, 265(2018) 670-80.
[52] K.D. Fong, T. Wang, H.-K. Kim, R.V. Kumar, S.K. Smoukov, Semi-Interpenetrating Polymer
Networks for Enhanced Supercapacitor Electrodes, ACS Energy Letters, 2(2017) 2014-20.
[53] W. Shi, Q. Yao, S. Qu, H. Chen, T. Zhang, L. Chen, Micron-thick highly conductive PEDOT films
synthesized via self-inhibited polymerization: roles of anions, Npg Asia Materials, 9(2017) e405.
[54] E. Smela, O. Inganas, I. Lundstrom, Controlled folding of micrometer-size structures, Science,
268(1995) 1735-8.
[55] E.W. Jager, O. Inganas, I. Lundstrom, Microrobots for micrometer-size objects in aqueous media:
potential tools for single-cell manipulation, Science, 288(2000) 2335-8.
[56] E.W. Jager, E. Smela, O. Inganas, Microfabricating conjugated polymer actuators, Science,
290(2000) 1540-5.
[57] A. Khaldi, C. Plesse, C. Soyer, E. Cattan, F. Vidal, C. Legrand, D. Teyssié, Conducting
interpenetrating polymer network sized to fabricate microactuators, Applied Physics Letters, 98(2011)
164101.
[58] A. Khaldi, A. Maziz, C. Plesse, C. Soyer, D. Teyssié, F. Vidal, E. Cattan, Patterning innovative
conducting interpenetrating polymer network by dry etching, 2014 IEEE/ASME International
Conference on Advanced Intelligent Mechatronics2014, pp. 1424-9.
[59] A. Khaldi, D. Falk, K. Bengtsson, A. Maziz, D. Filippini, N.D. Robinson, E.W.H. Jager, Patterning
Highly Conducting Conjugated Polymer Electrodes for Soft and Flexible Microelectrochemical Devices,
ACS Applied Materials & Interfaces, 10(2018) 14978-85.
[60] K.K. Lee, P.R. Herman, T. Shoa, M. Haque, J.D. Madden, V.X. Yang, Microstructuring of polypyrrole
by maskless direct femtosecond laser ablation, Adv Mater, 24(2012) 1243-6.
[61] N. Bityurin, 8[space][space]Studies on laser ablation of polymers, Annual Reports Section "C"
(Physical Chemistry), 101(2005) 216-47.

43
Chapter 2: PEDOT-based trilayer fabrication process

[62] A. Gursel, D. Valerie, R. Philippe, S. Geoff, Conducting polymer microactuators operating in air,
Journal of Micromechanics and Microengineering, 19(2009) 025017.
[63] M. Farajollahi, Fabrication and non-linear modeling of conducting polymer-based actuators :
toward catheter and tactile display applications: University of British Columbia; 2016.
[64] M. Anwar Mohammed, T. Saleh, D.W. Madden John, K. Takahata, Micropatterning Polypyrrole
Conducting Polymer by Pulsed Electrical Discharge, Macromolecular Materials and Engineering,
299(2013) 198-207.
[65] R. Brooke, M. Fabretto, P. Hojati-Talemi, P. Murphy, D. Evans, Evidence for ‘bottom up’ growth
during vapor phase polymerization of conducting polymers, Polymer, 55(2014) 3458-60.
[66] D. Evans, M. Fabretto, M. Mueller, K. Zuber, R. Short, P. Murphy, Structure-directed growth of
high conductivity PEDOT from liquid-like oxidant layers during vacuum vapor phase polymerization,
Journal of Materials Chemistry, 22(2012) 14889-95.
[67] A. Maziz, Microactionneurs à base de polymères conducteurs électroniques : Vers l’intégration
aux microsystèmes par de nouveaux procédés d’élaboration, France: Université de Cergy Pontoise;
2014.

44
Chapter 3: Electrochemomechanical characterization of the trilayer structure

Chapter 3: Electrochemomechanical characterization of

the trilayer structure

Contents

3.1 Introduction ........................................................................................................................... 46

3.2 Electro-chemical properties................................................................................................... 48
3.2.1 Ionic conductivity of the SPE and PEDOT layers ........................................................... 49
3.2.2 Electrical conductivity of the PEDOT electrodes........................................................... 52
3.2.3 Volumetric capacitance of the PEDOT electrodes ........................................................ 55
3.2.4 Possible short circuit between two PEDOT layers ........................................................ 58
3.3 Mechanical properties ........................................................................................................... 59
3.3.1 Young’s moduli of the SPE layer and of the trilayer actuator ....................................... 59
3.3.2 Damping ratio ............................................................................................................... 62
3.3.3 Blocking force characterization..................................................................................... 63
3.4 Empirical strain-to-charge ratio ............................................................................................. 64
3.4.1 Strain to charge ratio .................................................................................................... 64
3.4.2 Linear strain .................................................................................................................. 67
3.5 Conclusion.............................................................................................................................. 69

45
Chapter 3: Electrochemomechanical characterization of the trilayer structure

3.1 Introduction
In the previous chapter, a layer-by-layer method, combined with a vapor phase
polymerization process, were employed to fabricate a thin trilayer structure. In order to most
effectively exploit this process in actuators, having good models that predict response is
useful. For this reason, characterization of the actuator’s properties was done to gain
knowledge and identify the important parameters affecting actuation performance. In this
chapter, the geometric, electrochemical, mechanical and electrochemomechanical
properties of these trilayers are determined. This includes measuring thicknesses of the
conducting polymer containing layers and the separators, before and after swelling in ionic
liquid. Stiffness, resistances (ionic and electronic), capacitance and active deflection under
applied voltage are among the properties determined. From these measurements, key
properties are extracted for use in modeling: Young’s modulus, volumetric capacitance, ionic
and electronic conductivities and strain to charge ratio – in some cases as a function of
oxidation. These characteristics are then used to simulate polymer actuator and sensor
responses (Chapters 4 and 5).

The characteristics of the resulting actuators, which were fabricated following the process
mentioned in chapter 2, are studied in detail, as reported in the sections that follow. Before
proceeding to the details, a typical device response is described. Fig. 1a shows the current
response of the actuator under a step voltage. As can be seen, the output current quickly
drops when the PEDOT electrodes approach their fully charged state. The equivalent
displacement (Fig. 1b) is quite fast for first few seconds before it gets to a steady state. Fig.
1c shows the polarity of the applied voltage and the corresponding displacement direction of
the trilayer. The positive potential induces a contraction in one PEDOT electrode and
expansion of the other electrode. The amplitude of displacement is 3 mm of bending. The
output current and the displacement can be used later to extract the information about the
charge and strain.

46
Chapter 3: Electrochemomechanical characterization of the trilayer structure

Fig. 1. Actuator behavior under a step voltage a) The applied step voltage and the response current,
b) Bending response of the actuator, c) A picture of the bending actuator (length x width x thickness =
4 mm x 1 mm x 0.017 mm), the + and – indicate the polarity of the applied voltage.

47
Chapter 3: Electrochemomechanical characterization of the trilayer structure

Lots of factors can affect the strain, strain rate, and the blocking force of these actuators: for
example ionic conductivity of the CP electrodes and the SPE layer can limit the strain rate and
the bending speed of the actuator [1]. The electronic conductivity of the CPs can affect the
circularity of the curvature of the beam during its actuation and is a function of the oxidation
state, as demonstrated by Feldman [2-4]. The volumetric capacitance of the CP material
assesses the ability to accommodate the ions under a voltage excitation. This value had been
shown to vary with the scanning rate and the magnitude of the applied voltage [5, 6]. Since
the trilayer structure is thin and PEDOT is grown in from the two opposing surfaces, there is
a possible short circuit between the two CP electrodes. This can affect the the efficiency,
sensor signal decay rate and the self-heating of the actuator, and so is important to know.
The Young’s modulus of the CP material determines the blocking force of the actuator (for a
given strain). The Young’s modulus was shown to be dependent on the oxidation state of the
CP electrodes [4, 7].

In this chapter, the affecting factors discussed above will be taken into account during the
investigation on the ultrathin PEDOT:NBR/PEO:PEDOT structure. The electrochemical
properties of the actuators including ionic conductivity of the PEDOT electrodes and the
NBR/PEO, the electronic conductivity and the volumetric capacitance of the PEDOT
electrodes, as well as the possible short circuit between two PEDOT electrodes will be
investigated. The coupling between the electrochemical and the mechanical part,
representing by an empirical strain-to-charge ratio, is measured. The Young’s moduli of the
SPE, the PEDOT electrode as the function of the oxidation state, the damping ratio and the
linear strain of the actuator will be studied. These resulting parameters provide an insight the
actuator performance as well fulfilling the required values to establish the model predicting
actuator’s behavior.

3.2 Electro-chemical properties

Before beginning the full characterization, different sample configurations were fabricated to
adapt for different measurements (refer to section 2.6.1 for list of fabricated samples).

It is worth noting that trilayers of varying thickness were fabricated and tested in actuation,
to find a dimension range where significant bending actuation is achieved, for easy
characterization. Ten trilayer membranes were fabricated on ten separate silicon wafers, in
which the thicknesses of PEDOT electrodes were adjusted from 0.8 µm to 2.3 µm to
investigate the effect of the PEDOT’s thickness on the strain output of the trilayer actuator.
Bending actuation experiments were observed on 4 samples, where the thicknesses of PEDOT
electrodes vary from 1.2 µm to 2.3 µm. These exhibit bending under a voltage excitation. The
biggest displacement was seen in a sample having a thickness of PEDOT:NBR/PEO:PEDOT =
2.2 µm:7 µm: 2.2 µm (pre-swelling dimensions). After swelling in ionic liquid, the thicknesses
of the PEDOT electrodes and of the NBR/PEO layer are approximately 3.5 µm and 10 µm,

48
Chapter 3: Electrochemomechanical characterization of the trilayer structure

respectively. Therefore, this thickness ratio (PEDOT:NBR/PEO:PEDOT = 3.5 µm:10 µm:3.5 µm

after swelling in ionic liquid) is chosen for further investigation on the volumetric capacitance,
the possible short circuit between two PEDOT electrodes, the Young’s moduli, the damping
ratio, and the linear strain as well as the bending strain.

3.2.1 Ionic conductivity of the SPE and PEDOT layers

Ionic conductivity is one of a number of properties that limit current, and, as a result, charging
speed. [8]. Therefore, it is a factor that can limit strain rate and bending speed of the trilayer
actuator. PEDOT ionic conductivity was determined using an Electrochemical Impedance
Spectroscopy (EIS) measurement according to a procedure described previously [8, 9] and
summarized in Fig. 2a. A Solartron 1287A Potentiostat/Galvanostat combined with a
Solartron 1260A frequency response analyzer was used to obtain the frequency responses. A
4-point measurement across the ionic liquid and the membrane (a bilayer or a trilayer) is
used. We impose current between the working electrode (W.E) and the counter electrode
(C.E) and detect a local potential drop between two reference electrodes (R.Es) (Fig. 2a). The
W.E and the C.E are made of two pieces of glassy carbon, and the R.Es are the classic Ag/AgCl
wire filled 4M NaCl.

49
Chapter 3: Electrochemomechanical characterization of the trilayer structure

Fig. 2. a) EIS measurement using 4-point setup, b) NBR/PEO layer, bottom bilayer, and top bilayer
configuration, and c) a zoomed image of an area in the red rectangular A-A and its equivalent circuit
showing the current direction at low and high operating frequency.

It has been shown that there is an asymmetry between the top and the bottom PEDOT
electrodes in a trilayer actuator fabricated by layer-by-layer method (chap. 2, section 2.7)
[10]. As demonstrated, the roughness of the top PEDOT electrode is ten times higher than
that of the bottom PEDOT electrode. This fact may lead to asymmetric properties between
these two electrodes. Therefore, there is a need to characterize separately each electrode. A
top bilayer (PEDOT electrode is polymerized on the top of NBR/PEO) and a bottom bilayer
(PEDOT electrode is polymerized on a silicon wafer followed by a NBR/PEO layer) were
investigated to find out the variation in PEDOT ionic conductivity, shown in Fig. 2b. Before the
measurement, an NBR/PEO membrane, as well as the top bilayer and the bottom bilayer
samples were immersed into EMITFSI for one week [11], to ensure a full uptake of ions.

50
Chapter 3: Electrochemomechanical characterization of the trilayer structure

Because the resistance of each layer depends on their thickness, to separate the PEDOT ionic
resistance from other sources of resistance such as NBR/PEO resistance and ionic liquid
resistance, the thickness of specimens including the thickness of NBR/PEO and PEDOT (Table
1) was increased during the fabrication process. It is worth of noticing that the thickness of
swollen PEDOT layers were not measured directly but it was determined by SEM in its dry
state and then multiplied by 1.66, which is the swelling factor in EMITFSI deriving from
chapter 2 (section 2.7).

Fig. 3. Impedance values obtained from EIS measurements of a pure EMImTFSI, a top bilayer of
NBR/PEO-PEDOT, a bottom bilayer of PEDOT-NBR/PEO, and a single NBR/PEO layer.

The EIS experimental results are shown in Fig. 3. The resistance between the cannulas of the
ionic liquid alone (no film present) is 114 Ohms. The NBR/PEO matrix without PEDOT adds
another 8 Ohms. The PEDOT containing layers add another 2 to 3 Ohms, as can be seen from
the low frequency impedance. At high frequencies the effect of the conducting PEDOT layer
disappears completely as its parallel electronic conductivity bypasses the ionic resistance [8]
(as shown in Fig. 3c). Clearly the impedance is dominated by the ionic resistances of the
sample, and of the electrolyte. The ionic conductivity (δ) of the PEDOT and NBR/PEO layers is
determined by:

ℎ
𝛿=
𝑅×𝐴

where h is the thickness of either the PEDOT or the NBR/PEO layer (Table 1), and A is the
sample surface area in contacting with the electrolyte, which is equal to 0.78 cm2. Resistance
R, is calculated by a subtraction between NBR/PEO containing cell resistance and pure
EMImTFSI resistance (Fig. 3) to estimate the resistance of the NBR/PEO alone. To find the
ionic resistances of the PEDOT layers, the difference in resistance between low EIS frequency

51
Chapter 3: Electrochemomechanical characterization of the trilayer structure

(1 mHz – representing the resistance of all layers) and high EIS frequency (resistance at 1 kHz
– representing the resistance of NBR/PEO plus EMImTFSI) is calculated, for the same bilayer.

Table 1. Thickness in swollen state and ionic conductivity of the NBR/PEO alone, top bilayer, and
bottom bilayer

Thickness (swollen) Resistance (Ω) Ionic conductivity (S/m)

(µm)
Ionic liquid - - 0.9 (provided by supplier)
NBR/PEO layer 25.3 8.3 0.038±0.005
Top PEDOT layer of 5.0±0.4 2.6 0.025±0.002
the bilayer
Bottom PEDOT layer 7.7±0.4 3.3 0.030±0.002
of the bilayer
In the Table 1, the resistance of the NBR/PEO layer is obtained by subtracting the impedance
value of the “NBR/PEO matrix” data shown in Fig. 4a from the “pure EMITFSI”. Similarly, the
resistance of the PEDOT layer of the top and bottom bilayer is derived from the difference in
their impedance at low and high frequency (Fig. 4). The ionic conductivity values for the top
and bottom layers are similar to and slightly lower than the 0.04±0.002 S/m measured by
Dobashi et al. [8], in that case using aqueous LiTFSI as the electrolyte. As can be seen from
table 1, there is a slight difference between the ionic conductivity of the top and the bottom
PEDOT electrodes, perhaps indicating a small effect of the asymmetry on the ionic
conductivity. This difference is contributed by three main factors, an uncertainty of the
thickness measurement, a slight fluctuation (representing in the error bar in the table) of the
impedance at low and high frequencies, and an asymmetry in PEDOT electrodes themselves.
The ionic conductivity of the NBR/PEO (0.038 S/m) membrane is though lower than the values
reported by Vidal [12] (0.1 S/m) and Maziz [13] (0.13 S/m). These obtained values were
measured for a thick full IPN membrane (thickness>200 µm, where NBR is crosslinked) but
not a very thin sIPN as in our case.

3.2.2 Electrical conductivity of the PEDOT electrodes

Researchers have demonstrated that the electronic conductivity of PEDOT depends on the
potential applied [2-4]. Feldman et al. [2] shown that a conductivity of a 13.9 µm thick PPy
film in 0.1 M Et4NClO4/CH3CN quite constant at approximately 3 x 10-1 S/cm as the potential
of the film is decreased from 0.4 V to 0 V. When the potential continues to decrease to -0.6
V, the conductivity enormously drops to 10-6 S/cm. Warren et al. [3] and Farajollahi et al. [4,
14] found a similar trend where the conductivity of a 25 µm thick PPy film increases from 0.64
S/cm to 270 S/cm when the applied potential raises from -0.8 V to 0.4 V. This drop with
decreasing voltage can be explained: the PPy electrode is indeed reduced and the doping level
is decreased which reduces the charge carrier concentration, the mobility, and in turn the

52
Chapter 3: Electrochemomechanical characterization of the trilayer structure

conductivity. In my work, the electrical conductivities of the thin sIPN PEDOT/PEO electrodes
will be characterized to explore their dependences on the potential applied.

Fig. 4 presents a three-electrode system (Solartron potentiostat, ModuLab XM ECS) for cyclic
voltammetry analysis. The counter electrode was a platinum foil with a surface area greater
than that of the trilayer actuator electrode (at the working electrode). The reference
electrode was a standard wire Ag/AgCl filled with 4M NaCl solution. The potentiostat was
connected to both trilayer electrodes using platinum clamps and then fully immersed in
EMITFSI. This scheme is used to reduce or oxidize the specimens to a certain potential
(compared to Ag/AgCl reference electrode).

To measure the electrical conductivity as a function of oxidation state, a potential (versus

Ag/AgCl reference electrode) was applied to both sides of a trilayer structure in solution (Fig.
5). The dimensions of the specimen are length (L) х width (b) х total thickness (h) = 10 mm х
3 mm х 0.017 mm in the swollen state, the thickness of the PEDOT layers is hP = 3.5 µm and
the thickness of NBR/PEO is hS = 10 µm. The length and width of the sample were increased
to adapt them to the 4-line measurement setup in Fig. 5.

Fig. 4. Three-electrode system for setting the oxidation state of the PEDOT and for cyclic voltammetry
analysis.

To drive the sample to the desired potential using the setup in Fig. 4, the voltage is applied
and hold for 10 min, until the response current decays to 0 from its original position. The
potential of the sample is then measured to confirm its oxidation (reduction) state. After
reaching the desired potential, the sample is taken out the ionic liquid and laid down on the

53
Chapter 3: Electrochemomechanical characterization of the trilayer structure

four-line probe (Fig. 5) to measure its electrical conductivity. When the applied voltage is
larger than +1 V or smaller than -1 V, the potential of the PEDOT electrodes is not stable and
drops very quick once the sample is being taken out for measurement. This is the reason why
I choose the range of applied potential between -1 V and 1 V.

Fig. 5. A built in-house four-line probe to measure the electronic conductivity of the PEDOT
electrodes, where the distance between two copper lines is d = 2 mm.

A 4-point probe is commonly used to measure the electronic conductivity of a surface.

However, in the case of ultrathin polymer films, this device could damage the measurable
layer due to its sharp probes. A four-line probe built in-house made from copper foil, a low
resistance material, deposited on a glass slide to measure the electronic conductivity of a
surface is presented in Fig. 5. In this setup, a constant current is applied between two external
copper lines and the voltage between the two internal lines is then measured. The distance
between the two internal lines, the thickness of the electrode layer and its geometry, the
current applied, and the voltage measured are known and, therefore, the electronic
conductivity can be derived. The measurement was performed on the top and bottom
surfaces of the trilayer structure.

54
Chapter 3: Electrochemomechanical characterization of the trilayer structure

Fig. 6. PEDOT electrical conductivity as a function of oxidation state: a) on the bottom PEDOT surface
of the trilayer structure, b) on the top PEDOT surface of the trilayer structure

Fig. 6. shows the electrical conductivities of the bottom and the top PEDOT electrodes in a
trilayer configuration. At neutral state (the applied voltage U = 0 V), the electronic
conductivity of the bottom PEDOT layer is 59 S/cm, which is comparable to 200 S/cm
obtaining for a single PEDOT described in chapter 2 (section 2.4). As expected, the electrical
conductivity drops markedly - by a factor of 5 (Fig. 6a) and 9 (Fig. 6b) in this case - as the
PEDOT electrodes approach their reduced states (-1 V) from their oxidized states (1 V). This
dropping value is smaller in comparison to the value reported by Farajollahi et al. [14]
(dropped by factor of 24) for a 35 µm thick of PEDOT electrode in a 240 µm thick of trilayer
actuator. This change fits a third order function, which will later be used in the model. The
electrical conductivity of the bottom PEDOT electrode is about eight times higher than that
of the top PEDOT electrode, confirming the asymmetry of the top and the bottom electrodes.
This asymmetry stems from the actuator fabrication process, which is based on the layer-by-
layer method, where the bottom PEDOT layer in contact with a silicon wafer has a smoother
surface and a longer heat treatment period, while the top PEDOT layer formed on the
NBR/PEO surface is rougher and treated for a shorter period.

3.2.3 Volumetric capacitance of the PEDOT electrodes

The volumetric capacitance describes the amount of charge per volume of electrode that is
stored in response to a change of applied voltage. It also describes the ability to accommodate
ions of the conducting polymer film during its actuation, since all electronic charge is balanced
by ionic charge. Researchers have recently tried to determine the volumetric capacitance of
pure PEDOT or PEDOT:PSS thin films [15, 16]. They found that this capacitance not only
depends on the thickness and the density of PEDOT - determined by the VPP time and the
coating speed - but also on the type of dopant, as smaller dopants increase the volumetric
capacitance to a certain extent. In addition, Madden et al. [17, 18] have demonstrated a
variation in the volumetric capacitance value of the polypyrrole as a function of applied
voltage. Since the trilayer structure works as an actuator, a variation in the applied voltage is
expected. For our measurement, the working electrode in Fig. 5 was connected to both
electrodes of the trilayer actuator using platinum clamps and then fully immersed in
EMImTFSI. The dimensions of the trilayer are length (L) х width (b) х total thickness (h) = 5
mm х 1 mm х 0.017 mm in the swollen state, and the thickness of each layer is hP = 3.5 µm
and hS = 10 µm.

In the first experiment, the volumetric capacitance of the PEDOT electrodes was investigated
as a function of the scanning rate while the potential window was held constant. The potential
was swept from U1 = -0.6 V to U2 = 0.7 V vs. Ag/AgCl reference electrode at scan rates, 𝜗,
varying from 0.005 V/s to 1 V/s (Fig. 7a). This potential window is carefully chosen to cover
the oxidation and reduction peaks of the PEDOT electrodes during the redox process, but not

55
Chapter 3: Electrochemomechanical characterization of the trilayer structure

too high to accelerate other reactions due to the presence of impurity substances inside ionic
liquid.

Fig. 7. a) Cyclic voltammograms of PEDOT electrodes obtained in neat EMImTFSI within a potential
window of -0.6 V to +0.7 V with different scan rates: b) Volumetric capacitance of PEDOT electrodes
as function of the scanning rate.

The effective capacitance was derived from the expression [19]:

𝑼
∫𝑼 𝟐 𝒊(𝑼)𝒅𝑼
𝟏
𝑪𝑷𝑬𝑫𝑶𝑻 = (F) (1)
𝝑(𝑼𝟐 −𝑼𝟏 )

Effective volumetric capacitance was calculated from:

𝑪𝑷𝑬𝑫𝑶𝑻
𝑪𝑽 = (F/m3) (2)
𝟐𝑳𝒃𝒉𝑷

Here, hP is the thickness of the one PEDOT electrode.

As described in Fig. 7b, the accessible volumetric capacitance varies with the scan rate. This
value drops gradually from 90x106 F/m3 at low scanning rate of 5 mV/s to 57 x 106 F/m3 at the
fairly quick scanning rate of 0.1 V/s. This value decreases steadily to 12 x 10 6 F/m3 when the
scan rate increases to 1 V/s. We attribute the ability of the PEDOT samples in this case to
charge relatively rapidly in the presence of PEO pores, enabling faster ion transport [20].
Previous work has shown a value of volumetric capacitance of 145 x 106 F/m3 for PEDOT-paper
[5], between 49 x 106 F/m3 and 67.3 x 106 F/m3 for PEDOT:PSS [6], and 1 x 108 F/m3 [14] for
the PEDOT electrodes in a thick trilayer actuator. They have also observed a decrease in this
capacitance value with an increasing scan rate. Fig. 8b suggests that the volumetric
capacitance is still increasing as scan rate is decreasing, which could be due to long time
constants associated with ion insertion. In my study, the volumetric capacitance
measurement was conducted at even lower scanning rate of 0.5 mV/s, 1 mV/s, and 2 mV/s.
56
Chapter 3: Electrochemomechanical characterization of the trilayer structure

However, at these extremely low scan rates, the Pt clamp which is used to hold the trilayer as
the working electrode and the surface of the PEDOT electrodes also provide an apparent
capacitance behavior (another possibility to explain this increase is detailed in Appendix
A.3.1). This apparent capacitance is significant compared to that of PEDOT electrodes making
estimates of PEDOT volumetric capacitance difficult at these low scan rates.

In a second experiment, the potential window was varied from 0.5 Vpp to 3.3 Vpp while the
scanning rate is held at 20 mV/s, as depicted in Fig. 8.

Fig. 8. Volumetric capacitance as a function of the voltage window from 0.5 Vpp to 3.3 Vpp.

As can be seen in Fig. 8, the volumetric capacitance increases gradually when the potential
window increases from 0.5 Vpp to 2.2 Vpp. One way of picturing this enlargement is, as the
voltage increases, more and more holes or electrons are created in or supplied to the PEDOT
backbone chain leading to a gain in the number of ions inserted or extracted. This increment
approaches a state of complete oxidation (or reduction) state as the voltage window reaches
to 2.2 Vpp. The first possible contribution to this dependent is that at the oxidation state is
extreme at a specific potential indicating by a positive peak potential during the cyclic
voltammetry. Below this potential, the PEDOT electrodes are not fully oxidized and the
volumetric capacitance is small. When the potential goes beyond this potential, the
volumetric capacitance increases at higher speed.

Another possible explanation for this increment comes from the complex structure of
PEDOT/PEO electrodes, which contains PEDOT chains with different length. At low potential,
only long chains of PEDOT are oxidized while a larger voltage window will oxidize both long
and shorter chains. This results in a higher volumetric capacitance.

