Professional Documents
Culture Documents
Metadados
Data de Publicao
Resumo
2009-09-29
Due to the wide use of anisotropic solid structures, the study of
anisotropic piezoelectricity becomes increasingly important. The purpose
of this thesis is to extend the research of Alvarez-Dios & Viao [1993,
1996, 2003] and Trabucho & Viano [1996] to anisotropic elastic beams,
and present asymptotic models for anisotropic piezoelectric beams. These
models are derived in a rigorous way from a three-dimensional problem
for an anisotropic piezoelectric beam under an electric potential app...
O estudo do fenmeno piezoelctrico aumentou consideravelmente
devido crescente utilizao dos materiais piezoelctricos anisotrpicos
em aplicaes de engenharia. A presente tese tem como principais
objectivos, a determinao e justificao assimpttica de modelos de
vigas piezoelctricas anisotrpicas, para, desta forma, generalizar a
teoria de vigas elsticas propostas por lvarez-Dios & Viao [1993,
1996, 2003] e Trabucho & Viano [1996]. Os modelos desenvolvidos so
baseados na teoria da...
Tipo
doctoralThesis
http://repositorium.sdum.uminho.pt
Universidade do Minho
Escola de Cincias
UMinho | 2009
Julho de 2009
Universidade do Minho
Escola de Cincias
Tese de Doutoramento
Cincias
Trabalho efectuado sob a orientao do
Professor Doutor Juan Manuel Viao Rey
Universidade de Santiago de Compostela
Co-Orientador
Doutor Jorge Manuel da Silva Figueiredo
Universidade do Minho
Julho de 2009
Agradecimentos
iii
Abstract
of Alvarez-Dios
& Via
no [1993, 1996, 2003] and Trabucho & Via
no [1996] to anisotropic
elastic beams, and present asymptotic models for anisotropic piezoelectric beams.
These models are derived in a rigorous way from a three-dimensional problem for an
anisotropic piezoelectric beam under an electric potential applied on one of two types of
boundary: first type when the electric potentials are applied in both extremities of the
beam and second type when the voltage acts on its lateral surface. In order to analyse
the asymptotic behaviour of the solution of the three-dimensional piezoelectricity problem
for a beam which has diameter of the cross section much smaller than the length, the
asymptotic method was implemented by considering the diameter as a small parameter.
We start by briefly introducing the three-dimensional equations that describe the
linear piezoelectricity theory: a coupled system of mechanical and electrical equilibrium
boundary value partial differential equations; weak and strong formulations. After, we
expand the solution as an expansion with respect to the small parameter and we obtain
a sequence of problems which allow us characterize some terms of the development. As
a result of this analysis, we establish the following models for piezoelectric anisotropic
beams:
The one-dimensional beam theory from the three-dimensional problem for an
anisotropic piezoelectric beam of class 2 under an applied electric potential on both
ends is derived in Chapter 3. The characterization of the second-order displacements
is an essential step to achieve this model and to demonstrate the strong convergence.
v
electric potential acting on its lateral surface is derived in Chapter 4. The weak
convergence result is also discussed in this chapter for anisotropic piezoelectric beam
of the subclass 6mm, where it is concluded that the displacement vector field and
the electric potential weakly converge towards the leading terms of the displacement
- electric potential expansions.
A zeroth-order model for a transversely isotropic - 6mm symmetry class - piezoelectric shallow arch submitted to an electric potential at the both ends was determined
in Chapter 5 by the asymptotic expansion displacement - stress - electric potential
- electric displacement.
vi
Resumo
viii
material piezoelectrico pertence `a classe 2, e em resposta a um potencial electrico
induzido numa regiao do tipo I. A caracterizacao do segundo termo dos deslocamentos e essencial para a determinacao deste modelo e para a demonstracao da
convergencia forte. Estes calculos encontram-se detalhados no Captulo 3.
No Captulo 4, e justificado um modelo para uma viga piezoelectrica anisotropica
de classe 2 submetida a um potencial electrico numa area do tipo II. Demonstra-se
ainda que, se a viga e feita de uma material piezoelectrico pertencente `a classe de
simetria 6mm, entao o vector dos deslocamentos e o potencial electrico convergem
fracamente para os primeiros termos dos respectivos desenvolvimentos.
Um modelo de primeira ordem para vigas, debilmente curvas, constitudas por
material piezoelectrico transversalmente isotropico - classe 6mm - e com potencial
electrico aplicado nos extremos da viga, e apresentado no Captulo 5. A formulacao
mista, juntamente com o desenvolvimento do tipo deslocamento-potencial electricotensao-deslocamento electrico, mostrou ser uma solucao eficaz na obtencao do respectivo modelo e na demonstracao do resultado de convergencia forte.
Table of Contents
List of Figures
xiii
xv
1 Introduction
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Piezoelectricity theory
17
Piezoelectric material . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.1.2
Crystal classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.3
Matrix notation . . . . . . . . . . . . . . . . . . . . . . . . 21
2.1.3.2
2.1.3.3
2.2.2
2.2.3
Maxwells equations
2.2.4
Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 26
. . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.2
. . 28
2.3.2.1
Primal formulation
2.3.2.2
2.3.2.3
Mixed formulation . . . . . . . . . . . . . . . . . . . . . . 32
ix
. . . . . . . . . . . . . . . . . . . . . 28
Table of Contents
2.3.2.4
37
3.1.2
3.1.2.2
Mixed formulation . . . . . . . . . . . . . . . . . . . . . . 45
Weak convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.4.1
3.4.0.1
3.4.0.2
Strong convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
81
. . . . . . . . . . . . . 82
Characterization of u0 and u2 . . . . . . . . . . . . . . . . 93
4.3.0.3
Characterization of u1 and 0 . . . . . . . . . . . . . . . . 97
4.3.0.4
Model for a beam belonging to the class 6mm of piezoelectric crystals . . . . . . . . . . . . . . . . . . . . . . . . 105
4.4.2
4.4.2.2
Table of Contents
xi
173
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
Some perfectives for future research . . . . . . . . . . . . . . . . . . 174
178
List of Figures
4
9
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.1 The reference configuration of the shallow arch described in curvilinear and
Cartesian coordinates, and in a fixed domain. . . . . . . . . . . . . . . . . 128
xiii
Conventions
The notations follow in general these guidelines:
1. Tensors and vectors are denoted by boldface.
2. The standard Einstein summation is adopted, and Latin indices range over the
values 1,2,3 and Greek indices over the values 1,2.
3. The symbol h designates a parameter that is > 0 and approaches zero.
4. Superscripts h will be dropped when equal to one, so that
= 1 ,
= 1,
x = x1 = (x1 , x2 , x3 ),
eD = 1eD ,
...
Notations
We collect here the definitions of some of the most frequently used symbols; others are
defined where they first appear in the text.
Piezoelectricity
e:
E:
the electric vector field in which the components are represented by (Ei ) .
xv
xvi
List of Symbols
D:
U:
H:
C:
P :
F :
:
C
P :
, :
Lames constants.
