International Journal of Solids and Structures 41 (2004) 4473–4502

www.elsevier.com/locate/ijsolstr

On a model of layered piezoelectric beams

including transverse stress effect
a,b a,* c
Corrado Maurini , Jo€el Pouget , Francesco dell’Isola
a
Laboratoire d’Etudes Mecaniques des Assemblages (CNRS-FRE 2481), Universite de Versailles/Saint-Quentin-en-Yvelines,
Batiment Descartes, 45 Avenue des Etats-Unis, 78035 Versailles, France
b
Dipartimento di Meccanica e Aeronautica, Universita di Roma, ‘‘La Sapienza’’, via Eudossiana 18, 00184 Rome, Italy
c
Dipartimento di Ingegneria Strutturale e Geotecnica, Universita di Roma ‘‘La Sapienza’’, via Eudossiana 18, 00184 Rome, Italy.
Received 2 October 2003; received in revised form 1 March 2004
Available online 15 April 2004

Abstract
In this paper a Euler–Bernoulli-like model of layered piezoelectric beams is presented. It describes more accurately
than the others already presented in the literature both transverse (Poisson and piezoelectrically induced) cross-sec-
tional deformations and through-the-thickness variations of the electric field and electric displacement. A deductive
approach based on a mixed variational formulation is adopted and distributions of deformation, stress, electric field
and electric displacement are simultaneously prescribed. The attention is focused on the choice of the most fitting
assumptions to recover complex 3D cross-sectional field distributions. In particular, transverse interactions between
different layers are taken into account by enforcing specific conditions on transverse stress through the Lagrange
multipliers method. The estimate of electromechanical beam constitutive coefficients is discussed and comparison with
standard modelling approaches, which assume either vanishing transverse stresses or vanishing transverse strains, is
emphasized. For a sandwich piezoelectric beam and for a two-layer beam, expressions of the beam constitutive coef-
ficients are provided and the main features of the proposed model are highlighted by presenting the through-the-
thickness distribution of the 3D state fields associated to beam-axis deformations and applied voltage. As a main
peculiarity, the proposed beam model is able to coherently estimate the equivalent piezoelectric capacitance also when
the thickness of elastic and piezoelectric layers is comparable.

2004 Elsevier Ltd. All rights reserved.

Keywords: Euler–Bernoulli beam; Piezocomposites; Sandwich materials; Mixed variational formulation; Laminate; Piezoelectric
capacitance; Layered beams

1. Introduction

Composite materials made of active piezoelectric layers have been widely used for their sensor and
actuator functions and research in this area opens many applications in the domain of adaptive structures

*
Corresponding author. Tel.: +33-1-39-25-42-09; fax: +33-1-39-25-30-15.
E-mail address: [email protected] (J. Pouget).

0020-7683/$ - see front matter
2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijsolstr.2004.03.002
4474 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502

and structural control. Piezoelectric materials, and especially piezoelectric beam composites, are excellent
candidates for designing adaptive devices or smart materials, and numerous applications in the domain of
advanced technology have been proposed, running from aerospace or automotive (shape control of space
antennas or telescopes, vibration control of helicopter blades, noise reduction, etc.) to micropositioners or
microactuators (in medical apparatuses, measurement devices, robotics, etc.), as reported by Rao and
Sunar (1994), Tani et al. (1998), Chee et al. (1998), and Chopra (2002). The high performance of piezo-
electric composites becomes a requirement in the design of vibration control systems. Especially, piezo-
electric elements can be used as components for passive damping systems, thereby avoiding complex
control and feedback strategies (see e.g. Anderson and Hagood, 1994; Hagood and Von Flotow, 1991).
The present work attempts to present a consistent and efficient deductive approach to piezoelectric
laminated beams. Slender beam-shaped structures incorporating piezoelectric materials are often used in
engineering applications (robotic arms, airplane wings, rotor blades, etc.). One dimensional modelling
allows for a thumbnail description of their static and dynamic behavior. As the related one-dimensional
boundary-value problems usually can be solved analytically, it provides useful tools for structural design
and control applications. When deductive approaches are followed, one-dimensional models of piezo-
electric laminated beams are obtained from a three-dimensional formulation by assuming a priori distri-
bution of mechanical and electrical state fields. In such a micro–macro identification procedure, the beam
balance equations are found and the corresponding constitutive relations are expressed as functions of the
material and geometrical data at the three-dimensional level. The so-called induced strain models seem to
supply the simplest description of the actuation mechanism in laminated piezoelectric beams (see e.g. Al-
zahrani and Alghamdi, 2003). They can be classified into those assuming a simple extensional strain within
the piezoelectric elements (uniform strain model, see Park et al., 1996; Hong and Chopra, 1999), and those
accounting for extensional and bending deformations of the piezoelectric elements as prescribed by stan-
dard Bernoulli–Euler beam theory (see Crawley and de Luis, 1987; Crawley and Anderson, 1990). A
number of more refined and consistent approaches to piezoelectric laminated beam have been proposed
(reviews can be found in Chopra, 2002; Saravanos and Heyliger, 1999; Gopinathan et al., 2000). They
include the description of shear deformations, effects of bonding layers, full two-fold electromechanical
coupling (i.e. including both of direct and inverse piezoelectric effects), etc. Such models, although one-
dimensional, can become very complex by including several variables of microstructures to mimic the three-
dimensional behavior. However, as underlined also by Beckert and Pfundtner (2002), most of these models
are based on assumptions on the stress state that are too restrictive. For example, in unshearable models, it
is usually assumed (see e.g. Alzahrani and Alghamdi, 2003) that the stress tensor is in the form (refer to
Fig. 1 for notation and beam cross-sectional geometry)
T ¼ T11 ðe1  e1 Þ: ð1Þ
Due to this assumption, transverse interactions (in the width e2 direction) between different layers are
neglected and each layer is left free to deform in the transverse direction independently of the others,
without respecting the bonding condition. The consequence of this assumption is reflected in an incorrect
estimate of the one-dimensional constitutive equations. This issue is particularly relevant in layered beams
including piezoelectric layers polarized along the thickness (e3 direction). Indeed, when a potential differ-
ence is applied between the electrodes of a piezoelectric layer, it naturally tends to isotropically extend (or
shrink) in the plane orthogonal to the polarization axis (e1 –e2 plane). This behavior is in competition with
the Poisson effect in the elastic layer: when one tries to extend an elastic layer in one direction (say e1 ),
shrinking in the other direction (say e2 ) is typically induced. When piezoelectric and elastic layers are
bonded together to form a laminated piezoelectric beam, their contrasting behavior is conciliated by the
appearance of non-negligible normal stresses both in the axial direction (T11 stress) and transverse direction
(T22 stress). In Beckert and Pfundtner (2002) this problem is discussed: assumption (1) is improved by
including transverse ðT22 Þ stresses, in order to describe more accurately the complex interactions between
 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502 4475

e3 y
e2(n)
n

p zi e2(i)
hi i

e1 zi e2

1
e2(1)

Fig. 1. Generic layered piezoelectric beam: cross-section.

different layers in composite piezoelectric laminates. However, only actuation functions are considered and
no attention is paid to the determination of purely electrical properties such as the equivalent piezoelectric
capacitance. Its incorrect estimate can be misleading when piezoelectric elements are used as transducers, as
in passive vibration control devices similar to the electric vibration absorber studied by Hagood and Von
Flotow (1991). Indeed, in this case, an optimal design of the electric controller is strongly influenced by
mechanical, coupling, and electrical properties of the layered piezoelectric beam.
The aim of the present work is to discuss the consequences of transverse interactions between different
layers in piezoelectric laminates and to propose a novel efficient model including both of sensory and
actuation effects. The model thus proposed is based on (i) equivalent single-layer Bernoulli–Euler kine-
matics (linear strain distribution along the thickness direction of the beam); (ii) layerwise linear distribution
of mechanical stress with non-vanishing transverse normal stress (i.e. T22 ); (iii) layerwise linear distribution
of electric potential; (iv) layerwise constant distribution of the electric displacement. In particular, we try to
refine the description achieved by simplest modelling approaches without adding additional degrees of
freedom to describe more complex phenomena such as shear effects, bonding-layer effects, higher-order
distribution of electric field and displacement, etc. As a main peculiarity, the proposed model accounts for
transverse interactions between different layers by including a non-vanishing distribution of normal
transverse stresses. These additional fields are found among those verifying further integral force balance
laws, corresponding to transverse projections of force and moment balance laws for each beam cross-
section. These conditions will be referred to as weak conditions on transverse stress. Such a model, char-
acterized by assuming null transverse stress resultants (NSR model), is compared to standard models
assuming either pointwise null transverse stress (NS model), or null transverse sectional deformations (ND
model). One of the purposes of the present work is to examine and discuss how the newly introduced stress
field T22 and the associated weak conditions on it enhance the estimate of the electromechanical constitutive
properties of piezoelectric laminated beams.
The proposed beam model is deduced from the three-dimensional description by assuming at the same
time hypotheses on strain and stress, on electric field and electric displacement. To this end, a mixed
variational formulation of the Hellinger–Prange–Reissner type is adopted (see He, 2000; Teresi and Tiero,
1997; Washizu, 1982). Moreover, Lagrange multipliers are introduced to impose in the variational for-
mulation the different conditions on transverse stress.
The prerequisites for three-dimensional piezoelectricity (equations of motion, boundary conditions and
constitutive equations) and the associated mixed variational formulation are briefly recalled in the next
section. In Section 3, beam governing equations are deduced from the mixed variational formulation by
assuming peculiar distributions of three-dimensional state fields. In particular, governing equations for
models assuming either null transverse stress resultants (NSR), pointwise null transverse stress (NS), or null
transverse deformations (ND) are derived. In Section 4, comparisons between the different models are
made and the main properties of the NSR model are underlined. To this end, particular cross-sectional
4476 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502

configurations are considered: a three-layered sandwich piezoelectric beam with in-phase parallel connected
piezoelectric layers; a three-layered sandwich piezoelectric beam with counter-phase parallel connected
piezoelectric layers; and a two-layered piezoelectric beam. For the different conditions on the transverse
stress, analytical formulas and numerical illustrations of beam constitutive coefficients are provided. Their
dependence on material properties and the relative thickness ratio between different layers are analyzed.
The through-the-thickness distributions of stress, strain, electric field and electric displacement are also
presented. Lastly, Section 5 is devoted to the summary and discussion of the most pertinent results and
extensions of the present beam model are suggested.