57
Chapter 3: Electrochemomechanical characterization of the trilayer structure

The dependence of the volumetric capacitance on the potential window is approximated by

a high order function. The order is continuously added until the resulting function fits within
experimental uncertainty and the following sixth-order function is obtained:

𝐶𝑉 = 1.1 × 106 𝐹/𝑚3 𝑉 −6 𝑈 6 − 1.9 × 107 𝐹/𝑚3 𝑉 −5 𝑈 5 + 1.1 × 108 𝐹/𝑚3 𝑉 −4 𝑈 4 − 2.8 ×
108 𝐹/𝑚3 𝑉 −3 𝑈 3 + 3.4 × 108 𝐹/𝑚3 𝑉 −2 𝑈 2 − 1.7 × 108 𝐹/𝑚3 𝑉 −1 𝑈 + 6.9 × 107 (𝐹/𝑚3 ).
(3)

It is worth of noticing that this approximated function is only valid for the value of voltage
staying in the range of the potential window from 0.5 Vpp to 3.3 Vpp. A capacitance value
between two measured voltage window values can be interpolated from the equation (3).
The values are averaged over the potential range, and take into account changes in both
electrodes simultaneously.

To our knowledge, the voltage dependence of the specific capacitance of PEDOT material has
not been reported yet. However, this voltage dependence of other supercapacitor materials
such as manganese oxide/activated carbon in 2 mol/L KNO3 electrolyte has already been
studied [21] showing an increase in the capacitance as the operating voltage increases.
Equation (3) is to some extent agreed with the trend observed by Stoller [21].

The volumetric capacitance allows the extent of charge transferred to be estimated, since its
value is proportional to total charge transferred and the actuator volume. Along with the
resistance, the total capacitance also determines the rate of charging and the rate of
actuation. When multiplied by internal resistance, this RC product gives an estimate of
response time of the actuator – which will be investigated in the chapter 4, section 4.3.

3.2.4 Possible short circuit between two PEDOT layers

To investigate the possible short circuit between two PEDOT electrodes, an input step voltage
of Upp = 2 V, f = 0.005 Hz was applied to the trilayer actuator using a potentiostat with two-
electrodes connected. As shown in Fig. 9, at this low frequency the response current at first
reaches a peak and then quickly decays as the redox reactions in the electrodes is completed
and the actuation approaches its maximum bending state. The current is expected to
decrease to zero, which means no current passes through the trilayer. However, in this case
a small current (-15.75 µA) is still observed after 100 s. When the voltage is switched to the
other direction, the similar trend of the current is shown and the leakage current is 16.67 µA.

58
Chapter 3: Electrochemomechanical characterization of the trilayer structure

Fig. 9. The current response to an input step voltage at low frequency (f = 0.005 Hz)
𝑈𝑝𝑝 2
A rough estimation on the short resistance of the trilayer gives: 𝑅𝑠ℎ𝑜𝑟𝑡 = 𝑖 = 16×10−6 ≈
𝑙𝑒𝑎𝑘
𝑅𝑠ℎ𝑜𝑟𝑡 62×103 Ω
124 𝑘Ω. The leakage per area is then equal to: 𝑅 = = (3×2+5×1)𝑐𝑚2 = 11.2 𝑘Ω/𝑐𝑚2,
𝐴
where A is the total area of the actuator, including the rectangular shape (3 x 2 mm) fitting in
the clamping area and the beam area (5 x 1 mm). The thickness of the trilayer tested was 17
m.

3.3 Mechanical properties

In this section, the Young’s moduli of the SPE layer alone and of the trilayer actuator as a
function of the oxidation state were measured in a swollen state.

3.3.1 Young’s moduli of the SPE layer and of the trilayer actuator

A specimen of the SPE layer with the following dimensions of length (L) x width (b) x thickness
(hS) = 10 mm x 6 mm x 0.07 mm was tested using a Bose Electroforce 3000 dynamic
mechanical analyzer. Both ends of this thin film are attached using the Bose Electroforce 3000
dynamic mechanical analyzer. Then, a longitudinal displacement at a rate of 0.1 mm/s is
applied to the NBR/PEO sample and the force is recorded.

The relationship between the force F and the displacement ΔL gives a Young’s modulus ESPE
𝐹 𝐿
for the SPE material of 𝐸𝑆𝑃𝐸 = ∆𝐿 𝑏ℎ = 329 ± 50 𝑘𝑃𝑎. This value is consistent with the value
𝑠

reported by Festin [22] and Woehling [23] for the same type of material.

It must be recalled that Bahrami-Samani and Spinks [24, 25] have shown changes in shear
modulus of a polypyrrole thin films at various applied potentials, whereas Farajollahi [4] also
demonstrated the dependence of the Young’s modulus of a penetrated PEDOT layer on the

59
Chapter 3: Electrochemomechanical characterization of the trilayer structure

applied voltage. Consequently, we have decided to measure the variation of the Young’s
modulus of trilayer actuators versus the applied voltage. The trilayer actuator (length (L) x
width (b) x thickness (ht) = 10 mm x 6 mm x 0.017 mm) was subjected a Bose Electroforce
3000 dynamic mechanical analyzer and a potentiostat.

Fig. 10. Young’s moduli of the NBR-PEO layer alone - and of a trilayer structure as a function of
oxidation state - were measured using a Bose Electroforce 3000 dynamic mechanical analyzer

Fig. 10 describes the experimental setup used to measure the Young’s moduli of the trilayer
structure under voltage excitation, in a swollen state. The trilayer actuator is held in a Bose
Electroforce 3000 dynamic mechanical analyzer using a platinum clamp at one end and a
plastic clamp at the other end. The clamps and the trilayer actuator were immersed in
EMImTFSI ionic liquid. Different positive and negative potentials (vs. Ag/AgCl reference
electrode) (± 1.5, ± 1, ± 0.5 V) were applied to the specimens to oxidize or reduce the both
PEDOT electrode at the same time. After applying a voltage, the open circuit potential (vs.
Ag/AgCl) was measured to confirm the trilayer reached the desired voltage. The trilayer was
then subjected to a controlled displacement/load. The Young’s modulus was determined
from the slope of the force/displacement curve and the dimensions of the specimen. The
advantage of this experimental setup is that the Young’s modulus measurement can be
immediately performed after the trilayer reaches its oxidized (or reduced) state.
Consequently, the measurement can be carried out at low (-1.5 V) or high (1.5 V) potential
without an issue of the fast decaying voltage of the sample.

60
Chapter 3: Electrochemomechanical characterization of the trilayer structure

The Young’s moduli of the trilayer actuators are measured and the Young’s moduli of the
PEDOT layer is deduced in Fig. 7.

Fig. 11. Young’s modulus of the trilayer actuator and the PEDOT layer as a function of applied
voltage.

From the SPE Young’s modulus and the trilayer Young’s modulus, the Young’s modulus of the
(𝐸𝑡 ℎ𝑡 −𝐸𝑆 ℎ𝑆 )
PEDOT electrodes can be derived from the following equation [26] 𝐸𝑃 = (ht, hS,
2ℎ𝑃
and hP are the total thickness of the trilayer, the thickness of the NBR/PEO, and the thickness
of PEDOT electrode, respectively) and plotted on the same graph in Fig. 11. As we can see,
the Young’s modulus of both the trilayer and the PEDOT layer decreases slightly by 14 % from
10 MPa to 8.6 MPa and from 24.0 MPa to 20.5 MPa, respectively, when they are switched
from reduced state (- 1.5 V) to oxidized state (1.5 V).

A possible explanation for this reduction in modulus with increasing potential comes from the
ion insertion and extraction from the PEDOT electrodes. A negative potential initiates a
reduction of the polymer, leading to an insertion of EMI+ cations into PEDOT electrodes. These
ions balance the charge of the largely immobile TFSI- ions present in the matrix. This reduction
may induce a tight coupling between the PEDOT chains and would account for the increase in
Young’s modulus at negative potentials. In contrast, a positive potential may increase the
separation between the PEDOT chains and reduce the force between them leading to a
decrease in Young’s modulus. The observation here is consistent with the results reported by
Zheng [27] and Shoa [28]. In addition, this percentage reduction is quite consistent with the
value reported by Farajollahi [4].

61
Chapter 3: Electrochemomechanical characterization of the trilayer structure

3.3.2 Damping ratio

In the mechanical model, the natural frequency and damping coefficient are useful for
determining response at frequencies where inertia is important. These were determined
using a vibration experiment. In this experiment, the actuator was fixed at one end and free
at the other end. This actuator was then deflected using a 3 axis motorized stage with a
mounted by a force/displacement sensor (FT-RS1002 Microrobotic System, three degrees of
freedom, maximum speed: 5 mm/s coupled with an FT-S100 microforce sensing probe with a
measurement range of ±100 µN, resolution: 5 nN) and quickly released for free vibration.
Oscillation was measured directly using a Keyence laser sensor (Keyence LK-G32,
measurement range: ±4.5 mm, resolution: 0.05 μm), shown in Fig. 12.

Fig. 12. Experimental setup to measure the damping ratio of the trilayer actuator and the
blocking force

Damped oscillation is shown in Fig. 13.

62
Chapter 3: Electrochemomechanical characterization of the trilayer structure

Fig. 13. Beam vibrating as function of time.

This response suggests that the actuator is well represented as a second-order system [29].
After subtracting the damping coefficient was calculated using the formula below:
1
𝜁= 2
= 0.39 , (4)
√( 𝜋𝑡 ) +1
𝑙𝑛( )
𝑦

where t is overshoot amplitude, which is the distance from middle line of sinusoidal part to
highest peak and y is total displacement. There may be a second, longer time response, given
the longer rise time to steady state. This was not included in the model.

3.3.3 Blocking force characterization

The force generated at the tips of the actuator of length x width x thickness: 6 mm x 1 mm x
0.017 mm at various magnitudes and frequencies of a square wave applied voltage were
conducted using the setup described in Fig. 13, where the clamping is connected to a source
of voltage.

63
Chapter 3: Electrochemomechanical characterization of the trilayer structure

Fig. 14. Force generated as a function of the magnitude and the frequency of the applied voltage
from a trilayer actuator of length x width x thickness: 6 mm x 1 mm x 0.017 mm.

Fig. 14 shows an upward trend in the force generated at the tip of the beam as the applied
voltage is raised. This force decreases when the frequency increases. The blocking force can
𝐸𝑃 𝛼𝜌 ℎ 2 2ℎ𝑃 2
be calculated as the following equation [30]: 𝐹 = 𝑏 ( 2𝑆 ) [(1 + ) − 1], where 𝜌 =
𝐿 ℎ𝑆
𝑄
is the charge density. All the parameters in this equation are assumed to be constant
2𝐿𝑏ℎ𝑃
except for the total charge Q is a function of time. A high frequency gives a small value of
charge Q and in turn reduces the blocking force F. The force reaches zero or the reading is
inaccurate due to background noise when the frequency is 10 Hz. In this experiment, the
maximum force obtained at the tip of the trilayer was 11 μN at 0.1 Hz and 3.3 Vpp. It is worth
of mentioning that over 2.75 Vpp, the generated force almost reaches saturation despite the
increase in voltage, indicating that the two PEDOT electrode layers have been fully oxidized
and reduced.

3.4 Empirical strain-to-charge ratio

3.4.1 Strain to charge ratio

The coupling effect between strain and charge was first proposed by Baughman et al. [31] to
describe the change in polyacetylene thin film length corresponding to the insertion of ions.
Della Santa [32] and Madden [18] suggest a strain-to-charge ratio on the order of 10-10 m3/C
for polypyrrole-based CP actuators. In this work, we have determined the strain-to-charge
ratio of a PEDOT-based actuator. The experimental setup was to apply a triangular wave
voltage to the trilayer actuator allowing actuator to bend at slower speed and then to record
the output current and the bending response (Fig. 12).

64
Chapter 3: Electrochemomechanical characterization of the trilayer structure

From the actuator bending data, the strain difference produced by one PEDOT electrode
ℎ𝑡 1 1
(𝜀𝑎𝑐𝑡𝑢𝑎𝑙 ) can be derived via the following equation 𝜀𝑎𝑐𝑡𝑢𝑎𝑙 (𝑡)% = (𝑟 (𝑡) − 𝑟 ), where r0 and
2 𝑡 0

rt are the radius of the actuator in neutral and stimulated state, respectively, and ht is the
thickness of the trilayer actuator. Fig. 15 gives an example of how r0 and rt are obtained from
the bending measurement.

Fig. 15. Actuator at the a) neutral state where the bending radius is equal r0 = 12 mm, and b) excited
state where the bending radius is rt = 2.8 mm. The orange square on the background of the pictures
has the dimension of 1 x 1 mm.
𝑡 2ℎ 𝑤
Note that the strain difference equation above is used instead of the equation ɛ = 𝐿2 +𝑤 2. The

reason for this is that the latter is only valid for the case where the actuator is straight at its
initial position and is symmetrically bent under a voltage excitation. In our case, since the
actuator is initially curved, the resulting strain difference can be over or under estimated
(depending on whether the trilayer is initially bent with the same or the opposing curvature).

From this bending strain, the active strain of the PEDOT electrode, which is the linear strain
of a single PEDOT layer, is derived as following equation [33]:

ℎ𝑃 (ℎ4 𝐸 2 +8ℎ3 𝐸+12ℎ2 𝐸+8ℎ𝐸+24ℎ+12𝐸 2 +16)

𝜀𝑎𝑐𝑡𝑖𝑣𝑒 (𝑡) = 𝜀𝑎𝑐𝑡𝑢𝑎𝑙 (𝑡) × = 1.61 × 𝜀𝑎𝑐𝑡𝑢𝑎𝑙 (𝑡), (5)
6ℎ𝑡 (ℎ+1)(ℎ𝐸+2)

ℎ 𝐸
where ℎ = ℎ𝑆 and 𝐸 = 𝐸𝑆 . An error bar of 5 % obtained from the bending strain
𝑃 𝑃

measurements has been added.

Moreover, the volumetric charge density (ΔQ(t)/VPEDOT, where VPEDOT is the one PEDOT
electrode volume) (Fig. 17), resulting from the total insertion or extraction of ions per volume,
is calculated by integrating the current over time (Fig. 17).

65
Chapter 3: Electrochemomechanical characterization of the trilayer structure

Figure 16. Current and charge response to a triangular input voltage.

Fig. 17. Charge density and strain of the trilayer actuator as a function of time under a triangular
wave voltage excitation
𝜀𝑎𝑐𝑡𝑖𝑣𝑒 (𝑡)𝑉
The strain-to-charge ratio is finally obtained by the ratio: 𝛼 = . The Fig. 17 shows a
∆𝑄(𝑡)
quite constant ratio of the active strain to charge density giving a constant value of the train-
to-charge of 3.6 ± 0.5 x 10-10 (m3/C). The value is in the same range as other measurements
made in polypyrrole and PEDOT, as summarized in table 2. The strain to charge ratio will be
used in the model developed in Chapter 4.

66
Chapter 3: Electrochemomechanical characterization of the trilayer structure

Table 2. Summary strain-to-charge ratio for different materials and actuator configuration

Material Structure Dopant Solution Strain-to-charge ratio

value (m3C-1) x 10-10
PPy linear actuator DBS- NaDBS 1.7 [18]
PF6- (C2H5)4NPF6 5.3 [18]
BS- NaBS 0.3 [32]
trilayer TFSI- LiTFSI 5 [34]
PF6- TBAPF6 19.4 [35]

PEDOT a single layer

a trilayer TFSI- EMITFSI 2 - 4.56 [36, 37]
TFSI- LiTFSI 1.7 - 12 [37, 38]

DBS = Dodecylbenzene sulfonate, TBA: tetrabutylammonium, PF6: hexafluorophosphate

3.4.2 Linear strain

The relationship between the linear strain and the strain difference of the actuator is shown
in equation (5). In this section, the linear strain of the trilayer actuator is investigated by
applying a step voltage to a trilayer specimen to confirm this established relation. The trilayer
actuator (length (L) x width (b) x thickness (ht) = 10 mm x 6 mm x 0.017 mm – this length is the
part of the beam between clamps) was subjected a prestress of 1.5 gf by the Bose Electroforce
3000 dynamic mechanical analyzer. This constant load is applied to one end of the beam,
corresponding to a stress of 0.15 MPa, to keep the beam straight in the ionic liquid, as
described in Fig. 10. A step voltage of 2 Vpp vs. Ag/AgCl at f = 0.001 Hz was then applied to
both electrodes of the trilayer actuator. The uniaxial deformation of the beam was then
recorded as in Fig. 18. L2 and L1 are the length of the beam in its expanding and shrinking
state, respectively. A rough estimation gives us the value of linear strain of the trilayer
0.1𝑚𝑚
actuator 𝜀𝑎𝑐𝑡𝑖𝑣𝑒 ≈ = 1%, which is higher than the strain difference (0.56 %) between
10𝑚𝑚
two PEDOT electrodes obtaining in the previous section. Given the relation between the linear
strain and the strain difference is 𝜀𝑎𝑐𝑡𝑖𝑣𝑒 (𝑡) = 1.61 × 𝜀𝑎𝑐𝑡𝑢𝑎𝑙 (𝑡), this fits quite well with the
experimental results here: 1% ≈ 1.61 × 0.56%

67
Chapter 3: Electrochemomechanical characterization of the trilayer structure

Fig. 18. Uniaxial deformation of the trilayer actuator under a step voltage

68
Chapter 3: Electrochemomechanical characterization of the trilayer structure

3.5 Conclusion
In this chapter, the new electrochemomechanical system has been characterized. Its
electrochemical properties including the ionic conductivity of the PEDOT electrodes and SPE
layer, the PEDOT electronic conductivity as a function of oxidation state, and the dependent
PEDOT’s volumetric capacitance on the potential window and the scanning rate were all
study. In addition, the mechanical properties such as Young’s modulus of the PEDOT as a
function of the oxidation state, Young’s modulus of the SPE layer, the damping ratio were also
measured. Finally, the coupling between the electrochemical and the mechanical
representing via an empirical strain-to-charge ratio was determined.

Table 3 Summary of measured parameters of the ultrathin trilayer actuator

PEDOT electrodes electrical Top PEDOT electrode:

conductivity (S/cm) -2.62U3 + 2.76U2 + 9.89U + 7.49
Bottom PEDOT electrode:
-16.6U3 + 1.1U2 + 51.7U + 59.1
PEDOT electrodes ionic Top PEDOT electrode: 0.025
conductivity (S/m) Bottom PEDOT electrode: 0.030

PEDOT electrodes’ volumetric 1.1 × 106 𝑈 6 − 1.9 × 107 𝑈 5 + 1.1 × 108 𝑈 4

capacitance (F/m3) − 2.8 108 𝑈 3 + 3.4 × 108 𝑈 2 − 1.7
× 108 𝑈 + 6.9 × 107
NBR/PEO ionic conductivity 0.038
(S/m)
Strain-to-charge ratio (m3/C) 𝛼 = 3.6 ± 0.5 × 10−10
NBR/PEO Young’s modulus 329 ± 50
(kPa)
Trilayer actuator Young’s −0.09U3 − 0.08U2 − 0.26U + 9.54
modulus (MPa)
PEDOT Young’s modulus −0.23U3−0.19U2− 0.63U + 22.70
(MPa)
Damping ratio 0.39
Beam density (kg/m3) 2000

Maximum strain difference 0.56

(%)
1
Linear strain (%)

Table 3 summarizes the critical electrical, electrochemical, and mechanical properties of the
trilayer actuators. These obtained values allow to evaluate the non-linear dynamic model
established and will be discussed in the next chapter.

Besides, these obtained parameters provide us an inside on the physical structure of the
system and the working mechanism of the actuator. Most important, the key factors affecting

69
Chapter 3: Electrochemomechanical characterization of the trilayer structure

actuation behavior are all determined providing us a best guideline to improve actuator’s
performance.

In the future perspective, there are different key points needed to be improve. At first, to
study the optimum thickness of the PEDOT electrodes providing highest bending strain
output, we have varied the PEDOT electrode thickness in the trilayer configuration from 0.8
µm to 2.7 µm by changing the rotational speed of depositing the oxidant solution of the spin
coater from 3500 rpm to 1000 rpm, while the thickness of the SPE layer is constant at 7 µm.
However, the strain measurement showed only four out of ten samples were working, where
each sample is taken from one wafer. And the strain measurement is processed on the sample
with the total thickness of 17 µm. Nevertheless, emerging work is required to improve the
reproducibility of the fabrication process.

Moreover, the properties of the PEDOT and the SPE layers have been measured, however,
each property was investigated on one specific thickness of the sample. Though, there is a
need to confirm the obtained value by remeasuring on various thickness.

In addition, the unwanted asymmetric property between the top and the bottom PEDOT in
the electronic conductivity measurement can be reduced by adjusting the heat treatment
period during the fabrication of top bilayer and bottom bilayer.

70
Chapter 3: Electrochemomechanical characterization of the trilayer structure

[1] J.D. Madden, P.G. Madden, I.W. Hunter, Polypyrrole actuators: modeling and performance, 2001,
pp. 72-83.
[2] B.J. Feldman, P. Burgmayer, R.W. Murray, The potential dependence of electrical conductivity and
chemical charge storage of poly(pyrrole) films on electrodes, Journal of the American Chemical
Society, 107(1985) 872-8.
[3] M.R. Warren, J.D. Madden, A structural, electronic and electrochemical study of polypyrrole as a
function of oxidation state, Synthetic Metals, 156(2006) 724-30.
[4] F. Meisam, S. Farrokh, N. Naser, F. Adelyne, P. Cédric, T.M.N. Giao, V. Frédéric, D.W.M. John,
Characterization and dynamic charge dependent modeling of conducting polymer trilayer bending,
Smart Materials and Structures, 25(2016) 115044.
[5] J.A. Lee, M.K. Shin, S.H. Kim, H.U. Cho, G.M. Spinks, G.G. Wallace, M.D. Lima, X. Lepró, M.E. Kozlov,
R.H. Baughman, S.J. Kim, Ultrafast charge and discharge biscrolled yarn supercapacitors for textiles
and microdevices, Nature Communications, 4(2013) 1970.
[6] Y. Liu, B. Weng, J.M. Razal, Q. Xu, C. Zhao, Y. Hou, S. Seyedin, R. Jalili, G.G. Wallace, J. Chen, High-
Performance Flexible All-Solid-State Supercapacitor from Large Free-Standing Graphene-PEDOT/PSS
Films, Scientific Reports, 5(2015) 17045.
[7] G. Alici, P. Metz, G.M. Spinks, A mathematical model to describe bending mechanics of polypyrrole
(PPy) actuators, Proceedings, 2005 IEEE/ASME International Conference on Advanced Intelligent
Mechatronics2005, pp. 1029-34.
[8] Y. Dobashi, A. Fannir, M. Farajollahi, A. Mahmoudzadeh, A. Usgaocar, D. Yao, G.T.M. Nguyen, C.
Plesse, F. Vidal, J.D.W. Madden, Ion Transport in Polymer Composites with Non-Uniform Distributions
of Electronic Conductors, Electrochimica Acta, 247(2017) 149-62.
[9] D.S. Yoo, A. Mahmoudzadeh, E.C.W. Fok, K. Walus, J.D.W. Madden, Multiple time constant
modelling of a printed conducting polymer electrode, Electrochimica Acta, 56(2011) 4711-6.
[10] T.N. Nguyen, K. Rohtlaid, C. Plesse, G.T.M. Nguyen, C. Soyer, S. Grondel, E. Cattan, J.D.W. Madden,
F. Vidal, Ultrathin electrochemically driven conducting polymer actuators: fabrication and
electrochemomechanical characterization, Electrochimica Acta, 265(2018) 670-80.
[11] D. Mecerreyes, Applications of Ionic Liquids in Polymer Science and Technology, Berlin: Springer;
2015.
[12] G. Petroffe, L. Beouch, S. Cantin, P.-H. Aubert, C. Plesse, J.-P. Dudon, F. Vidal, C. Chevrot,
Investigations of ionic liquids on the infrared electroreflective properties of poly(3,4-
ethylenedioxythiophene), Solar Energy Materials and Solar Cells, 177(2018) 23-31.
[13] A. Maziz, C. Plesse, C. Soyer, C. Chevrot, D. Teyssié, E. Cattan, F. Vidal, Demonstrating kHz
Frequency Actuation for Conducting Polymer Microactuators, Advanced Functional Materials,
24(2014) 4851-9.
[14] M. Farajollahi, Fabrication and non-linear modeling of conducting polymer-based actuators :
toward catheter and tactile display applications: University of British Columbia; 2016.
[15] N. Kurra, R. Wang, H.N. Alshareef, All conducting polymer electrodes for asymmetric solid-state
supercapacitors, Journal of Materials Chemistry A, 3(2015) 7368-74.
[16] C.M. Proctor, J. Rivnay, G.G. Malliaras, Understanding volumetric capacitance in conducting
polymers, Journal of Polymer Science Part B: Polymer Physics, 54(2016) 1433-6.
[17] M. Farajollahi, G. Alici, M.S. Sarwar, J.D.W. Madden, Conducting Polymers as EAPs: Physical
Description and Simulation, in: F. Carpi (Ed.) Electromechanically Active Polymers: A Concise
Reference, Springer International Publishing, Cham, 2016, pp. 353-83.
[18] J.D.W. Madden, Conducting polymer actuators: Massachusetts Institute of Technology; 2000.
[19] M. Ghidiu, M.R. Lukatskaya, M.-Q. Zhao, Y. Gogotsi, M.W. Barsoum, Conductive two-dimensional
titanium carbide ‘clay’ with high volumetric capacitance, Nature, 516(2014) 78.
[20] J. Xi, X. Qiu, J. Li, X. Tang, W. Zhu, L. Chen, PVDF–PEO blends based microporous polymer
electrolyte: Effect of PEO on pore configurations and ionic conductivity, Journal of Power Sources,
157(2006) 501-6.