Y, :
deformation map.
id :
x:
u(x) :
ij
.
ekl
ij
.
Ek
Di
.
Ej
n:
0 :
0 :
Common notations
x, x:
u, u:
, :
= :
the closure of .
i :
i = /xi , x = (xi ) .
C(), C() :
List of Symbols
Ls () :
xvii
the Lebesgue space of s-integrable functions u with the norm
Z
1/s
s
kukLs () =
|u| dx
, 1 s < .
L () :
| | , k k0, :
H m () :
u(x)v(x)dx.
H m () = {u L2 () : u L2 (), || = 1 + + n m}.
which is a Banach space for the norm
1/2
Z
P
2
kukH m () = kukm, =
| u(x)| dx
.
||m
H01 (),
1/2
(N ) :
D() :
D () :
div :
Pd
i=1
ui (x)
xi
for a vector
v2
v1
D ()2 , curl = x
,
, curl u = x
x
.
x1
2
1
2
vn v :
.
x2
x3 x3
x1 x1
x2
P
2
the Laplace operator, u(x) = di=1 u(x)
for x Rd .
2x
i
vn v :
dD (eD ) :
curl (N = 3):
V0 () :
Vw,0() :
xviii
List of Symbols
0 () :
2 () :
() :
X0 () :
X1 () :
X2 () :
dx :
p = (pi ) :
f = (fi ) :
g = (gi ) :
Piezoelectric beam
:
d :
0 :
L:
= (0, L).
h = h (0, L).
h0 = h {0} :
hL = h {L} :
h0 :
hdD :
heD :
dN (eN ) :
uh = (uhi ) :
h :
h :
an extension of h0 in H 1 (h ).
List of Symbols
= .
xix
h :
d :
dh :
ph = (phi ) :
f h = (fih ) :
g h = (gih ) :
nh = (nhi ) :
eh = (ehij ) :
E h = (Eih ) :
h = (ijh ) :
h
h
the stress tensor, typically given by ijh = Cijkl
ehkl Pkij
Ekh .
D h = (Dih ) :
h h
the electric displacement, typically given by Dkh = Pkij
eij + hkl Elh .
u(h) = (uhi ) :
ui (h) : R :
(h) : R :
A() :
the area of .
Z
I =
x2 d.
I :
X , Y , Z, L :
X , Y , Z :
J, Y :
J:
xx
List of Symbols
bhij (xh ) :
oh :
n
h :
i h
:
n o
h
x
h = (
xhi ) :
u
h = (
uhi ) :
h :
h ( (0, L)) :
h = h (0 ) :
h = h (L ) :
L
h = h (D ) :
D
h = h (D ) :
eD
h.
the displacement vector at a point x
h in the body
An arbitrary admissible displacement vector is denoted v
h .
h.
the electric potential vector at a point x
h in the body
n o
h .
the lateral surface of the set
the left end of beam.
h .
dN
h =
h
h .
eN
jh vih (
x ):
h = (
e
ehij ) :
h = (E h ) :
E
ph = (
phi ) :
f = (fih ) :
g
h = (
gih ) :
eh = (ehi ) :
= (Eih ) :
E
h = (
ijh ) :
h
h h
the stress tensor, typically given by ijh = Cijkl
ehkl Pkij
Ek .
ih ) :
h = (D
D
h = P h ehij + h E h .
the electric displacement, typically given by D
k
kij
kl l
n
h = (
nhi ) :
(t , n , b ) :
h (x3 ), h (x3 ) :
e = (eij ) :
List of Symbols
E = (Ek ) :
xxi
the electric field vector.
Chapter
Introduction
1.1
Background
Piezoelectric Solid (PS) has been defined in the literature as a solid that produce an
electric field when deformed and, conversely, undergo deformation when subjected to an
electric field [see e.g. Hwu et al., 2004]. The electromechanical coupling is the key physical
property of the piezoelectric materials and was investigated by Crawley & Luis [1987] via
experiments and analytic models. These piezoelectric materials are widely used as sensors
and actuators in many important applications, such as structural, aerospace, robotic, various medical purposes, etc. With the increasing successful technological applications of
piezoelectric actuators and sensors, piezoelectric solids have attracted considerable attention from scientific researchers, in order to understand the basic phenomena responsible
for their particular properties.
In mathematical terms the electromechanical coupling is described through a set of
partial differential equations of second order, in which the displacement components and
electric potential are taken as the essential unknowns (see e.g. Tiersten [1969], Jackson
[1975], Nye [1985], Ikeda [1990]). Due to the complexity of the coupling effect between
mechanical and electrical fields it is very difficult, if not impossible, to find an exact solution to the piezoelectric problem. Many simplifying assumptions, such as simplification
of geometry and restrictions on the piezoelectric materials electromechanical behavior,
have been made to obtain approximative solutions. Several approximations have been
developed by Robbins & Reddy [1991], Crawley & Luis [1987], Bisegna & Maceri [1996b],
Saravanos & Heyliger [1999], Vidoli & Batra [2000] and Vidoli et al. [2000] for modelling
of piezoelectric structures.
One way to generate models for piezoelectric problems in thin piezoelectric solids, in
which one or more dimensionals are small, compared to the others, is to reduce the original
problem to a new one, in a lower dimensional space, where the small dimensions disappear.
1
Chapter 1. Introduction
Latter papers by Collard & Miara [2003] and Sabu [2002, 2003] published twodimensional models for boundary value problems considering piezoelectric shells. A twodimensional nonlinear shell model was proposed by Collard & Miara [2003], where they
derived two-dimensional membrane equations and flexural models written on the middle
surface of the shell, and verified that the coupling between the electric limit displacement field and the limit electric potential inherent to piezoelectricity appears only in the
membrane model. The two-dimensional theories for the vibrations of thin piezoelectric
flexural and shallow shells were proposed by Sabu [2002, 2003], by considering the limit,
as the thickness goes to zero, the eigenvalue problem for piezoelectric respective flexural
and shallow shells.
Pioneering work for modelling of thin linearly isotropic piezoelectric beams was performed by Via
no et al. [2005a,b], where a purely mechanical Bernoulli-Navier beam theory emerges. Recent models of the phenomena piezoelectric beams were described by
Figueiredo & Leal [2006], Viriyasrisuwattana et al. [2007] and Weller & Licht [2008]. A new
mathematical model for a linearly nonhomogeneous anisotropic thin beam was proposed
by Figueiredo & Leal [2005], which is a system of coupled equations, with generalized
Bernoulli-Navier equations and reduced Maxwell-Gauss equations but it was not possible to prove the uniqueness of the solution for the limit equations. A one-dimensional
asymptotic models for a linearly piezoelectric slender beams can be found in Weller &
Licht [2008].