2. Mixed variational formulation of 3D linear piezoelectricity

In this section, the basic equations of three-dimensional linear piezoelectricity are recalled and a mixed
variational formulation for the associated boundary value problem is proposed. Moreover, the application
of the mixed variational formulation to structural modelling is briefly discussed.

2.1. Equations of linearly piezoelectricity

Consider a piezoelectric body that is identified by means of its reference configuration B. In the quasi-
electrostatic approximation, the actual kinematical state of the system is described by the mechanical
displacement field u and by the electric potential u. Let us denote by oB the boundary of B, by ou B and
of B the parts of oB where the displacements u0 and the tractions f 0 are imposed, by ou B and oq B the parts
on which the electric potential u0 and the charge density q0 are imposed (ou B [ of B ¼ ou B [ oq B ¼ oB
and ou B \ of B ¼ ou B \ oq B ¼ ;). Let S be the linearized strain tensor, T the Cauchy stress tensor, D the
electric displacement vector, and E the electric field vector inside the piezoelectric body.
In the quasi-electrostatic approximation, a strong formulation of the equations of linearly piezoelec-
tricity is provided by the following equations (see e.g. Eringen and Maugin, 1990)


r Tþb¼0
force balance : on B; ð2aÞ
r D¼0


S ¼ SymðruÞ
kinematic compatibility : on B; ð2bÞ
E ¼ ru


u ¼ u0 on ou B;
essential boundary conditions : ð2cÞ
u ¼ u0 on ou B;


Tn ¼ f 0 on of B;
natural boundary conditions : ð2dÞ
Dn ¼ q0 on oq B;
( 
oF 
S¼ oT D
¼ sD T þ gt D
constitutive equations : oF  on B: ð2eÞ
E¼ oD T
¼ gT þ bT D

In the equations above n denotes the external unit normal to the boundary oB, b is an external force per
unit volume applied on B (and eventually including inertial forces), rð Þ denotes the spatial gradient with
respect to the reference position, r ð Þ the corresponding divergence operator, and Symð Þ extracts the
symmetric part of a second order tensor. The constitutive equations are given in the so-called T–D form
through the free energy density function for linear piezoelectricity (see Ikeda, 1990)
 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502 4477

1 1 T
FðT; DÞ ¼ sD T T b D D þ gT D; ð3Þ
2 2
where sD is the mechanical compliance fourth order tensor for null electric displacement, g is the piezo-
electric coupling third order tensor and bT is a second order tensor modelling the inverse of the electric
permittivity for null mechanical stress. For a transversely isotropic piezoelectric material, the tensors
appearing in the constitutive equations are represented in the classical Voigt notation by means of the
following matrices
2 T 3
b11 0 0
6 7
bT ¼ 4 0 bT11 0 5;
0 0 bT33

2 3
sD
11 sD
12 sD
13 0 0 0
6 sD sD sD 0 0 0 7
6 12 11 13 7
6 sD sD sD 0 0 0 7
6 7 ð4Þ
sD ¼ 6 13 13 33
7;
6 0 0 0 sD 0 0 7
6 44 7
4 0 0 0 0 sD 0 5
44
0 0 0 0 0 2ðsD
11 sD
12 Þ

2 3
0 0 0 0 g15 0
g¼4 0 0 0 g15 0 0 5:
g31 g31 g33 0 0 0

2.2. Mixed variational formulation and reduced models

Let us define the affine space Vu of kinematically admissible displacement and strain tensors and the
space Vu of admissible electric potential and electric field vectors as follows:
Vu fðu; SÞ : u ¼ u0 on ou B and S ¼ SymðruÞ on Bg; ð5aÞ

Vu fðu; EÞ : u ¼ u0 on ou B and E ¼ ru on Bg: ð5bÞ

Moreover, let us denote by VT and VD the vector spaces of symmetric stress tensor fields and of electric
displacements vector fields defined on B.
Let us consider the following functional (Hellinger–Prange–Reissner functional for piezoelectricity (see
Yang and Batra, 1995, Batra and Vidoli, 2002))
Z Z Z
H½ðu; SÞ; ðu; EÞ; T; D ¼ ðFðT; DÞ T S D E þ b uÞ dB þ f 0 u dS q0 u dS ð6Þ
B of B oq B

defined over the space

V ¼ Vu  Vu  VT  VD : ð7Þ
Under suitable regularity conditions, it can be verified that the solution to the problem of linear piezo-
electricity (2) is characterized by rendering stationary the Hellinger–Prange–Reissner functional (6) over V.
In particular, the equilibrium equations (2a) and the natural boundary conditions (2d) are found by
imposing vanishing first variations of functional (6) with respect to u and u, and the constitutive equations
4478 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502

(2e) with respect to T and D. On the other hand, the compatibility conditions (2b) and the essential
boundary conditions (2c) are included in the definitions (5a) and (5b) of the functional spaces Vu and Vu .
In a model derived by a mixed variational formulation, it is often useful to distinguish between the
generalized deformations (S, E) obtained from the generalized displacements ðu; uÞ through the kinematic
compatibility (2b) and the generalized deformations ðS; EÞ obtained from the generalized stresses ðT; DÞ
though the constitutive equations (2e). Here and henceforth, the latter will be differentiated from the former
by a superscripted bar.
By following a direct approach, reduced models of a given refined one are deduced by assuming specific a
priori ‘‘types’’ of state fields. The accuracy of the description supplied by the reduced model relies on how
judiciously the aforementioned ‘‘types’’ are chosen. By following the mixed variational formulation above,
restrictive hypotheses on the generalized displacement (u; u), as well as on generalized stresses (T, D), are
straightforwardly included in the model. These hypotheses are introduced in the variational principle as
constraints on the space V on which the variations of the mixed functional must be taken. In particular,
when these constraints are linear, the constrained variational problem can be formulated as the search for
stationary points of the given functional on a linear subspace W of V. Alternatively, the constraints on the
admissible state fields can be imposed in the variational formulation by the Lagrange multiplier method.
The latter approach is particularly useful when the constraints naturally appear in an implicit form.
Moreover, the physical interpretation of the Lagrange multipliers furnishes a deeper understanding of the
problem.

3. Beam models

In this section, the three-dimensional mixed variational formulation is used to deduce beam models of
slender piezoelectric laminates by assuming specific hypotheses on the cross-sectional distribution of the
electromechanical state fields. In particular, by focusing on Euler–Bernoulli-like models, we investigate how
assumptions on cross-sectional strains and stresses influence the estimate of the beam constitutive coeffi-
cients. Indeed, standard beam models (see e.g. Crawley and de Luis, 1987; Crawley and Anderson, 1990;
Park et al., 1996) assume very drastic hypotheses about the distribution of transverse normal stress or
strain, which although working well for mono-layered beams, fail to predict some phenomena appearing
when different layers are interacting. After introducing basic definitions and key hypotheses (Section 3.1), a
beam model able to properly describe the basic features of cross-sectional deformation and normal stress
distributions is presented (Section 3.2). Moreover, in order to compare the proposed model (NSR model) to
standard approaches, the governing equations for models retaining standard assumptions on transverse
stress and strain are derived. Namely, in Sections 3.3 and 3.4, models assuming vanishing transverse normal
stress (NS model) and vanishing transverse deformation (ND model) are considered.

3.1. Definitions and hypotheses

3.1.1. Geometry and materials

Let us consider a multilayered straight-axis piezoelectric beam that is made by stacking up piezoelectric
and elastic layers. We decompose the reference domain B as the Cartesian product of the beam axis A and
the beam normal cross-section S. We denote by Si with i 2 I ¼ f1; . . . ; ng, the cross-sectional part
occupied by the ith layer and by n the total number of layers ([i2I Si ¼ S). Moreover, we denote by Ip
and Ie (Ip [ Ie ¼ I) the set of indices i associated with the piezoelectric and elastic layers, respectively.
The following geometric and material properties are assumed: (i) the cross-sectional part Si is rectangular
and it is characterized by the width ai and thickness hi ; (ii) each layer is materially homogeneous and either
orthotropic or transversely isotropic with respect to an axis oriented along its thickness (in particular the
 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502 4479

piezoelectric layers are polarized along the thickness); (iii) the upper and lower surfaces of the piezoelectric
layers are covered by a conductive layer with negligible mechanical properties, the lateral ones are bare; (iv)
the electrodes of the piezoelectric layers are parallel connected one to each other, and the whole beam is
electrically accessible only through two external electric terminals; for each piezoelectric layer, we define a
constant xi ¼ 1 defining the electric connection scheme between the electrodes of the ith layer and the
external terminals (xi ¼ 1 in-phase connection, xi ¼ 1 counter-phase connection).
A global reference frame C ¼ fo; e1 ; e2 ; e3 g and n local reference frames CðiÞ ¼ foðiÞ o þ zi e3 ; e1 ; e2 ; e3 g
are fixed once for all, as sketched in Fig. 1. The coordinate representation in C of a generic point p 2 B is
denoted by ðx; y; zÞ. The local e3 -coordinate in CðiÞ is denoted by zi ¼ z zi , where zi is chosen in such a way
that
Z
zi dS ¼ 0: ð8Þ
Si

3.1.2. Hypotheses
With the proposed mixed variational formulation, the reduced model generalized forces and the asso-
ciated equilibrium equations are determined by hypotheses for ðu; uÞ. On the other hand, hypotheses for (S,
D) affect the associated constitutive prescription. Here, in order to introduce the minimum number of
kinematical descriptors, an elementary Euler–Bernoulli-like electromechanical kinematics is adopted. At
the same time, the estimate of the associated constitutive prescriptions is improved by carefully selecting the
cross-sectional distribution of mechanical stress and electric displacement.
In order to derive efficient and accurate beam models of layered piezoelectric composites accounting for
the two-fold electromechanical coupling, we assume the following hypotheses for the distribution of the
electromechanical state fields over a beam cross-section.