71
Chapter 3: Electrochemomechanical characterization of the trilayer structure

[21] M.D. Stoller, R.S. Ruoff, Best practice methods for determining an electrode material's
performance for ultracapacitors, Energy and Environmental Science, 3(2010) 1294-301.
[22] N. Festin, Conducting interpenetrating polymer network actuator sensor for biomimetic
perception system: Université de Cergy Pontoise; 2012.
[23] V. Woehling, New Developments in electroactive materials based on electronic conductive
polymers : Towards integration into biomedical systems: Université de Cergy Pontoise; 2016.
[24] M. Bahrami-Samani, C.D. Cook, J.D. Madden, G.M. Spinks, P.G. Whitten, Quartz crystal
microbalance study of volume changes and modulus shift in electrochemically switched polypyrrole,
Thin Solid Films, 516(2008) 2800-7.
[25] G.M. Spinks, L. Liu, G.G. Wallace, D. Zhou, Strain Response from Polypyrrole Actuators under Load,
Advanced Functional Materials, 12(2002) 437-40.
[26] G. Alici, B. Mui, C. Cook, Bending modeling and its experimental verification for conducting
polymer actuators dedicated to manipulation applications, Sensors and Actuators A: Physical,
126(2006) 396-404.
[27] W. Zheng, Analysis of high performance polypyrrole actuators: University of Wollongong; 2011.
[28] T. Shoa, T. Mirfakhrai, J.D.W. Madden, Electro-stiffening in polypyrrole films: Dependence of
Young's modulus on oxidation state, load and frequency, Synthetic Metals, 160(2010) 1280-6.
[29] H. Jack, Engineer On A Disk, 2010.
[30] P.G.A. Madden, Development and Modeling of Conducting Polymer Actuators and the Fabrication
of a Conducting Polymer Based Feedback Loop: MASSACHUSETTS INSTITUTE OF TECHNOLOGY; 2003.
[31] R.H. Baughman, L.W. Shacklette, R.L. Elsenbaumer, E.J. Plichta, C. Becht, Micro Electromechanical
Actuators Based on Conducting Polymers, in: P.I. Lazarev (Ed.) Molecular Electronics: Materials and
Methods, Springer Netherlands, Dordrecht, 1991, pp. 267-89.
[32] A.D. Santa, D.D. Rossi, A. Mazzoldi, Characterization and modelling of a conducting polymer
muscle-like linear actuator, Smart Materials and Structures, 6(1997) 23.
[33] P. Du, X. Lin, X. Zhang, A multilayer bending model for conducting polymer actuators, Sensors and
Actuators A: Physical, 163(2010) 240-6.
[34] C.H. Nguyen, G. Alici, G.G. Wallace, Modelling trilayer conjugated polymer actuators for their
sensorless position control, Sensors and Actuators, A: Physical, 185(2012) 82-91.
[35] A.D. Price, Fabrication, Modelling and Application of Conductive Polymer Composites: University
of Toronto; 2012.
[36] M. Bentefrit, S. Grondel, C. Soyer, A. Fannir, E. Cattan, J.D. Madden, T.M.G. Nguyen, C. Plesse, F.
Vidal, Linear finite-difference bond graph model of an ionic polymer actuator, Smart Materials and
Structures, 26(2017) 095055.
[37] A. Fannir, Actuators based on conducting polymers with a linear deformation in open air and
compatible with a space environment: Université de Cergy Pontoise; 2017.
[38] M. Farajollahi, A. Usgaocar, Y. Dobashi, V. Woehling, C. Plesse, F. Vidal, F. Sassani, J.D.W. Madden,
Nonlinear Two-Dimensional Transmission Line Models for Electrochemically Driven Conducting
Polymer Actuators, IEEE/ASME Transactions on Mechatronics, 22(2017) 705-16.

72
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior

Chapter 4: Linear dynamic and nonlinear dynamic

model to predict PEDOT-based trilayer actuation
behavior

Contents

4.1 Motivation.................................................................................................................. 74
4.1.1 Objectives........................................................................................................... 74
4.1.2 Proposed methodology...................................................................................... 74
4.2 State of art ................................................................................................................. 74
4.2.1 A summary of Black box, white box, grey-box models for CP actuators ........... 75
4.2.1.1 Black-box model ............................................................................................. 75
4.2.1.2 Grey-box model .............................................................................................. 75
4.2.1.3 White-box model........................................................................................... 78
4.2.2 Why the choice of the Bond Graph language? .................................................. 80
4.3 Dynamic Bond Graph modeling ................................................................................. 84
4.3.1 Actuation description......................................................................................... 84
4.3.2 Word Bond Graph model ................................................................................... 86
4.3.3 BG submodels .................................................................................................... 87
4.3.3.1 Electrical model .............................................................................................. 87
4.3.3.2 Electromechanical coupling ........................................................................... 91
4.3.3.3 Mechanical model .......................................................................................... 92
4.3.4 BG global models ............................................................................................... 98
4.4 Simulation tests ....................................................................................................... 100
4.4.1 Software implementation ................................................................................ 100
4.4.2 Comparison between the linear and nonlinear simulations ........................... 101
4.5 Comparison to experimental tests .......................................................................... 104
4.5.1 Time domain responses ................................................................................... 105
4.5.2 Frequency responses ....................................................................................... 108
4.6 Parameter sensitivity and power performance analysis ......................................... 111
4.7 Conclusion ................................................................................................................ 115

73
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
4.1 Motivation
4.1.1 Objectives

The goal of this chapter is to develop an electrochemomechanical model to simulate the

performance of trilayer actuators. This model must on one hand, be able to predict the
dynamic behavior under a voltage excitation, and on the other hand allow an evaluation of
the critical parameters affecting actuation and the power performance.

4.1.2 Proposed methodology

In order to establish a suitable model having an ability to describe the CP-based trilayer
actuators, a critical literature review on the modeling of CP actuators is the first step. From
an engineering point of view, a grey-box model based on Bond Graph language [1, 2] was
selected, as I expected these selections can provide a visual view for the reader and later on
facilitate the control of the trilayer actuators.

In a second step, a nonlinear dynamic model to predict the nonlinear mechanical behavior
and the complexity of the structuring materials of ultrathin trilayer actuators is etablished.
The nonlinear mechanical behavior of these trilayer actuators stems from changes that occur
as the voltage is increased [3-5]. During operation, oxidation in one PEDOT electrode layer
and reduction in the other electrode, accompanied by a drop in potential along the length of
the electrodes, are observed. These changes in state modify the electrical, coupling, and
mechanical properties. Here, the nonlinear properties, including electrical conductivity,
volumetric capacitance of the PEDOT-based electrodes, and Young’s modulus, were all
characterized and incorporated into the model. At the same time, a simpler version - a linear
model - to predict small beam displacement, is derived from the nonlinear model. This linear
model includes (1) an RC electrical circuit to represent the device charging, (2) an
electromechanical coupling matrix that is based on the accepted relationship between the
strain and the volumetric charge density, and (3) a mechanical model based on the modal
superposition method to treat Euler-Bernoulli dynamic beam equations for beam
displacement. A comparison of the simulation results with the nonlinear model shows the
limitations of such an approach. A number of experimental validations have been conducted
in both the time and frequency domains, highlighting the predictive properties of the
nonlinear model. Finally, from the proposed model, an evaluation of the sensitivity of the
actuator parameters and of the power performance helps to improve the understanding of
actuator efficiency and its crucial properties.

4.2 State of art

Within this section, different approaches on the modeling of CP actuators will be reviewed
from the perspective of black-box, grey-box, and white-box approaches. Languages used in
these modeling will also be classified.

74
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
4.2.1 A summary of Black box, white box, grey-box models for CP actuators

Models are hypotheses that can be used to help describe, explain and/or predict the
phenomena in a novel technology. Therefore, models are proposed and simulations are
compared with experiments on a laboratory – scale. The agreement between the simulations
and corresponding experimental data on one hand confirms the hypotheses proposed earlier
and improves our understanding on the investigated technology on the other hand. A
successful model can be a powerful tool for designer, since it allows prediction, in advance,
of the behavior of the device under design [6, 7]. Once this step is solved, the model may also
to be used to control the performance of the systems.

The literature review on the past and current status of the CP actuator modeling theory is
divided into three sections: black-box or data-driven models [8], white-box or principle
models [7], and grey-box models [6] - which are a trade-off between the white-box and the
black-box models.

4.2.1.1 Black-box model

In order to facilitate the real-time control of the CP polypyrrole-based actuator positioning,

Fang et al. [8] have proposed a simplified model, based on the knowledge of the actuation
bandwidth as well as on typical values of the physical parameters, from the full infinite-
dimensional model which was proposed by Madden [9, 10]. In this reduced third-order
transfer function, the author assumed that the trilayer operates at low frequency, and
diffusion coefficient will decay overtime due to solvent evaporation. However, this model is
a purely empirical model which is obtained via output-input data, and must be re-identified
for different samples. John et al. [11] proposed another transfer function approach by fitting
the experimental data of polypyrrole-based actuator displacement as a function of frequency
and applied voltage. This model was validated under a dynamic input voltage signal in both
time and frequency domains. However, these models are a purely empirical model which is
obtained via output-input data, and must be re-identified for different samples.

4.2.1.2 Grey-box model

The grey-box model of CP actuators relies on the idea of using strain-to-charge ratio and was
first introduced by Della Santa et al. [12] for the case of a single layer of PPy doped with
benzensulphonate anions (BS−) in solvent of sodiumbenzensulphonate and in a mixture of
acetonitrile and de-ionized distilled water. The actuator model was divided into three sub-
models, where the electrochemical model is simply the total amount of charge exchanged
during a stimulation cycle and the coupling between electrochemical part and mechanical
part is denoted by αech. This electrochemical expansion coefficient is obtained from fitting a
straight line to observed isotonic (constant force) length changes as a function of exchanged
charge. It has the units of % C/mm. The mechanical sub model is based on the parallel

75
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
combination of two Maxwell elements (composed of an elastic and a viscous body in series
[12]) and one elastic body. These models are intended to describe the viscoelastic properties
of the conducting polymer films. The author has experimentally confirmed that the proposed
model was able to predict the producing stress as well as the strain of PPy film in an electrolyte
bath. However, this model only validated for a small deformation.

Madden [10] introduced the diffusive elastic model, which again uses the strain to charge
ratio to describe actuation, but adds to it by considering the factors that limit the rate of
charging. The idea proposed is that diffusion of ions into the polymer film is a major rate
limiting factor, in addition to the resistances of the electrolyte and of the conducting polymer
itself. This model contains an RC transmission line to describe the diffusion of ions through
the thickness of the CP electrode. The electrical equivalent circuit is shown in Fig. 1a, where
ZD is the total impedance of the RC transmission line.

Fig. 1. a) The equivalent electrical circuit of the diffusive elastic model, which includes: a
double layer capacitance C, electrolyte resistance R, and the impedance associated with ionic
charge transport dominated by diffusivity ZD, b) Example of an equivalent electrical circuit
including a finite diffusion element Cjk and Rijk, a resistance Rej, and capacitance at the
interface between the solution and the polymer

The model was later expanded to consider the distributed nature of charging in two
dimensions, with a circuit model describing this situation shown in Fig. 1b. In this model, the
distributed charge storage is represented by capacitances, Cmn. Their rate of charging (and
hence the local rate of strain) is determined by electronic resistances, Rem, of the conducting

76
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
polymer film along its length (up the page), and ionic resistance, Rimn, representing ion
transport within the conducting polymer. The rightmost column of capacitors and resistors
describe the double layer capacitance between the electrolyte and the polymer and the ionic
resistance of the electrolyte, respectively. The model is designed to describe a polymer films
in electrolyte. In the case of a trilayer, the circuit diagram is reflected about the right hand
edge. When the electrical resistance, Rej, representing the electrical resistance along the
length of the CP electrodes, is negligible, a two-dimensional transmission line is used as in Fig.
1b [13, 14]. In this model, the RC transmission lines extends from left to right to model the
effective diffusion of ions through polymer thickness by capacitive coupling with ionic
resistance of CP. The electrical resistance is considered to be negligible in comparison to ionic
resistance through the thickness of the conducting polymer film. Each horizontally oriented
one dimensional RC transmission line is separated from the next one by an CP electrical
resistance Rej, that accounts for the voltage drop that occurs as current passes along the
length of the conducting polymer film. This discrete model is useful in that each element
represents the properties of the device at a particular location, and each can be given a
voltage dependent value.

Madden [10] incorporated electromechanical coupling based on the concept that the
volumetric charge density is proportional to the induced material strain. The relation is given
as following equation:
𝜎
𝜀 = 𝛼𝐶𝑉 𝑉 + 𝐸,

where ɛ is the induced strain, α is the strain-to-charge ratio, CV is the volumetric capacitance,
and the applied stress and Young’s modulus are σ and E, respectively. The terms on the right
describe strain due to application of an external load, while the middle term describes the
electromechanical coupling.

Madden’s model was at first implemented in a linear, 1D configuration. However, conducting

polymers have long been known to show changes in electronic and ionic conductivity as a
function of oxidation state (see e.g. [15, 16]). Farajollahi et al. [3, 17] integrated the electronic
and ionic conductivities of conducting polymers, and Young’s modulus of the PEDOT-based
trilayer actuator as a function of oxidation state into the 2-D RC transmission line model using
a state-space representation. The model has the ability to predict dynamic response of the
beam via radius of bending curvature at specific positions through time. However, the model
is not suited to predict actuation strain as a function of the frequency of applied voltage, and
generally needs to be rewritten for non-step voltage inputs.

Alici [18, 19] proposed several variations of a bending curve model to predict force and
bending output of CP actuators. The key commonality in these models are the
electromechanical coupling, α, determined experimentally at each operating voltage by

77
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
observing the actuator bending curvature of an unloaded beam. The determined α can be
used to predict maximum blocking force and establish a nonlinear second-order differential
equation relating the displacement of the beam and the applied voltage, as well beam
mechanical properties. To a certain extent, the model showed a good agreement with
experimental results, but it was not able to predict the tip excursions exceeding 900, and the
transient actuation response.

In Moghadam et al.’s [20, 21] model, while the electrical model and the electromechanical
coupling were derived from a linear transmission line model [13], the mechanical model was
either based on Euler-Bernoulli beam theory and Hamilton’s principle, or the dynamic rigid
finite element method, to predict bending actuation. These approaches offer efficient
methods in case of large displacement and complex shapes. However, the use of a linear
electrical model limits their advantage.

In a recent study, C.H. Nguyen et al. [14] showed an advanced mathematical model including
a 2D transmission line extracting from the Madden’s model, the mechanical, and viscoelastic
properties of a CP trilayer actuator that is a the Kelvin–Voigt model. The proposed model was
experimentally validated in both frequency and time domains of a linear displacement.
Afterwards, the authors implemented the model into an inversion-based feedforward
controller intended for biomedical applications and proved the ability to control the
actuator’s displacement of the proposed controlling system.

4.2.1.3 White-box model

In the white box model, the nature of the actuation mechanism including the movement of
charge under a potential and ions concentration is related via thermodynamic and
transport/kinetic theories. Thermodynamics helps predict the final equilibrium state, while
ion transport, reaction kinetics and the dynamics of motion determine the time it takes to
reach equilibrium. There is no conducting polymer actuator model that is completely white
box as they do not describe how volume changes from first principles. In this section, some
aspects of the charging and actuation that have been treated from a fundamental perspective
are discussed.

On the transport side the Wang and Smela [22] developed a finite element model for charge
transport by diffusion and migration in a single layer of polypyrrole in electrolyte, based on
transport equations for ionic and electronic charge coupled with the Nernst-Planck equation
in term of reaction time t, as follows:

, (1)

78
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
where i indicates the species under consideration, which can be anionic, cationic species, or
the electronic charge distribution in CP chains. Ci is the concentration, zi is valence charge
number, µi is the ionic mobility of species i, and ɸ is the local electric potential. The two terms
on the right represent diffusion and migration. In this equation, convection is neglected as it
is assumed that the charge concentration is diluted and there is no interaction among the
charge carriers. Therefore, convection flux can be separated from diffusion flux. Poisson’s
equation is used to couple net charge density Q (C/cm3) and the concentration Ci:

𝜖0 ∇ ∙ (𝜖𝑟 ∇∅) = 𝑄 = ∑𝑛𝑖=1 𝑧𝑖 𝐶𝑖 . (2)

Here 𝜖0 is the permittivity of a vacuum, and 𝜖𝑟 is the dielectric constant of the material. The
model allowed to answer several important questions: the migration does contribute to ion
transport at extreme potential, and the presence or the absence of a net charge in the
polymer does not make any change to the concentration. However, this model does not link
the local net charge density to mechanical behavior and neither does it consider any actuation
phenomena. A number of gray box models are able to incorporate similar models of
transport. For example, the transmission line model in Fig. 1b incorporates the responses of
Equations 1 and 2 for the special cases that migration is dominating transport, and that charge
storage is capacitive. In the limit as the elements in Fig. 1b are reduced in size, the response
becomes diffusive.

Otero [23] has introduced a first-principle model based on the electrochemical and polymeric
principles as applied to conducting polymers. He combined the theory of reaction rate for
oxidation processes, the kinetic relations among the activation energy, and the increment of
the mechanical work. The model describes the evolution of the potential under flow of a
constant current, as follows:

∆(𝑃𝑉) 𝑅𝐾 𝑖
𝑎
𝑈 = 𝑈0 + 𝑖𝑎 𝑍 + ∆𝑈(𝑛 − 1) − (1−𝛼)𝑛𝐹 + (1−𝛼)𝑛𝐹 × {ln (𝐹𝑉 ) − 𝑑 ln[𝐴− ] −
𝑖 𝑡
𝑒 ln ([𝑃𝑜𝑙 ∗ ]𝑖𝑛𝑖𝑡𝑖𝑎𝑙 − 𝐹𝑉
𝑎
) − ln 𝑘𝑎0 }, (3)

where 𝑖𝑎 is the anodic current, [𝑃𝑜𝑙 ∗ ] is the evolution concentration of the active centers of
the polymeric chains storing a positive charge after oxidation in the film during the time of
current flow (starting at [𝑃𝑜𝑙 ∗ ]𝑖𝑛𝑖𝑡𝑖𝑎𝑙 ), Z is electrochemical cell impedance, R is the universal
gas constant, F is the Faraday constant, n is the number of charges involved in a chain
reaction, V is the volume of the reduced and dry reactive CP film inside the electrolyte, ka is
the rate constant. The voltage across the cell, U, from left to right, is the sum of the initial
potential, Uo, a voltage drops due to the internal impedance, a shift due to work done on the
sample, and finally terms resulting from the over potential needed to drive the reaction away
from the equilibrium potential, resulting from the shift in oxidation state. Since the
equilibrium electrochemical response of PPy and other conducting polymers is known to be
more complicated than one described by a single redox reaction with one equilibrium

79
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
potential, the response in Equation 3 is considered instead to be the sum of many closely
spaced redox reactions, indexed by integer n, representing the many oxidation states of the
polymer. In the end, these many states are often accounted for by fitting factors (in this case
an assumed separation between states of ΔU), greying the model. The model includes both
actuating and sensing magnitudes, as any change through mechanical work (ΔPV),
temperature K, and electrolyte concentration [A-] can be sensed via the potential evolution
U. The authors confirmed the predicting ability of the model in both actuating and sensing
using a polypyrrole-based bilayer in LiClO4 electrolyte. A close correlation has been observed
between the simulation and experiment at low frequency of 0.01Hz, however it is not
experimentally validated for the intermediate and high frequencies yet. A difference between
this work and the grey box work of Madden and Farajollahi is that the thermodynamics of the
conducting polymer are instead accounted for by a capacitance in the work of those authors,
rather than related to standard reduction potentials. The end effect though is the same, as
the reduction potentials are fit, rather than being known. Another difference is the explicit
incorporation of the mechanical work done into the electrochemical model.

Venugopal et al.[24] introduced a thermodynamic chemomechanical constitutive model that

they used to describe polypyrrole doped with dodecylbenzene sulfonate incorporated into a
bilayer actuator in aqueous NaCl solution. The model is developed from work and energy
formulations for chemomechanical actuation, given by:

1/𝐸 𝜉𝑐
∆𝜀 ∆𝜎
{ }=[ 𝑅𝐾 ] { }, (4)
∆𝜇𝑖 𝜉𝑚 ( 𝐶 0 ) (1 − 𝑦𝑖0 ) ∆𝐶𝑖
𝑖

where Δ𝜀 is mechanical strain, Δµi is the chemical potential resulting from electrochemical
gradients of ionic species, 𝜎 is the applied stress, and Ci is moles/molar concentration of
diffusing species ‘i’. 𝜉𝑚 and 𝜉𝑐 represent chemomechanical coupling coefficients due to
chemical gradients and an applied mechanical stress, respectively, which are calculated under
the assumption of zero dead space in CP layers, indicating a tight packing of polymer chains.
𝜉𝑐 is akin to the strain to charge ratio, , introduced above. It is a strain per molar change in
concentration that in this case is computed based on the occupied volume of species inserted
into the conducting polymer. 𝜉𝑚 is the inverse coefficient with the same units, describing the
change in potential energy due to the application of a stress. E is the Young’s modulus, and
𝑦𝑖0 is mole fraction of species ‘i’ contained originally in the system. Two hypotheses were
stated: the actuation parameters are regulated by the finite number of redox sites in the CP
chains, and the strain developed is limited by saturation concentration of electrolyte. The
experimental strain presented in this paper showed a correlation with the modeling data,
however the relative error is considerable and may not suitable for actuation behavior
prediction.

4.2.2 Why the choice of the Bond Graph language?

80
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer actuation behavior

Table 1. Summarize on various model approaches of CP-based actuators

Ref. Box Model characteristic Lumped/distributed/co Model Model validation

model ntinuous presentation
Fang et Black Robust self-turning regulator allowing - 3rd order transfer Experimentally validated at
al. [8] box highly adaptive prediction the bending function low operating frequency f <
1Hz during 4h of continuously
operating
John et Black - - th
6 order transfer Experimentally validated in
al. [11] box function frequency domain up to 100 Hz
during 4 seconds of operation
Della Grey For a thin PPy film Lumped approach Experimentally validated in
Santa et box Electrochemical model: charge density as time domain
al. [12] a function of time
Electrochemomechanical coupling:
empirical strain-to-charge
Mechanical model: Maxwell elements in
parallel
Madden Grey For a thin PPy film Distributed Transfer function Experimentally validated in
[10] box Electrochemical model: elastic diffusive time and frequency domains
model
Electrochemomechanical coupling:
empirical strain-to-charge
Mechanical model: Elastic material
Alici [18, Grey For a PPy-based trilayer Lumped in Second-order Experimentally validated for
19] box Electrochemical model: simplified elastic electrochemical model differential actuation at steady state and
diffusive model Continuous in equation small displacement
Electrochemomechanical coupling: mechanical model
empirical strain-to-charge

81
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer actuation behavior

Mechanical model: second order

differential equation
Moghad Grey For a PPy-based trilayer Lumped in Second-order Experimentally validated for
am et al. box Electrochemical model: simplified elastic electrochemical model differential time and frequency domain for
[20, 21] diffusive model Distributed in equations a linear displacement
Electrochemomechanical coupling: mechanical model
empirical strain-to-charge
Mechanical model: rigid finite element
method
C. Grey For a PPy-based trilayer Lumped Transfer function Experimentally validated for
Nguyen box Electrochemical model: simplified elastic time and frequency domain
et al. diffusive model
[14] Electrochemomechanical coupling:
empirical strain-to-charge
Mechanical model: beam bending
equation
Wang White For a PPy film Distributed Partial differential Simulation validated for
and box Electrochemical model: charge transport equations electrochemical model
Smela by diffusion and migration
[22]
Otero White For a PPy bilayer film Lumped Derivation of Experimentally validated for
[23] box Electrochemical model: reaction rate for Buttler–Volmer time domain at low operating
oxidation process equation frequency. Not fully white as
Electrochemomechanical coupling: redox potentials are fit.
thermodynamic equations
Venugo White For a PPy bilayer film Lumped 2x2 coupling Experimentally validated with
pal et box A chemomechanical constitutive model matrix considerable error in strain
al.[24] based thermodynamics prediction

82
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
As summarized in table 1, modeling can be based on white, grey and black box models
depending on the understanding and the ability to describe a system. In the case of the black
box model, a transfer function is usually built from experimental data, whereas the white box
models are generally constituted from partial differential equations, which allow the research
and development of analytical solutions in space and time of continuous systems.

Between these two extremes, there are many other methods that can be proposed and that
are well adapted to the grey-box model. They are usually divided in two categories: the
lumped element approach and the distributed element approach [2], both methods being
considered as an approximation of the continuous problem since they are represented by
ordinary differential equations and depend only of the time. The distributed model is more
accurate and complex as it uses infinitesimal elements to represent the system along the
space. In contrast, the lumped approach concentrates the properties of the systems at
singular points. Among the possible elements, one can cite, either the physical elements as
the equivalent circuit, the mass-spring-damping system, the transmission line, or the
numerical elements as in the finite element, finite difference, and the finite volume method…
Of course, in case of multi-domain analysis, the modeling can be hybrid and based on a
combination of these different methods depending on which aspects one emphasizes:
mathematical or physical, accurate or not complex.

Since ionic EAP actuators inherently involve mechanical and electrical energy flow, combined
with the chemical domain, a thorough model for explaining the behavior of this actuator class
must include coupling between these multiple energy domains. BG (Bond Graph)
methodology is well suited for describing coupled domains [2, 25]. It allows easy access to the
magnitudes of energy transformed (e.g. between electrical and mechanical), stored power,
and dissipated power inside these actuators. The BG language has shown also its flexibility in
working with different energy domains and feasibility to describe interactions between
systems. These characteristics provide an approach for design of actuators and further, the
design of complete actuated robots, while retaining the physical structure of the model [26].

There are few works in modeling ionic EAP actuators using a BG. Among these, Byung-Ju Yi
concentrated on the diffusion in the separator and electrode layers of ionic polymer-metal
composites (IPMCs), but not on describing the mechanical response. The model was not
verified by experiments [27]. Nishida et al. [26] demonstrated a model of IPMCs by using Bond
Graphs to describe distributed port-Hamiltonian (DPH) systems on multiple spatial scales. The
coupling among different energy domains in IPMC can be modeled by DPH systems with
unidirectional energy flows without calculating analytic solutions. The advantage of this
approach is that it is a powerful tool in optimizing the design and control method of the
investigated system. However, the disadvantages of the distributed port-Hamiltonian is that
it normally requires understanding on mathematical model and currently, there is no
commercial software to simulate the port-Hamiltonian model. Bowers [28] built a BG model

83
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
of bilayer actuators operating in a liquid environment. Recently, Bentefrit et al. [29] presented
a finite-difference BG model of a trilayer actuator. However, this model focuses on linear
behavior and therefore can only predict the mechanical displacement of the actuator when
the applied voltage is relatively small. In this work, the aim is to extend the BG modeling of
conducting polymer actuators to nonlinear systems, and to incorporate both actuation
(Chapter 4) and sensing (Chapter 5). The models are applied to actuators developed using the
new layer-by-layer process.

4.3 Dynamic Bond Graph modeling

In this section, a nonlinear model and its simplification – a linear model to predict trilayer
actuator behavior - are proposed. The nonlinear model represents an application and
substantial extension of the linear transmission line model developed by Madden [10]. The
nonlinear model is intended to predict the nonlinear behavior of the actuator. It consists of a
1D RC electrical sub-system that is simplified from a 2D transmission line. Complex processes
such as diffusion-like transport of ions through the internal PEDOT nano-structure, the redox
process in the polymer, and the formation of a double layer between the interface of the
PEDOT and electrolyte, are all considered. An empirical coupling ratio relating the
instantaneous charge input and the output moment is implemented. The deformation of the
polymer is converted to bending using a mechanical model based on the rigid finite element
method. This approach is valid for large displacement and takes mass of the actuator into
account.

The proposed linear model comprises a capacitor in series with a resistor to represent the
electrical response. This is a simplification of the 1D RC transmission line circuit used in the
full model. An electromechanical coupling coefficient then relates the amount of charge input
to the free strain of the polymer. The beam geometry and mechanical properties are used to
estimate deflection and the total force applied to the free end of the beam as the output. The
mechanical model in this simplified case is based on a linear dynamic Euler-Bernoulli beam.