In this work, we take advantage of previous experience, varying the applied boundary
electric conditions as suggested by Weller & Licht [2002] and using the asymptotic method
mentioned above, and we find various models for linearly piezoelectric beams and shallow
arches as the diameter of its cross section approaches zero. We complete the asymptotic
analysis, proving weak and strong convergence results and characterizing completely the
limit models, closing several open questions form above mentioned papers.
1.2
Chapter 1. Introduction
conditions. We assume that the beam is clamped along the boundary hdD , and subjected
to an electric potential h0 on heD . The beam is also subjected to applied body forces of
density (fih ) : h R3 acting inside h , and on surface forces (gih ) on hdN . We denote a
typical point in h by xh = (xh , x3 ).
h
h
We denote by C h = (Cijkl
), P h = (Pikl
) and h = hij , respectively, the elastic tensor
field, the piezoelectric tensor field, and the dielectric tensor field that characterize the
material.
In the framework of small deformations and linear piezoelectricity, the three-dimensional static equations for the piezoelectric solid h are the following: Find a displacement
vector field uh : h R3 and an electric potential h : h R such that (see Chapter
2 with h = 1)
h
h
ijh uh , h = Cijkl
ehkl (uh ) Pkij
Ekh (h ) in h ,
and
h h
Dkh uh , h = Pkij
eij (uh ) + hkj Ej (h )
jh ijh uh , h = fih ,
ijh uh , h nhj = gih
uhi = 0,
in h ,
in hdN ,
on hdD ,
in h ,
ih Dih uh , h = 0,
Dkh uh , h nhk = 0
h = h0
in h ,
in heN ,
on heD ,
where ehij (uh ) = 1/2(ih uhj + jh uhi ) denote the components of the linearized strain tensor,
Eih (h ) is the components of the static electric vector field defined by Eih (h ) = i h ,
and (nhi ) is the unit outer normal vector along h . This problem can be put in a variational
form, which consists in finding (uh , h ) V0h (h ) H 1 (h ) such that h = h0 on heD
(b) Cross-section.
h h h
h
h
Cijkl ekl (u ) Pmij
Em
h ehij (v h )dxh
h
Pmij
ehij (uh )
hmi
Eih
h
Em
dx =
h
dN
gih vih dh ,
for all
h
V0h
:=
V0h (h )
n
o
1 h 3 h
h
h
= v H ( ) : v = 0 on dD ,
h h0 := h0 (h ) = h H 1 (h ) : h = 0 on heD .
h h h
h
h
Cijkl ekl (u ) Pmij
Em
h ehij (v h )dxh
h
Pmij
ehij (uh ) + hmi Eih h
fih vih
h
Pkij
Ek (h )ehij (v h )
h
Em
h dxh
hkl El (h )Ekh ( h )
dx +
h
dN
gih vih dh .
The existence and uniqueness of solution (uh , h ) and (uh , h ) follows from the classical
Korns and Poincares inequalities, by assuming some regularity to the data (see Proposition 1 and Corollary 1).
In Chapter 3, we consider that the beam is made of a piezoelectric crystal material of
class 2 which the components of the elastic, piezoelectric and dielectric material satisfy
(cf. (3.4))
h
h
Ciijj
6= 0, Cklkl
6= 0,
for k 6= l,
h
h
h
P3
6= 0, P3
6= 0, P333
6= 0, h 6= 0,
h33 6= 0.
We also assume that the beam is weakly clamped (clamped in mean) along at one end h0 ,
i.e. the displacement field uh being such that
Z
h
0
uhi
d = 0,
h
0
which we denote by hui = 0 on h0 , the other end being controlled by surface forces
Chapter 1. Introduction
h h h
h
h
Cijkl ekl (u ) Pmij
Em
h ehij (v h )dxh
h
Z
h h h
h h h
+
Pmij eij (u ) + hmi Eih h Em
dx
h
Z
Z
Z
h
=
fih vih dxh +
gih vih dh +
phi vih dh , (v h , h ) Vw,0
h0 ,
h
h
N
h
L
h
where Vw,0
and h2 are defined by (see Section 3.3)
n
o
3
h
h
Vw,0
: = Vw,0
(h ) = v H 1 () : hvi = 0 on dD ,
h2 : = h2 (h ) = h H 1 (h ) : h h h0 .
(1.1)
(1.2)
The existence and unicity of solution is also obtained by classical results of elliptic variational equations.
=
In Section 3.2 we define an equivalent problem, but now posed over set
(0, L),
and with
which is independent of h. We denote by x = (x1 , x2 , x3 ) a generic point in ,
we associate the point xh
h through the bijection
each point x ,
xh = (hx1 , hx2 , x3 )
h.
h : x = (x1 , x2 , x3 )
(1.3)
(xh ) = h2 (h)(x),
D (h)(x) = h1 Dh (xh ),
(h)(x)
= h (xh ),
h
3 (h)(x) = h1 3
(xh ),
h
33 (h)(x) = 33
(xh ),
h
h
Cijkl
(xh ) = Cijkl (x), Pkij
(xh ) = Pkij (x), hij (xh ) = ij (x), for all xh h ,
h (xh3 ) = (x
3 ),
where (x
3) =
1
(L
L
+ .
It is found in this fashion that the scaled unknown (u(h), (h)) satisfies a variational
problem of the form (Proposition 2 in Section 3.2)
(v, ) V ,
0,w
0
where the bilinear forms a4 , a2 and a0 and the linear forms l are defined in Chapter 3
and are independent of h.
The specific form of this variational problem suggests that we use the method of formal
asymptotic expansion, i.e., we let
u(h) = u0 + h2 u2 + h.o.t.,
(h) = 0 + h2 2 + h.o.t,
ui V0,w ,
0 ,
p 0 , with p 1,
(i) The vector field u0 is a Bernoulli-Navier displacement field, i.e.: the functions u0
depend only on x3 and displacement u03 takes the form u03 (x1 , x2 , x3 ) = 3 (x3 ) x (x3 ),
with 3 H 1 (0, L) satisfying 3 (0) = 0, and H 2 (0, L) satisfying (0) = (0) = 0.
(ii) The scalar function 0 (x1 , x2 , x3 ) = z3 (x3 ) with z3 H 1 (0, L), satisfying z3 (0) =
00 and z3 (L) = 0L .
(iii) The vector field (i , z3 ) solves a one-dimensional coupled boundary value problem
Chapter 1. Introduction
C33
A() 3 P3 A() z3 = F3 in (0, L),
in (0, L),
3 A() z3 P3 A() 3 = 0
(4)
C33
I = F + M in (0, L),
(0) = 0, (0) = 0,
C I (L) = M L ,
C33
I (L) = FL M (L),
33
C33
= C3333 C33 [C C33 + C12 C1233 ] = Ac33 33 ,
fi d +
M3 =
FiL
ML
gi d,
M =
(x2 f1 x1 f2 ) d +
x f3 d +
x g3 d,
N
(x2 g1 x1 g2 ) d,
pi d,
x p3 d,
M3L
(x2 p1 x1 p2 ) d.
that satisfy:
h
h
C333
= C3
= 0,
h
h
h
P
= P33
= P33
= 0,
h3 = 0,
and that the beam is weakly clamped at the extremities h0 hL , i.e. hvi = 0 on h0 hL .