(H1) Mechanical displacement. Basic equivalent-single-layer Euler–Bernoulli kinematics:

uðx; y; zÞ ¼ ðuðxÞ zw0 ðxÞÞe1 þ wðxÞe3 ; ð9Þ
where uðxÞ and wðxÞ are the beam axis displacements along e1 and e3 , respectively.
(H2) Electric potential. Layerwise linear distribution of the electric potential and parallel interconnection
of the different layers
1 zi
uðx; y; zÞ ¼ þ xi V; ð10Þ
2 hi
where V is the electric potential difference across the two electric terminals of the beam.
(H3) Mechanical stress. Layerwise linear distribution of plane-stress with vanishing shear term:
Tðx; y; zÞ ¼ ðr1;i ðxÞ zi f1;i ðxÞÞðe1  e1 Þ þ ðr2;i ðxÞ zi f2;i ðxÞÞðe2  e2 Þ; ð11Þ
where, layer by layer, constant (r’s) and linear (f’s) contributions to the beam stress in the axial (e1 )
and transverse (e2 ) directions are defined.
(H4) Electric displacement. Layerwise constant distribution along the thickness direction:
Dðx; y; zÞ ¼ D3;i ðxÞe3 : ð12Þ

Particular attention is devoted to the discussion of the consequences (and the plausibility) of hypotheses
about the distribution of transverse normal stress and strain T22 ¼ Te2 e2 and S22 ¼ Se2 e2 . Models
assuming the following three different further conditions on transverse normal stress T22 are analyzed and
compared.
4480 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502

(C1) Null transverse stress resultants (NSR). The following conditions of vanishing stress resultants are im-
posed:
( R
Ny ¼ S Tðx; y; zÞe2 e2 dS ¼ 0;
R ð13Þ
My ¼ S zTðx; y; zÞe2 e2 dS ¼ 0:

(C2) Null transverse stress (NS). The following condition of pointwise null transverse stress is enforced:

T22 ¼ Tðx; y; zÞe2 e2 ¼ 0: ð14Þ

(C3) Null transverse deformation (ND). We allow the stress fields of the type specified by (H3) to vary
freely. In this case the distribution of the transverse normal stress T22 will be determined in order
to retain condition of null transverse deformation implied by hypothesis (H1).

The different hypotheses on the electromechanical state fields and their influence on the properties of the
associated beam models deserve some comments.

(i) Notwithstanding the elementary Euler–Bernoulli kinematics in hypothesis (H1), the beam constitu-
tive prescriptions will include the influence of cross-sectional deformations and transversal interactions
between different layers through hypotheses on transverse stress (i.e. (H3) combined with one between
conditions (C1), (C2), and (C3)). Indeed, the influence of a sectional distension along the thickness is
implicitly taken into account by enforcing null normal stress T33 . Furthermore, the influence of sectional
extensional deformations along e2 are introduced by constraining the admissible transverse normal stress
T22 .
(ii) The linear distribution of the electric potential assumed in (H2) specifies the electric kinematics only
in terms of the potential difference V . However, because of hypothesis (H4), the beam constitutive equa-
tions are derived by assuming a layerwise constant electric displacement. As extensively detailed in Sze et al.
(in press), the hypothesis of constant electric displacement is the best suited for fitting the three-dimensional
distribution of the electric state fields without introducing additional electric degrees of freedom. In this
way, as can be checked by writing down the constitutive equation for the electric potential, a through-the-
thickness linear contribution to the electric field (corresponding to a quadratic electric potential) is
implicitly associated with flexural strains. One of the advantage of the proposed mixed variational for-
mulation is to straightforwardly include in the model the hypothesis of layerwise constant electric dis-
placement.
(iii) Hypotheses (H3) and (H4) introduce in the beam constitutive equations the effects of cross-sectional
deformations and quadratic distribution of the electric potential. However, the electrical and mechanical
shear-like effects associated with the axial variations of cross-sectional displacements and electric potential
are neglected. In other words: the piezoelectric laminate is modelled as beam axis on which orthogonal
cross-sections are positioned; for given axial displacements these cross-sections, being constrained to re-
main orthogonal to the axis, are left free to deform in their own plane so as to fit the conditions on the
distribution of transverse stress; moreover, for a given potential difference V , the electric potential inside the
piezoelectric layers is left free to vary along the thickness in order to satisfy the condition (H4) of constant
electric displacement; however, since all the shear-like electrical and mechanical effects are discarded, all
shear interactions between two adjacent beam cross-sections are neglected and only those exerted through
axial normal stress are retained.
(iv) Conditions (C1) and (C2) restrict the admissible stress distributions over a section. Condition (C2)
implies (C1). In the case of a homogeneous single-layer beam, given hypothesis (H3), also the inverse is true
and the two conditions are equivalent. This is not the case of a multi-layered beam. In the following, we
refer to the NS condition (C1) also as to the strong condition on transverse stress and to the NSR condition
 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502 4481

(C2) also as the weak condition on transverse stress. The weak condition can be interpreted as an integral
version of the force balance in the transverse direction. As it will be shown in the following, in a multi-
layered beam transverse interactions between different layers are present and the transverse normal stress is
not vanishing pointwise. In this case, the strong condition (C2) is not physically grounded, while the
weakened version (C3) is still legitimized by global balance laws.

In the following, the three beam models accounting for assumption (H1)–(H4) and corresponding to
conditions (C1), (C2) or (C3) are deduced by a mixed variational formulation based on the Hellinger–
Prange–Reissner functional (6). The additional conditions (13) and (14) on transverse normal stress
are imposed on the variational problem by the Lagrange multipliers method. The following external
actions are included in the description: (i) a force distribution on the beam bases having a force resul-
tant F ¼ N e1 þ T e3 and a moment resultant M ¼ Me2 ; (ii) a body force per unit volume bðx; y; zÞ ¼
bðx; y; zÞe3 ; (iii) either a voltage V or a total charge Q imposed on the electric terminal of the set of piezo-
electric layers.

3.1.3. Beam state fields

When hypotheses (H1) and (H3) are assumed, the three-dimensional distribution of the mechanical
displacement and the electric field is given as function of the following axial fields
U ¼ ½u w ; ð15aÞ

V ¼ ½V : ð15bÞ

Moreover, the strain tensor and the electric fields in the ith layer are in the form
Sðx; y; zÞ ¼ ðe1;i ðxÞ zi j1;i ðxÞÞe1  e1 ; ð16aÞ

Eðx; y; zÞ ¼ E3;i e3 ; ð16bÞ

where
e1;i ðxÞ ¼ u0 ðxÞ zi w00 ðxÞ; j1;i ðxÞ ¼ w00 ðxÞ ð17Þ
and
xi
E3;i ¼ V: ð18Þ
hi
Hence, the 3D distribution of the mechanical deformation and electric field is specified by the axial fields
(here and henceforth the explicit dependence on the space variables is omitted)
S ¼ fe1;i ; j1;i gni¼1 ; E ¼ fE3;i gni¼1 : ð19Þ
We say that the fields ðU; SÞ are compatible if they are related through the compatibility relations (17) and if
the displacement u induced by U through (9) verifies the geometric boundary condition u ¼ u0 on ou B. In
the same way, we say that ðV; EÞ are compatible if they are related through the compatibility relations (18)
and if the electric potential u induced by V through (10) verifies the geometric boundary condition u ¼ u0
on ou B. In what follows, we denote by Wu the functional space of compatible mechanical kinematical fields
ðU; SÞ and by Wu that of compatible electric kinematical fields ðV; EÞ.
By the assumed distribution for the mechanical stress and the electric displacement, the beam dynamic
state is specified through the generalized stress descriptors
n
T ¼ fr1;i ; f1;i ; r2;i ; f2;i gi¼1 ; ð20Þ
4482 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502

D ¼ fD3;i gni¼1 : ð21Þ

Let us define WT as the functional space of the suitably regular beam stress descriptors T and WD as the
functional space of the beam electric displacement descriptors D. Hence, the fields in the space
W Wu  Wu  WT  WD characterize the electromechanical state of the reduced beam model.
The definitions above will be extensively used for the formulation of the variational problem associated
with the different beam models.

3.2. Model with null transverse stress resultants (NSR model)

The variational problem for the beam model corresponding to hypotheses (H1)–(H4) and to the NSR
condition on transverse stress is formulated by rewriting the reduced beam version of the Hellinger–
Prange–Reissner functional (6) by taking into account hypotheses (H1)–(H4) and successively by enforcing
the weak conditions (13) on transverse stresses by the Lagrange multiplier method. Associated beam
balance and constitutive equations are found by looking for the stationary point of the so obtained beam
functional. In particular, the Euler equations found by varying mechanical and electrical kinematical fields
(U; V) furnish mechanical and electrical equilibrium equations; those found by varying mechanical and
electrical dynamic fields (T; D) furnish mechanical and electrical constitutive equations. The NSR conditions
are recovered by imposing the beam functional to be stationary with respect to variations of the introduced
Lagrange multipliers.

3.2.1. Variational formulation

Let us define the following beam model free energy
Z
Fbeam ¼ FðT; DÞ dS: ð22Þ
S

By making use of hypotheses (H1)–(H4) it becomes

X1 X X1 T
Fbeam ðT; DÞ ¼ ðAi sab;i ra;i rb;i þ Ji sab fa;i fb;i Þ þ Ai g3a;i ra;i D3;i b33;i Ai D23;i ; ð23Þ
i2I
2 i2Ip i2Ip
2

where
Z Z
Ai ¼ dS ¼ ai hi ; Ji ¼ z2i dS ¼ ai h3i =12 ð24Þ
S S

and an implicit summation over a, b 2 f1; 2g is implied. The constitutive coefficients sab;i , g3a;i , and bT33;i are
defined with the standard notation for piezoelectric materials except for dropping the superscript D (null
electric displacement) on s and adding a subscript i to distinguish the constitutive properties of different
layers. Due to material symmetries g32;i ¼ g31;i , s12;i ¼ s21;i , s11;i ¼ s22;i .
Hence, by substituting hypotheses (H1)–(H4), the Hellinger–Prange–Reissner functional H defined in
Eq. (6) becomes,
Z XZ XZ
Hbeam ½ðU; SÞ; ðV; EÞ; T; D ¼ Fbeam ðT; DÞ dx þ Ai D3;i E3;i dx ðAi r1;i e1;i þ Ji f1;i j1;i Þ dx
A i2Ip A i2I A
Z
þ ðbN u þ bM w0 þ bT wÞ dx þ ½N u þ Mw0 þ T woA QV ; ð25Þ
A
 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502 4483

where
Z Z Z
bN ¼ f 0 e1 d‘ þ b e1 dS; N¼ f 0 e1 dS;
of S S S
Z Z Z
bM ¼ zf 0 e1 d‘ ðzb e1 Þ dS; M¼ zf 0 e1 dS; ð26Þ
of S S S
Z Z Z
bT ¼ f 0 e3 d‘ þ b e3 dS; T ¼ f 0 e dS;
of S S S

and Q is the total charge imposed on the two electric terminals of the multilayered beam. The reduced beam
functional (25) is defined over the space W Wu  Wu  WT  WD .
The variational formulation is completed by imposing the weak conditions (13) on transverse stress by
the Lagrange multiplier method. Hence, the following modified functional is introduced

H ~ ðKÞ
~ beam ¼ Hbeam þ H ð27Þ
beam

with
Z Z Z Z
~ ðKÞ ¼
H kðxÞ Tðx; y; zÞe2 e2 dS dx lðxÞ zTðx; y; zÞe2 e2 dS dx: ð28Þ
beam
A S A S

If the following generalized transverse strains are defined

e2;i ¼ kðxÞ zi lðxÞ;
ð29Þ
j2;i ¼ lðxÞ;
~ ðKÞ can be rewritten as
then H beam
Z X
~ ðKÞ
Hbeam ¼ ðAi e2;i r2;i þ Ji j2;i f2;i Þ dx: ð30Þ
A i2I

Finally, the constrained variational problem associated with the NSR model consists of looking for the
stationary point of the modified functional H~ beam over the extended space

f
W Wu  Wu  WT  WD  WK ; ð31Þ

where WK is the space of admissible Lagrange multipliers K ¼ fk; lg.