4.3.1 Actuation description

An ultrathin trilayer configuration operating in air [30] is studied here. In this configuration, a
solid polymer electrolyte (SPE), made of nitrile butadiene rubber (NBR) /poly(ethylene oxide)
(PEO), accommodates cations 1-Ethyl-3-methylimidazolium (EMIm+) and anions
bis(fluorosulfonyl)imide (TFSI-) and is sandwiched between two PEDOT containing CP
electrodes. When an excitation voltage is applied between the two electrodes, the positive
potential extracts electrons (oxidation process) from the one PEDOT electrode [31]. Cations
leave the electrode to balance charge. At the other electrode, cations are injected by a
diffusion-like process into the CP layer, neutralizing the reduced polymer backbone [32, 33].
In addition, a voltage drop is expected along the trilayer due to the existence of the electrical
resistance along the length of the PEDOT electrodes. The ion transfer leads to swelling of one

84
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
of the conducting layers and contraction of the other, resulting in bending of the sandwich
structure up or down (Fig. 2b).

Fig. 2. a) a CP actuator in neutral state and its dimensions (shown in the magnified section in
the middle), and b) a voltage excitation induces a tips displacement w(L, t) and an angular
rotating angle 𝜃(𝑥, 𝑡) of the actuator, which is equivalent to an external force F(L, t) or a
moment M(x, t), respectively. The Young’s moduli of the PEDOT electrodes and the NBR/PEO
layer are EP and ES, respectively.

Fig. 2a describes the actuator dimensions, where L, hp, hs, ht, and b are the length of the
actuator, the PEDOT electrode’s thickness, the NBR/PEO’s thickness, the total thickness, and
the width of the actuator, respectively.

Table 2. Notations for continuous, lumped, and distributed modeling approaches

Continuous Lumped Distributed

Voltage input 𝑈 𝐶 (𝑥, 𝑡) 𝑈 𝐶 (𝑡) 𝑈𝑖𝐶 (𝑡)
Current input 𝑖 𝐶 (𝑥, 𝑡) 𝑖 𝐶 (𝑡) 𝑖𝑖𝐶 (𝑡)
Charge 𝑄 𝐶 (𝑥, 𝑡) 𝑄 𝐶 (𝑡) 𝑄𝑖𝐶 (𝑡)
Capacitance 𝐶(𝑥, 𝑡) 𝐶(𝑡) 𝐶𝑖 (𝑡)
Force 𝐹(𝑥, 𝑡) 𝐹(𝑡) 𝐹𝑖 (𝑡)
Displacement 𝑤(𝑥, 𝑡) 𝑤(𝑡) 𝑤𝑖 (𝑡)
Rotating angle 𝜃(𝑥, 𝑡) 𝜃(𝑡) 𝜃𝑖 (𝑡)
Angular velocity 𝜃̇(𝑥, 𝑡) 𝜃̇(𝑡) 𝜃𝑖̇ (𝑡)
Moment 𝑀(𝑥, 𝑡) 𝑀(𝑡) 𝑀𝑖 (𝑡)

85
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
A voltage excitation U(t) induces a displacement w(L, t) at the tips of the beam and a rotating
angle θ(x, t) at position x away from the clamping point. From the tips displacement value,
the curvature K(t), the strain difference ɛactual(t), and the linear strain ɛactive(t) can be derived
(refer to chapter 3, section 3.4 for the detail derivation on the strain difference and the linear
strain). Three modeling approaches are now discussed: continuous, lumped and distributed.
To facilitate the modeling process later on, I distinguish the notations among three cases in
table 2. The continuous variables are both time, t, and position, x, dependent, while in the
distributed case position is replaced by an element index, i. The lumped elements have no
spatial dependence.

4.3.2 Word Bond Graph model

Creation of a bond graph, like other models, often starts with a block diagram known as a
word bond graph. The proposed “word BG” model of this actuator consists of three sub-
systems: the electrochemical subsystem, the electrochemomechanical coupling, and the
mechanical sub-system. The application of a voltage and current leads to charge accumulation
and current, which are passed to the electro-chemo-mechanical subsystem, where the
insertion of charge is converted to strain and stress. The resulting stresses and strains then
produce a moment and a bending of the device. The connections between these subsystems
are shown in Fig. 3 and are represented by “half arrows” indicating the assumed
instantaneous positive flow of energy in the actuator. Chapter 5 treats the reverse case for
the sensor and produces a combined model.

Fig. 3. Trilayer actuator Words BG model.

In the nonlinear model, the global system input is voltage, 𝑈(𝑡), whereas the outputs are
current 𝑖(𝑡), and beam angular velocity 𝜃̇𝑖 (𝑡). The electrochemical model inputs the time and
𝑑𝑄𝑖𝐶 (𝑡)
space variation in charge of the actuator during the operation, . Local charging is then
𝑑𝑡
related to strain and mechanical deformation through the electrochemomechanical coupling
system described below. The mechanical system is driven by the modal actuation moment
𝑀𝑖 (𝑡) generated from coupling. In addition, each pair of input and output variables provides
information about power transmission, described as follows:

86
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
𝑄 𝐶 (𝑡) 𝑑𝑄𝑖𝐶 (𝑡)
( 𝑣𝑜𝑙𝑡𝑎𝑔𝑒, 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 ) = ( 𝑈(𝑡), 𝑖(𝑡)) ; ( 𝑖 , ), and
𝐶 (𝑡) 𝑖 𝑑𝑡

(𝑚𝑜𝑚𝑒𝑛𝑡, 𝑟𝑜𝑡𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑎𝑛𝑔𝑢𝑙𝑎𝑟 𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦 ) = ( 𝑀𝑖 (𝑡), 𝜃̇𝑖 (𝑡) ).

In the linear model, angular velocity 𝜃̇𝑖 (𝑡) and actuation moment 𝑀𝑖 (𝑡) are replaced by beam
displacement 𝑤(𝑥, 𝑡) and actuation force 𝐹(𝐿, 𝑡), respectively. Other elements remainthe
same as in the nonlinear model.

4.3.3 BG submodel descriptions

Once the word BG is obtained, the BG describing each sub-system can be derived from the
equations involving in the electrical, the coupling, and the mechanical variables.

4.3.3.1 Electrochemical model

As reported in section 4.2.1.2, a grey-box electrochemical model is used in this study. An RC

equivalent circuit was proposed to explain the diffusion through the electrode thickness and
the electrical resistance of PEDOT was added along the length of the electrodes, leading to a
2D transmission line model of a trilayer actuator [10, 34]. The transmission lines here present
a combination of storage and migration. The effect of ion diffusion due to a difference in
concentration is neglected. In a recent research [15], nuclear magnetic resonance (NMR)
spectroscopy has been used to determine the diffusion coefficient of the ions PF6 ̶ in PPy. The
result found (5x10-9 cm2/s) is an order of magnitude smaller than the value extracted from
the EIS measurement. This suggests that field driven ion migration is dominant in the
impedance spectroscopy case.

The diffusion transmission line (Fig. 4a) is described by capacitors Cij, i = 1,3, j = 1, …, m, and
Rij are the ionic resistances of the electrode layer. R1ie, R3ie are the electrical resistances along
the length of the PEDOT electrodes. RSj are the ionic resistances of the SPE layer, Rshort is the
possible short circuit resistance between the electrodes (1stand 3rd layers), and Cdli are the
double layer capacitances formed at the interface between the PEDOT electrodes and the SPE
layer.

Conducting polymers store substantial charge. Getting the charge in and out is typically the
rate limiting factor (though inertia can also limit rate of deformation at high frequencies, as
will be seen). The capacitors in the model represent charge storage, and the resistors act to
slow charging. The model in Figure 4a can be simplified substantially because not all rate
limiting factors are significant, in part since the ratio between the thickness and the length of
the electrodes is small (<1/103). The magnitudes of the polymer RC charging time constant
along the length, the diffusion time constant through the thickness of the electrodes, the

87
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
double layer charging time, and the RC time constant determined by ionic transport through
the SPE layer – are now estimated to determine which factors dominate.

Estimates are based on time constants that are the product of a resistance and an effective
capacitance [35]. The more important factors are selected and the 2D transmission line model
is simplified, with a small effect on the accuracy. These rate limiting factors are calculated for
a beam having a typical geometry of length (L) х width (b) х total thickness (h) = 5 mm х 1 mm
х 0.017 mm in the electrolyte swollen state, and thicknesses of a PEDOT electrode: hP= 3.5
µm and of SPE layer hS= 10 µm.

The double layer charging time constant is determined by:

ℎ𝑆 10−5 𝑚×0.2 𝐹/𝑚2

𝜏𝑑𝑙 = 𝑅𝑆 × 𝐶𝑑𝑙 = 𝛿 𝑏 × 𝐿 × 𝐶𝐴 = (0.038 𝑆/𝑚)
= 53 𝜇𝑠, (5)
𝑆 ×𝑏×𝐿

where RS and hS are the resistance and the thickness of the SPE layer, respectively, b is the
width of the actuator, L is the length of the actuator, CA (F/m2) is the double layer capacitance
per area [36], and δs (S/m) is the ionic conductivity of the SPE layer, which is characterized in
chapter 3 (section 3.2). This time is very short compared to typical actuation response times
as the double layer capacitance are tiny compared to the bulk storage, and so the double
layer capacitances can be neglected.

‘Effective diffusion time through electrodes’ thickness is calculated using the following
formula:

2 2
ℎ𝑃 (3.5×10−6 𝑚) ×(52×106 𝐹/𝑚3 )
𝜏𝐷 = = = 0.0073 𝑠, (6)
𝐷 0.025 𝑆/𝑚

where hP is the thickness, and D is the effective diffusion coefficient which is determined by
𝛿𝑖 (S/m), the ionic conductivity within the electrode, divided by CV (F/m3), the volumetric
capacitance of the electrode (the values of 𝛿𝑖 and CV are referred to chapter 3 (section 3.2)),
in the common case that it is the ionic resistance that limits the charging rate [10]. The time
is quite short, so we shall see that this effect does not dominate for this particular electrode
thickness, and so simplifications can be made.

The resistance to ion transport through the SPE layer RS is another rate limiting factor to be
considered:

ℎ𝑆 𝐶𝑉 ×𝐿×𝑏×ℎ𝑃 (10−5 𝑚)52×106 𝐹/𝑚3 (3.5×10−6 m)

𝜏𝑆 = 𝑅𝑆 × 𝐶 = 𝛿 = = 0.0240 𝑠, (7)
𝑆 ×𝑏×𝐿 2 2×(0.038 𝑆/𝑚)

where the factor of “two” accounts for the two capacitors in series.

88
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
The polymer charge time constant resulting from the electronic resistance along the length
of the electrodes is:

2
𝐿 (5×10−3 𝑚) 52×106 𝐹/𝑚3
𝜏𝐿 = 𝑅𝑃 × 𝐶𝑉 × 𝐿 × 𝑏 × ℎ𝑃 = 𝛿 𝐶 × 𝐿 × 𝑏 × ℎ𝑃 = = 0.52 𝑠, (8)
𝑒 ℎ𝑃 𝑏 𝑉 2500𝑆/𝑚

where RP is the electrical resistance, VP is the volume of the PEDOT electrode, 𝛿𝑒 is the PEDOT
electrical conductivity (refer to chapter 3, section 3.2). This time constant is relatively large,
indicating that the resistance along the length is important and should be included in the
model.

89
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
Fig.4. a) 2D diffusion transmission line across the thickness and the length of a trilayer
actuator, and simplified electrochemical model of the trilayer actuator for b) nonlinear
model, and c) linear model

As can be seen from four values above, the polymer charge time constant (0.52 s) and the
resistance to ion transport through the SPE layer (0.024 s) play important roles in comparison
to the diffusion time through the thickness (0.0073 s) and the double layer charging time
constant (53 µs). The 2D transmission line model shown in Fig. 4a can then be simplified to
the 1D transmission line shown in Fig. 4b. The investigated trilayer actuator is divided into (n
+ 1) elements corresponding to the space discretization.

To establish even a simpler electrochemical model, we could also assume in the same case
that the electronic conductivity of the PEDOT electrodes is high resulting in a small τL
(equation 4). The transmission line in Fig. 4b is reduced to only two capacitors in series with
a resistor and all in parallel with a resistor, as shown in Fig. 4c.

Fig. 5. BG representation of a) the nonlinear electrochemical model, b) the linear

electrochemical model.

The BG representations shown in the Fig. 5a and 5b are used to describe the electrochemical
model of the 1D transmission line shown in Fig. 4b and 4c, respectively. Resistance and
capacitance are described by the one-port R-elements and the one-port C-elements in the

90
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
electrochemical system, respectively. The one-port C elements are shown for simplicity, - but
in the full electrochemomechanical model, these are two-port ₵-elements in which the charge
state of the capacitor is coupled to the stress-strain state of the actuator. As defined in all BG
models, 0-junctions behave such that all the voltage, force or torque values are equal across
the bonds, and total current values in and out of the junction are equal zero. In the 1-
junctions, the current, speed or angular velocity is the same through each bond and the sum
of voltages is equal to zero. These are equivalent to parallel (0) and series (1) electrical circuit
configurations.

4.3.3.2 Electrochemomechanical coupling

BG language is especially advantageous for modeling the coupled electrochemical -

mechanical effects compared to equivalent circuits [37] since it is possible to develop a new
element depending on the physical laws. The electrochemical-mechanical coupling effects are
represented by two-port ₵-elements in BG having the ability to describe both energy transfer
between different domains and energy storage in the electrodes.

In Fig. 6, the electromechanical coupling effect of the nonlinear model is depicted by a two-
port ₵-element with current 𝑑𝑄𝑖 /𝑑𝑡 and total moment 𝑀𝑖 (including an external mechanical
moment and a moment generated by an input current) implemented on both sides. Based on
the relationship between total moment Mi and charge per volume 𝜌 [38], the coupling matrix
of a 2 × 2 element is established as follows:
1
0
𝑈𝐶 𝐶𝑖 𝑄𝑖𝐶
[ 𝑖] = [𝛼𝐸 4𝑏𝑠𝑖𝑛(𝜃𝑖 ) 3 3 ], ][ (9)
𝑀𝑖 𝑃 (ℎ𝑆 +ℎ𝑃 )
𝐸 (
ℎ𝑆
) [(1 +
2ℎ𝑃
) − 1 +
𝐸𝑆
] 𝜃𝑖
2𝐿𝑖 3𝐿 𝜃 𝑃 2 ℎ 𝐸
𝑖 𝑖 𝑆 𝑃

where, 𝛼 is an empirical strain-to-charge ratio [39], and 𝜃𝑖 is the rotational angle of the ith
element relative to the direction of (i-1)th element. Li is the length of element i. In this coupling
matrix, the bottom left element describes the effect of charge on output moment, via the
strain-to-charge ratio, while the bottom right element relates the beam bending angle to the
output moment. The relation on the bottom left stems from the simple relationship that active
free strain is proportional to charge density, with the strain to charge ration being the constant
of proportionality. The strain is then scaled by modulus of the polymer layer to find stress. The
bottom right term has been derived from the equations for force generation, F, by a cantilever
𝐸𝑃 𝛼𝜌 ℎ 2 2ℎ𝑃 2 2𝑏𝐸𝑃 𝐾 ℎ𝑆 3 2ℎ𝑃 3 𝐸
beam [38]: 𝐹 = 𝑏 ( 2𝑆 ) [(1 + ) − 1] + ( 2 ) [(1 + ) − 1 + 𝐸𝑆 ], where
𝐿 ℎ𝑆 3𝐿 ℎ𝑆 𝑃
𝑄 2𝑤
𝜌 = 2𝐿𝑏ℎ is the charge density, and 𝐾 = 𝐿2 +𝑤2 is the curvature of the actuator (refer to
𝑃

Appendix A.4.1 for the detail derivation).

91
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior

Fig. 6. Coupling between electrical sub-system and mechanical sub-system.

To obtain a simple electrochemomechanical coupling description for the linear model, where
we only have a single capacitor and two capacitors, the discretized voltage 𝑈𝑖𝐶 , moment 𝑀𝑖 ,
rotation angle 𝜃𝑖 , and charge 𝑄𝑖𝐶 are replaced by total voltage across the capacitor 𝑈𝐶 , force
at the tips of the beam 𝐹(𝐿, 𝑡), the tips displacement 𝑤(𝐿, 𝑡), and total charge on the
capacitor 𝑄𝐶 (𝑡), respectively. This coupling matrix is described below:
1
𝐶 0
𝑈 (𝑡) 𝐶(𝑡) 𝑄 𝐶 (𝑡)
[ ] = [𝛼𝐸 (ℎ +ℎ ) 3 3 ][ ], (10)
𝐹(𝐿, 𝑡) 𝑃 𝑆 𝑃 4𝑏𝐸𝑃 ℎ𝑆
( 2 ) [(1 +
2ℎ𝑃 𝐸 1
) − 1 + 𝐸𝑆 ] 𝐿2 +𝑤2 𝑤(𝐿, 𝑡)
2𝐿 3𝐿 ℎ𝑆 𝑃

A common property of the coupling matrix in equation (9) and (10) is that the factor of “0” on
the top right of the coupling matrix represents the sensing coupling. An input displacement
or stress induce an output voltage. Literature [40-43] has shown that the trilayer transducer
used in our study shows bidirectional operation. An applied voltage can induce a bending
output and an input displacement can produce a voltage output. Since the electromechanical
coupling is small, the influence of the sensing part on actuation is small and is therefore
neglected here. Sense voltages are in the range of 1 mV while actuation voltages applied are
typically 1 V or more. Within the scope of this chapter the focus is on actuation and for now
the sensing aspect is ignored. However, this aspect can easily be introduced, as is done in
Chapter 5.

The bottom right factors of the coupling matrix in equation (9) and (10) provide feedback,
which changes the total moment/force based on the output rotating/displacement value.

4.3.3.3 Mechanical model

The generated moment from coupling effects is then applied to the trilayer actuator resulting
in the bending. Since the ratio between thickness and length of the ionic EAP actuator is small
((2ℎ𝑃 + ℎ𝑆 )/𝐿 ≪ 1), this actuator can be considered as a thin beam.

In addition, evidence of the large displacement of the trilayer actuator used in this study and
shown in Fig. 16 (section 4.5.1) suggests that a nonlinear model should be developed to
describe the mechanical system. Several beam modeling methods exist, such as, for a small

92
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
displacement: the Euler-Bernoulli beam theory [44], the Timoshenko beam theory, and the
Rayleigh-Ritz beam theory [45], and for a large displacement such as the Galerkin finite
element method to solve second order nonlinear differential equations [46], as well as the
rigid finite element method (RFEM) [47]. RFEM allows us to take into account the nonlinear
dynamic mechanical behavior of the actuator easily, and to account for inertia [25, 48, 49].
Effects of mass have been left out of previous models [50, 51] since conducting polymer
actuators have historically been slow. However the PEDOT-NBR/PEO trilayer similar to those
presented in this work have been shown to be fast in their responses [52], so inclusion of
inertial effects in the proposed model is important to predict the actuation at high operating
frequencies.

Fig. 7. Rigid Finite Element Model (RFEM) of a trilayer actuator, in which it is divided into
(n+1) rigid elements: from RFE0 to RFEn, the elements each have a mass and are linked by a
spring-damper element SDE1 to SDEn, shown in their initial positions. The integrated
coordinator includes X0 axis along the length of the actuator, X1 axis perpendicular to the X0
and X2 axis perpendicular to the X0X1 surface.

So, if we consider RFEM, the actuator is divided into (𝑛 + 1) rigid elements from RFE0 to RFEn,
where the length of each element is calculated according to the following rule: Li = L/n, i = 1,
…, n-1 and L0 = Ln = L/2n. Each element is rigid and has a mass. Consecutive elements are
linked together by a torsional spring-damping element (SDE), which allows a rotation around
its axis X2 (perpendicular to X0X1 surface) (Fig. 7). The general Lagrange principle at each SDE
is written as follows:

𝑑 𝜕𝑇 𝜕𝑇 𝜕𝐺 𝜕𝐷
( ) − 𝜕𝜃 + 𝜕𝜃 + 𝜕𝜃̇ = 𝑀𝑖 , (11)
𝑑𝑡 𝜕𝜃𝑖̇ 𝑖 𝑖 𝑖

where i = 1, …, n, and T, G, and D are the kinetic energy, potential energy, and dissipation
energy, respectively, of the system of (n+1) elements. The terms in the equation are now
described mathematically, following the treatment presented by Wittbrodt et al. [49] and
Moghadam et al. [53] [54].

93
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior

Fig. 8. Trilayer actuator in a bent state. The bending angle between the element (i-1) and the
element i is 𝜃𝑖 .

Considering a differential element dmi on RFEi, the kinetic energy of this element is:

1
𝑑𝑇𝑖 = 2 𝑡𝑟{𝑟̇𝑖 𝑟̇𝑖𝑇 }𝑑𝑚𝑖 , (12)

𝑇 (𝑖) (𝑖)
where 𝑟𝑖 = [𝑋0(𝑖) (𝑖)
𝑋1 ] . 𝑋0 , and 𝑋1 are the components of displacement in two axes of
the ith element, in the local coordinates, as depicted in Fig. 8. The total kinetic energy of the
system is then:

1 1
𝑇𝑖 = 2 ∫𝑚 𝑡𝑟{𝑟̇𝑖 𝑟̇𝑖𝑇 }𝑑𝑚𝑖 = 2 𝑡𝑟{𝐵̇𝑖 𝐻𝑖 𝐵̇𝑖𝑇 }, (13)
𝑖

𝑚𝑖 𝐿𝑖 2 /3 0 𝑚𝑖 𝐿𝑖 /2
̇
where pseudo-matrix 𝐻𝑖 = ∫𝑚 𝑟̌𝑖 𝑟̌𝑖 𝑑𝑚𝑖 = [ 0 𝑚𝑖 ℎ2 /12 0 ], (14)
𝑖
𝑚𝑖 𝐿𝑖 /2 0 𝑚𝑖

mi is the mass of the RFEi, h is the thickness of the actuator, and Li is the length of the RFEi.

The matrix 𝐵𝑖 = 𝐵̌1 𝐵̌2 … 𝐵̌𝑖 , enables the description of the total rotation of an element and its
𝐶𝑜𝑠 𝜃𝑖 −𝑆𝑖𝑛 𝜃𝑖 𝐿𝑖
position as a function of time. Here, 𝐵̌𝑖 = [ 𝑆𝑖𝑛 𝜃𝑖 𝐶𝑜𝑠 𝜃𝑖 0 ], which represents the
0 0 1
rotational matrix of the element the RFEi, expressed in terms of angles, 𝜃𝑖 , relative to the
previous beam segment RFEi-1, and the length of each segment, Li (Fig. 24).

Then the total kinetic energy of the system is obtained by the sum of T i:

𝑇 = ∑𝑛𝑖=1 𝑇𝑖 . (15)

94
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
1 𝐸𝐼
The total potential energy of the system is 𝐺 = 2 ∑𝑛𝑖=1 𝑘𝑖 𝜃𝑖2 (16), where 𝑘𝑖 = , which is the
𝐿𝑖
stiffness of each RFEi element [55].
1 𝜂𝐼
The total dissipation energy is 𝐷 = 2 ∑𝑛𝑖=1 𝑐𝑖 𝜃̇𝑖2 (17), where 𝑐𝑖 = , 𝜂 is the normal damping
𝐿𝑖
material constant, I is the inertial moment, and E is the Young’s modulus of the trilayer
actuator.

The external moment is determined by the electrochemomechanical coupling established

from equation (5).

The final Lagrange equation is obtained by inserting equations (17), (16), and (15) into (11).

𝑑 𝜕𝑇
In order to describe the Lagrange principle in a BG, the component ( ) needs to be
𝑑𝑡 𝜕𝜃̇ 𝑖

rewritten as the following chain:

𝑑 𝜕𝑇 𝜕 𝜕𝑇 𝜕𝜃𝑗 𝜕 𝜕𝑇 𝜕𝜃̇𝑗
( ) = ∑𝑛𝑗=1 [𝜕𝜃 (𝜕𝜃̇ ) + 𝜕𝜃̇ (𝜕𝜃̇ ) ], (18)
𝑑𝑡 𝜕𝜃̇ 𝑖 𝑗 𝑖 𝜕𝑡 𝑗 𝑖 𝜕𝑡

and the component 𝐵̇𝑖 is described as:

𝑑𝐵 𝜕𝐵 𝜕𝜃
𝐵̇𝑖 = 𝑑𝑡𝑖 = ∑𝑖𝑗=1 [𝜕𝜃 𝑖 𝜕𝑡𝑗 ]. (19)
𝑗

This set of Lagrange equations can be rewritten in the following form:

𝜃̈1 = 𝑓(𝜃̇1 , 𝜃̇2 , … , 𝜃̇𝑛 , 𝜃1 , 𝜃2 , … , 𝜃𝑛 , 𝑡)

𝜃̈2 = 𝑓(𝜃̇1 , 𝜃̇2 , … , 𝜃̇𝑛 , 𝜃1 , 𝜃2 , … , 𝜃𝑛 , 𝑡), (20)
⋮
̈𝜃𝑛 = 𝑓(𝜃̇1 , 𝜃̇2 , … , 𝜃̇𝑛 , 𝜃1 , 𝜃2 , … , 𝜃𝑛 , 𝑡)

or in a compact form:

Ȋ(𝜃, 𝜃̈). 𝜃̈ + R𝜃̇ + 𝐶𝜃 + 𝑒(𝜃, 𝜃̇) = 𝑀 (21)

In our case, since the actuator is initially curved, the initial condition can be provided by setting
the vector of the rotational angles of the RFEs at the time t = t0: θ(𝑡0 ) =
[𝜃1 (𝑡0 ) 𝜃2 (𝑡0 ) … 𝜃𝑛 (𝑡0 )]𝑇 . Assuming that the actuator is stationary at the time t = t 0:
𝑇
θ̇(𝑡0 ) = [𝜃̇1 (𝑡0 ) 𝜃̇2 (𝑡0 ) … 𝜃̇𝑛 (𝑡0 )] = [0 0 … 0]𝑇 . Finally, the results form a matrix
of rotational angles at each SDE during the time span t = t0, …, tm, where time has been divided
into m:

95
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
𝜃(𝑡0 ) 𝜃1 (𝑡0 ) 𝜃2 (𝑡0 ) … 𝜃𝑛 (𝑡0 )
𝜃(𝑡1 ) 𝜃 (𝑡 ) 𝜃2 (𝑡1 ) … 𝜃𝑛 (𝑡1 )
[ ] = [ 1⋯1 ⋯ … … ]. (22)
⋮
𝜃(𝑡𝑚 ) 𝜃1 (𝑡𝑚 ) 𝜃2 (𝑡𝑚 ) … 𝜃𝑛 (𝑡𝑚 )

Knowing the rotational angles as a function of time, the position of the RFEs and the actuator
shape can be easily derived.