As shown in Figure 1.2, we assume that the beam has electric potential h0 on heD =
h
h
h
eD
(0, L) with eD
h and measure(eD
) > 0. Due to the force loading and electric
potential, the pair (uh , h ) is derived from h = h + h with (uh , h ) satisfying the
xh = (hx1 , hx2 , x3 )
h,
h : x = (x1 , x2 , x3 )
(1.4)
and we take a new scaling to electrical part. In other words, we use the appropriate
scalings
(h)(x)
= h1 h (xh ),
D (h)(x) = Dh (xh ),
h (xh ) = h(h)(x),
(h)
+ (h).
h
eD
h0
h
h
10
Chapter 1. Introduction
In this way, the scaled principle virtual work reads: Find (u(h), (h))
X 0 such that
Z
ij (h)eij (v) dx +
Dk (h)Ek () dx =
fi vi dx +
gi vi d,
(v, ) X 0 ,
where (h) = (h)(u(h), (h)) and D(h) = D(h)(u(h), (h)) are given by
(h) = h4 C e (u(h)) + h2 C33 e33 (u(h)) h1 P3 E3 ((h)),
3 (h) = 2h2 C33 e3 (u(h)) h1 P3 E ((h)),
33 (h) = h2 C33 e (u(h)) + C3333 e33 (u(h)) hP333 E3 ((h)),
D (h) = 2h1 P3 e3 (u(h)) + E ((h)),
D3 (h) = h1 P3 e (u(h)) + hP333 e33 (u(h)) + h2 33 E3 ((h)),
and X 0 = V0 0 or X 0 = V0,w 0 . Inserting the developments mentioned above into
the previous problem results in a set of variational equations that must be satisfied for
all h > 0 and consequently the terms at the successive powers of h must be zero. This
procedure allows to show that the negative terms can be cancelled, and consequently to
get the characterization of the terms u0 (Theorem 10), u1 (Theorem 11) and u2 (Theorem
14) of the development of u(h). Summarizing, we show that:
(i) The first term in the asymptotic expansion of the scaled displacement field
0
Z L
Z L
Z L
Y I dx3 =
F dx3
M dx3 , for all H02 (0, L),
0
0
0
1
H0 (0, L),
3
Z L
Z L
Y A()33 dx3 =
F3 3 dx3 ,
for all 3 H01 (0, L),
0
where Y =
det C
,
det M det N
(ii) The leading term in asymptotic expansion of the scaled electric potential is determined by the pair (r, 0 ) where
1
(r, )(x1 , x2 ) =
L
0
11
Z
Z
=
P3 0 d 0 d, for all (, ) T.
The spaces Q(), S(), the function w and the torsion constant J > 0 are defined in
Section 4.3.0.3. In particular, we have that, for a beam of class 6mm material, the electric
potential satisfies the following Laplaces equation,
0 S() such that a.e. x3 [0, L],
Z
Z
P14
P14
0
P14
+ 11 d =
P14
+ 11 0 d,
C44
C44
12
Chapter 1. Introduction
where h (x3 ) are given functions verifying h C 3 [0, L], and n and b are the normal
and the binormal vectors of the Frenet trihedron (t , n, b ) associated to the curve C h
h =
h
defined by h (see Section 5.1). The shallow arch is clamped on the portion
0
dD
h
h
electrical potential 0 (
x3 ) on eD is imposed. Then, the body undergoes a mechanical
h
h
h
displacement field u
= (
uh ) :
R3 and an electrical potential h :
R satisfying
i
Z
h
h
h
h
h
h
k ij d
Cijkl
kl + Pkij D
x +
+
h) X
h
for all (
h, d
1
Z
Z
h
h
h h
ij eij (
v ) d
x +
h
Z
h
h
h
h
h
h
l dk d
Pkij
ij + kl D
x =
dhk Ekh (h ) d
xh ,
h E h (h ) d
D
xh =
k k
+
ijh ehij (
uh ) d
xh
fih vih d
xh +
h
gih vih d
h,
phi vih d
h
for all (
v h , h ) X
0,w
h
h
where Cijkl
, Pkij
and hij are the elastic, piezoelectric and dielectric tensor field that cha-
kl
+ ij D
j
Let = (0, L). We define the scaled displacement field u(h), the scaled electric
potential (h), the scaled stress tensor (h) and the scaled electric displacement D(h)
h } ,
by, for all x
h = h h (x) {
u (h)(x) = h uh (
xh ),
u3 (h)(x) = uh3 (
xh ),
(h)(x)
= h (xh ),
h
h
h
(h)(x) = h2
(
xh ), 3 (h)(x) = h1
3
(
xh ), 33 (h)(x) =
33
(
xh ),
h
h (
D (h)(x) = h1 D
x ),
h
h (
D3 (h)(x) = D
3 x ).
13
f3h (
xh ) = f3 (x),
gh (
xh ) = h2 g (x),
g3h (
xh ) = h g3 (x),
ph (
xh ) = h2 p (x),
ph3 (
xh ) = h p3 (x),
h } ,
for all x
h = h (x) {
and
h ,
for all x
h = h (x)
N
h ,
for all x
h = h (x)
L
h (
xh3 ) = (x
3 ),
h (
xh3 ) = h (x3 ) for all xh3 [0, L].
As we referred in the last chapter, the constants of the material satisfy the following
conditions:
h
Cijkl
(xh ) = Cijkl (x),
h
h } .
for all x
h = h (x) {
Pkij
(xh ) = Pkij (x),
h h
ij (x ) = ij (x).
As a consequence of these scalings and assumptions, the scaled unknowns satisfy the
mixed scaled problem, described in Section 5.3,
( , d) X 1 ,
(v, ) X 0,w ,
2 = 2 () = H 1 () : 0 ,
14
Chapter 1. Introduction
we find that the leading term satisfy, ( 0 , D 0 ), ((u0 , 0 ) X 1 X 2,w
0
2
0
2
a
(
,
D
),
(
,
d)
=
a
(
,
D
),
(
,
d)
H,0
H,2
a#
(h, ) ( 2 , D2 ), ( , d) ,
( , d) X 1 ,
b ( 0 , D 0 ), (v, ) = l (v, ),
(v, ) X 0,w .