The transverse strains ðe2;i ; j2;i Þ are the analog of ðe1;i ; j1;i Þ in the width direction and they can be re-
garded as the constant and linear part of the deformation along ðe2  e2 Þ. Similarly, the Lagrange multi-
pliers ðk; lÞ are the analog of ðu0 ; w00 Þ in the transverse direction. They are associated with the extensional (k)
and flexural (l) sectional deformation in the transverse direction, exactly as ðu0 ; w00 Þ are in the axial
direction.

3.2.2. Balance equations

The balance equations and the natural boundary conditions are derived by imposing that the first
variation of the functional (27) in Wu  Wu is null. Hence, the following balance equations are found:
8 0
< bN þ Nx ¼ 0;
>
bM bT þ Mx00 ¼ 0;
0
ð32Þ
>
: R
A
q dx Q ¼ 0;
4484 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502

where the following force resultants have been defined

XZ X
Nx ¼ T ðe1  e1 Þ dS ¼ Ai r1;i ð33aÞ
i2I Si i2I

XZ X X
Mx ¼ zT ðe1  e1 Þ dS ¼ zi Ai r1;i þ Ji f1;i ; ð33bÞ
i2I Si i2I i2I

X
q¼ xi ai D3;i : ð33cÞ
i2Ip

The natural boundary conditions are derived by imposing that the following conditions are satisfied for all
the admissible variations du and dw of u and w in Wu
(
½ðNx N Þ duoA ¼ 0;
ð34Þ
½ðMx MÞ dw0 þ ðT þ Mx0 þ bM Þ dwoA ¼ 0:
As expected because of the assumed kinematics, the balance equations above correspond to those of a
standard Euler–Bernoulli beam model.

3.2.3. Constitutive equations

Local constitutive equations. Local constitutive equation valid layer by layer are found by imposing that
the first variation of the functional (27) with respect to ðT; DÞ is null. They are given by
8
>
> sD r þ g3a D3;i ¼ ea;i ;

< ab b;i
sab rb;i ¼ ea;i ;
piezoelectric layers sDab f b;i ¼ j a;i ; elastic layers ð35Þ
i2Ip >
> i2I s ab fb;i ¼ ja;i :
: g r þ bT D ¼ EðlÞ ; e
3a a;i 33 3;i 3;i

The constitutive equations above can be inverted and rewritten in the form
8
> E (
< ra;i ¼ ~cab eb;i ~e3b E3;i ;
>
D ra;i ¼ ~cab eb;i ;
piezoelectric layers fa;i ¼ ~cab jb;i ; elastic layers ð36Þ
i2Ip >
> i2Ie f b;i ¼ ~
cab jb;i :
: D ¼ ~e e þ ~S E
3;i 3b b;i 33 3;i

Explicit expressions for the new constitutive coefficients are given in Appendix A.1. As it can be easily
checked, their expressions correspond to those of the constitutive coefficient of piezoelectric and elastic
materials in which the normal stress along e3 (namely, T33 ) is imposed to be null and the electric dis-
placement constant along the thickness. This is consequence of assumption (H3) where T33 ¼ Te3 e3 ¼ 0
and assumption (H4) where oD=oz ¼ 0.
Conditions on transverse stresses. The NSR conditions on transverse stresses given by Eqs. (13) are re-
trieved by imposing that the first variation of the beam functional (27) with respect to the Lagrange
multipliers ðk; lÞ is null. By introducing the following force resultants
X X
Ny ¼ Ai r2;i ; My ¼ Ji f2;i zi Ai r2;i ; ð37Þ
i2I i2I

they appear in the form

Ny ¼ 0;
ð38Þ
My ¼ 0:
 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502 4485

Global constitutive equations. Let us collect the force resultants in the axial and transverse direction in the
following vectors:
   
Nx Ny
Rx ¼ ; Ry ¼ : ð39Þ
Mx My
Moreover, let us introduce the vectors collecting the axial fields representing beam deformations in the two
directions as follows
 0  
u k
dx ¼ ; d y ¼ ¼ K; ð40Þ
w00 l
where the Lagrange multipliers have been explicitly associated with transversal extension (k) and flexion (l)
of the beam cross-section. Hence, by substituting the local constitutive equations (36) and the kinematic
compatibility equations (16), (17), (18) in the definitions of the force resultants (33)–(37), the global beam
constitutive equations in the following form are found:
2 3 2 32 3
Rx Kxx Kxy etdV dx
4 Ry 5 ¼ 4 Kxy Kxx etdV 54 dy 5; ð41Þ
q edV edV qV V
where
   
kNu kNw kN k kN l
Kxx ¼ ; Kxy ¼ ; edV ¼ ½ eNV eMV : ð42Þ
kNw kMw kN l kMl
Explicit expressions for the constitutive constants appearing above in terms of three-dimensional material
and constitutive properties are reported in Appendix A.2.
The conditions on transverse stresses (38) impose that
Ry ¼ Kxy dx þ Kxx dy etdV V ¼ 0: ð43Þ
Hence, the following expressions for the transverse deformations are found
dy ¼ Kxx1 Kxy dx þ Kxx1 etdV V : ð44Þ
Finally, the following constrained global constitutive equations are obtained by substituting the expressions
(44) for the transverse deformations in the constitutive equations (41)
  " ðNSRÞ #
ðNSRÞ t  
Rx Kxx ðedV Þ dx
¼ ðNSRÞ tðNSRÞ
; ð45Þ
q edV dV V

where
8
>
> K ðNSRÞ ¼ Kxx Kxy Kxx1 Kxy ;
< xx
ðNSRÞ
edV ¼ edV edV Kxx1 Kxy ; ð46Þ
>
>
: ðNSRÞ ¼  þ e K 1 et :
qV qV dV xx dV

If the different layers are symmetrically disposed with respect to a given plane or if cE12 =cE11 ¼ c12 =c11 , it is
possible to choose the position of the e2 axis is such a way that
kNw ¼ kN l ¼ 0; ð47Þ
and the matrices Kxx and Kxy are diagonal. Therefore, the constrained beam constitutive equations (45) are
written as
4486 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502
2 3 2
ðNSRÞ ðNSRÞ
32 3
Nx kNu 0 eNV u0
6 7 6 76 00 7
4 Mx 5 ¼ 6 74 w 5;
ðNSRÞ ðNSRÞ
4 0 kMw eMV 5 ð48Þ
q ðNSRÞ ðNSRÞ ðNSRÞ V
eNV eMV qV
where the constitutive coefficients are given by the following simple expressions:
ðNSRÞ
kNu ¼ kNu ð1 kN2 k =kNu
2
Þ; ð49aÞ

ðNSRÞ 2 2
kMw ¼ kMw ð1 kMl =kMw Þ; ð49bÞ

ðNSRÞ
eNV ¼ eNV ð1 kN k =kNu Þ; ð49cÞ

ðNSRÞ
eMV ¼ eMV ð1 kMl =kMw Þ; ð49dÞ

ðNSRÞ
qV ¼ qV ð1 þ e2NV =ðkNu qV Þ þ e2MV =ðkMw qV ÞÞ: ð49eÞ

Moreover, the relations (44) expressing the cross-sectional transverse deformations in terms of axial ones
and applied electric potential are written explicitly as
2 3
    u0
k kN k =kNu 0 eNV =kNu 4 00 5
dy ¼ ¼ w : ð50Þ
l 0 kMl =kMw eMw =kMw
V
The expressions above for cross-sectional transverse extension and bending show how, in order to enforce
the NSR conditions (38), the constitutive equations of the NSR model (45) account for cross-sectional
deformations. This feature is provided by the mixed variational formulation, where the beam constitutive
behavior is prescribed through the hypotheses on generalized stress descriptors. In this way, despite the
assumed elementary Euler–Bernoulli kinematics, beam cross-sectional deformations are implicitly allowed
for by suitably restricting admissible stress states.

3.3. Model with pointwise null transverse stress (NS model)

3.3.1. Variational formulation

The strong conditions (14) on transverse stress naturally restrict the space of admissible stress to the
subspace Wc T of WT made of the generalized stress T characterized by r2;i ¼ f2;i ¼ 0. Hence, the variational
problem in this case could be stated as finding the stationary point of (25) over Wu  Wu  W c T  WD .
However, the NS conditions (14) are imposed here by the Lagrange multiplier method, in order to
emphasize analogies and differences with respect to the model assuming the NSR conditions (13). Hence,
the following modified functional is introduced
b beam ¼ Hbeam þ H
H b ðKÞ ; ð51Þ
beam

where, in view of hypothesis (H3), the additional term enforcing the NS condition (14) is given by
Z X
b ðKÞ
H beam ¼ ðki ðxÞr2;i ðxÞ þ li ðxÞf2;i ðxÞÞ: ð52Þ
A i2I
 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502 4487

By introducing the functional space of admissible Lagrange multipliers Wc K ¼ fk1 ; l1 ; . . . ; kn ; ln g, the

variational problem associated with the NS model is finally formulated as finding the stationary set of the
functional Hb beam over

c
W c K:
Wu  Wu  WT  WD  W ð53Þ

3.3.2. Balance equations

The equilibrium equations and the natural boundary conditions are found by taking the variation of the
functional (51) in Wu  Wu . As it can be easily checked, they are the same of those found for the model
assuming the weak conditions on transverse stress (Eqs. (32) and (34)).