Fig. 9. Mechanical BG model.

Based on equation (21), the mechanical BG model is described as in Fig. 9. Here, the input
power, represented by the applied moment, M, and the angular velocity 𝜃̇, is divided among
n-port R-elements, n-port C-elements,n-port Ȋ-elements, and a multiport source element
MSe, which indicates the damping constant, the inverse of the spring constant, the mass, and
nonlinear dynamic rotation 𝑒(𝜃, 𝜃̈). The multi-BG notation (n bonds) is used here since it turns
out to be a natural and a concise way of representing the power behavior of macroscopic
multiport systems.

For a simple linear model, the Euler-Bernoulli dynamic beam theory is selected to present the
mechanical model instead of the rigid finite element method used in the nonlinear describing
previously. The Euler-Bernoulli theory provides an accurate model, low-order, lumped-
parameter representation of distributed elements to predict the small displacement of a
bending beam.

The Euler-Bernoulli equation (21) can be also derived from the Lagrange principle (equation
7) by assuming that the dissipation energy is equal to zero (D = 0). The potential energy of a
single beam element is then calculated as below:

96
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
2
𝜕2 𝑤(𝑥,𝑡)
𝐺 = 𝐸𝐼 ( ) , (23)
𝜕𝑥 2

and its kinetic energy is given by:

𝜕𝑤(𝑥,𝑡) 2
𝑇 = 𝜌𝐴 ( ) . (24)
𝜕𝑡

Yielding equation (23) and equation (24) into equation (11) gives equation (25).

Karnopp [56] introduced the first model BG to describe results of solving Euler-Bernoulli beam
equation using the separation of variables method. A similar method can be used to derive
the CP actuator model in this paper, but with different boundary conditions. The Euler-
Bernoulli equation for a dynamic beam is developed as follows:

𝜕4 𝑤(𝑥,𝑡) 𝜕2 𝑤(𝑥,𝑡)
𝐸𝐼 + 𝜌𝐴 = 𝐹(𝐿, 𝑡). (25)
𝜕𝑥 4 𝜕𝑡 2

Here 𝐸𝑆 is SPE layer’s Young’s modulus, while 𝐴 and 𝑏 are cross section and width of the
beam, respectively.

The set of boundary conditions for a CP beam with clamping at one end are:
𝜕𝑤(𝑥,𝑡)
𝑤(𝑥, 𝑡)|𝑥=0 = 0; | =0 . (26)
𝜕𝑥 𝑥=0

In the case of an applied point force at the other free end:

𝜕2 𝑤(𝑥,𝑡) 𝜕 3 𝑤(𝑥,𝑡)
| = 0 ;−𝐸𝐼 | = 𝐹(𝐿, 𝑡) . (27)
𝜕𝑥 2 𝑥=𝐿 𝜕𝑥 3 𝑥=𝐿

The separation of variables method [56] begins by assuming the displacement 𝑤(𝑥, 𝑡) can be
separated into a function of 𝑥 only , 𝑌(𝑥) and a function of time, 𝑡, 𝜇(𝑡):

𝑤(𝑥, 𝑡) = 𝑌(𝑥) 𝜇(𝑡). (28)

Substituting (7) into (5), the solution for this equation is in form of mode shape functions:

𝑌𝑛 (𝑥) = (𝑐𝑜𝑠 (𝑟𝑛 𝐿) − 𝑐𝑜𝑠ℎ( 𝑟𝑛 𝐿))(𝑠𝑖𝑛( 𝑟𝑛 𝑥) + 𝑠𝑖𝑛ℎ( 𝑟𝑛 𝑥))

−(𝑠𝑖𝑛( 𝑟𝑛 𝑥) − 𝑠𝑖𝑛ℎ( 𝑟𝑛 𝑥))(𝑐𝑜𝑠 (𝑟𝑛 𝐿) + 𝑐𝑜𝑠ℎ( 𝑟𝑛 𝐿)) , (29)

and the contribution of 𝑛𝑡ℎ mode is:

𝑚𝑛 𝜇𝑛̈ (𝑡) + 2𝜁𝑛 𝜔𝑛 𝑚𝑛 𝜇̇ 𝑛 (𝑡) + 𝑚𝑛 𝜔𝑛2 𝜇𝑛 (𝑡) = 𝐹(𝑡)𝑌𝑛 (𝐿), (30)

𝐸𝐼 (𝑟𝑛 𝐿)4
where 𝑟𝑛 𝐿 yields equation: 𝑐𝑜𝑠 (𝑟𝑛 𝐿)𝑐𝑜𝑠ℎ( 𝑟𝑛 𝐿) = 1, mode frequencies 𝜔𝑛2 = 𝜌𝐴 ,
𝐿4
𝐿
mode mass 𝑚𝑛 = ∫0 𝜌𝐴𝑌𝑛2 𝑑𝑥, and model damping 𝑏𝑛 = 2𝜁𝑛 𝜔𝑛 𝑚𝑛 in which damping ratio 𝜁𝑛
is found by experiment. The model stiffness is given by: 𝑘𝑛 = 𝑚𝑛 𝜔𝑛2 .

97
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
Once mode shape functions 𝑌𝑛 (𝑥) and contribution functions 𝜇𝑛 (𝑡) have been determined,
equation (28) can be rewritten as: 𝑤(𝑥, 𝑡) = ∑∞
𝑛=1 𝑌𝑛 (𝑥)𝜇𝑛 (𝑡), or in differential form:

𝑤̇ (𝑥, 𝑡) = ∑∞
𝑛=1 𝑌𝑛 (𝑥)𝜇𝑛̇ (𝑡). (31)

From the equations (26), (27), (30), the BG representation of the mechanical model can be
established as in the Fig. 10. It is worth noticing that the value of 𝑌𝑖 (0) and 𝑌𝑖 (𝐿) obtaining
from the equation (29), together with the model mass 𝑚𝑖 and model stiffness 𝑘𝑖 can be all
calculated once the dimensions, Young’s modulus, and density of the trilayer actuator are
determined.

Fig. 10. BG representation of the mechanical model of the trilayer actuator, the block 1 (red
color) indicates the point force F(L, t) obtaining from the electromechanical coupling matrix
applying to the free end of the beam, the block 2 (green color) presents the boudary
condition where one end of the beam is fixed, the block 3 (pink color) is a BG presentation of
the equation (25), and finally the block 4 (orange color) introduces the method to extract the
tips displacement of the beam w(L, t).

4.3.4 BG global models

In previous sections, the electrochemical model, the electrochemomechanical coupling, and

the mechanical model have been established. To form the global model, the electrochemical

98
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
model is combined with the electrochemomechanical coupling in BG. The moment is exported
from the coupling, and then transmitted into the mechanical model. For simplicity, the
elements in the model are organized and divided into small sub-systems as depicted in Fig.
11.

Fig. 11. CP actuator global model: 1. Voltage source, 2. Electrochemical model, 3. Coupling
between electrochemical and mechanical part, 4. Mechanical model, 5. Boundary conditions:
initial stress, 6. Displacement output.

It is worth noticing that the number of the RFE elements is (n + 1) whereas the SDE elements
is n. Theoretically, the higher the number of RFE elements is, the more precise the simulation
will be in comparison with the analytical method but at the expense of computational time.
A rough estimation has shown that the relative error is 1.14 % and 0.41 % at first resonant
frequency [47] with respect to the analytical method solution when the number of RFE is 4 (n

99
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
= 3) and 6 (n = 5), respectively. Within our proposed model, we have chosen two cases where
4 RFEs and 6 RFEs are used to investigate the accuracy of our model.

To establish the global linear model, the physical model, the electromechanical coupling, and
the mechanical model obtained in previous section are connected. The elements in the model
are organized and divided into small sub-systems as depicted in Fig. 12. The power
transmitted through the different sub-system is followed which proves one advantage of the
BG approach.

Fig. 12. Linear BG global model of CP actuators, where a source of voltage is employed as the
input and the tips displacement of the beam is the output of the model.

Dissipation during one cycle is 12 µJ, which includes both electrochemical and mechanical
dissipation. If all of this heat remains in the sample, the resulting increase in temperature is
approximately 0.02 0C. As the actuator is operated continuously during 100 cycles, the
increasing temperature would have reached up to 2 0C. However, since the trilayer is small
and thin, the heat convection between the interface of the actuator and surrounding air is
assumed to be sufficient to cool down the actuator and thermal effects are ignored.

However, in Bond Graph language, the coupling between thermal dissipation and thermal
expansion could be modeled by introducing an electro-thermal-mechanical element.
Continuing in this sense, another coupling in parallel to electro-chemo-mechanical coupling
will be added into the current model.

4.4 Simulation tests

4.4.1 Software implementation

To implement the BG models, the 20-sim mechatronic modeling package, version 4.6 (Student
License), was employed. This commercial software provides a simple method to graphically

100
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
build, simulate, and analyze the model behavior in multiple domains, and create control
systems.

20-sim offers various toolboxes such as block diagrams, physical components, and BGs, and
flexible links among others, enabling a quick and intuitive design.

Figure 13. Interface of 20-sim simulation software

An example of 20-sim user interface is shown in Fig. 13, where the linear model of the trilayer
actuator is organized into submodels.

4.4.2 Comparison between the linear and nonlinear simulations

In the previous section, two models, including a linear model and a nonlinear model, were
built to predict the actuation behavior of the trilayer actuators. This section will deal with the
advantages and disadvantages of both models by comparing simulations.

It is worth noting that to simulate the nonlinear model, parameters obtained from chapter 3
(section 3.5) are employed and processed into the built model. These also include the
potential dependence of the electrical conductivity, the volumetric capacitance, and the
Young’s modulus of the PEDOT electrodes. For the linear model, by assuming the potential
is V = 0, the required values can be extracted and described in table 3.

101
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
Table 3 Summary of derived parameters for the linear model

Length x width x PEDOT’s 5 x 1 x 0.0035 x 0.01

thicknesses x NBR/PEO
thickness (mm)
PEDOT electrodes’ volumetric 6.8 × 107
capacitance (F/m3)
NBR/PEO ionic conductivity 0.038
(S/m)
Strain-to-charge ratio (m3/C) 𝛼 = 3.6 ± 0.5 × 10−10
NBR/PEO Young’s modulus 329 ± 50
(kPa)
Trilayer actuator Young’s 9.54
modulus (MPa)
PEDOT Young’s modulus 12.98
(MPa)
Damping ratio 0.39
Beam density (kg/m3) 2000

In the first step, the output currents from the linear model and the nonlinear model (with 6
RFEs) under a step voltage of 2 Vpp and 0.1 Hz are compared (Fig. 14). Considering half cycle
of the applied voltage, it is clear that during the first few seconds (< 2s) the current output
from the 6 RFEs decays quicker than the linear model. An inspection on the nonlinear model
shows that this quicker decay is due to the incorporation of PEDOT electronic resistance into
the model, as a higher PEDOT electronic resistance induces a faster decay. At a longer
response (>2s), the output currents from two models match quite well.

Fig. 14. Comparison between the output current of the linear model and nonlinear (6 RFEs)
responses to a step voltage of 2 Vpp at f = 0.1 Hz

102
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
In next step, the bending displacement simulations of two models are compared (Fig. 15). At
a small magnitude of the applied voltage (Vpp = 0.5 V, 1 V), these models show a similar
actuation behavior, as expected. When the magnitude is increased (Vpp = 1.5 V, 2 V) the
difference between two models is getting bigger. This can be explained as the Euler –
Bernoulli beam theory cannot predict the beam displacement (w (x, t)) in x direction but only
in y direction. The fact that the linear model cannot predict the large displacement is clearly
shown in this comparison. However, the nonlinear model (with only 6 RFEs) can better
accommodate large bending.

Fig. 15. Comparison between the linear model and the nonlinear model in bending
displacement of the actuator in response to the step input voltage with various magnitudes

Two comparisons above visually show the disadvantages of the linear model against the
nonlinear one. A full comparison between these two models is summarized in the table 4.

103
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
Table 4. Pros and cons of the linear model in comparison to the nonlinear model

Criterion Linear model Nonlinear model

Model establishing Easy Difficult
Required parameters 9 13 (3 are potential
dependent) - much longer
characterization
Time needed to run the Short time (30 s) Long time (10 min)
model
Ability to implement Impossible Possible
nonlinear properties of
materials
Ability to implement Impossible Possible
asymmetric properties
between the PEDOT
electrodes
Large displacement Impossible Possible with high accuracy
prediction
Ability to predict voltage Impossible Possible
drop along the length of
the beam
Dynamic behavior Possible Possible
prediction
Accessibility to Possible Possible
information on energy
and efficiency
Ability to implement for a Possible (but may become Possible
complex structure complex in the process)

In general, the proposed nonlinear model is an advance relative to the linear model as it
overcomes all the weaknesses available in the linear model. However, establishing and
characterizing the necessary parameters for the nonlinear model requires much more time.
The linear can be considered to use i) where a small displacement is expected, and ii) a small
voltage magnitude is applied.

In this study, since the electrochemomechanical properties of PEDOT depend on its oxidation
state, as demonstrated, and the trilayer actuator shows a large displacement responding to a
voltage excitation, the proposed nonlinear model is selected for further study in the next
sections.

4.5 Comparison to experimental tests

In the previous section, a nonlinear BG model was proposed to predict actuation behavior
and the parameters required for the proposed model were characterized. To evaluate the
agreement between the BG model and the experimental results, a set of experiments has

104
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
been performed. The geometry of the trilayer actuator used in these characterizations is
equal to length x width x thickness = 5 mm x 1 mm x 0.017 mm, unless otherwise is stated.
First, the simulation and experimental current outputs will be compared, followed by the
mechanical responses. Then an analysis of performance is made to identify the critical
parameters affecting it. Finally, the energy efficiency will be investigated.

4.5.1 Time domain responses

Fig. 16 shows a trilayer actuator under a square wave voltage excitation. The asymmetry
between the top and the bottom PEDOT electrode layers is demonstrated via an initial curved
beam state (Fig. 16b). This curved state is due to stresses inside the trilayer. To account for
this bending in the neutral state in comparisons with model preditions below, the model was
used to find the relative change in curvature, which is added to the initial curvature of in the
simulation response.As seen in Fig. 16a, beam deflects leftwards when -1.1 V is applied to the
left side (bottom) electrode and +1.1 V is applied to the right side top PEDOT electrode layer.
As the voltage direction is reversed, the beam bends back until it is straight (Fig. 16c).

Fig. 16. Trilayer actuator under voltage excitation: a) Top PEDOT electrode voltage: -1.1 V,
Bottom PEDOT electrode: 1.1 V, b) neutral state, c) Top PEDOT electrode: 1.1 V, Bottom
PEDOT electrode: -1.1 V.

The input current and displacement of the beam were captured and compared with model
predictions. As demonstrated in the previous section, the PEDOT electrical conductivity, the
PEDOT volumetric capacitance, and the Young’s modulus of the trilayer actuator are
dependent on the applied potential, and this dependence was fitted with a polynomial to
improve the convergence when solving the BG model. All voltage dependencies were
incorporated into the proposed BG model.

A triangular wave allowing more precise position control followed by a square wave voltage
providing a fast actuation response was applied to the trilayer actuator (Fig. 17a). The
recorded current was then compared with the output current from the electrochemical sub-
model.

105
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior

Fig. 17. a) Triangular wave and rectangular wave applied voltage, b) Measured and
simulation current response

As shown in Fig. 17b, the simulated current (4 RFEs and 6 RFEs simulated current are
indistinguishable, so only the result of 6 RFEs is plotted) matched the experimental results
quite well, and there is an asymmetry in the output current between two halves cycles of the
applied voltage. This asymmetry is due to the dependence of the electronic conductivity of
the PEDOT electrodes on the applied potential. In the case of a step voltage, the response
time and the amplitude of the current of the models were slightly faster than in the
experiment. This is due to the limited sampling rate (5 points/s) of the device used to measure
the current.

106
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior

Fig. 18. Modeled voltage drops along the length of the beam due to the existence of
electrical resistance of PEDOT electrodes. The black curve is the voltage seen at the fixed end
of the cantilever where electrical contact is made. The dashed green line is the predicted
voltage as seen at 0.9 of the length, and the others are in between.

Fig. 18 depicts the modeled drop in voltage along the length of the beam. In the step
response, a significant reduction in the voltage is observed at the position x = L/10 in
comparison with the applied voltage (at the position x = 0), close to the clamping point, at
short times (< 0.3 s). For longer times (> than half a second), the drop becomes negligible. The
drop increases further away from the clamping point, as expected. However, in the ramp
response, it is clear that the reduction in voltage is not significant (around 7% at the tip of the
beam), and the maximum voltages are shifted to the right showing a delay or a slower
actuation response as we go further away from the clamping point. An asymmetric voltage
between the two half cycles of the applied voltage is also observed, which is due to the
variation in PEDOT electronic conductivity following the change in voltage.

To evaluate the ability of the model to track the mechanical behavior of the trilayer actuator,
the beam displacement was recorded using a camera to determine the radius of curvature
and the bending strain as a function of time and position along the length of the beam. The
bending induced strain was determined from the radius of curvature, which is calculated from

107
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
the rotation angle 𝜃𝑖 of each RFE. In the first set of experiments, the magnitude of the square
wave voltage was varied from 0.5 Vpp to 3 Vpp, while the frequency was held at 0.1 Hz. Fig.
19 shows quite good agreement between the measured and the theoretical strain in both
cases (where the beam was divided into 4 RFEs (3 SDEs) or 6 RFEs (5 SDEs)), except for the
point where the voltage is 3 Vpp. This difference is due to a change in the electrical and the
mechanical properties of the trilayer actuator as a result of the irreversible degradation in
polymer chains inside the electrodes [57]. However, this change was not considered in the
proposed model. Continuously operating the actuator at this high voltage will significantly
decrease its lifetime. As it has been observed by the authors, after tens of cycles of deflection
at this high applied voltage, the magnitude of displacement will quickly decrease to zero.

Fig. 19. Comparison between the simulations and the experiments on the strain as a function
of the magnitude of applied voltage at a constant frequency of 0.1 Hz

4.5.2 Frequency responses

An advantage of the proposed model is the ability to incorporate the damping coefficient and
the mass of the actuator. This becomes particularly important at high frequencies and short
times, as is now shown.

In a second set of experiments, the frequency of the applied square wave voltage was varied
between 0.05 Hz and 15 Hz while the magnitude was constant at 2.2 Vpp (Fig. 20a). Fig. 20a
shows the magnitude of the error between the model and the simulation as a function of
frequency. In the simulations, the damping coefficient was varied to study its effect on the
goodness of fit. The value measured in Chapter 3 is 0.39, which also gives the best match to
the data. To simplify these two figures, the 4 RFEs curves have been removed from the graph
as they are very similar to the 6 RFEs curves. Once again, a good agreement between the
simulations (indicated by “Simulation 6 RFEs: η = 0.39” curve in Fig. 20a, b and experiments is
observed. It is worth noting that this damping ratio η = 0.39 was obtained via a vibration

108
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
measurement [58]. However, when the frequency increases to over 5 Hz, a difference can be
seen (a relative error of approximately 15%). This inaccuracy may come from the method
used to measure the strain based on capturing actuation (limited frame rate: 24fps) and
imaging analysis which is not very accurate at high frequency.

Fig. 20. a) Comparison between the simulations and the experiments on the strain as a
function of the frequency of applied voltage at a constant magnitude of ±1.1 V, b) The
relative error of the simulations compared to the measured data.

Damping shows a significant effect on the strain of the actuator at intermediate and high
frequency (> 0.5 Hz). In the case of overdamping, where the assumed damping ratio (η = 3.9)
is ten times higher than the measured value [58], the strain drastically drops as the frequency
is increased and the relative error is around 100%. A small increase in strain is seen when the
damping ratio is assigned as a nearly undamped (η = 0.039) situation, and the first resonant
frequency of the actuator is clearly visible in this case as there is an increase in strain around
10 Hz (Fig. 20a, b). When mass of the actuator is removed from the model, there is a clear

109
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
overestimation of strain by the simulation (by approximately 100%), demonstrating the
importance of including mass in the simulations at moderate to high frequencies.

The model has the ability to predict the dynamic mechanical response of the beam. A voltage
of 2.2 Vp at 0.1 Hz was applied to the trilayer. The beam was previously held at -2.2 V (top vs.
bottom electrode) in the almost straight position depicted in Fig. 16c. At the time 0 s, the
voltage reserved the direction (Fig. 16a) and the dynamic behavior of the beam was captured
at time steps of 0 s, 1 s, 3 s, and 5 s, and then compared to the simulation results. As shown
in Fig. 21, the predicted results follow the beam movement at those time points.

Fig. 21. Comparison between the experiment and the simulation in terms of the dynamic
mechanical behavior of the beam subjected to a voltage excitation of 2.2 Vpp at 0.1 Hz.

In general, the nonlinear model shows a good agreement with the experimental results. In
comparison with previous studies, the biggest advantage of our proposed model is the ability
to implement the nonlinear behavior of the PEDOT electrical conductivity and volumetric
capacitance and the Young’s modulus of the trilayer actuator due to its dependence on the
oxidation state. The integration of the mass and the damping effect into the proposed model
allows prediction of the actuation behavior at high frequency – which has become important
as high actuation rates are reached [52]. The ability to extract the nonlinear dynamic
mechanical displacement of the beam is another benefit of the model, which allows us to
track beam motion as a function of time.

110
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
4.6 Parameter sensitivity and power performance analysis
The previous sections have confirmed the ability to predict actuator performance of the
model in both time and frequency domains. Another advantage of the model is that it can
provide an insight into actuator performance as a function of its critical parameters, which
have been characterized. For example, Fig. 22 gives an assessment of how sensitive the active
strain difference (proportional to the deflection) is to Young’s modulus, PEDOT electrical
conductivity, SPE ionic conductivity, and the length of the actuator at both low (Fig. 22a) and
high frequency (Fig. 22b). This result suggests that for the actuator geometry studied,
reducing elastic modulus will have a significant effect on deflection, while the influence of the
length is very important at high frequency. Not surprisingly, conductivity has little effect on
strain at low frequencies of operation, but this effect becomes large at high frequencies.

Fig. 22. Key factors affecting the actuator displacement under a square wave voltage (Vpp =
2.4 V), a) at low frequency f = 0.1 Hz, and b) at high frequency f = 16 Hz. For each parameter,

111
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
three columns from left to right represent for its real value (blue column), its value increased
by 20% (orange column), and its value reduced by 20% (grey column).

The value of these parameters have been adjusted by ± 20% of their values while the others
were maintained constant. It is clear that at low frequency (f = 0.1 Hz) of applied voltage, an
increase of the PEDOT electrical conductivity would lead to very little increase in strain
difference while a stiffer beam would reduce displacement. SPE ionic conductivity shows a
small effect on the beam displacement, as expected. At high frequency (f = 16 Hz), while the
effect of the Young’s modulus remains similar, the PEDOT conductivity has a great influence
on the strain difference, and the increase of length reduces deformation. The latter are
expected due to the high currents seen by the trilayer at short times, and hence the large
voltage drops along the length.

In addition, the proposed BG model allows ready access to information about the energy
stored, dissipated, and transmitted – in this case electrochemical and mechanical energies.
Considering half a period of the applied voltage (0.1 Hz at 2 Vpp), Fig. 23 shows the evolution
of the electrochemical (left) and the much smaller mechanical (right) component energies
with time, including both stored and dissipated contributions. As expected for a system
composed of resistors with capacitors in series, the input energy increases with time until the
energy storage is complete (Fig. 23a). Mechanical energy also rises and then saturates, with
the same time constant (Fig. 23b).

112
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior

Fig. 23. Evolution of energy over half a period of applied voltage: a) In the electrochemical
part, b) In the mechanical part.

Fig. 24 replots the stored and dissipated energy components of energy as fractions of the total
input energy. From this plot it is evident that the mechanical energy represents about 0.2%
of the total. 40 % of the input electrical energy is stored, which is not surprising since a 50:50
split is what would be expected from a simple series RC circuit that is charged with a step
voltage input. A slow ramp input, in which currents are kept small to minimize iR drop, would
lead to a significantly higher component of stored energy. On the mechanical side, kinetic
energy and dissipation loss accounting for tiny portion, which depends on speed, rise quickly
and then decays again as the actuator accelerates and then decelerates. The stored
mechanical energy is elastic potential energy that accumulates as the beam bends, and is
more than 99% of the total. Dissipation losses can even be reduced, once again, by actuating
slower. The relatively small fraction of electrochemical energy converted to mechanical is
consistent with previous reports showing a small electromechanical coupling coefficient [10,
41, 59]. The efficiency however need not be so small, especially in slow charging situations
where much of the input electrical energy can be recovered during discharge. Previous work
shows that overall efficiency can exceed 10 %, if electrical energy is recovered [10].

113
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior

Fig. 24. Instantaneous electrical to mechanical coupling.

Using the proposed model, one can use to model to explore how to improve the energy
conversion from electrical into mechanical. Here, a slow ramp voltage is used instead of the
step voltage. The simulation results show that an input voltage of 100 mV/s with a frequency
of f = 0.05 Hz increases the electrochemical stored energy and mechanical potential energy
up to 60 % and 0.4 %, respectively. The switch from a step to a ramp doubles the coupling in
much the same way as charging an RC circuit with a ramp rather than a step increases charging
efficiency by reducing the ohmic losses. Furthermore, a slower ramp of 5 mV/s with a
frequency f = 0.0025 Hz is predicted to increase the electrochemical stored energy to 90%
and the mechanical potential energy to 3%.

The conversion to mechanical energy provides a value for the electrochemical internal
coupling. In practice we want the mechanical work to be applied to an external load. The
energy in the trilayer beam needs to be transferred to an external mass, spring or other type
of load, which further reduces the overall electromechanical coupling, since not all of the
elastic potential energy in the beam will be passed on.

Coupling is poor in these actuators. Most of the input energy is stored electrochemically, with
only a small fraction converted to work. The good news is that, if we can recover the stored
electrical energy in each charge/discharge cycle, the efficiency can be much higher than the
electromechanical coupling. For example, in the full charge/discharge cycle at 5 mV/s, 90 %
of energy is stored electrochemically. Assuming we can recover 90 % of this stored energy on
discharge, then we have lost 14 % of the input energy. The upper bound on efficiency is then
the output mechanical energy percent divided by the lost energy percent (3%/14%). This gives
an overall electrical to mechanical efficiency of 21 %. This percentage will be reduced by the
coupling between the beam and the external load, but nonetheless energy recovery
dramatically improves efficiency.

114
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
Furthermore, the electrochemical lost can be reduced by increasing the thickness of the
separator (NBR/PEO) to avoid the contact between two PEDOT electrodes. And an increase
in the strain-to-charge ratio by investigating different materials for the electrodes and
electrolyte will efficiently increase the electromechanical coupling.