H
H
In Section 5.5.2.1 we show that the cancelation of the factors of hq , 4 q 1,
implies that the formal expansion of the tensor and electric displacement does not contain
any negative powers of h, and consequently we derive an expression for the components
u2 of the displacement u2 , taking into account that u0 and 0 belong respectively to the
spaces VBN
and + 3 defined by
VBN
= {v : R3 : v (x1 , x2 , x3 ) = (x3 ), V02 (0, L),
and
v3 (x1 , x2 , x3 ) = 3 (x3 ) b (x1 , x2 , x3 ) (x3 ), 3 V01 (0, L) ,
+ 3 = + H 1 () : (x1 , x2 , x3 ) = z(x3 ), z H01 (0, L) ,
b = / (1 )2 + (2 )2
b1 (x1 , x2 , x3 )
b2 (x1 , x2 , x3 )
1/2
b21 + b22 = 1,
b1 (x3 ) b2 (x3 )
b2 (x3 )
b1 (x3 )
x1
x2
15
and then lead to show that the leading terms can be fully identified solving a onedimensional problem. In Section 5.4 is proved that there is a constant C such that
|| := E3 ()
3
0,
C kk1, ,
3 ,
and therefore we also obtain the existence and uniqueness of the solution of the previous
Z L
Z L
I
C33 dx3 + A ()
C33 3 + + P3 q3 dx3
0
0
=
(F + M ) dx3 M (L) (L) + FL (L) + ML (L),
V02 (0, L),
0
Z L
A
()
C
+
P
q
33
3
3 3 3 dx3
0
Z L
A
()
P
q
z3 dx3
3
3
3
3
Z L
=
F3 3 dx3 + F3L3 (L),
(3 , z3 ) V01 (0, L) H01 (0, L).
0
where the admissible spaces V01 (0, L), V02 (0, L) and + H 1 (0, L) defined by
V01 (0, L) = H 1 (0, L) : (0) = 0 ,
V02 (0, L) = H 2 (0, L) : (0) = (0) = 0 ,
+ H 1 (0, L) : = z H 1 (0, L) : z H01 (0, L) ,
C33
= 33 Ad33 ,
P3 = P333 Ad33 ,
3 = C3333 Ad33 ,
1
.
Ad33 =
C3333 33 + P333 P333
Chapter
Piezoelectricity theory
For a better description of the piezoelectrical phenomena we discuss in this chapter some
relevant theoretical aspects. In recent years a vast literature has flourished describing
the piezoelectric effect [see e.g. Ikeda, 1990; Nye, 1985; Taylor et al., 1985; Royer &
Dieulesaint, 2000].
Section 2.1.1 begins with a short description of the piezoelectric phenomena and a
review of the constitutive equations that describe the piezoelectrics response, based on
thermodynamical principles. Simplifications of the coupling matrices due to crystal symmetries are also presented in this section. The field equations governing linear piezoelectric
solids are given in Section 2.2. Section 2.3.2 gives the primal and mixed variational formulation for electromechanical boundary value problem as well as the existence, uniqueness
and regularity results for the weak solutions.
2.1
2.1.1
Material laws
Piezoelectric material
A piezoelectric material has the ability to produce an electric field when subjected to
mechanical stress [see e.g. Rovenski et al., 2007]. This phenomenon is known as the
direct effect (the word comes from the Greek piezein which means to press ) and
was first discovered in 1880 by Jacques and Pierre Curie. The converse (or indirect)
piezoelectric effect, by which certain materials deform when subjected to an electric field,
was mathematically deduced from fundamental thermodynamical principles in 1881 by
Lippmann and confirmed experimentally by the Curie brothers.
17
18
2.1.2
Crystal classes
The two piezoelectric effects are specific to the piezoelectrics crystals structure that
lacks a center of symmetry [Boor & Tabaka, 2008]. Of the total 32 crystal classes, 21 are
noncentrosymmetric and with the exception of one class, all of these are piezoelectric. The
32 classes are divided into seven groups: triclinic, monoclinic, orthorhombic, tetragonal,
trigonal, hexagonal and cubic. There is a relationship between these groups and the elastic
nature of the material: triclinic represents anisotropic material, orthorhombic represents
orthotropic material and cubic are usually isotropic materials.
Using Maugin and Hermanns notation, the 20 piezo-classes are: triclinic class 1;
monoclinic classes 2 and m; orthorhombic classes 222 and mm2; tetragonal classes 4, 4,
422, 4mm, and 42m; trigonal classes 3, 32, and 3m; hexagonal classes 6, 6, 622, 6mm,
and 62m; and cubic classes 23 and 43m.
19
applied force, and therefore, the crystal may possess polarization or net electric charge.
More detail about piezoelectric material equations can be found in Nye [1985], Ikeda
[1990] and Royer & Dieulesaint [2000].
2.1.3
For piezoelectric materials the constitutive equations, which relate the strain tensor e
and the electric field vector E to the stress tensor and the electric displacement D,
are derived from thermodynamic potentials and they are described in detail in several
texts [see e.g. Maugin, 1988; Tiersten, 1969; Mechkour, 2004; Royer & Dieulesaint, 2000].
Following these works we can observe that the constitutive equations can assume different
forms depending on the state field. We take the strain tensor components eij and the
electric field components Ek as independent variables; that is to say, the state of the
crystal, and in particular the stress tensor components ij and the electric displacement
components Dk , are determined when the quantities eij and Ek are given. Accordingly,
we may write
ij
Dk
Dk
ij
ekl +
Ek ,
Dk =
eij +
El ,
(2.1)
ij =
ekl
Ek
eij
El
which are known by piezoelectric constitutive relationship and were first given by Voigt.
One way is to define a function
H(e, E) = U ED,
(2.2)
where U represents the internal energy density. From the first law of thermodynamics for
piezoelectric continuum we have
dU = ij deij + Ek dDk .
(2.3)
To obtain U with respect to the variables eij and Ek , we build the first derivative of
(2.2) and apply (2.3)
U
Ei
Di
eij
Ei
H
=
Di Ei
= ij
Di
.
t
t
t
t
t
t
(2.4)
(2.5)
20
H
,
eij
Di =
H
.
Ei
(2.6)
(2.7)
ij
kl
2H
2H
=
= Cijkl = Cklij =
=
,
eij ekl
ekl
eij
ekl eij
(2.8)
Di
Dj
2H
2H
=
= ij = ji =
=
.
Ei Ej
Ej
Ei
Ej Ei
(2.9)
In a similar way,
Dk = Pkij eij + kj Ej ,
(2.10)
1 i, j, k 3,
where C = (Cijkl ) is the forth-order elasticity tensor, P = (Pkij ) is the third-order piezoelectric tensor and = (ij ) is the second-order dielectric tensor.
When the matrices in equations (2.10) are denoted by single letters we finally have
the following compact expression:
= Ce P E,
D = P e + E.
Now, substituting (2.10) into equation (2.6), and integrating both equations of (2.6)
with respect to mechanical and electrical strains, respectively, yields,
1
H(e) = Cijklekl eij Pkij Ek eij
2
(2.11)
and
1
H(E) = Pkij Ek eij ij Ei Ej .
2
Combining both conservative fields, the electric enthalpy reads
1
1
H(e, E) = Cijkl ekl eij Pkij Ek eij ij Ei Ej .