3.3.3. Constitutive equations

Local constitutive equations. By taking the variation of the functional (51) in WD  WT , local consti-
tutive equations in the form (35) and (36) are found also in this case; however the following different
definitions for transverse deformations in terms of Lagrange multipliers holds:


e2;i ¼ ki ;
ð54Þ
j2;i ¼ li :

In this case, two Lagrange multipliers for each layer are defined. They can be interpreted as transverse
cross-sectional extension and bending of each layer.
Condition on transverse stresses. The constraints on transverse stresses (14) are retrieved by imposing that
n
the variations of the functional (51) with respect to fki ; li g1¼1 are null. They are


r2;i ¼ 0;
ð55Þ
f2;i ¼ 0:

These conditions, being local, can be imposed directly on the local constitutive equation (35) and solved for
n
the Lagrange multipliers fki ; li g1¼1 . In this way, the following constrained local constitutive equations are
found
8
>
> r ¼ ^cE11 e1;i ^e31 E3;i ; (
< 1;i r1;i ¼ ^c11 e1;i ;
piezoelectric layers f1;i ¼ ^cD 11 j1;i ; elastic layers ð56Þ
i2Ip >
> i2Ie f1;i ¼ ^c11 j1;i ;
: S
D3;i ¼ ^e31 e1;i þ ^33 E3;i ;

where the explicit expressions of the constitutive coefficients in terms of the three-dimensional material and
geometric properties are given in Appendix A.1. As it can be easily checked, the constitutive constants with
a superimposed hat correspond to those for piezoelectric and elastic materials under a uniaxial stress state
in the e1 -direction (i.e. for T22 ¼ T33 ¼ 0). Moreover, the following expressions for the transverse sectional
deformations of each layer are found:
8 8
>
> ~cE12 ~e31 > ~c12
>
< e2;i ¼ e þ E E3;i ;
E 1;i
>
< e2;i ¼ e1;i ;
~c11 ~c11 ~c11
piezoelectric layers elastic layers ð57Þ
i2Ip >
> ~cD i2Ie >
> ~c12
>
: j2;i ¼ 12
j1;i ; : j2;i ¼ j1;i :
~cD11
~c11
4488 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502

The main difference with respect to the model in which weak conditions on transverse stress are enforced,
consists of neglecting the interactions between different layers in the transverse direction. Indeed, in this
case, each layer is left free to extend and bend in the transversal direction independently of the other ones.
Global constitutive equations. By substituting the local constitutive equations (56) in the stress resultant
definitions, global constrained constitutive equations in the following form are found
2 3 2 ðNSÞ ðNSÞ
3
ðNSÞ 2 0 3
Nx kNu kNw eNV u
6 7 6 6 ðNSÞ ðNSÞ
7
ðNSÞ 76 00 7
4 Mx 5 ¼ 4 kNw kMw eMV 54 w 5: ð58Þ
q ðNSÞ ðNSÞ ðNSÞ V
eNV eMV qV
Explicit expressions for the constitutive coefficients above are given in Appendix A.2.

3.4. Model with null transverse strain (ND model)

3.4.1. Variational formulation

The condition of null transverse deformation is satisfied when no further hypotheses of transverse stress
distribution are imposed. The corresponding variational formulation of the problem consists of looking for
the stationary set of the Hellinger–Prange–Reissner functional (25) over the space W.
The equilibrium equations, the natural boundary conditions, and the constitutive relations are found by
imposing the variations of (25) in Wu  Wu and WT  WD equal to zero. The only difference with respect
to the models assuming further conditions on transverse normal stress is that the transverse deformations
ðe2;i ; j2;i Þ appearing in the constitutive equations in the form (35) are null in each layer. Hence, the cor-
responding global constitutive relations are given directly by (41) for dy ¼ 0: They are in the form
2 3 2 ðNDÞ ðNDÞ
3
ðNDÞ 2 0 3
Nx kNu kNw eNV u
6 7 6 76 00 7
4 Mx 5 ¼ 6 eMV 7
ðNDÞ ðNDÞ ðNDÞ
4 kNw kMw 54 w 5 ð59Þ
q ðNDÞ ðNDÞ ðNDÞ V
eNV eMV qV
with the constitutive constants given in Appendix A.2 (Eqs. (A.11)).

4. Results and comments

In this section, we discuss the main differences between the proposed NSR model and the standard NS
and ND models by focusing on the following points:

(1) the comparison of the estimates of the beam mechanical, electrical, and coupling constitutive coeffi-
cients as functions of the thickness ratio between the piezoelectric and elastic layers;
(2) the comparison of the through-the-thickness distribution of the 3D state fields associated with a given
deformation state of the beam model (which is specified by assigning ðu0 ; w00 ; V Þ).

To this end, two particular configurations of piezoelectric laminated beams are considered:

(i) a sandwich three-layered beam composed of a central elastic layer on which two identical piezoelectric
layers are symmetrically bonded (see Fig. 2); in particular, both of the configurations with in-phase and
counter-phase electric connections between the piezoelectric layers are examined (extensional-electric
coupling and flexural-electric coupling, respectively).
(ii) a two-layered beam composed of a piezoelectric and an elastic layer (see Fig. 3).
 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502 4489

e3

hp 3 PZT-5H e2(3)

2
e1 e2 = e2(2)
hb Aluminum

hp 1 PZT-5H e1(1)

Fig. 2. Three-layered sandwich piezoelectric beam: cross-section.

e3
e1(2)
hp 2 PZT-5H

e1 e2 = e2(1)
hb 11 Aluminum

Fig. 3. Two-layered piezoelectric beam: cross-section.

In the following, we assume that the different layers all have the same width a. We denote by hp the
thickness of the piezoelectric layers, by hb the thickness of the elastic layers, and we introduce the thickness
ratio
hb
g¼ : ð60Þ
hp
Elastic layers made of aluminum, and piezoelectric layers made of the piezoelectric ceramic PZT-5H are
considered. The corresponding numerical values of the relevant material properties are reported in Table 1.
The following dimensionless parameters are introduced:
~ ¼ ~c11 ;
b ~ ¼ ~c12 ;
b ^ ¼ ^c11 ;
b
11 E 12 E 11
~c ~c12 ^cE11
s11ffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffi ð61Þ
~e231 ~e231 ^e231 ~cE
~c11 ¼ ; ~
c 12 ¼ ; ^
c 11 ¼ ; ~m ¼ 12 :
~eS33~cE11 ~eS33~cE12 ^eS33^cE11 ~cE11
They represent the stiffness ratios between the elastic and piezoelectric materials (b ~ ;b~ ;b^ ), the dimen-
11 12 11
~
sionless coupling parameters (~c11 ; c12 ), and the piezoelectric Poisson coefficient (~m).

Table 1
Numerical values of the relevant piezoelectric and elastic constitutive coefficients for PZT 5-H and aluminum
s11 (m2 N 1 ) s12 (m2 N 1 ) d31 (m V 1 ) bT33 (F 1
m)
Elastic layer (aluminum) 14.5 · 10 12
)4.78 · 10 12
– –
Piezoelectric layer (PZT-5H) 16.1 · 10 12
)4.57 · 10 12
)320 · 10 12
2.98 · 107
4490 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502

4.1. Constitutive coefficients

We begin by comparing the beam constitutive coefficients relating the generalized beam stress ðNx ; Mx ; qÞ
to the beam generalized deformations ðu0 ; w00 ; V Þ as obtained by the ND, NS and NSR models. In par-
ticular, for specific cross-sectional configurations (the in-phase and counter-phase connected sandwich
beams and the two layered beam), the influence of the different hypotheses on transverse stress on the beam
constitutive behavior as a function of the thickness ratio g is discussed.

4.1.1. Three-layered sandwich beam

In this case (see Fig. 2), the layer configuration is symmetric, beam extensional and flexural modes are
ðNDÞ ðNDÞ
mechanically uncoupled (kNw ¼ kN l ¼ 0), and the beam constitutive equations can be evaluated from
reduced expressions of the type (49). For the ND model, the following constitutive parameters are found
(see expressions in Appendix A.2):
ðNDÞ ~ Þ;
kNu ¼ ~cE11 ahp xg ðb 11 ð62aÞ

ðNDÞ ~cE11 ah3p

kMw ¼ ~ ; ~c Þ;
v ðb ð62bÞ
12 g 11 11
ðNDÞ
eNV ¼ ðx1 þ x2 Þa~e31 ; ð62cÞ
ðNDÞ
eMV ¼ ðx1 x2 Þa~e31 hp ð1 þ gÞ=2; ð62dÞ

ðNDÞ 2a S
qV ¼ ~ ; ð62eÞ
hp 33
where the following functions giving the dependence on the thickness ratio g are defined:
xg ðbÞ ¼ 2 þ gb; vg ðb; cÞ ¼ 2ð1 þ c2 Þ þ 6ð1 þ gÞ2 þ g3 b: ð63Þ
For in-phase connected piezoelectric layers ðx1 ¼ x2 ¼ 1Þ, the potential difference between the electric
terminals of the piezoelectric transducers is coupled only to the beam extensional mode. On the other hand,
for counter-phase connected piezoelectric layers ðx1 ¼ x2 ¼ 1Þ, the piezoelectric coupling involves only
the beam flexural mode.
The NSR and NS models furnish different estimates of the electromechanical constitutive properties of
the layered beam. We detail in Tables 2 and 3 the corresponding expressions for the mechanical extensional
and flexural stiffnesses (kNu and kMw ), coupling coefficients (eNV ; eMV ), and capacitance per unit length qV .
These quantities are reported as ratio to those of the ND model. The cases of in-phase and counter-phase
electrical connections are reported separately. By assuming the numerical values of the constitutive

Table 2
Constitutive coefficients of a sandwich beam with in-phase parallel-connected piezoelectric layers
NS model NSR model
ðNDÞ
kNu =kNu ^ Þ
-g ðb ~ Þ
-2g ðb
11 12
ð1 ~m2 Þ 1 ~m2
~ Þ
-g ðb 2
- ðb ~ Þ
11 g 11

ðNDÞ
~ Þ
-g ðb 12
eNV =eNV 1 ~m 1 ~m
~ Þ
-g ðb 11

2~c211
qv =ðNDÞ
qv 1 þ ~c211 1þ
-g ðb11 Þ

Ratios between the constitutive coefficients of the NS and NSR models and those of the ND model are reported.
 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502 4491

Table 3
Constitutive coefficients of a sandwich beam with counter-phase parallel-connected piezoelectric layers
NS model NSR model
ðNDÞ
kMw =kMw ^ ; ^c Þ
vg ðb ~ ; ~c Þ
v2g ðb
11 11 12 12
ð1 ~m2 Þ 1 ~m2
v ðb~ ; ~c Þ v2 ðb~ ; ~c Þ
g 11 11 g 11 11
~ ; ~c Þ
vg ðb
ðNDÞ 12 12
eMV =eMV 1 ~m 1 ~m
v ðb~ ; ~c Þ
g 11 11

þ 1=gÞ2
6~c211 ð1
qv =ðNDÞ
qv 1 þ ~c211 1þ
~
vg ðb11 ; ~c11 Þ

Ratios between the constitutive coefficients of the NS and NSR models and those of the ND model are reported.