4.7 Conclusion
In this chapter, a nonlinear dynamic model and a simpler dynamic linear model have been
proposed to simulate the actuation of ultrathin trilayer actuators. A comparison between
these two models has shown that the linear model is simple to build and easier to characterize
all of its require parameters. However, it is most suitable to predict small displacements, and
losses accuracy at large displacements and voltages.

The proposed nonlinear model showed advances to the existing work on trilayer actuators in
that it can precisely predict a nonlinear dynamic actuation with an ability to implement
nonlinear electrochemical and mechanical properties of the materials: PEDOT electrical and
ionic conductivities; volumetric capacitance values; and Young’s modulus variation. In
addition, it accounted for the mass and damping associated with the beam. All physical
characteristics used in the model were measured.

The concordance between the simulation and experimental results in both time and
frequency domains confirms the ability to predict the bending of the trilayer actuator.
Moreover, the approach is represented in Bond Graph language, providing a simple way to
access the energy information, evaluate the critical parameters affecting the performance of
the actuator, and allowing an insight into the phenomena occurring in the ionic actuator.

In addition, the model can also be used to optimize the design of the transducer before it is
fabricated. The simulation software allows the input parameters such as dimensions, and
material properties (conductivities, Young’s modulus, …) to be varied over defined ranges to
help define the range that will produce the desired response. These values will become a
guideline for the designer to tune the transducer properties. The tool can also be used to
determine which material properties most limit response in order to drive material
development or enable the evaluation of emerging materials. For example, a new material
may feature a higher strain to charge ratio, but also a lower ionic conductivity. The model is
capable of exploring the implications of this change.

For the future perspective, the proposed nonlinear model is theoretically suitable to predict
even higher frequency actuation. In addition, the model was built based on Bond Graph
language, providing at first a possibility to integrate it into a real actuator controlling system
and second, to develop a more complex model to control not only a single beam but also a
complicated geometry. These possibilities can bring more power to the model.

115
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
[1] D. Karnopp, R.C. Rosenberg, Analysis and simulation of multiport systems; the bond graph
approach to physical system dynamics, Cambridge, Mass.,: M.I.T. Press; 1968.
[2] D. Karnopp, System dynamics : modeling and simulation of mechatronic systems, Hoboken: Wiley;
2012.
[3] F. Meisam, S. Farrokh, N. Naser, F. Adelyne, P. Cédric, T.M.N. Giao, V. Frédéric, D.W.M. John,
Characterization and dynamic charge dependent modeling of conducting polymer trilayer bending,
Smart Materials and Structures, 25(2016) 115044.
[4] G.M. Spinks, L. Liu, G.G. Wallace, D. Zhou, Strain Response from Polypyrrole Actuators under Load,
Advanced Functional Materials, 12(2002) 437-40.
[5] C.H. Nguyen, G. Alici, G.G. Wallace, Modelling trilayer conjugated polymer actuators for their
sensorless position control, Sensors and Actuators A: Physical, 185(2012) 82-91.
[6] F. Carpi, E. Smela, Biomedical applications of electroactive polymer actuators, Chichester, West
Sussex: John Wiley & Sons; 2009.
[7] F. Carpi, Electromechanically active polymers, 1st ed., Place of publication not identified: Springer;
2016.
[8] Y. Fang, X. Tan, G. Alici, Robust Adaptive Control of Conjugated Polymer Actuators, IEEE
Transactions on Control Systems Technology, 16(2008) 600-12.
[9] J.D. Madden, P.G. Madden, I.W. Hunter, Polypyrrole actuators: modeling and performance, 2001,
pp. 72-83.
[10] J.D.W. Madden, Conducting polymer actuators: Massachusetts Institute of Technology; 2000.
[11] S. John, G. Alici, C. Cook, Frequency response of polypyrrole trilayer actuator displacement, SPIE
Smart Structures and Materials + Nondestructive Evaluation and Health Monitoring, SPIE2008, p. 8.
[12] A.D. Santa, D.D. Rossi, A. Mazzoldi, Characterization and modelling of a conducting polymer
muscle-like linear actuator, Smart Materials and Structures, 6(1997) 23.
[13] J.D.M. Tina Shoa, Chi-Wah Eddie Fok, Tissaphern Mirfakhrai, Rate Limits in Conducting Polymers,
Advances in Science and Technology, 61(2008) 7.
[14] C.H. Nguyen, G. Alici, G.G. Wallace, Modelling trilayer conjugated polymer actuators for their
sensorless position control, Sensors and Actuators, A: Physical, 185(2012) 82-91.
[15] C.-H. Tso, J.D. Madden, C.A. Michal, An NMR study of PF6− ions in polypyrrole, Synthetic Metals,
157(2007) 460-6.
[16] M.J. Ariza, T.F. Otero, Ionic diffusion across oxidized polypyrrole membranes and during oxidation
of the free-standing film, Colloids and Surfaces A: Physicochemical and Engineering Aspects, 270-
271(2005) 226-31.
[17] M. Farajollahi, A. Usgaocar, Y. Dobashi, V. Woehling, C. Plesse, F. Vidal, F. Sassani, J.D.W. Madden,
Nonlinear Two-Dimensional Transmission Line Models for Electrochemically Driven Conducting
Polymer Actuators, IEEE/ASME Transactions on Mechatronics, 22(2017) 705-16.
[18] G. Alici, B. Mui, C. Cook, Bending modeling and its experimental verification for conducting
polymer actuators dedicated to manipulation applications, Sensors and Actuators A: Physical,
126(2006) 396-404.
[19] G. Alici, N.N. Huynh, Predicting force output of trilayer polymer actuators, Sensors and Actuators
A: Physical, 132(2006) 616-25.
[20] A.A.A. Moghadam, M. Majid, T. Keivan, T. Masoud, Analytical dynamic modeling of fast trilayer
polypyrrole bending actuators, Smart Materials and Structures, 20(2011) 115020.
[21] A.A.A. Moghadam, W. Hong, A. Kouzani, A. Kaynak, R. Zamani, R. Montazami, Nonlinear dynamic
modeling of ionic polymer conductive network composite actuators using rigid finite element method,
Sensors and Actuators A: Physical, 217(2014) 168-82.
[22] X. Wang, B. Shapiro, E. Smela, Development of a Model for Charge Transport in Conjugated
Polymers, The Journal of Physical Chemistry C, 113(2009) 382-401.
[23] J.G. Martinez, T.F. Otero, Mechanical awareness from sensing artificial muscles: Experiments and
modeling, Sensors and Actuators B: Chemical, 195(2014) 365-72.

116
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
[24] V. Venugopal, H. Zhang, R. Northcutt, V.B. Sundaresan, A thermodynamic chemomechanical
constitutive model for conducting polymers, Sensors and Actuators B: Chemical, 201(2014) 293-9.
[25] M. Szczotka, A modification of the rigid finite element method and its application to the J-lay
problem, Acta Mechanica, 220(2011) 183-98.
[26] G. Nishida, K. Takagi, B. Maschke, T. Osada, Multi-scale distributed parameter modeling of ionic
polymer-metal composite soft actuator, Control Engineering Practice, 19(2011) 321-34.
[27] B.J. Yi, I. Busch-Vishniac, Modeling of EAPs as multiple energy domain systems: A bond graph
approach, Proceedings of SPIE - The International Society for Optical Engineering2006.
[28] T.A. Bowers, MSc thesis title: Modeling, Simulation, and Control of a Polypyrrole-Based
Conducting Polymer Actuator: Massachusetts Institute of Technology; 2004.
[29] M. Bentefrit, S. Grondel, C. Soyer, A. Fannir, E. Cattan, J.D. Madden, T.M.G. Nguyen, C. Plesse, F.
Vidal, Linear finite-difference bond graph model of an ionic polymer actuator, Smart Materials and
Structures, 26(2017) 095055.
[30] T.N. Nguyen, K. Rohtlaid, C. Plesse, G.T.M. Nguyen, C. Soyer, S. Grondel, E. Cattan, J.D.W. Madden,
F. Vidal, Ultrathin electrochemically driven conducting polymer actuators: fabrication and
electrochemomechanical characterization, Electrochimica Acta, 265(2018) 670-80.
[31] G.A. Snook, P. Kao, A.S. Best, Conducting-polymer-based supercapacitor devices and electrodes,
Journal of Power Sources, 196(2011) 1-12.
[32] F. Vidal, C. Plesse, D. Teyssié, C. Chevrot, Long-life air working conducting semi-IPN/ionic liquid
based actuator, Synthetic Metals, 142(2004) 287-91.
[33] C.M. Proctor, J. Rivnay, G.G. Malliaras, Understanding volumetric capacitance in conducting
polymers, Journal of Polymer Science Part B: Polymer Physics, 54(2016) 1433-6.
[34] T. Shoa, J.D. Madden, N.R. Munce, V. Yang, Analytical modeling of a conducting polymer-driven
catheter, Polymer International, 59(2010) 343-51.
[35] J.D. Madden, P.G. Madden, I.W. Hunter, Conducting polymer actuators as engineering materials,
SPIE's 9th Annual International Symposium on Smart Structures and Materials, SPIE2002, p. 15.
[36] J.C.E. Marin S. Halper, Supercapacitors: A Brief Overview, MITRE Nanosystems Group, 2006.
[37] I.J. Busch-Vishniac, H.M. Paynter, Bond graph models of acoustical transducers, Journal of the
Franklin Institute, 328(1991) 663-73.
[38] P.G.A. Madden, Development and Modeling of Conducting Polymer Actuators and the Fabrication
of a Conducting Polymer Based Feedback Loop: MASSACHUSETTS INSTITUTE OF TECHNOLOGY; 2003.
[39] J.D.W. Madden, PhD thesis title: Conducting Polymer Actuators, Massachusetts Institute of
Technology2000.
[40] T.F. Otero, Biomimetic Conducting Polymers: Synthesis, Materials, Properties, Functions, and
Devices, Polymer Reviews, 53(2013) 311-51.
[41] T. Shoa, J.D.W. Madden, T. Mirfakhrai, G. Alici, G.M. Spinks, G.G. Wallace, Electromechanical
coupling in polypyrrole sensors and actuators, Sensors and Actuators, A: Physical, 161(2010) 127-33.
[42] N. Festin, C. Plesse, C. Chevrot, D. Teyssié, L. Josselin, P. Pirim, F. Vidal, Actuation and Sensing
properties of Electroactive Polymer Whiskers, Procedia Computer Science, 7(2011) S4-S7.
[43] Y. Wu, G. Alici, J.D.W. Madden, G.M. Spinks, G.G. Wallace, Soft Mechanical Sensors Through
Reverse Actuation in Polypyrrole, Advanced Functional Materials, 17(2007) 3216-22.
[44] S.A. Sina, H.M. Navazi, H. Haddadpour, An analytical method for free vibration analysis of
functionally graded beams, Materials & Design, 30(2009) 741-7.
[45] A. Labuschagne, N.F.J. van Rensburg, A.J. van der Merwe, Comparison of linear beam theories,
Mathematical and Computer Modelling, 49(2009) 20-30.
[46] G.R. Bhashyam, G. Prathap, Galerkin finite element method for non-linear beam vibrations,
Journal of Sound and Vibration, 72(1980) 191-203.
[47] E. Wittbrodt, I. Adamiec-Wójcik, S.a. Wojciech, SpringerLink ebooks - Engineering (2006),
Dynamics of flexible multibody systems : rigid finite element method, Berlin: Springer; 2006.
[48] A.A. Shabana, Flexible Multibody Dynamics: Review of Past and Recent Developments, Multibody
System Dynamics, 1(1997) 189-222.

117
Chapter 4: Linear dynamic and nonlinear dynamic model to predict PEDOT-based trilayer
actuation behavior
[49] I.A.-W. Edmund Wittbrodt, Stanislaw Wojciech, The Rigid Finite Element Method, Dynamics of
Flexible Multibody Systems: Rigid Finite Element Method, Springer Berlin Heidelberg, Berlin,
Heidelberg, 2006, pp. 35-82.
[50] G. Alici, B. Mui, C. Cook, Bending modeling and its experimental verification for conducting
polymer actuators dedicated to manipulation applications, Sensors and Actuators, A: Physical,
126(2006) 396-404.
[51] C.H. Nguyen, G. Alici, G. Wallace, An Advanced Mathematical Model and its Experimental
Verification for Trilayer Conjugated Polymer Actuators, IEEE/ASME Transactions on Mechatronics,
19(2014) 1279-88.
[52] A. Maziz, C. Plesse, C. Soyer, C. Chevrot, D. Teyssié, E. Cattan, F. Vidal, Demonstrating kHz
Frequency Actuation for Conducting Polymer Microactuators, Advanced Functional Materials,
24(2014) 4851-9.
[53] A.A.A. Moghadam, W. Hong, A. Kouzani, A. Kaynak, R. Zamani, R. Montazami, Nonlinear dynamic
modeling of ionic polymer conductive network composite actuators using rigid finite element method,
Sensors and Actuators A: Physical, 217(2014) 168-82.
[54] A.A.A. Moghadam, A. Kouzani, R. Zamani, K. Magniez, A. Kaynak, Nonlinear large deformation
dynamic analysis of electroactive polymer actuators, Smart Structures and Systems, 15(2015) 22.
[55] S. Wojciech, I. Adamiec-Wójcik, Nonlinear vibrations of spatial viscoelastic beams, Acta
Mechanica, 98(1993) 15-25.
[56] D.C. Karnopp, Chapter 10: Distributed-parameter systems, SYSTEM DYNAMICS: Modeling,
Simulation, and Control of Mechatronic Systems, JOHN WILEY & SONS2012, pp. 470-512.
[57] J. Ding, D. Zhou, G. Spinks, G. Wallace, S. Forsyth, M. Forsyth, D. MacFarlane, Use of Ionic Liquids
as Electrolytes in Electromechanical Actuator Systems Based on Inherently Conducting Polymers,
Chemistry of Materials, 15(2003) 2392-8.
[58] N.T. Nguyen, C. Plesse, F. Vidal, C. Soyer, S. Grondel, J.D.W. Madden, E. Cattan, Microfabricated
PEDOT trilayer actuators: synthesis, characterization, and modeling, SPIE Smart Structures and
Materials + Nondestructive Evaluation and Health Monitoring, SPIE2017, p. 13.
[59] J.D.W. Madden, N.A. Vandesteeg, P.A. Anquetil, P.G.A. Madden, A. Takshi, R.Z. Pytel, S.R.
Lafontaine, P.A. Wieringa, I.W. Hunter, Artificial muscle technology: Physical principles and naval
prospects, IEEE Journal of Oceanic Engineering, 29(2004) 706-28.

118
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

Chapter 5: Sensing ability and sensing model of the

PEDOT-based trilayer actuators

Contents

5.1 Introduction ............................................................................................................. 120

5.2 Theories on the mechanoelectrical effects ............................................................. 120
5.3 Sensing modeling ..................................................................................................... 126
5.3.1 Mechanoelectrical coupling ............................................................................. 126
5.3.2 Global model .................................................................................................... 128
5.4 Experimental setup .................................................................................................. 128
5.5 Results ...................................................................................................................... 130
5.5.1 Comparison between the model simulation and experimental results .......... 131
5.6 Discussion................................................................................................................. 135
5.6.1 Force in response to a step displacement ....................................................... 137
5.7 Conclusion ................................................................................................................ 138

119
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

5.1 Introduction
CP-based trilayer is an active material that exhibits an interesting bidirectional
electromechanical coupling where an input voltage results in the displacement of the film and
a voltage is produced when a displacement is applied to the film: mechanical deformation of
the transducer by external mechanical loads causes movement of ions, and the generation of
voltages. In this chapter, various theories on the mechanoelectrical effect will be reviewed
and a sensing model for CP’s will be described. The predictions of this model will then be
compared with the experimental sensing results, showing that the complete
electromechanical model represents an effective approach to describe both sensing and
actuation. In conclusion, we will discuss the potential of the trilayer PC based as a sensor.

5.2 Theories on the mechanoelectrical effects

In general, the mechanism of mechanical sensing is based on the insertion and expulsion of
ions into and out of the CP film. Researchers have developed two different configurations for
CP – based sensor devices (Fig. 1). One is a free-standing film of CPs operating in electrolyte
[1], and another one is the trilayer structure operating in open air where a polymer electrolyte
is sandwiched between two CP electrodes [2].

The sensing ability of CPs has attracted some researchers to use them in few applications. A
system containing two strips of trilayer structuring of CP - solid polymer electrolyte - CP is
employed to mimic the whisker’s functionality [3].

Fig. 1. a) Schematic of CP free-standing film sensing mechanism before and after subjected to a linear
force, and b) Schematic of a bending trilayer sensor [4]

The mechanoelectrical effect of CP was first observed by Takashima et al. [5] in 1997, for a
free standing film of polyaniline in 1 M HBF4 aqueous solution. In his sensing approach, redox
potential was maintained constant while a time varying tension was applied to the sample.

120
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

The authors demonstrated that the induced charge depends upon the magnitude of the
tensioning load and the degree of doping. The energy conversion efficiency from mechanical
to electrical was less than 0.01 %. One of the possible mechanisms proposed by the authors
behind this induced charge is the stretching of the film resulting in the stretching of the main
chain of the CP, and thus in the change of the density of states near the chemical potential
induced redox current.

In 2007, Takashima et al. [1] performed a more intensive study on the sensing ability of
polypyrrole composite films. A PPy film electrodeposited in dodecylbenzene sulfonic acid
(DBS) and in bis(trifluoromethane sulfonyl)imide (TFSI) produced a negative and a positive
current, respectively, when subjected to an increased tensile load. The switching polarity of
the induced current agrees well with the film actuation polarity, indicating that the
electromechanical actuation is the phenomenological counterpart of the mechanoelectrical
sensing. The authors proposed a model based on Gibb’s free energy theory and Fick’s law to
describe the flux of ions into and out of CP film. From this model, the authors claim that the
maximum current is proportional to the concentration of mobile ion in the film and to the
applied load. The re-establishing of Donnan’s equilibrium is used to explain the current
response.

At the same time, Wu et al. [2] introduced an experimental study on sensing ability of a
trilayer structure of PPy/A – PVDF – PPy/A in open air in both reduced and oxidized states,
where A is either DBS- or ClO4-. Fig. 2 summarizes the results of experiments, where a
sinusoidal voltage is obtained at open circuit in response to a sinusoidal displacement. The
effect of small mobile ions ClO4- and large immobile ions DBS- are clearly shown via the
difference in the polarity of the output voltage. In this paper, authors have suggested two
different mechanisms. In the first mechanism, so called “Deformation Induced Ion Flux” has
been proposed to explain the phenomena observed in this experiment (Fig. 3). In this theory,
a mechanical deformation affects the ion concentration within CP and, therefore, shifts the
balance of the ions between the CP and the electrolyte. An ion flux results - with only smaller
mobile ions displacing, while large dopants move at much lower speed. Consequently, the
mechanical deformation results in a separation of ionic charge and causes a voltage
difference.

The proposed model was confirmed via another set of experiments, where the authors used
ionic liquid BMI.PF6 (1-butyl-3-methyl-imidazolium hexafluorophosphate) and propylene
carbonate (PC) mixtures as supporting electrolyte for the PPy/PF 6 – Pt/PVDF/Pt - PPy/PF6
sensor. Depending on the ratio between BMI.PF6 and PC, an “out of phase” voltage at
amplitude of 0.07 mV (molar ratio BMI.PF6 : PC = 10 : 1) or an “in phase” voltage output at
amplitude of 0.04 mV (molar ratio BMI.PF6:PC = 1:10) is obtained.

121
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

Fig. 2. Effect of counter-ion and redox state on the voltage output of polypyrrole mechanical sensor
[2] (Reprinted with permission from John Wiley and Sons)

Based on the elastic energy change produced by inserting or removing an ion from the
polymer by a change in stress, the input stress (σ) is proportional to the sensor voltage output
(V) via an empirical strain-to-charge ratio (α): U = α x σ [1,2].

122
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

Fig. 3. Schematic diagram illustrating the net ion transport expected in both the “Deformation
Induced Ion Flux” model and based on elastic energy considerations [2] (Reprinted with permission
from John Wiley and Sons). In the first case compression of the polymer leads to an increase in ion
concentration within the polymer, and a net flux of ions out. In the second, it is the compressive
stress that squeezes ions out.

Shoa et al. [6] also observed a similar sensing behavior of a PPy thin film in an aqueous
solution of NaPF6, using a two electrodes experimental setup. The voltage generation in such
a thin film is proposed to result from the tensile stress. This stress expands the PPy network,
and thus reduces the force that required for ions to be inserted into the PPy network. The
elastic insertion forces are key in this explanation. The authors have proposed a model to
predict the magnitude and polarity of the output voltage, where the relationship between
elastic modulus of the polymer, the capacitance, and the mechanical capacitance resulting
from elastic energy stored during charging are all considered. The key input to the model is
the strain to charge ratio, , which is related to the amount of volume change per ion inserted.
This in turn determines how much elastic energy is needed to insert ions – which is in addition

123
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

to any electrostatic and chemical potential energy added as the polymer is charged and the
oxidation state is changed. The basic prediction of the model is that the voltage U change at
open circuit resulting from a change in the applied stress, , is U =  [7].

In 2012, Festin [8] demonstrated the sensing response of 220 µm thick PEDOT-based trilayer
actuators, where the PEDOT chains interpenetrate an IPN film made of NBR/PEO. The author
has experimentally observed a linear relationship between the applied stress (mechanical
strain) and the output voltage measured in open circuit state. A switch in the output voltage
polarity was also reported when the ionic liquid EMITFSI is used instead of LiClO4.PC.

Recently, Madden and his team [9, 10] have proposed a mechanism model for piezoionic
effect. In their experimental setup, a thin film of ionic conducting polymer, such as
polyacrylamide, poly(vinylidenefluoride - hexafluoropropylene) (PVDF-HFP), polar
(polyurethane), highly polar poly(ethylene oxide) (PEO), charged poly(ethylene glycol) (PEG)
or PEG derivatives, is placed between two rigid electrodes, in which the distance between
those two is 1 cm. A voltage and a current ranges of 10-2 V and of 10-7 A, respectively, is
observed when an externally applied local pressure at a portion of the layer (Fig. 4).

Fig. 4. Experimental scheme and induced voltage mechanism of a thin ionic conducting polymer film
under an applied stress [10]

This induced voltage and current could be a result of the redistribution of mobile ions due to
a change in the local ions concentration, based on the Donnan potential. A possible equation
for predicting the electrical potential (U) due to the difference in local concentration (∑[𝐶 + ])
is:

𝑅𝐺 [𝐶 + ]𝑥
∆𝑈 = ln ( + )
𝑧𝐹 [𝐶 ]𝑥+∆𝑥

where G is the temperature, R is the ideal gas constant, z is the valence charge of the ion, F is
the Faraday’s constant, [𝐶 + ]𝑥 and [𝐶 + ]𝑥+∆𝑥 are the concentration of the ith mobile ions in the
polymer at location x and (x + Δx). In this study, the speed of rising voltage and current, the

124
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

polarity and amplitude of the generated voltage are also discussed showing their dependent
on type and size of mobile ions, and mechanical properties of the host polymer.

These results obtaining by Madden suggest that any single or interpenetrating polymer
network, which is able conduct the ions – and does not necessary to conduct electrons, could
produce a voltage and a current under a suitable mechanical excitation. In our case, since the
fabricated NBR/PEO is a sIPN, which conducts ions as demonstrated previously via ionic
conducting measurement, this NBR/PEO layer in the trilayer configuration plays an important
role in the sensing ability of the trilayer.

In Woehling’s study [11], the contribution of PEDOT electrodes in sensing ability of the trilayer
was demonstrated. Fig. 5 shows that an input deflection induces an output voltage in both
trilayer structures: Gold:NBR/PEO:Gold and PEDOT:NBR/PEO:PEDOT, swollen in EMITFSI or
EMITFSI-PC at various concentration from 0.01 M to 3 M. In case of the pure EMITFSI and low
concentration EMITFSI-PC, the magnitude of the output voltage from gold:NBR/PEO:gold is
25% smaller than that of second case. This suggests the output voltage of the trilayer
structure composed of PEDOT:NBR/PEO:PEDOT includes the contribution of two factors: the
voltage induced by NBR/PEO alone (similar to the voltage generated in an IPMC), and the
voltage produced by conducting polymer electrodes [6].

Fig. 5. Schematic experiments and comparison of output voltage in two different trilayer
configurations: Gold-NBR/PEO-Gold, and PEDOT-NBR/PEO-PEDOT with the dimension of L = 15mm, b
= 5mm, ht = 0.17 mm at various EMITFSI-PC concentrations, under a 2% applied strain [11] (Retrieved
from https://tel.archives-ouvertes.fr/tel-01413677. By permission from author.). In this case a
positive voltage suggests anion motion.

The results from Fig. 5 also show that the output voltage’s polarity depends on the EMITFSI –
PC concentration. The magnitude of output voltage is highest at the concentration extremes
of pure EMITFSI and at the lowest (0.01 M) EMITFSI concentration. As the concentration of
EMITFSI in PC is decreased, the magnitude is reduced. When the EMITFSI concentration is 2.5
M, the output voltage switched its polarity and increased as the concentration continues to
decline.

125
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

Following this literature review, it can be concluded that there are three potential models
that may explain sensing in conducting polymers: (1) an IPMC-like flux due to a pressure
gradient within the ionically conductive separator; (2) a concentration (Donnan potential)
mechanism, in which change in polymer volume induces a change in ion concentration and a
resulting voltage generation, and (3) an elastic energy induced charge expulsion mechanism.
All three predict that the more mobile ions will be involved in the transport, and hence result
in voltage changes of the same sign. Whatever the mechanism, it does appear that the sensing
and actuation mechanisms are related, as expected by microscopic reversibility. For this
reason, I will choose to represent the sensing effect in section 5.2 via the empirical strain to
charge ratio. Within this chapter, a simple sensing relation – which is the inverse of the
actuation mechanism – will be implemented into the linear actuation model developed in
chapter 4. Furthermore, a single model is developed for operating in both sensing and
actuation modes. Then, a set of experiments will be performed and compared to a simulation
in section 5.3 and 5.4 to confirm the ability of the model to predict the output voltage
magnitude and polarity. Finally, several aspects on model limitations will be discussed in
section 5.5.

5.3 Sensing modeling

In this section, a simple sensing model is developed, comprising three sub models: an
electrochemical submodel containing a capacitor in series with a resistor (refer to section
4.2.2.1), a mechanoelectrical coupling (described in 5.2.1), and a mechanical model based on
Euler-Bernoulli dynamic beam theory (refer to chapter 4.2.2.3). In order to follow the same
methodology than in chapter 4, it is again represented using a Bond Graph formalism.