2
2
(2.12)
(2.13)
21
Other symmetries are implied by the symmetry of the strain tensor, such as
(Hc21 )
(2.14)
For stable materials, both C and are positive definite, i.e., there exists a constant c1 > 0
(Hc22 )
Cijklij kl c1
3
X
ij di dj c1
(ij ) ,
i,j=1
3
X
(di )2 ,
(2.15)
i=1
2.1.3.1
Matrix notation
A=
P112
P212
P122
P222
P131
P231
P132
P232
P133
P233
11
12
12
22
13
23
P311
P312
P322
P331
P332
P333
13
23
33
is a matrix of coefficients on the right-side of equations (2.10). Consequently, the behaviour law (2.10) reads:
=A
e
E
A=
C P T
P
(2.16)
The simplifications introduced by (2.8)-(2.9) and (2.14) mean that the corresponding
matrix A has at most 45 independent coefficients as follows:
11
12 13
22 23 ,
sym.
33
(2.17)
22
C=
C2323 C2333
C3333
(2.18)
P
T
C
(2.19)
where
= ( + P C 1 P T )1 ,
=
= C 1 (I P T
C 1 ).
P C 1 and C
P
P
(2.20)
The blocks of the constitutive matrix A1 that characterize the constitutive equation
P
,
are of the form
(2.19) has similar expressions to the matrix A, i.e., the blocks C,
(2.17)-(2.18).
The effect of increasing crystal symmetry and the choice of reference axes allows us to
reduce the number of independent components needed to specify completely the properties
of the crystal.
In this research work we will consider, at most, the monoclinic crystal structural class 2. This means that it can be anything up to crystal symmetry class 2, including the
most popular piezoceramics and piezopolymers - the Lead Zirconate Titanate (PZT) and
the Polyvinylidene Fluoride (PVDF).
In next sections, we give a brief summary of the properties of the material that crystalize in the monoclinic crystal system (class 2 and class m) and in the class 6mm of the
hexagonal crystal system.
2.1.3.2
As already mentioned, there are 3 crystallographic classes in the monoclinic group: 2/m,
m and 2. The class 2 and the class m for which the corresponding elasticity tensor has
at most thirteen non-null components [see e.g. Royer & Dieulesaint, 2000]
C1111 , C1112 , C1122 , C1133 , C1212 , C1222 , C1233 , C2222 , C2213 , C1313 , C1323 , C2323 , C3333 ,
23
2.1.3.3
A transversely isotropic structures of hexagonal symmetry - 6mm symmetric class is characterized by ten non-zero independent matrix elements consisting of 5 independent elastic constants, 3 independent piezoelectric constants and 2 independent dielectric
constants. For these materials, the reduced matrix form of the above constitutive relationships can now be written as:
C=
C11
C13
C16
C11 C13
2
0
C11
0
0
0
0
0
C16
C44
0
C44
sym.
,
0
C66
0 0 0 P14 0
0
11
0
0
P = 0 0 0
0 P14 0 , =
11 0 ,
P31 0 P31 0
0 P36
sym.
33
where the non-zero components are given by the following relations
C11 = C1111 = C2222 ,
C44 = C1313 = C2323 ,
(2.21)
(2.22)
(2.23)
24
2.2
The mechanical behavior of any material is governed by certain physical laws, which relate
stress, strain, electric fields and electric displacement together. In any points, including
points on the boundary, it must satisfy three basic equations which are Cauchys equations
of motion, kinematics equations and constitutive equations (defined before).
2.2.1
We consider a piezoelectric body continuum occupying a region in a stress-free configuration (in the absence of any electric field or mechanical load), henceforth called the
reference configuration of the body. The goal of this section is to describe the deformation
in response to given forces.
Let = id + u : x 7 (x + u(x)) R3 be a standard C 1 -deformation, with
u = u(x) denoting the mechanical displacement at x . Let us denote the electric
potential at point x as (x). We let x = (xi ) denote a generic point in the set
. The associated generalized deformations are the linearized strain e and the electric
field E, which are expressed as a function of u and through the following equations of
kinematical compatibility:
e (u) := (eij (u)) =
E () := (i ) ,
1
(i uj + j ui ) ,
2
(2.24)
(static),
(2.25)
where i = /xi , x = (xi ) . The expression (2.25) will be deduced in Section 2.2.3.
2.2.2
One the basic principles of mechanics is the balance of momentum. In the presence of a
body force f , this law takes the form (static case)
div = f ,
(2.26)
i ij = fj .
(2.27)
or, in componentwise,
2.2.3
25
Maxwells equations
In this section, we deduce the main Maxwells equations for the electric field variables in
the absence of magnetic fields, free currents and electric charges.
The Maxwells equations of electromagnetism are written as
curl H = t D + J,
(2.28)
curl E = t B,
(2.29)
div B = 0,
(2.30)
div D = 0,
(2.31)
where E denotes the electric field as before, D the electric displacement, H the magnetic
field, B the magnetic induction and J the current density. The magnetic filed and the
magnetic induction are related by
B = 0 (H + M ),
where M is the magnetic field (will be neglected) and 0 is the dielectric impermeability.
The electric field and the electric displacement are related by
D = 0 E + P ,
where P is the electric polarization and 0 is the (symmetric positive definite) permittivity
tensor. From equation (2.30) we deduce the existence of A such that
B = curlA.
The previous equation together with (2.29) imply that E +
derives of one scalar electrical potential , and therefore
E =
A
t
A
t
(2.32)
26
(2.33)
in which the electric field is represented by the negative gradient of the electric potential
. We also assume that the body is a perfect dielectric, i.e., non charge electric in (2.31).
Thus, the only relevant Maxwell equation is
div D = 0.
(2.34)
Combining all the above, we have the so-called field equations for linear piezoelectric
problems:
E = ,
e = 21 ( u + ( u)T ),
2.2.4
= Ce P E,
D = P e + E,
div = f ,
div D = 0.
(2.35)
Boundary conditions
System (2.35) is not a well-posted problem unless we provide it with appropriate boundary
conditions.
Let be a region occupied by a piezoelectric body. The set is assumed to be an
open bounded subset of R3 with a Lipschitz continuous boundary. On the surface of
the solid, mechanical and electrical boundary conditions are applied. We consider that
the surface = is composed by
= dN dD ,
dN dD = ,
eN eD = ,
27
and
ij nj = gi
on dN ,
(2.36)
= 0
on eD and
Dk nk = q
on eN .
In our case, we also assume that q = 0. Note that the lower subscripts eN and eD in
eN and eD refer to electric (e), Neumann (N) and Dirichelet (D) boundary conditions,
respectively, while the lower subscripts dN and dD in dN and dD refer to displacement
(Neumann and Dirichelet) boundary conditions.
eD
g
dD
dN
g
n
0
eN
Figure 2.2: Mechanical boundary conditions Figure 2.3: Electrical boundary conditions
2.3
In this section, we derive the Boundary Value Problem (BVP) for the piezoelectric problem
described before and its variational formulation.