0.90
0.91

(ND)
kNu / k(ND)
Nu

kMw / kMw
0.86
0.90

0.82
0.89
0 5 10 15 20 0 5 10 15 20

0.72 0.72
(ND)

(ND)
eNV /eNV

eMV /eMV

0.70 0.66

0.68
0.60
0 5 10 15 20 0 5 10 15 20

1.8 1.8
(ND)

(ND)
qV

qV

1.4 1.4
qV /

qV /

1.0 1.0
0 5 10 15 20 0 5 10 15 20
η η
(a) (b)

Fig. 4. Beam constitutive coefficients for the in-phase (a) and counter-phase (b) connected three-layered piezoelectric beam as a
function of the thickness ratio g ¼ hb =hp . The continuous lines refer to the NSR model, the dashed lines to the NS model. The
constitutive coefficients are reported as ratios with respect to those estimated by the ND model.

coefficients in Table 1, the same quantities are plotted in Fig. 4 as a function of the thickness ratio g for a
fixed thickness of the piezoelectric layer.
From expressions (62), Tables 2 and 3, and Fig. 4, we draw the following conclusions regarding the
constitutive coefficients estimated by the different models.
Electric capacitance per unit length (eqV ). Both for the in-phase and counter-phase electric connection
between the two piezoelectric layers, the predicted value of eqV strongly depends on the assumed model. In
particular, once the thickness hp of the piezoelectric layers is fixed, eqV is independent of the thickness ratio g
between the piezoelectric and elastic layers in the ND and NS models. In the ND model, eqV is given by the
4492 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502

electric capacitance for null transverse strain, and in the NS model by that for null transverse stress; the
ratio between these two values depends only on the coupling coefficient ~c11 (see Table 2). On the other hand,
in the NSR model, the electric capacitance per unit length varies with g: for the in-phase electric connection
it is equal to the electric capacitance for null transverse strain for thick elastic substrates (g ! 1); it in-
creases with g going towards the capacitance for null transverse stress for very thin elastic substrates
(g ! 0). This behavior has a physical justification: for g ! 1 the very thick elastic substrate behaves as a
rigid element and the piezoelectric layers bonded on it cannot deform in the transverse direction; for g ! 0,
the elastic layer becomes negligible and the two in-phase connected piezoelectric layers can freely deform in
the transverse direction. For g ! 1 the same considerations also hold for the counter-phase electric
connection, while a different phenomenon must be considered for g ! 0. In this case, when the thickness of
the elastic layer is negligible, the two counter-phase connected piezoelectric layers are in the so-called bi-
morph configuration: when an electric potential is applied, one tends to extend, the other to shrink. As a
consequence of the bonding condition, transverse interactions between the two layers arise and transverse
stresses are different from zero also for g ! 0. For this reason, when g ! 0 the electric capacitance per unit
length goes toward a value different from that of the NS model.
Mechanical stiffness (kNu ; kMw ). Both for in-phase and counter-phase electric connections, the mechanical
stiffnesses estimated by the NS and NSR are remarkably different from that given by the ND model. As
evident from relations in Tables 2 and 3, their ratio mainly depends on the value of the Poisson ratio m. This
is a well known effect: while in the NS and NSR models the beam cross section is left free to deform in its
plane, in the ND model it is constrained to be rigid. As a consequence, the ND model overestimates the
extensional and flexural stiffnesses when the beam’s lateral boundary is stress-free. On the contrary, for the
assumed material properties of the different layers, the estimates provided by the NSR and NS models are
similar, although the NS model neglects any interaction between different layers in the transverse direction
(each layer is left free to deform independently from the other ones). This may be the reason why the NS
condition is usually accepted for modelling mechanical laminates.
Coupling coefficient (eNV ; eMV ). For a given thickness of the piezoelectric layers, the coupling coefficients
predicted by ND and NS are independent of the thickness of the elastic layer. On the other hand, the NSR
model accounts for a dependence on the thickness ratio g. Moreover, the ND model sensibly overestimates
(+30%) the coupling coefficient in comparison to the NS and NSR models. To give an illustrative expli-
cation of the underlying phenomena, let us consider the case of the in-phase connected beam and let us write
the eNV coefficient as the induced electric charge per unit length for a given axis extension u0 ¼ S11 , under the
condition of null axis bending and null applied voltage. For a plane-stress state (T33 ¼ 0), we find

q 
eNV ¼ ¼ 2a~e31 ð1 þ S22 =S11 Þ: ð64Þ
S  00
11 w ¼V ¼0
ðNDÞ
In the ND model, since S22 is forced to be zero, the coupling coefficient is given by
ðNDÞ
eNV ¼ 2a~e31 :
On the other hand, in the NS model, since the transversal stress T22 is assumed to be zero, we find
ðNSÞ ðNSÞ
S22 ¼ ~mE S11 ; eNV ¼ 2a~e31 ð1 ~mE Þ;
where ~mE is the in-plane Poisson coefficient of the piezoelectric material for T33 ¼ 0 and E3 ¼ 0. Hence, for
ðNSÞ ðNDÞ
positive Poisson coefficients, eNV < eNV : In the NSR case, the transverse deformations of the piezoelectric
layers are influenced also by the deformation of the elastic layer: if the Poisson ratio ~m of the elastic layer is
greater that the one of the piezoelectric layer (as in the numerical case we considered), then the elastic layer
will induce a stronger shrinking for a given axial extension. Hence, for ~mE < ~m,
ðNSRÞ ðNSÞ ðNSRÞ ðNSÞ
S22 < S22 ; eNV < eNV :
 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502 4493
ðNSRÞ ðNSÞ
On the contrary, if ~m < ~mE then eNV > eNV . The relations above can be verified also by analyzing the ratio
between the analytical expressions furnished in Table 2. Such a difference between the coupling coefficient
of the NSR and NS model is small because exclusively due to the difference in the Poisson ratio of different
layers, which is usually small. In the case of counter-phase connected piezoelectric layers, the underlying
phenomena are essentially the same, but some further complications arise because of the though-the-
thickness linear contributions to the stress and strain which are associated with the beam flexion. As for
the equivalent capacitance, for g ¼ 0 also the coupling coefficients of the in-phase connected beam given by
the NS model and the NSR model coincide. Indeed, in this case the beam is completely equivalent to a
single layer piezoelectric beam and the two conditions (NS) and (NSR) are equivalent. This is not the case
for the counter-phase connection. In this case, when g ¼ 0, the sandwich beam reduces to a bimorph
piezoelectric pair, where transverse interactions between different layers are important.

4.1.2. Two-layered beam

For the asymmetric layer configuration, the potential difference between the electrodes of the piezo-
electric layers is intrinsically coupled both to the beam extension and bending, which, in general, are also
mechanically coupled to each other. We report in the following table the constitutive coefficients appearing
in the unconstrained beam constitutive equations (41), evaluated in a reference frame positioned as in
Fig. 3.
ðNDÞ ~ Þ;
kNu ¼ kNu ¼ ahp~cE11 ð1 þ gb 11

ðNDÞ ð1 þ gÞ
kNw ¼ kNw ¼ ah2p~cE11 ;
2
ðNDÞ 2 ~ Þ;
kMw ¼ kMw ¼ ~cE11 J2 ð3ð1 þ gÞ þ 1 þ ~c11 þ g3 b 11
ðNDÞ
eNV ¼ eNV ¼ a~e31 ;
ðNDÞ
qV ¼ qV ¼ a~S33 =hp ;
ðNDÞ
ð65Þ
eMV ¼ eMV ¼ a~e31z2 ;
!
A1~c12
kN k ¼ A2~cE12 1þ ;
A2~cE12
ð1 þ gÞ
kN l ¼ ah2p~cE12 ;
2
kMl
2 ~ Þ:
¼ ~cE12 J2 ð3ð1 þ gÞ þ 1 þ ~c11 þ g3 b 12

As the explicit evaluation of the constitutive coefficients appearing in Eq. (48) becomes rather cum-
bersome due to the presence of mixed extension and bending stiffness, we limit ourselves to showing some
plots for these coefficients obtained in the case of the material properties specified by Table 1. In Fig. 5, the
ratio between the constitutive coefficients of the NS and NSR models to those of the ND model are re-
ported. In this case, the one related to the mechanical coupling between extension and bending (kNw ) is not
vanishing. Comments similar to those detailed for the three-layered beam hold also in this case, and an
analogous interpretation of the dependence on the thickness ratio between the piezoelectric and the elastic
layer can be drawn.