5.3.1 Mechanoelectrical coupling

As mentioned in the introduction, Shoa’s model [6] and Wu’s model [2] proposed a simple
relation between the applied stress (σ) and the output voltage (U) via an empirical strain-to-
charge ratio (α) U = σ x α. This ratio shows a bidirectional electrochemical effect. When the CP
film is considered as an actuator, the strain to charge represents the relation between the
volume change and the inserted charge density (units of m3/C). Whereas for the sensor case,
this ratio describes the amount of electrical potential produced by an applied stress (V/Pa –
which is equivalent since a Volt is a Joule per Coulomb).

We demonstrate here the flexibility of two-port ₵-element which is used to couple both
electromechanical (actuation) and mechanoelectrical (sensing) relationships by a single
matrix of 2 x 2 elements, shown in equation (1):
1 1.61×(ℎ𝑠 +ℎ𝑝 )𝛼𝐸𝑃
𝑈𝐶 2𝐶 𝐿2 +𝑤 2 𝑄𝐶
[ ] = [𝛼𝐸 (ℎ +ℎ ) 3 2ℎ 3
][ ]. (1)
𝐹 𝑃 𝑆 𝑃 4𝑏𝐸𝑃 ℎ𝑆
( 2 ) [(1 + 𝑃 𝐸𝑆 1
) − 1 + 𝐸 ] 𝐿2 +𝑤2 𝑤2
2L 3𝐿 ℎ𝑆 𝑃

126
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

Here Q is the charge per volume applied to produce actuation and w is the displacement
imposed externally on the trilayer (and that is being sensed). UC is the sense voltage (change
in open circuit potential) plus the voltage determined by the state of charge. F is the force
needed to deflect the beam by w plus the force generated by the application of the charge.
In the equation (1), the actuation side has been derived from chapter 4 (section 4.3.3.2),
𝛼𝐸𝑃 (ℎ𝑆 +ℎ𝑃 ) 4𝑏𝐸𝑃 ℎ𝑆 3 2ℎ𝑃 3 𝐸 1
where 𝐹 = ×𝑄+ ( 2 ) [(1 + ) − 1 + 𝐸𝑆 ] 𝐿2 +𝑤2 × 𝑤. Here Ep and ES are
2L 3𝐿 ℎ𝑆 𝑃

the elastic moduli of the conducting polymer and the separator layers, respectively. hs and hp
are the thicknesses of the separator layer and the PEDOT electrode, respectively, and L is the
length of the beam. On the sensing side from top row of the matrix in Equation 1, we start
with the relationship:

𝑈 𝐶 |𝑄=0 = 𝛼 × 𝜎𝑃 = 𝛼 × 𝜀𝑒𝑥𝑡𝑒𝑟𝑛𝑎𝑙 × 𝐸𝑃 (2)

where 𝜀𝑒𝑥𝑡𝑒𝑟𝑛𝑎𝑙 is the linear strain applied to the PEDOT electrodes as a result of
displacement, wext.

From the chapter 3 (section 3.4.1), based on beam bending stiffness of a trilayer, the relation
between the linear strain 𝜀𝑒𝑥𝑡𝑒𝑟𝑛𝑎𝑙 and the strain difference (𝜀𝑎𝑐𝑡𝑢𝑎𝑙 ) of one PEDOT electrode
is:
ℎ𝑃 (ℎ4 𝐸 2 +8ℎ3 𝐸+12ℎ2 𝐸+8ℎ𝐸+24ℎ+12𝐸 2 +16)
𝜀𝑎𝑐𝑡𝑖𝑣𝑒 = 𝜀𝑎𝑐𝑡𝑢𝑎𝑙 × = 1.61 × 𝜀𝑎𝑐𝑡𝑢𝑎𝑙 (3)
6ℎ𝑡 (ℎ+1)(ℎ𝐸+2)

ℎ 10𝜇𝑚 𝐸 329𝑘𝑃𝑎
where ℎ = ℎ𝑆 = 3.5𝜇𝑚 = 2.9 and 𝐸 = 𝐸𝑆 = 12.9𝑀𝑃𝑎 = 0.0255.
𝑃 𝑃

The relation between the strain difference (𝜀𝑎𝑐𝑡𝑢𝑎𝑙 ) and the beam deflection (w) is:
2(ℎ𝑠 +ℎ𝑝 )𝑤
𝜀𝑎𝑐𝑡𝑢𝑎𝑙 = 𝐿2 +𝑤 2
(4)

Replace (4) and (3) into (2), we have:

2(ℎ𝑠 +ℎ𝑝 )𝑤
𝑈 𝐶 |𝑄=0 = 1.61 × 𝛼 × 𝐸𝑃 × (5)
𝐿2 +𝑤 2

We also have:
𝑄
𝑈 𝐶 |𝑤=0 = 𝐶 , (6)
𝑃

From (5) and (6), we obtain the sensing relation as following:

𝑄 (ℎ𝑠 +ℎ𝑝 )𝑤
𝑈 𝐶 = 𝑈 𝐶 |𝑄=0 + 𝑈 𝐶 |𝑤=0 = 2𝐶 + 1.61 × 𝛼 × 𝐸𝑃 × (7)
𝑃 𝐿2 +𝑤 2

It is worth noting that the output voltage (UP) is the peak-to-peak voltage obtained between
two PEDOT electrodes according to the strain difference also between two PEDOT layers.

127
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

From equation (2) and (3), the coupling matrix in equation (1) is derived.

5.3.2 Global model

From the electrochemical model containing a single capacitor in series with a single resistor,
the two-port ₵-element, and the mechanical model based on Euler-Bernoulli dynamic beam
theory (refer to chapter 4), a unique Word BG model could be established and as shown in
Fig. 6.

Fig. 6. Global Word BG model, where (1) and (3) represent the input voltage and the tip
displacement, respectively, as the trilayer structure in actuation mode; the (4) and (5) are the input
displacement and the voltage output, respectively, when the trilayer is operated as a sensor, (2) is
the global BG model

It is worth noting that in comparison to the linear actuation model, in this sensing model an
applied displacement (4) or a force is introduced as an input to the model instead of an input
voltage (1), since we are investigating the mechanoelectrical ability of the trilayer structure.
The bidirectional electromechanical coupling is described in the equation (1) and the detailed
description of most of the global BG model (2) was established in chapter 4 (section 4.2.2).
Only the parameters of the coupling ₵-element have been modified as shown in equation 1.

5.4 Experimental setup

The trilayers used in this measurement were fabricated following the procedure described in
chapter 2 (section 2.5). The resulting trilayers were swollen in ionic liquid during one week
before the experiment. To investigate the mechanoelectrical effect of the trilayer structure,
an experimental setup is depicted as in Fig. 7a. One side of the trilayer sensor is placed on a
thin plastic film – which has much higher stiffness than the trilayer. The film and trilayer stick
together thanks to the tacky surface of the ionic liquid containing trilayer. The film forms a
fourth layer (it is now a quadruple layer), adding stiffness, and increasing resonance frequency
of the sensor to about 110 Hz calculated value (from as low as 2.5 Hz for the longer trilayer
alone). One end of the trilayer is clamped between two metal electrodes which are connected
to a National Instruments data acquisition system via a homemade amplifier (gain can tune
between 1 and 1000, input impedance is 10 MΩ) to collect the voltage signal. At the other

128
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

end, it is held by the moving part of the Bose ElectroForce 3100 Series. The input displacement
is precisely applied to the tip of the trilayer as depicted in Fig. 7a. The computer is
programmed to apply a specified magnitude and frequency of a sinusoidal or a step
displacement, and record the actual displacement from the feedback sensor on the Bose
system. It simultaneously records the sensor voltage signal. An example of the resulting
voltage responses to an input displacement of magnitude of 2 mm and frequency of 3 Hz is
shown in Fig. 7b. Notice that the output voltage is more stable after few cycles of applying
displacement.

129
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

Fig. 7. a) Experimental setup to measure the voltage response to a sinusoidal displacement, and b)
An output voltage response to an input displacement of the magnitude of 2 mm and frequency of 3
Hz on a trilayer of 7 mm x 1 mm x 0.017 mm.

In my study, the trilayer structure containing ionic liquid can be seen as two separated phases:
the sIPN as a porous phase, and ionic liquid as a fluid phase. Literatures have shown a
poroelastic behavior of a gel network under a compression [12, 13]. Poroelasticity describes
the relation between the porous scaffold and the fluid flow containing in this structure. When
a mechanical load (a compression) is applied to this composite structure (solid and fluid), the
porous part experiences a pressure changing leading to the fluid motion. As the applied
displacement is kept constant, the responded force (as a function of time) is decreased before
reached plateau. This phenomenon has been demonstrated via experiments and modeling by
the authors.

Fig. 8 illustrates the experiment to measure the force response of a trilayer structure to a step
displacement to evaluate the degree of ionic liquid flows through the trilayer structure – or a
stress relaxation. A three–axis nanomanipulation unit is used to apply a step displacement,
and a force/displacement sensor (FT-RS1002 Microrobotic System, three degrees of freedom,
maximum speed: 5 mm/s coupled with an FT-S100 microforce sensing probe with a
measurement range of ±100 µN, resolution: 5 nN) records the force response as a function of
time, and a micro camera (Dino-Lite AM7000/AD7000, resolution: 5 MPX) is used to record
the bending displacement of this trilayer.

Fig. 8. Experimental setup to measure the decaying of blocking force of the actuator in response to a
step displacement

5.5 Results
In order to confirm the agreement between the simulation results and the experiments on
the output voltage, it is recorded as a function of the magnitude and the frequency of a
sinusoidal displacement. The voltage response is also investigated in case of an input step
displacement. Finally, the blocking force of a specimen is measured under a step
displacement. All the measurements have been conducted on the trilayer structures with the

130
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

dimension of length (L) х width (b) х total thickness (h) = 7 mm х 1 mm х 0.017 mm in the
swollen state, and the thickness of each layer: hP = 3.5 µm and hS = 10 µm. It is worth noting
that the clamped area is a rectangular of 3 mm x 2 mm and it is not taking into account in my
calculation.

5.5.1 Comparison between the model simulation and experimental results

As discussed in the introduction, the mechanoelectrical mechanism is contributed from two

factors: the movement of ions in the PEDOT electrodes and in NBR/PEO layer, could be
distinguishes from the actuation mechanism. In order to evaluate the sensing model, the
sense strain-to-charge ratio needs to be remeasured as we expect it is different from the value
used in the actuation model, which is equal to 3.6 ± 0.5 x 10-10 m3/C (refer to chapter 3, section
3.4.1).

A trilayer beam is subjected to a sinusoidal displacement. In the first case, the magnitude of
the displacement is held at 2 mm while the frequency is varied between 0.5 Hz and 100 Hz.
The polarity of the output voltage is determined as described in Fig. 9a, where the top
(compressed) PEDOT electrode produces a negative potential relative to the bottom
electrode. This suggests a motion out of this electrode by the cation. The magnitude of the
measured voltage output and the estimated empirical strain-to-charge ratio are a steady 0.75
mV and 4.9 x 10-10 m3/C, when the frequency increases from 0.5 Hz to 20 Hz, shown in Fig.
9b. These values gradually reduce to 0.14 mV and 7.4 x 10-11 m3/C when the frequency reaches
100 Hz. The contribution of different factors could explain this reduction. At first, when the
trilayer sensor is operating at the frequency higher than 30 Hz, the feedback displacement
data showed a fluctuation in magnitude, which indicates this frequency approaches the
resonant frequency of the Bose system and perhaps overestimates the displacement
magnitude (the real magnitude is smaller than the set magnitude). Therefore, the output
voltage is expected to reduce. The second possibility comes from the nature of the sensing
mechanism, which is based on the flow of ions into and from the CP electrodes, and the ion
mobility can be a limiting factor at this high frequency. Shoa et al. [6] also observed the drop
in voltage at high frequency.

If voltage and displacement obtained in this measurement are normalized in term of thickness
of the sensor, a coefficient can be calculated as below:
𝑄 𝑄
∆𝑉 = 𝛼 × 𝜎  𝑉𝑜𝑙𝑢𝑚𝑒 = 𝛼 × 𝐶𝑉 × 𝜎 𝐴𝑟𝑒𝑎 = 𝛼 × 𝑡 × 𝐶𝑉 × 𝜎 (8)

In the PVDF piezoelectric [14], the coefficient d33 is derived as following:

𝑄
𝑄 = 𝑑33 × 𝐹  𝐴𝑟𝑒𝑎 = 𝑑33 × 𝜎 (9)

From (8) and (9), we have:

131
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

𝑑33 2 × 10−11 𝐶/𝑁 2 × 10−2

= = = 1.5 × 10−4
𝛼 × 𝑡 × 𝐶𝑉 3.6 × 10−10 𝑚3 /𝐶 × 7 × 10−6 𝑚3 × 5 × 107 𝐹/𝑚3 126

This value suggests that the sensitivity of the trilayer at this thickness is better than that of
piezoelectric.

A range of the strain-to-charge ratio obtaining from Fig. 9b is implemented into the model.
The agreement between the simulation and experiments is confirmed via another set of
experiments on a different sample with the same dimensions, shown in Fig. 9c.

132
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

Fig. 9. a) Schematic of a trilayer transducer under a displacement excitation and the produced
voltage polarity b) Voltage magnitude and the strain-to-charge ratio as a function of frequency when
the magnitude of the sinusoidal displacement is constant at 2mm, c) Comparison between the
simulation and the experiment of the magnitude of output voltage

In a second experiment, the frequency of the sinusoidal displacement is held constant at 3 Hz

while the magnitude is changed between 0.04 mm and 3 mm. The linear relation between
the output voltage and the magnitude of displacement, which was reported by Wu [2] and
Woehling [11], is shown again in Fig. 10a, despite a small fluctuation. This voltage reaches the
maximum at 0.76 mV when the magnitude approaches 2 mm, and then it decreases to 0.66
mV as the magnitude continues to increase. The strain-to-charge ratio has been derived in

133
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

this case and its average value is 4.76 x 10-10 m3/C, which is similar to the previous experiment.
A good fit between the simulation and experiment is demonstrated in Fig. 10b, despite the
last point where the magnitude of displacement is biggest. There are few possibilities
contributing to the decreasing voltage output at the largest displacement. Firstly, a large input
displacement can induce a non-circular curvature of beam, in which the position closest to
moving part (Fig. 7a) is more straight and the beam can show a wavy bending. In addition, the
contacting area between the electrical contact and the electrodes can be changed under a
large displacement because of a high stiffness of the plastic lead.

Fig. 10. a) Voltage magnitude and the strain-to-charge ratio as a function of magnitude when the
frequency of displacement is constant at 3 Hz, b) Comparison between simulation and experiment of
the magnitude of output voltage

Fig. 11 described another experiment where a step displacement was applied to the tip of the
trilayer beam. The voltage output at first increases with the same speed as the step
displacement and reaches the maximum value of 0.66 mV as the displacement peaks 3 mm

134
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

(ℎ𝑠 +ℎ𝑝 )𝑤𝑒𝑥𝑡 (10+3.5)10−6 𝑚×3×10−3 𝑚

(induce an average strain difference of 2 = = 0.069%
𝐿2 +𝑤𝑒𝑥𝑡 (72 +32 )10−6 𝑚

between two PEDOT electrodes). Immediately after the increase, the voltage quickly decays
as the displacement remains constant. The simulation result follows quite well the
experiment, though the decrease of the voltage in this case is the result of charging the
capacitance or the PEDOT electrodes in other words. It is worth noting that the parameters
used in the sensing model remain the same to those of the linear actuation model (chapter
4, section 4.2). These parameters include the strain-to-charge ratio obtained in previous
characterization (8.45 x 10-10 m3/C), the capacitance value (chapter 3, section 3.2.3), the
leakage resistance (chapter 3, section 3.2.4), the NBR/PEO ionic resistance (chapter 3, section
3.2.1). A closer look to the model parameters shows that the speed of the decay depends on
the level of the leakage resistance and the capacitance, where a smaller leakage and
capacitance values induce a faster decay. In contrast, a bigger leakage resistance and
capacitance result in a long decay of the voltage – or simply this is the decay in an RC circuit
when the supplied voltage is cut off. Even though the model is able to predict the peak voltage
and the shape of the decay, it does not reflect the real mechanism of the output voltage. A
possible qualitative explanation will be proposed in the appendix A.5, where the flux of
cations and anions through the trilayer is considered as a main contribution to the shape peak
and the later decay of the output voltage.

Fig. 11. Comparison between the simulation and measured voltage responding to an applied step
displacement

It takes 1.7 s for the sensor to lose 90% of its maximum magnitude voltage, indicating the
sensor cannot be used to measure a static or a low frequency (< 0.3 Hz) displacements. This
is similar to piezoelectric sensors, which are inherently unable to measure static excitations
[15]. In the case of the present sensor, the drop in sensor response is the result of a short
between the two electrodes. The PEDOT is in contact across the separator. If this can be

135
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

removed, it should be possible to record displacements at much lower frequencies. There still
may be some parasitic reactions that slowly lead to self-discharge, ultimately determining the
low frequency limit of the device.

Recent work (not published yet) conducted by K. Rothlaid (U. Cergy-Pontoise 2018) has shown
that the use of PEDOT doped with polystyrene sulfonate (PSS) results in a stable output
voltage for at least for few minutes, with no decay observed. This alternative material can be
a suitable selection for the sensing application in static and low frequency case.

5.6 Discussion
In the previous section, the strain-to-charge ratio (α) was estimated as the function of
frequency and magnitude of the input sinusoidal displacement. The obtained value at the
frequencies below 30 Hz and at strains smaller 0.23% is approximately 8.4x10-10 m3/C. This
value is bigger than the strain-to-charge value obtaining for the electromechanical effect,
which equals to 3.6 ± 0.5 x 10-10 m3/C.

There may be a contribution of different factors during the measurement resulting in a

different strain-to-charge ratio. At first, in the experiments above, the input displacement is
applied at the position of the tip of the trilayer. It is a point load. This leads to the fact that
the stress distribution is not homogeneous along the length of the trilayer, in which the
largest stress is induced at the position close to the fixed end and smallest stress is observed
at the position close to the flexible end. A simple stress distribution along the length of the
beam is derived as following:
𝑤×3𝐸𝑡 𝐼𝑡 ×(𝐿−𝑥)
𝜎(𝑥) = ,
𝐿3 𝑆

where w is the input displacement, Et and It is the Young’s modulus and inertial moment of
the beam, respectively, x indicates the position along the length of the beam (L) away from
the clamping point, and 𝑆 = 𝑏ℎ3 /12 is section modulus of the beam.

As a result, the strain to charge ratio should be considered as a function of position along the
length of the beam. The use of constant value of the strain to charge ratio can be explained
as an average of this value for a specific PEDOT volume.

The second contribution to the measurement error may come from the use of the plastic lead
during the measurement. Even though this plastic foil is thinner than the trilayer structure,
the presence of this foil will change the neutral axis of the trilayer during the bending, in which
two PEDOT electrodes will not be induced a symmetric strain in opposite sign. Therefore, the
derived strain from the input displacement is needed to be adapted. The plastic lead (Young’s
modulus ≈ 3GPa) is much stiffer than the trilayer sensor (Young’s modulus ≈ 13MPa),
therefore in this experiment, we assumed that the neutral axis of the four layers (the trilayer
and the plastic lead) close to that of the plastic lead. The strain difference between two PEDOT

136
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

(ℎ𝑠 +ℎ𝑝 )𝑤
electrodes under an input bending can be derived as 𝜀𝑎𝑐𝑡𝑢𝑎𝑙 = , this relation is used
𝐿2 +𝑤 2
in equation (4).

Apart from these two measurement errors, this different strain to charge ratio – if it is real –
suggests a possibility of existing a different operating mechanism in the mechanoelectrical
effect of the trilayer structure.

In Appendix A.5.1, a model is proposed to explain for the sensing mechanism of the trilayer
actuator, however without the experimental confirmation, this model becomes an opening
conclusion.

5.6.1 Force in response to a step displacement

Here stress relaxation is investigated. Whatever the sensing working mechanism of the
trilayer structure may be, the movement of ions inside the trilayer under a mechanical
excitation (a displacement or a force) is essential for all the possibilities. One question is
whether or not there is a substantial flow of ionic solution. If there is, this will lead to a
reduction in force over time when a displacement is applied. Such a drop can also be due to
stress relaxation in the film of course.

Fig. 12a shows a two-steps input displacement to a trilayer sample. The specimen is pushed
forward and pulled backward by a silicon tip of a displacement/force sensor. The resulting
force is recorded as a function of time, shown in Fig. 12b. The force increases to approximate
2 µN as the step displacement of - 0.5 mm is applied, and then it slowly reduces and reaches
plateau after 30 s. The same trend is observed when a step of + 0.5 mm is applied, however
a bigger voltage magnitude was recorded in this case. The force drop suggests the existence
of ion flow upon an application of displacement or just simply the viscoelastic response of the
material. The responding force reaches the plateau indicating this flow is getting slower and
will stop at some point or again just the viscoelastic behavior after a certain amount of time
that is bigger than the relaxation time constant of the material.

137
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

Fig. 12. a) Experiment of applying step displacement, and b) The decay of force response of the beam
as a function of time.

5.7 Conclusion
In this chapter, a simple relation between the induced stress and the output voltage has been
implemented in the linear actuation model with the purpose to build a unique linear model
operating in both actuation and sensing modes. The strain-to-charge ratio was
recharacterized and the concordance between the simulation and experiment of the output
voltage has confirmed the ability to predict the sensing function of the trilayer, despite a small
error. The model shows its advance in predicting quite well the voltage decay under a step
displacement.
The future perpective of this chapter is to optimize the experimental design. Applying
homogeneous stress along the length of the trilayer would reduce any variation in sense
voltage along the length. The second is to reduce the asymetric strain between two PEDOT
electrodes, which could be done by using a symmetric bending set-up. These improvements
will ensure the reliablity of the experimental results and provide a clue on the existing of novel
sensing mechanism.

138
Chapter 5: Sensing ability and sensing model of the PEDOT-based trilayer actuators

[1] W. Takashima, K. Hayasi, K. Kaneto, Force detection with Donnan equilibrium in polypyrrole film,
Electrochemistry Communications, 9(2007) 2056-61.
[2] Y. Wu, G. Alici, J.D.W. Madden, G.M. Spinks, G.G. Wallace, Soft Mechanical Sensors Through
Reverse Actuation in Polypyrrole, Advanced Functional Materials, 17(2007) 3216-22.
[3] N. Festin, C. Plesse, C. Chevrot, D. Teyssié, L. Josselin, P. Pirim, F. Vidal, Actuation and Sensing
properties of Electroactive Polymer Whiskers, Procedia Computer Science, 7(2011) S4-S7.
[4] T. Wang, M. Farajollahi, Y.S. Choi, I.T. Lin, J.E. Marshall, N.M. Thompson, S. Kar-Narayan, J.D.W.
Madden, S.K. Smoukov, Electroactive polymers for sensing, Interface Focus, 6(2016).
[5] W. Takashima, T. Uesugi, M. Fukui, M. Kaneko, K. Kaneto, Mechanochemoelectrical effect of
polyaniline film, Synthetic Metals, 85(1997) 1395-6.
[6] T. Shoa, J.D.W. Madden, T. Mirfakhrai, G. Alici, G.M. Spinks, G.G. Wallace, Electromechanical
coupling in polypyrrole sensors and actuators, Sensors and Actuators, A: Physical, 161(2010) 127-33.
[7] J.D.W. Madden, Conducting polymer actuators: Massachusetts Institute of Technology; 2000.
[8] N. Festin, Conducting interpenetrating polymer network actuator sensor for biomimetic perception
system: Université de Cergy Pontoise; 2012.
[9] M.S.u. Sarwar, Y. Dobashi, E. Glitz, M. Farajollahi, S. Mirabbasi, S. Naficy, G.M. Spinks, J.D.W.
Madden, Transparent and conformal 'piezoionic' touch sensor, SPIE Smart Structures and Materials +
Nondestructive Evaluation and Health Monitoring, SPIE2015, p. 9.
[10] M.S. John Madden, Yuta DOBASHI, Edmond Cretu, Shahriar Mirabbasi, Ettore Glitz,Meisam
Farajollahi Method and sensor for pressure sensing based on electrical signal generated by
redistribution of mobile ions in piezoionic layer, in: U.S.P. Application (Ed.) US Application Publication,
University of British Columbia United State, 2015, p. 40.
[11] V. Woehling, New Developments in electroactive materials based on electronic conductive
polymers : Towards integration into biomedical systems: Université de Cergy Pontoise; 2016.
[12] S. Cai, Y. Hu, X. Zhao, Z. Suo, Poroelasticity of a covalently crosslinked alginate hydrogel under
compression, Journal of Applied Physics, 108(2010) 113514.
[13] W. Hong, X. Zhao, J. Zhou, Z. Suo, A theory of coupled diffusion and large deformation in polymeric
gels, Journal of the Mechanics and Physics of Solids, 56(2008) 1779-93.
[14] F. Bernard, L. Gimeno, B. Viala, B. Gusarov, O. Cugat, Direct Piezoelectric Coefficient
Measurements of PVDF and PLLA under Controlled Strain and Stress, Proceedings, 1(2017).
[15] G. Gautschi, Piezoelectric Sensorics: Force Strain Pressure Acceleration and Acoustic Emission
Sensors Materials and Amplifiers: Springer-Verlag Berlin Heidelberg; 2002.

139
Chapter 6: Conclusion and outlook

Chapter 6: Conclusion and outlook

Developments in conducting polymer transducers operating in air make them more effective
in terms of generated force, displacement and frequency of operation, among others.
Recently, they have been grown at acceptable scales for potential integration into
microsystems. The ultimate goal is to have a material whose properties are well known and
can be modeled in order to be predictive in the manufacture of a soft microsystem that
includes flexible transducers. That is why this thesis was focused on their characterization,
and modeling. To the best of my knowledge, this is the first effort to fully characterize and
model such a very thin trilayer transducer, where the total thickness of the transducer is only
17 µm - after being swollen in ionic liquid, and it is the most complete combined effort at
characterization and modeling yet undertaken for conducting polymer actuators.

The first step in this work was to improve on a recently developed clean room compatible
microfabrication process for conducting polymer trilayers. Although the yield remains low
with this process, the relevant physical characteristics of a successful device were fully
characterized, including factors affecting ionic and electronic transport, coupling between
charge and strain, material stiffnesses, and damping. In actuation a strain difference of
approximately 1%, a blocking force of 12 µN, and a sensing voltage of 0.7 mV were obtained
from a transducer of length x width = 6 mm x 1 mm. This deflection and sensing voltage match
the expected response, as described by a non-linear model.