2.3.1
Combining equations (2.26), (2.34) with previous mechanical and electrical boundary
conditions (2.36), we arrive at the BVP for piezoelectricity:
R3 and the scalar
BVP 1 In a domain R3 , find the displacement field u :
R satisfying
:
j ij (u, ) = fi in ,
ij (u, ) nj = gi on dN ,
ui
= 0 on dD ,
(2.37)
28
k Dk (u, ) = 0 in ,
Dk (u, ) nk = 0 on eN ,
= 0 on eD .
(2.38)
Hypotheses 1 Throughout this work, we are going to make the following standard assumptions on the data:
,
Cijkl L
f [L2 ()]3 ,
2.3.2
,
Pkij L
,
ij L
0 H 1/2 (eD ).
(2.39)
(2.40)
There are two different ways to formulate the variational formulation of the BVP 1,
defined in previous section: the primal variational principle and the mixed variational
principle. We begin by introducing the appropriate spaces and then derive a variational
(weak) formulation of our system of partial differential equations.
2.3.2.1
Primal formulation
Let H 1 () be the usual Sobolev space of functions, whose generalized derivatives of order
at most 1 are squared integrable, that is, they belong to L2 (). Let the essential spaces
for the piezoelectric problem be given by
n
o
3
V0 := V0 () = v H 1 () : v = 0 on dD ,
0 := 0 () = H 1 () : = 0 on eD .
(2.41)
(2.42)
(2.43)
29
||v||V0 = |e(v)|0, =
||||0 = ||||H 1 () ,
3
X
|eij (v)|20,
i,j=1
!1/2
v V0 (),
0 ().
(2.44)
(2.45)
Let v V0 be the test vector function. Take the scalar product of the first equation of
the system (2.37) with the test vector function v V0 , and integrate it over , one have
Z
ij
(u, )vi dx =
xj
fi vi dx.
(2.46)
Apply the generalized divergence theorem to the left hand side of (2.46), and get
Z
vi
dx =
vi ij nj d
vi ij nj d + ij
xj
dN
dD
fi vi dx.
(2.47)
Since v V0 , which implies v = 0 on dD , the second term on the right hand side is
zero. Then using the Newmann mechanical boundary condition, the previous equation
becomes
Z
Z
Z
vi
vi gi d + ij
dx =
fi vi dx.
(2.48)
xj
dN
1
= ij
2
vi
vj
+
xj
xi
= ij eij (v),
(2.49)
ij eij (v)dx =
gi vi d +
dN
fi vi dx.
(2.50)
fi vi dx +
gi vi d.
(2.51)
dN
Analogously, let 0 . Multiplying both sides of the first equation of (2.38) and
integrating it over a domain it leads to
Z
30
Combining the Greens formula with Neumanns electric boundary conditions (2.38).
Z
(Di ni )d
eN
Di i dx =
Di i dx,
(2.52)
Adding equations (2.51) and (2.52), we obtain the variational formulations to the problem
(2.37)-(2.38) [see Haenel, 2000]:
(
where
a((u, ) , (v, )) =
(v, ) V0 0 ,
(2.53)
(2.54)
l (v, ) =
fi vi dx +
gi vi d.
dN
Now we give an alternative variational formulation which derives from the nonhomogeneous condition (2.36).
From hypothesis 0 H 1/2 (eD ), there exists an extension of 0 in H 1 (), i.e. a
function H 1 () such that = 0 on eD .
We define
= 0 ,
(2.55)
Dk (u, ) = Dk (u, )
+ kl El (),
(2.56)
k Dk (u, )
= k (kl El ())
,
(2.57)
= 0 on eD ,
(2.58)
31
and
ij (u, )
nj = gi + Pkij Ek ()n
j
Dk (u, )
nk = kl El ()n
k
on dN ,
(2.59)
on eN .
(2.60)
We first consider the equilibrium equations (2.57). Multiplication of both sides of the
equations (2.57) by respectively test functions v V0 and 0 and integration over a
domain lead to
[j ij (u, )
vi + k Dk (u, )
] dx
ij (u, )
eij (v) dx
Dk (u,)
Ek () dx
ij (u, )n
j vi d +
dN
Dk (u, )n
k d
eN
(2.61)
Pkij Ek ()v
i nj d +
dN
kl El ()n
k d
eN
Substituting the stress tensor and the electric displacement by expressions (2.10) and
using (2.59)-(2.60) we deduce that [see e.g. Haenel, 2000; Mechkour, 2004] the solution
(u, ) of (2.53)-(2.54) is derived from = + with (u, )
satisfying:
V0 0 such that
Find (u, )
a((u, ),
(u, )) = l2 (v, )
(2.62)
(v, ) V0 0
where the bilinear form a(, ) was defined in (2.54) and the linear form l2 () reads
l2 (v, ) =
dN
gi vi d.
(2.63)
32
(v, ) V0 2 .
(2.64)
Let us first introduce some results to the linear and bilinear forms (see Haenel [2000] and
Mechkour [2004]), which allow us to guarantee the existence and uniqueness of solution to
the problem (2.53)-(2.54) by applying the Lax-Milgrams Lemma (see Viriyasrisuwattana
et al. [2007], Weller & Licht [2008]).
3
Lemma 1 Assume that f = (fi ) [L2 ()] and g = (gi ) [L2 (dN )] . Then the linear
form l2 : V0 0 R defined in (2.63) is continuous.
Lemma 2 Assume that meas (dD ) > 0, meas (eD ) > 0. Then the bilinear form a(, )
Proposition 1 Assume that meas (dD ) > 0, meas (eD ) > 0, f = (fi ) [L2 ()] ,
3
g = (gi ) [L2 (dN )] . Then (2.62)-(2.63) has a unique solution (u, )
V 0 0 .
Corollary 1 Assume the hypotheses of Proposition 1 and also 0 H 1/2 (eD ). Then
the variational problem (2.53)-(2.54) has one and only one solution (u, ) V0 2 .
Moreover, = + ,
where (u, )
is the only solution of problem (2.62)-(2.63) and
H 1 () is an extension of 0 .
2.3.2.3
Mixed formulation
X 1 := X 1 () = [L
()]9s
L (),
k(, )kX1
X 0 := X 0 () = V0 0 ,
k(, )kX0
X 2 := X 2 () = V0 2 ,
k(, )kX2
h
i1/2
2
2
,
= | |L2 () + | |L2 ()
h
i1/2
2
2
,
= || ||V0 + || ||H 1 ()
h
i1/2
2
2
.
= || ||V0 + || ||H 1 ()
33
To deduce the mixed variational formulation of BVP 1, we write the constitutive law
for piezoelectric material in the inverse formulation as follows (cf. (2.19)-(2.20))
eij = Cijkl kl + Pkij Dk ,
Ei = Pikl kl + ik Dk .
(2.65)
From properties of coerciveness and symmetry of tensors Cijkl and ij , and symmetry
of Pkij , the following properties are derived:
(Hm
ij = ji , Pmij = Pmji .