4.2. Comparison of 3D fields underlying the ND, NS and NSR models

The kinematical state of the three beam models above (ND, NS and NSR) is specified through the fields
(u; w; V ). They determine beam axis extension, bending, and electric potential, respectively. For a given
4494 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502

0.72
0.91
(ND)

(ND)
kNu / kNu

eNV/ e NV
0.70
0.90
0.68

0.89
0 5 10 15 20 0 5 10 15 20

0.72
0.90
(ND)

(ND)
kMw / kMw

eMV/ e MV
0.86 0.66

0.82 0.60
0 5 10 15 20 0 5 10 15 20
1.8
0.918
(ND)

(ND)
eNw / eNw

qV

0.917 1.4
/ qV

0.916
1.0
0 5 10 15 20 0 5 10 15 20
η η
(a) (b)

Fig. 5. Beam constitutive coefficients for two-layered piezoelectric beam as a function of the thickness ratio g ¼ hb =hp . The continuous
lines refer to NSR model, the dashed lines to the NS model. The constitutive coefficients are reported as ratios with respect to those
estimated by the ND model.

physical situation, their axial distribution is determined as a solution of the associated 1D boundary value
problem, consisting of balance equations (32), suitable boundary conditions, and the specific form of the
constitutive relations (i.e. Eqs. (59) for the ND model, Eqs. (58) for the NS model and Eqs. (48) for the
NSR model). Each model associates given beam generalized deformations (u0 ; w00 ; V ) to different distribu-
tion of the three-dimensional state fields (T; D; S; E). Such distributions are determined as follows: (i)
generalized stresses (ra;i ; fa;i ; D3;i ) are found by local constitutive equations in the form (36) or (56); (ii) the
three-dimensional stress state (T; D) is obtained from the latter through Eqs. (11) and (12); (iii) finally, the
3D generalized deformations can be found through 3D constitutive equations in the form (2e). Here and
henceforth, we will refer to the 3D deformations and electric filed found in this way as constitutive induced
deformation and electric field (S; E). Indeed, they are the deformation and electric field determined by the
stress and electric displacement (T; D) through constitutive equations (2e). As a peculiarity of the reduced
model deduced by a mixed variational formulation, in general they differ from the mechanical strain S and
the electric field E compatible with the kinematical hypotheses (9) and (10). The (S; E) fields are those used
to evaluate the internal energy (3).
In the following, the distribution of the 3D state fields associated with the proposed NSR model is
compared to the ones related to the standard NS and ND models. In particular, the field distributions
corresponding to the following beam kinematical states are analyzed:
 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502 4495

• beam extension: ðu0 ! S0 ; w00 ! 0; V ! 0Þ,

• beam bending: ðu0 ! 0; w00 ! S0 =z0 ; V ! 0Þ,
• applied voltage ðu0 ! 0; w00 ! 0; V ! V0 Þ.

The distribution for a generic beam kinematical state (u; w; V ) can be regarded as a linear combination of
these three situations. In order to deal with non-dimensional deformations, stress, electric fields and electric
displacement, the following scaling quantities are introduced: S0 (characteristic mechanical deformation), pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
E E S
z0 ¼ hp (characteristic thickness dimension),
pffiffiffiffiffiffiffiffiffiffiffi T0 ¼ ~c11 S0 (characteristic mechanical stress), E0 ¼ S0 ~c11 =~e33
E ~S
(characteristic electric field), D0 ¼ S0 ~c11 e33 (characteristic electric displacement), and V0 ¼ E0 z0 (charac-
teristic voltage).
The cases of a two-layered beam and a three-layered beam are considered; the thickness ratio between
the piezoelectric and the elastic layers (g ¼ hb =hp ) is fixed to 2. For sake of brevity, for the three-layered
beam only the more interesting configuration of a counter-phase connection between the piezoelectric
layers is considered. In this case, the electric potential across the electrodes is coupled only to beam
bending. We report in Figs. 6 and 7 the distribution of the non-zero component of the three-dimensional
strain, stress, electric field and electric displacement for imposed beam bending ðu0 ! 0; w00 ! S0 =z0 ; V ! 0Þ
and applied voltage ðu0 ! 0; w00 ! 0; V ! V0 Þ, respectively. For the two-layered beam in Fig. 5, extension,
bending, and applied voltage are all coupled, both by mechanical and piezoelectric effects. We report in
Fig. 8–10 the corresponding three-dimensional field distribution for imposed beam extension
ðu0 ! S0 ; w00 ! 0; V ! 0Þ, bending ðu0 ! 0; w00 ! S0 =z0 ; V ! 0Þ, and applied voltage ðu0 ! 0; w00 ! 0;
V ! V0 Þ, respectively. The distributions associated to the ND, NS and NSR models are compared (dashed,

Fig. 6. 3D-field distribution for the three-layered sandwich piezoelectric beam with a counter-phase electric connection and a thickness
ratio g ¼ hb =hp ¼ 2. A mechanical bending is imposed ðu0 ! 0; w00 ! S0 =z0 ; V ! 0Þ.
4496 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502

Fig. 7. 3D-field distribution for a counter-phase connected three-layered sandwich piezoelectric beam with a counter-phase electric
connection and a thickness ratio g ¼ hb =hp ¼ 2. An electric potential is imposed ðu0 ! 0; w00 ! 0; V ! V0 Þ.

dotted and continuous lines, respectively). Components of strain and stress tensors, electric fields and
electric displacement vectors are identified by usual index notation.

4.2.1. Comments
The analysis of Figs. 6–10 highlights some important features of the through-the-thickness field distri-
butions related to the different beam models, giving also a deeper insight into the phenomena evoked when
discussing the beam constitutive coefficients and their dependence on the thickness ratio between the pie-
zoelectric and the elastic layers.

(i) The NSR model accounts for transverse interactions between different layers and transverse deformations
related to cross-section bending and extension. When a potential difference is applied across the electrodes
of the piezoelectric layers, and the axial deformations are imposed to be null, layerwise linear, but dis-
continuous, transverse stresses T22 appear. These stresses conciliate the piezoelectrically induced trans-
verse deformations of piezoelectric layers with the strain state in the elastic layer and the perfect bonding
condition. Moreover, the NSR model accounts for uniform cross-section transverse bending and exten-
sion, as evident from the linear continuous distribution of S22 in Fig. 7 (simple bending of the three-lay-
ered sandwich beam) and in Fig. 10 (combined bending and extension for the two-layered beam).
Similar phenomena are revealed also for imposed axial deformations and null potential difference V :
when a bending deformation w00 is imposed on the three-layered beam, the NSR model account for a
transverse sectional bending and layerwise linear transverse stresses T22 (see S22 and T22 plots in Fig.
6). Similar considerations hold also for imposed extension and bending of the two layered beam (see
S22 and T22 plots in Figs. 8 and 9).
 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502 4497

Fig. 8. 3D-field distribution for a two-layered sandwich piezoelectric beam with a thickness ratio g ¼ hb =hp ¼ 2. A mechanical
extension is imposed ðu0 ! S0 ; w00 ! 0; V ! 0Þ.

(ii) The NS model neglects transverse interactions between different layers and does not take into account the
bonding condition in the transverse direction. By condition (14), each layer is left free to deform along
the width, independently of the others (see the dotted lines in the S22 plots of Figs. 6–10). For example,
when a potential difference is applied, while the central elastic layer does not deform along the width,
the upper piezoelectric layer extends, the lower one shrinks uniformly, and the bonding condition is
not satisfied at the interface (see the S22 plot in Fig. 7).
(iii) The ND model neglects transverse sectional deformations and transverse stress transfer between different
layers. In each layer transverse stresses are determined by the condition of transverse rigidity, indepen-
dently of the state of other layers (see the dashed lines of the T22 plots in Figs. 6–10). For example,
when a potential difference is applied to the three-layered beam (Fig. 7), transverse T22 stresses arise
in order to assure null transverse deformation; these stresses are present only in the piezoelectric layers
and they do not influence the state of the central elastic layer.
(iv) All the three models coherently leave the cross section free to deform in the thickness direction.By assum-
ing vanishing normal stress T33 (all three models assume a stress distribution in the form (11)), a sec-
tional distension along the thickness direction is implicitly accounted for (see the layerwise linear
distribution of S33 in Figs. 7–10). When no forces are prescribed on the beam lateral boundary, each
layer either shrinks or extends along the thickness independently of all the others. This component of
the deformation is naturally induced both by the standard Poisson effect and piezoelectric coupling
(so-called 33 coupling).
(v) All three models coherently account for a linear contribution to the electric field in bent piezoelectric lay-
ers. Since by hypothesis (12) the electric displacement has been assumed to be layerwise constant, the
4498 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502

Fig. 9. 3D-field distribution for a two-layered sandwich piezoelectric beam with a thickness ratio g ¼ hb =hp ¼ 2: A mechanical bending
is imposed ðu0 ! 0; w00 ! S0 =z0 ; V ! 0Þ.

linear part of the mechanical deformation constitutively induces a linear contribution to the electric
field (quadratic contribution to the electric potential). Hence, when the piezoelectric layers are bent,
linear contributions to the electric field are present (see E3 plots in Figs. 6–10). As a related effect, when
the potential difference applied across the electrodes is null (condition V ¼ 0), piezoelectric layers are
characterized by two different mechanical stiffnesses (see local constitutive equations (36) and (56)): the
constant contribution to the mechanical deformation is characterized by the mechanical stiffness for a
null electric field (superscript E) and the linear contribution by the mechanical stiffness for a null elec-
tric displacement (superscript D). Indeed, for V ¼ 0, the constant part of the electric potential is null,
while the linear part is arbitrary; at the same time, the linear part of the electric displacement is null
because of hypothesis (12), while the constant one is arbitrary. Hence, the constant part of the mechan-
ical deformation is affected by condition E3 ¼ 0 and the linear one by condition D3 ¼ 0.
(vi) The intensity of the electric displacement for a given electric potential and axial deformation depends on
the transverse stress and strain state. It is at a maximum when transverse stresses are null and piezoelec-
tric layers can freely extend or shrink along the width; it is at a minimum when transverse deformations
are constrained to be null; it attains a value between these extremes when neither transverse stresses nor
transverse deformations are null. As an example, in the D3 plots of Figs. 7 and 10, the value of the
electric displacement predicted by the NSR model is in between those of the NS and ND models, being
null neither transverse stress nor transverse strain. This effect is directly related to the difference in the
estimates of the equivalent piezoelectric capacitance per unit length eqV in the beam constitutive equa-
tions: the NSR model, by catching non-trivial transverse interactions between different layers, fur-
nishes a more realistic estimate of the electric displacement intensity for a given electric potential
and axial deformation (and consequently of eqV ).
 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502 4499

Fig. 10. 3D-field distribution for a two-layered sandwich piezoelectric beam with a thickness ratio g ¼ hb =hp ¼ 2: An electric potential
is imposed ðu0 ! 0; w00 ! 0; V ! V0 Þ.