The conducting polymer PEDOT was used as the active material. Since the 3,4-
ethylenedioxythiophene (EDOT) polymerization is a complicated and sensitive process and
PEDOT is a complex material showing nonlinear electrochemomechanical properties, the
fabrication, characterization, and modeling processes are confronted with various issues.
Overcoming these technical challenges led to some main scientific and engineering
contributions as listed here:

- The characterization and improvement of the trilayer fabrication process. In this

trilayer configuration, a semi-NBR/PEO is sandwiched between two PEDOT/PEO
electrodes, produced by stacking layer by layer. This fabrication technique allows us
to tune the thickness of each layer as desired. In a departure from previous work,
changes have been made, such as the optimization of the PEDOT electrodes, resulting
from the increase in vapor phase polymerization time from 30 min to 50 min,
switching to a vacuum hotplate (Sawatec HP-200) allowing more precise temperature
and vacuum control than was achieved using a vacuum bell, and performing all the
fabrication process in a clean room environment. The percentage of the PEO in the
PEDOT layer was also changed to 10%, from 20 %, to ensure a balance between high
electronic conductivity (around 200 S/cm) and high volumetric charge density (leading

140
Chapter 6: Conclusion and outlook

to an increase from 2.3 x 107 C/m3 to 1.0 x 108 C/m3) of the PEDOT layer. In addition, I
have investigated the asymmetry in surface roughness between the top and the
bottom PEDOT layers, which might have consequences on the electrical properties of
the PEDOT electrodes, and finally on the mechanical properties (displacement, force
…) of the actuator.
- Full characterization of the critical electrochemomechanical properties of the trilayer
transducers has been done. The electrochemical properties including the ionic
conductivity of the PEDOT electrodes and of the SPE layer, the PEDOT electronic
conductivity as a function of oxidation state, and the dependence of PEDOT’s
volumetric capacitance on the potential window were all studied. In addition, the
mechanical properties such as Young’s modulus of the PEDOT as a function of the
oxidation state, Young’s modulus of the SPE layer, and the damping ratio. Finally, the
coupling between the electrochemical and the mechanical were represented via an
empirical strain-to-charge ratio, the linear strain was determined and the blocking
force measured as a function of applied potential. These obtained parameters provide
us an insight into the physical structure of the system and the working mechanism of
the actuator. Most importantly, the key factors affecting actuation behavior are all
determined, providing us a best guideline to improve the actuator’s performance.
- Development of a nonlinear dynamic model – and along with it a much simpler
dynamic linear version of the model – capable of simulating the actuation of ultrathin
trilayer actuators. A comparison between these two models has shown that the linear
model is simple to build and it is easier to obtain all of its required parameters.
However, it is most suitable to predict small displacements, providing a less precise
prediction than the non-linear model for larger applied voltages and displacements.
The nonlinear model showed advances over the existing work on trilayer actuators in
that it can precisely predict a nonlinear dynamic actuation with an ability to implement
nonlinear electrochemical and mechanical properties of material: PEDOT electrical
and ionic conductivities; volumetric capacitance values; and Young’s modulus
variation. In addition, it accounted for the mass and damping associated with the
beam. All physical characteristics used in the model were measured. The concordance
between the simulation and experimental results in both time and frequency domains
confirms the ability to predict the bending of the trilayer actuator. In addition, the
approach is represented in a Bond Graph language, providing simple way to access the
energy information, evaluate the critical parameters affecting the performance of the
actuator, and allow an insight into the phenomena occurring in the ionic actuator.

- Development of a unique linear model operating in both actuation and sensing modes.
A simple relation between the induced stress and the output voltage has been
implemented in the linear actuation model. The strain-to-charge ratio was
recharacterized in sensing mode and the concordance between the simulation and

141
Chapter 6: Conclusion and outlook

experiment of the output voltage has confirmed the ability to predict the sensing
function of the trilayer, despite a small error. Another advance of the model is its
predictionof the sensing voltage decay rate under a step displacement.

Recommendations for future work

In addition to the contributions listed above, this thesis work also draws out some new
challenges in all three domains including fabrication, characterization, and modeling.

- At first, a further optimization on the transducers’ fabrication process can be

performed to improve the yield and reproducibility. The VPP process in this thesis has
used the conventional oxidant solution consisting of 55% Fe(TOs)3 in butanol.
However, literature1 has shown that this high percentage produces a highly acidic and
reactive oxidant solution, resulting in uncontrollable polymerization and structural
defects in the deposited film. Investigation on optimizing the percentage of oxidant
solution, or adding base inhibitors such as pyridine to control the reaction rate are
suggested to further improve the electrochemical properties of the PEDOT electrode.
In addition, an asymmetric geometry between the top and the bottom PEDOT
electrodes was observed and resulted in an asymmetry in the electronic conductivity
of PEDOT layers as well an initially curved state of the trilayer beam. This can be
possibly overcome by fabricating PEDOT electrodes with different thicknesses to
compensate for the pre-stress in the beam.
- Secondly, in the characterization, the properties of the PEDOT and the SPE layers have
been measured, however, each property was investigated on one specific thickness of
the sample. There is a need to confirm that the obtained values are consistent at other
thicknesses. If they are unchanged with thickness, then one set of values can be used
for all device geometries. It is also interesting to consider the extension of this model
to other trilayer structures, whether conducting polymer-based or IPMC, as well as to
other shapes. Even in IPMCs it is possible that the same characteristics can be used to
understand their sensing and actuation responses.
- In addition, the nonlinear model is theoretically suitable to predict high frequency
actuation (f > 100 Hz). In addition, the model was built based on Bond Graph language,
providing at first a possibility to integrate it into a real actuator control system, and
second to develop a more complex model to control not only a single beam but also a
complex geometry. These possibilities can show the power of the model, which is not
fully demonstrated here yet.
- Finally, in the bidirectional electromechanical model, there is need to optimize the
experimental design. In the testing method used, in which the beam is bent using a
point load applied near the tip, the moment is not uniform along the length, and so

142
Chapter 6: Conclusion and outlook

the sensing voltage is expected to vary also. It would be useful to investigate the effect
of this variation along the length on the time response of the sensor. The second is to
reduce the asymmetric strain between two PEDOT electrodes, which could be done
by using a symmetric bending setup. These improvements will ensure the reliability of
the experimental results and provide a clue on the origins of the novel sensing
mechanism.

This thesis proposes methods of characterization and modeling flexible transducers that
should facilitate the engineering work to design microsystems for example micro-robotics.

1 W. Shi, Q. Yao, S. Qu, H. Chen, T. Zhang, L. Chen, Micron-thick highly conductive PEDOT films synthesized via self-inhibited polymerization:
roles of anions, Npg Asia Materials, 9(2017)

143
Appendix

A.2 Chapter 2: PEDOT-based trilayer fabrication process

A.2.1 Optimization of electrochemical properties of PEDOT electrodes
Optimization of LBL synthesized electrodes was carried out in order to obtain highly
conductive and electroactive of conducting polymer electrodes (CPE) (PEDOT layers) which
are critical parameters for obtaining efficient bending type trilayer microactuators. High
electronic conductivity will be necessary to promote fast charge transport along the length of
the device and will be involved in the curvature homogeneity of the final beam. Besides
electronic conductivity, also volumetric charge density as an indicator of electroactivity, is a
critical parameter since the electrode strain is directly proportional to the number of
inserted/expelled ions [1]. In other words, if PEDOT chains are highly conductive but not
accessible for insertion/expulsion of ions of the chosen electrolyte, the resulting deformation
will be small. For this purpose, the volumetric charge density of each synthesized CPE has
been minutely characterized and needs to be as high as possible.

Optimization on PEDOT/PEO electrode properties has been performed first as a function of

mPEG content in the oxidant solution and in the present section all the results for the PEDOT
electrodes are given in the swollen state (EMImTFSI), if not stated otherwise.

It was found that while increasing the mPEG content in the oxidant solution, the thickness of
the electrodes (Fig. 1a) remained in the same range from 0% - 20% mPEG before a drastic
increase. A possible reason for the increase in thickness is the pre-polymerization of mPEG in
the flask. Indeed, peroxide based initiators, such as DCPD, and Fe 3+ ions can react and
generate radicals at room temperature [2]. The presence of radicals will lead to a non-desired
polymerization of mPEG in the oxidant solution. While this phenomenon seems rather limited
for a low content of mPEG, and a low content of DCPD, it becomes obvious for high content
of mPEG, since it was found that the bulk polymerization occurs with 50 wt% of mPEG. As the
mPEG and DCPD contents are increased in the oxidant solution, the concentration of
monomers and generated radicals are both increasing the polymerization kinetics and
promote an increase of viscosity in the flask. The thickness of spin-coated oxidant solution
layer on the substrate is then higher and leads to the increase in thickness of PEDOT electrode
layers.

144
Appendix

Fig. 1. Results of mPEG content in the oxidant solution: thickness (a-■), electronic conductivity
(b-●), and volumetric charge density (b-▲) at a scan rate of 20 mV/s and a voltage window of
2 V. Spin coating speed/acceleration/duration of oxidant solution: 2500 rpm/1000 rpm s -1/30
s; EDOT VPP time: 30 min; EDOT VPP temperature: 40C.

Adding up to 20 wt% mPEG into the oxidant solution leads also to an improvement of the
electrical and electrochemical properties of the PEDOT electrodes (Fig. 1b). Adding only 5 wt%
mPEG to the oxidant solution increases the electronic conductivity from 103 S/cm to 164
S/cm. These results are consistent with the literature indicating that adding glycol based
additives [3-5] into the oxidant solution helps to affect positively the synthesis of highly
conductive and electroactive PEDOT layers, so in our case, mPEG is behaving as a reactive
additive. In contrast, the excess of PEG-based monomers (above 20 wt% mPEG) in the oxidant
solution leads to a decrease of both electrical and electrochemical properties, probably due
to the increase of non-electrically conducting phase (mPEG) in the electrodes. On the other
hand, the reason for the decrease of electrical and electrochemical properties above 20 wt%
mPEG in the electrodes can also be explained by the pre-polymerization of mPEG in the flask,
which forms thicker and more disorganized electrodes and can result in the loss of charge
transport along the electrodes.

145
Appendix

Fig. 2. Thickness (a-■), electronic conductivity (b-●) and volumetric charge density (b-▲) of
PEDOT electrodes as a function of rotation speed of the spin coater. mPEG content: 10%; EDOT
VPP time: 30 min; EDOT VPP temperature: 40C. Acceleration 1000 rpm s-1, duration 30 s.

Decreasing the rotation speed of the oxidant solution during the spin coating step led to a
thicker oxidant solution layer on the substrate and therefore thicker CPE layers (Fig. 2a).
Changing the rotation speed did not have a tremendous effect on the electroactivity (Fig. 2b)
of the electrode layers, which remain high at all speeds (between 7.9 x 107 C/m3 and 9.6 x 107
C/m3). A same behavior is also observed with the electronic conductivity i.e. the rotation
speed in the range of 1000 rpm to 2000 rpm hardly influences the electronic conductivity
values which remain high (around 200 S/cm). Depending on the desired electrode thickness,
any rotation speed can be chosen while maintaining the same synthesis conditions (10%
mPEG, EDOT VPP for 30 min at 40°C) to obtain highly conductive PEDOT electrodes with high
electroactivity.

Fig. 3. Effect of EDOT VPP time on PEDOT electrode’s thickness (a-■/♦), electronic conductivity
(b-●), and volumetric charge density (b-▲). mPEG content: 10 wt%; spin coating
speed/acceleration/duration of oxidant solution: 2500 rpm/1000 rpm s-1/30 s; EDOT VPP
temperature: 40C.

The effect of EDOT VPP time on PEDOT electrode properties is described in Fig. 3. While
increasing the polymerization time, it can be seen that the thickness of the swollen electrodes
(Fig. 3a) is decreasing. Those results were surprising but can be explained when measuring
the thickness of dry/non-swollen electrodes. Indeed, the thickness of the dry electrodes is
increasing as a function of VPP time as expected. This result points out a decrease of swelling
ability of the electrodes while polymerization time is increased, probably related to higher
stiffness of the CPE layer. At the same time, the electronic conductivity is decreasing gradually
from 145 S/cm (30 min) to less than 50 S/cm (180 min) (Fig. 3b). This trend has already been
described and explained by the fact that longer polymerization times lead to higher structural

146
Appendix

disorder or randomly orientated disconnected islands in the polymer. These defects limit
charge transport and result in higher resistance on the PEDOT layer [6, 7].

Fig. 4. Influence of EDOT VPP temperature on PEDOT electrode’s thickness (a-■), electronic
conductivity (b-●), and volumetric charge density (b-▲). mPEG content: 10 wt%; spin coating
speed/acceleration/duration of oxidant solution: 2500rpm/1000rpm s-1/30s; EDOT VPP time:
30 min.

Increasing the EDOT VPP temperature also increases the thickness of the PEDOT electrodes
because of the higher concentration of EDOT vapor reacting with the surface of the oxidant
solution (Fig. 4a). The higher temperature also allows partial polymerization of PEO during
the VPP process which ensures the swelling ability of the electrodes in EMImTFSI, thus
avoiding the decrease in thickness as observed while the VPP time was increased. Between
30C and 50C the electronic conductivity (Fig. 4b) lies around 100-150 S/cm and a significant
decrease to 40 S/cm and 4.6x107 C/m3 is observed when the temperature is increasing to
80C. The VPP temperature has a great influence on the morphology and the lattice structure
of the PEDOT layers. At higher temperature (above 50°C), the polymerization rate is faster,
causing structural disorder in the electrode. It has been shown previously that the optimal
VPP temperature for fabricating PEDOT layers is 46±1°C. Above this temperature the
electronic conductivity of CPE layers decreased because the lattice structure of the PEDOT
films had almost no orientation [8].

It can be concluded that to obtain PEDOT electrodes with high electronic conductivity and
high electroactivity using VPP, it is necessary to add PEO precursors to the oxidant solution
while not increasing their concentration above 20%. The rotation speed can be chosen
according to the desired CPE thickness without affecting the electronic conductivity. In
addition, the polymerization time and temperature must not exceed 90 min and 50°C,
respectively, in order to maintain both, high electronic conductivity and electrochemical
properties. The optimal parameters resulted in high conductivities (around 200 S/cm) and
high electroactivity (between 2.3 x 107 C/m3 and 1.0 x 108 C/m3). PEDOT electrode thicknesses
can be tuned between 0.8 and 5.3 m. The optimization of the electrodes in this section

147
Appendix

helped to determine the right parameters for fabricating highly conductive and electroactive
PEDOT electrodes.

A.2.2 Surface measurement method

Fig. 5. Surface roughness measurement apparatus

The Fig. 5 depicts four different steps to measure the roughness on four surfaces. The first
three surfaces (Fig. 5 a, b, c) were simply scanned as they were fabricated. In the last
measurement (Fig. 5 d), the trilayer was flipped over and put back on a silicon wafer before
scanned.

148
Appendix

A.3 Chapter 3: Electrochemomechanical characterization of the

trilayer structure
A.3.1 Qualitative explanation the apparent capacitance of the PEDOT
electrodes at extreme low scan rate

Fig. 6 Cyclic voltammograms of PEDOT electrodes obtained in neat EMImTFSI within a

potential window of -0.6 V to +0.7 V with different scan rates. The vertical axis indicates the
current (A) scaled by the scanning rate (V/s).

As a function of rate, there are three regions of response in the CVs above. In the first region,
scan rate is so fast that we never see the full capacitance. The response looks more resistive
with a loop representing capacitance (1000 mV/s to 300 mV/s). For the second region, scan
rate is slow enough that we see the capacitance, but the time constant is still significant
compared to the scan time. As a result, there is a rise time before reaching a constant current.
Constant current is indicative of the full capacitance. This effect can be seen by looking at the
curves in the positive scan direction. Curves from 200 mV/s down to 20 mV/s reach a fairly
constant current at positive voltages on the upward scan (a constant 0.004 A/(V/s)). As the
scan rate slows, the rise in current is shifted to a lower voltage, making the loop area larger,
and increasing the apparent capacitance. And finally, the rate of charging eventually becomes
so slow that we don’t see a purely capacitive response any more. This is seen in the range of
10 mV/s down to 5 mV/s. The current is no longer flat as we approach 0.7 V. There is a rise in
current that looks like that from a parasitic reaction (curving upwards with increasing
potential). This shifts the entire curve upwards. We can also see this influencing the current
at negative voltages also. To compensate for this, we can integrate over a full cycle (integrate

149
Appendix

the current on the way up, then integrate it on the way down. Subtract the charge on the way
down from the charge on the way up, and divide by scan rate and twice the voltage range).

A.4 Chapter 4: Linear dynamic and non-linear dynamic model to

predict PEDOT-based trilayer actuation behavior
A.4.1 The coupling matrix derivation method
From [9], we have:

𝐸𝑃 𝛼𝜌 ℎ 2 2ℎ𝑃 2 2𝑏𝐸𝑃 𝐾 ℎ𝑆 3 2ℎ𝑃 3 𝐸

𝐹= 𝑏 ( 2𝑆 ) [(1 + ) − 1] + ( 2 ) [(1 + ) − 1 + 𝐸𝑆 ], (1)
𝐿 ℎ𝑆 3𝐿 ℎ𝑆 𝑃

𝑄
where F is the total force produced by the trilayer actuator, 𝜌 = 2𝐿𝑏ℎ is the charge density,
𝑃
2𝑤
and 𝐾 = 𝐿2 +𝑤2 is the curvature of the actuator.

The first term of the equation (1) can derived as following:

𝐸𝑃 𝛼𝜌 ℎ 2 2ℎ𝑃 2 𝐸𝑃 𝛼 𝑄 𝛼𝐸𝑃 (ℎ𝑃 +ℎ𝑆 )

𝑏 ( 2𝑆 ) [(1 + ) − 1] = 𝑏ℎ𝑃 (ℎ𝑃 + ℎ𝑆 ) = 𝑄 (2)
𝐿 ℎ𝑆 𝐿 2𝐿𝑏ℎ𝑃 2𝐿2

1 2sin(𝜃)
From the Figure above, we also have: 𝐾 = 𝑅 = (3)
𝐿

The second term of the equation (1) can be derived:

2𝑏𝐸𝑃 𝐾 ℎ𝑆 3 2ℎ𝑃 3 𝐸 2𝑏𝐸𝑃 2sin(𝜃) ℎ𝑆 3 2ℎ𝑃 3 𝐸

( 2 ) [(1 + ) − 1 + 𝐸𝑆 ] = ( 2 ) [(1 + ) − 1 + 𝐸𝑆 ] (4)
3𝐿 ℎ𝑆 𝑃 3𝐿 𝐿 ℎ𝑆 𝑃

150
Appendix

Yield (4) and (2) into (1) and apply for RFEi element, we obtained:

𝛼𝐸𝑃 (ℎ𝑃 +ℎ𝑆 ) 2𝑏𝐸𝑃 2sin(𝜃𝑖 ) ℎ𝑆 3 2ℎ𝑃 3 𝐸

𝐹𝑖 = 𝑄𝑖𝐶 + ( 2 ) [(1 + ) − 1 + 𝐸𝑆 ] (5)
2𝐿2𝑖 3𝐿𝑖 𝐿𝑖 ℎ𝑆 𝑃

Multiple the right and left term by 𝐿𝑖 , we have:

𝛼𝐸𝑃 (ℎ𝑃 +ℎ𝑆 ) 4𝑏𝐸𝑃 sin(𝜃𝑖 ) ℎ𝑆 3 2ℎ𝑃 3 𝐸

𝑀𝑖 = 𝐹𝑖 × 𝐿𝑖 = 𝑄𝑖 + ( 2 ) [(1 + ) − 1 + 𝐸𝑆 ] × 𝜃𝑖 (6)
2𝐿𝑖 3𝐿𝑖 𝜃𝑖 ℎ𝑆 𝑃

The equation (6) represents the electrochemomechanical coupling matrix used in equation
(9), section 4.3.3.2.

A.5 Chapter 5: Sensing ability and sensing model of the PEDOT-based

trilayer actuators
A.5.1 A possible qualitative sensing model
Earlier experiments have suggested some facts that are:

- Sign of voltage in sensing suggests cations’ movement is much faster in comparison to

the negative charges’ flow (section 5.4.1).
- Sign of actuation during application of a charge suggests the movement direction of
the cations deciding the bending side of the actuator. No stress relaxation of the
actuator is observed during the input voltage is being applied, indicating that only one
type of ions involves in the actuation mechanism (section 3.1).
- Decay in voltage seen in step response suggests the existing of ion flow involving of
both anions and cations in the mechanoelectrical phenomenon of the trilayer
structure.
- Much higher strain to charge ratio in sensing may indicate multiple mechanisms.

These facts suggesting a possible sensing mechanism of the trilayer structure which is
described in Fig. 8.

Since PEO is well known for its ionic conductivity, the penetration of the PEO through all three
layers: PEDOT – NBR – PEDOT increases the mobility of both cations EMI+ and anions TFSI- in
the whole structure. PEO plays a role as a local tank to store the ions and a medium for the
transport of the ions. In the other words, mobile anions and cations are available in NBR/PEO
membrane as well in the PEDOT layer. In addition, relatively immobile anions (TFSI-) exist in
PEDOT electrodes in the doped state to balance charge on the PEDOT backbones. This is
demonstrated in the electromechanical phenomenon, where, based on the direction of
bending, it appears that anion flux dominates. Fig. 7 describes the 3D molecular structure of
both EMI cation and TFSI anion, where the size of those two is comparable. However, in the
pure form of the ionic liquid, EMI- and TFSI- ions tend to form a cluster of positive and negative

151
Appendix

charges whose net charge is “ ̶ ” resulting in bigger negative charges and smaller positive ions
[10-13].

Fig. 7. 3D molecular structure and size of EMI cation and TFSI anion (Adapted with permission
from Ref. [14]. American Chemical Society, 2008.)

An input step displacement makes the beam bend and creates a change in volume (Fig. 8).
The upper side of the beam from the neutral axis suffers a negative strain and a reduction in
volume leading to an increase in local ion concentration and an increase in elastic potential
energy. To balance this change in concentration, the mobile charges start to migrate to lower
concentration areas. Due to a possible interaction between the TFSI- and PEO backbones and
the difference in size between the EMI+ and negative ionic clusters, the negative charges tend
to move slower than the cations, creating a total negative charge or a negative potential in
the upper side. In contrast, the lower side increases its volume and produces a total positive
charge or positive potential. These two simultaneous phenomena produce a sharp rise
voltage at the beginning. In Fig. 8, negative ionic clusters are bigger and move slower, denoted
by shorter velocity arrows. And also depending on the position away from the neutral axis, an
increase in induced stress leads to an increase in diffusion speed when we go away from the
neutral axis.

After a certain amount of time, due to the ion redistribution, the local difference in
concentration of ions is reduced and leads to the decay of the voltage to zero. In term of
thermodynamic, the rebalance decreases the energy of the trilayer system to a more stable
state, which reduces the force of the trilayer.

152
Appendix

Fig. 8. A qualitative model on sensing mechanism of the trilayer structure: an input

displacement causes a flux of cations and anion through the PEDOT electrodes and NBR/PEO
layer, the direction of arrows indicates the moving direction while the length of arrows implies
the ions velocity

In this sensing model, the strain is induced in all three layers - PEDOT electrodes and the
NBR/PEO layers - when a displacement is applied to the trilayer, resulting in the ion flow all
through the three layers. In other words, this displacement induces more charges to move
resulting in a higher output voltage and finally higher strain to charge ratio in case of sensing.

However, in the actuation model, an input voltage only induces the strain on two PEDOT
electrodes.

153
Appendix

[1] J.D.W. Madden, Conducting polymer actuators: Massachusetts Institute of Technology; 2000.
[2] J.F. Perez-Benito, Iron (III)− Hydrogen Peroxide Reaction: Kinetic Evidence of a Hydroxyl-Mediated
Chain Mechanism, The Journal of Physical Chemistry A, 108(2004) 4853-8.
[3] M. Mueller, M. Fabretto, D. Evans, P. Hojati-Talemi, C. Gruber, P. Murphy, Vacuum vapour phase
polymerization of high conductivity PEDOT: Role of PEG-PPG-PEG, the origin of water, and choice of
oxidant, Polymer, 53(2012) 2146-51.
[4] K. Zuber, M. Fabretto, C. Hall, P. Murphy, Improved PEDOT conductivity via suppression of
crystallite formation in Fe (III) tosylate during vapor phase polymerization, Macromolecular Rapid
Communications, 29(2008) 1503-8.
[5] M. Fabretto, C. Jariego-Moncunill, J.-P. Autere, A. Michelmore, R.D. Short, P. Murphy, High
conductivity PEDOT resulting from glycol/oxidant complex and glycol/polymer intercalation during
vacuum vapour phase polymerisation, Polymer, 52(2011) 1725-30.
[6] H. Kim, K. Jeong, C.-J. Yu, H.-S. Nam, H. Soh, J. Lee, The effects of the surface morphology of poly
(3, 4-ethylenedioxythiophene) electrodes on the growth of pentacene, and the electrical performance
of the bottom contact pentacene transistor, Solid-State Electronics, 67(2012) 70-3.
[7] A. Ugur, F. Katmis, M. Li, L. Wu, Y. Zhu, K.K. Varanasi, K.K. Gleason, Low‐Dimensional Conduction
Mechanisms in Highly Conductive and Transparent Conjugated Polymers, Advanced Materials,
27(2015) 4604-10.
[8] D. Wu, J. Zhang, W. Dong, H. Chen, X. Huang, B. Sun, L. Chen, Temperature dependent conductivity
of vapor-phase polymerized PEDOT films, Synthetic Metals, 176(2013) 86-91.
[9] P.G.A. Madden, Development and Modeling of Conducting Polymer Actuators and the Fabrication
of a Conducting Polymer Based Feedback Loop: MASSACHUSETTS INSTITUTE OF TECHNOLOGY; 2003.
[10] J. Hou, Z. Zhang, L.A. Madsen, Cation/Anion Associations in Ionic Liquids Modulated by Hydration
and Ionic Medium, The Journal of Physical Chemistry B, 115(2011) 4576-82.
[11] B.A. Marekha, O.N. Kalugin, M. Bria, R. Buchner, A. Idrissi, Translational Diffusion in Mixtures of
Imidazolium ILs with Polar Aprotic Molecular Solvents, The Journal of Physical Chemistry B, 118(2014)
5509-17.
[12] S. Katsuta, K. Imai, Y. Kudo, Y. Takeda, H. Seki, M. Nakakoshi, Ion Pair Formation of
Alkylimidazolium Ionic Liquids in Dichloromethane, Journal of Chemical & Engineering Data, 53(2008)
1528-32.
[13] S. Katsuta, R. Ogawa, N. Yamaguchi, T. Ishitani, Y. Takeda, Ion Pair Formation of 1-Alkyl-3-
methylimidazolium Salts in Water, Journal of Chemical & Engineering Data, 52(2007) 248-51.
[14] C. Largeot, C. Portet, J. Chmiola, P.L. Taberna, Y. Gogotsi, P. Simon, Relation between the ion size
and pore size for an electric double-layer capacitor, J Am Chem Soc, 130(2008) 2730-1.

154