5. Conclusions

In this paper, models of layered piezoelectric beams accounting for direct and inverse electromechanical
coupling were discussed. An equivalent-single-layer Euler–Bernoulli model was deduced from a three-
dimensional continuum model by a mixed variational principle. The estimates obtained by simplest
modelling approaches are refined without adding additional degrees-of-freedom (such as shear deforma-
tions, multilayer kinematics, higher-order electric field distribution, etc.). The model is based on hypotheses
(9)–(12) and weak condition (13) on transverse stresses (null transverse stress resultants). Its main pecu-
liarities reside in the following points: (i) to account for transverse interactions between different layers and
sectional bending and extension in the transverse direction (thanks to the non-standard weak condition (13)
on transverse stress); (ii) to include as state variables all the electric and mechanical state fields and to
account for full two-fold piezoelectric coupling between them; (iii) to account for bending of the piezo-
electric layers by including coherent distributions of electric field and displacement (by imposing in the
mixed variational formulation a layerwise constant distribution of the electric displacement through
hypothesis (H4)).
The attention was focused on the influence of hypotheses on transverse normal stress and strain on the
deduction of the beam constitutive coefficients from three-dimensional material and geometric properties
and on the reconstruction of the through-the-thickness distribution of three-dimensional state fields. It was
shown that, differently from models assuming either null transverse deformations or null transverse stress,
the model assuming null transverse stress resultants takes into account transverse interactions between
different layers and the related non-trivial transverse strain and stress distribution. As main benefit, such a
4500 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502

model correctly estimates the beam constitutive coefficients relating axial deformations and potential dif-
ference across piezoelectric layers to axial stress resultants and stored charge. In particular, one of the main
weaknesses shown by standard one-dimensional models is overcome by improving the estimate of equiv-
alent piezoelectric capacitance. Indeed, this property is strongly dependent on the assumed distribution of
transverse strain and stress.
The theoretical analysis developed in the present paper was motivated by some experimental falsifica-
tions of the ND and NS models. More precisely, a preliminary set of measurements which we performed
indicated that the piezoelectric capacitance of layered beams were not correctly predicted. As a natural
extension of the present work, future investigations will involve the experimental validation of the presented
NSR model and its estimates of electromechanical constitutive properties. To this end, experimental setups
aimed at the characterization of the electrical and mechanical properties of layered piezoelectric beams
must be performed. On the other hand, the most prominent theoretical refinement to be included in the
model is the description of non-perfect bonding between different layers and the main phenomena related to
the presence of the bonding layer.

Appendix A. Constitutive coefficients

A.1. Local constitutive equations

The constitutive coefficients appearing in the form (36) of the local constitutive equations are expressed
as a function of ðsE11 ; sE12 ; d31 ; bT33 Þ as follows. They correspond to the constitutive coefficients of the 3D
piezoelectric constitutive equations in the so-called S–E form for a plane-stress condition in the e1 –e2 plane
and a uniaxial electric displacement along e3 .
sE11
~cE11 ¼ 2 2
; ðA:1Þ
ðsE11 Þ ðsE12 Þ

sE12
~cE12 ¼ 2 2
; ðA:2Þ
ðsE11 Þ ðsE12 Þ

d31
~e31 ¼ ; ðA:3Þ
sE11 þ sE12
2
1 d31
~S33 ¼ 2 ; ðA:4Þ
bT33 sE11 þ sE12

~cD cE11 þ ~e231 =~eS33 ;

11 ¼ ~ ðA:5Þ

~cD cE12 þ ~e231 =~eS33 :

12 ¼ ~ ðA:6Þ

On the other hand, in Eqs. (56) the constitutive coefficients for an uniaxial stress along e1 and an uniaxial
electric displacement along e3 appear. The are given by
1
^cE11 ¼ ~cE11 ð1 ð~cE12 =~cE11 Þ2 Þ ¼ ; ðA:7Þ
sE11

d31
^e31 ¼ ~e31 ð1 ~cE12 =~cE11 Þ ¼ ; ðA:8Þ
sE11
 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502 4501
!
~e231 1 2
d31
^S33 ¼ ~S33 1 þ ¼ ; ðA:9Þ
E ~S
~c11 33 bT33 E
s11

2
^cD cD
11 ¼ ~11 ð1 ð~cD cD
12 =~ cE11 þ ^e231 =^S33 :
11 Þ Þ ¼ ^ ðA:10Þ
For the elastic layers, the coefficients (~c11 ; ~c12 ) and ^c11 are found by expressions analog to (A.1) and (A.2)
where piezoelectric compliances must be substituted by the corresponding elastic compliances.

A.2. Global constitutive equations

The constitutive coefficients appearing in Eq. (41) are found by substituting the local constitutive
equations (36) and the relations of kinematical compatibility in the definition of the stress resultants (33)
and (37). The following expressions are found:
ðNDÞ
X X
kNu ¼ kNu ¼ Ai~cE11 þ Ai~c11 ;
i2Ip i2Ie
ðNDÞ
X X
kNw ¼ kNw ¼ Aizi~cE11 þ Aizi~c11 ;
i2Ip i2Ie
ðNDÞ
X X
kMw ¼ kMw ¼ ðAiz2i ~cE11 þ ~cD
11 Ji Þ þ ~c11 ðAiz2i þ Ji Þ;
i2Ip i2Ie
ðNDÞ
X X
eNV ¼ eNV ¼ Ai xi~e31 =hi ¼ ai xi~e31 ;
i2Ip i2Ip
ðNDÞ
X
eMV ¼ eMV ¼ xi~e31zi ai ; ðA:11Þ
i2Ip
ðNDÞ
X
qV ¼ qV ¼ ai~S33 =hi ;
i2Ip
X X
kN k ¼ Ai~cE12 þ Ai~c12 ;
i2Ip i2Ie
X X
kN l ¼ Aizi~cE12 þ Aizi~c12 ;
i2Ip i2Ie
X X
kMl ¼ ðAiz2i ~cE12 þ Ji~cD
12 Þ þ ~c12 ðAiz2i þ Ji Þ:
i2Ip i2Ie

Similarly, the constitutive constants appearing in Eq. (58) are found by the local constitutive equations (56).
They are given by
ðNSÞ
X X
kNu ¼ Ai^cE11 þ Ai^c11 ;
i2Ip i2Ie
ðNSÞ
X X
kNw ¼ zi Ai^cE11 þ zi Ai^c11 ;
i2Ip i2Ie
ðNSÞ
X X
kMw ¼ ð^cD z2i Ai^cE11 Þ þ
11 Ji þ  ^c11 ðJi þ z2i Ai Þ;
i2I i2Ie
ðNSÞ
Xp ðA:12Þ
eMV ¼ zi Ai^e31 xi =hi ;
i2Ip
ðNSÞ
X
eNV ¼ Ai^e31 xi =hi ;
i2Ip
ðNSÞ
X
qV ¼ ai =hi^S33 :
i2Ip
4502 C. Maurini et al. / International Journal of Solids and Structures 41 (2004) 4473–4502

References

Alzahrani, B.A., Alghamdi, A.A.A., 2003. Review of the mechanics of materials models for one-dimensional surface-bonded
piezoelectric actuators. Smart Mater. Struct. 12, N1–N4.
Anderson, E.H., Hagood, N.W, 1994. Simultaneous piezoelectric sensing/actuator: analysis and application to controlled structures.
J. Sound Vibrat. 174 (5), 617–639.
Batra, R.C., Vidoli, S., 2002. Higher-order piezoelectric plate theory derived from a three-dimensional variational principle. AIAA J.
40 (1), 91–104.
Beckert, W., Pfundtner, G., 2002. Analysis of the deformational behaviour of a bimorph configuration with piezoelectric actuation.
Smart Mater. Struct. 11, 599–609.
Chee, C.Y.K., Tong, L.S., Grant, P., 1998. Review on the modelling of piezoelectric sensors and actuators incorporated in intelligent
structures. J. Intell. Mater. Syst. Struct. 9 (1), 3–19.
Chopra, I., 2002. Review of state of art of smart structures and integrated systems. AIAA J. 40 (11), 2145–2187.
Crawley, E.F., de Luis, J., 1987. Use of piezoelectric actuators as elements of intelligent structures. AIAA J. 25, 1375–1385.
Crawley, E.F., Anderson, E.H., 1990. Detailed models of piezoceramic actuation of beams. J. Intell. Mater. Syst. Struct. 1, 12–25.
Eringen, A., Maugin, G.A., 1990. Electrodynamics of Continua I. Springer, New York.
Gopinathan, S.V., Varadan, V.V., Varadan, V.K., 2000. A review and critique of theories for piezoelectric laminates. Smart Mater.
Struct. 9, 24–48.
Hagood, N.W., Von Flotow, A., 1991. Damping of structural vibrations with piezoelectric materials and passive electrical networks.
J. Sound Vibrat. 146, 143–168.
He, J.-H., 2000. Generalized Hellinger–Reissner principle. J. Appl. Mech. 67, 326–331.
Hong, C.H., Chopra, I., 1999. Modeling and validation of induced strain actuation of composite coupled plates. AIAA J. 37, 372–377.
Ikeda, T., 1990. Fundamentals of Piezoelectricity. Oxford University Press, New York.
Park, C., Walz, C., Chopra, I., 1996. Bending and torsion models of beams with induced-strain actuators. Smart Mater. Struct. 5,
98–113.
Rao, S.S., Sunar, M., 1994. Piezoelectricity and its use in disturbance sensing and control of flexible structures: a survey. Appl. Mech.
Rev. 47, 113–123.
Saravanos, D.A., Heyliger, P.R., 1999. Mechanics and computational models for laminated piezoelectric beams, plates, and shells.
Appl. Mech. Rev. 52, 305–320.
Sze, K.Y., Yang, X.-M., Fan, H., in press. Electric assumptions for piezoelectric laminate analysis. Int. J. Solids Struct. 41, 2363–2382.
Tani, J., Takagi, T., Qui, J., 1998. Intelligent material systems: application of functional materials. Appl. Mech. Rev. 51, 505–521.
Teresi, L., Tiero, A., 1997. On variational approaches to plate model. Meccanica 32, 143–156.
Washizu, K., 1982. Variational methods in Elasticity and Plasticity. Pergamon Press, New York.
Yang, J.S., Batra, R.C., 1995. Mixed variational principles in nonlinear piezoelectricity. Int. J. Nonlinear Mech. 30 (5), 719–726.