1 ) Cijkl = Cklij = Cjikl ,
3
33
(Hm
,
2 ) There exists c > 0 such that, for any d = (di ) R , = (ij ) R
Cijkl ij kl c
3
X
(ij ) , ij didj c
i,j=1
3
X
(di )2 .
(2.66)
i=1
The constitutive equation (2.65) is now considered. The weak form of this equation,
obtained by multiplying both sides of the equation (2.65) by a test function ( , d) X 1
Cijkl kl + Pkij Dk ij dx +
eij (u)ij dx
Pkij ij + kl Dl dk dx
Ek ()dk dx = 0.
where
Cijkl kl + Pkij Dk ij dx
Z
+
Pkij ij + kl Dl dk dx
Z
Z
bH (( , d), (u, )) =
eij (u)ij dx
Ek ()dk dx
Z
Z
lH (v, ) = fi vi dx
gi vi d,
dN
(2.67)
(2.68)
(2.69)
34
where
l1,H ( , d) =
( , d) X 1
Ek ()d
k dx,
fi vi dx +
2.3.2.4
(2.70)
(2.71)
gi vi d.
dN
To show the existence and uniqueness of solution for the above mixed formulation, it is
enough to show that:
( , d) K 0 ,
(2.72)
where
K 0 = {( , d) X 1 : bH (( , d), (v, )) = 0
(v, ) X 0 } .
(2.73)
Poincares and Korns inequalities, we prove the existence of 2 > 0 satisfying the above
inf-sup condition (2.73). The elliptic property (2.72) is obvious since the ellipticity of C
35
and
Z
Cijkl kl ij + kl dl dk dx
c k k2L2 () + kdk2L2 () , ( , d) X 0 .
aH (( , d), ( , d)) =
We are now in a position to apply Babuska-Brezzis Theorem (see e.g. Girault &
Raviart [1986]): there exists a unique pair of functions ((, D), (u, ))
X1 X0
satisfying (2.70)-(2.71).
Chapter
38
3.1
Here, we are going to study the BVP 1, introduced in Section 2.3, for a family of a linearly
piezoelectric beams.
3.1.1
continuous. Then, the area of cross section h is Ah = h2 A and the diameter of order h
is assumed very small when compared with L.
An arbitrary point of h will be denoted by xh = (xh1 , xh2 , xh3 ) and the unit outer
normal vector to the boundary h = h by nh = (nhi ). The coordinate system Oxh1 xh2 xh3
will be assumed a principal system of inertia associated to h , which means that
Z
xh
d =
xh1 xh2 d h = 0.
(3.1)
hL =
h {L} ,
39
xh2
gh on hN
h0
xh1
h,0
0
ph on hL
xh3
h,L
0
Figure 3.1: Schematic representation of the solid domain and their mechanical and electrical boundary conditions.
h
0
uhi
d = 0,
h
0
(3.2)
(3.3)
ii) The surface forces g h are acting on hN and ph on hL . This corresponds to put
hdN = hN hL .
iii) An electric potential h0 is applied on heD = h0 hL . More specifically, we denote
h,L
h
h
h0 h = h,0
=
(,
,
0)
and
= h0 (, , L).
0
0
0 h = 0
0
Remark 2 As we will see, the use of the average clamping condition (3.3) instead of
a strong clamping condition such as uh = 0 on h0 allows to avoid the boundary layer
problem (see Lions [1973]) that arises when a strong clamping condition is used - which
turns out to be related to the fact that in general a strong clamping of the beam is physically
impossible (see e.g. Trabucho & Via
no [1996] and references therein).
h
h
We denote by Cijkl
, Pkij
and hkl the components of elasticity tensor, piezoelectric
tensor and dielectric tensor of the material, where h is made of a monoclinic piezoelectric
h
h
material of class 2, i.e., the components Cijkl
, hkl and Pkij
satisfy the following conditions:
h
h
h
h
h
C3
= C333
= h3 = P
= P33
= P33
= 0.
(3.4)
h
h
The conditions C3
= C333
= 0 reflet that the longitudinal axis Ox3 is a principal directional of piezoelectricity (see Lekhnitskii [1981]), Royer & Dieulesaint [2000])
Therefore the stress tensor h = (ijh ) : h R9s and the electric displacement vector
Dh = (Dih ) : h R3 are related to the linear strain tensor, ehij (uh ) = 12 (ih uhj +jh uhi ), and
40
the gradient of the electric potential, Eih (h ) = ih h , through the following behavior
laws:
h
h
h
h
(uh , h ) = C33
eh33 (uh ) + C
eh (uh ) P3
E3h (h ),
h
h
h
3
(uh , h ) = 2C33
eh3 (uh ) P3
Eh (h ),
h
h
h
h
33
(uh , h ) = C3333
eh33 (uh ) + C33
eh (uh ) P333
E3h (h ),
Dh (uh , h ) = 2P3
eh3 (uh ) + h Eh (h ),
h h h
h
h
D3 (u , ) = P3
eh (uh ) + P333
eh33 (uh ) + h33 E3h (h ),
in h ,
in h ,
in h ,
(3.5)
in h ,
in h .
h
31
(uh , h )
h h h
32 (u , )
h h h
D1 (u , )
D2h (uh , h )
h
h
h
h
2C3131
2C3132
P131
P231
h
h
h
h
2C3232
P132
P232
2C3231
=
h
h
2P131
2P132
h11
h12
h
h
2P231
2P232
h21
h22
eh31 (uh )
h h
e32 (u )
h h
E1 ( )
E2h (h )
(3.6)
and
h
11
(uh , h )
h h h
12 (u , )
h h h
22 (u , )
h h h
33 (u , )
D3h (uh , h )
h
h
h
h
h
C1111
2C1112
C1122
C1133
P311
h
h
h
h
h
C1222
C1233
P312
C1211 2C1212
h
h
h
h
h
C2222
C2233
P322
= C2211 2C2212
h
h
h
h
h
C3311 2C3312
C3322
C3333
P333
h
h
h
h
P311
2P312
P322
P333
h33
eh11 (uh )
h h
e12 (u )
h h
e22 (u ) .
h h
e33 (u )
h
h
E3 ( )
Equivalently,
eh31 (uh )
h h
e32 (u )
h h
E1 ( )
E2h (h )
h
h
h
h
2C3131
2C3132
P131
P231
h
h
h
h
2C3232
P132
P232
2C3231
=
h
h
2P131
2P132
h11
h12
h
h
2P231
2P232
h21
h22
h
31
(uh , h )
h h h
32 (u , )
h h h
D1 (u , )
D2h (uh , h )
(3.7)
41
and
eh11 (uh )
h h
e12 (u )
eh (uh ) =
22
h h
e33 (u )
h
h
E3 ( )
h
h
h
h
h
C1111
2C1112
C1122
C1133
P311
h
C1211
h
2C1212
h
C1222
h
C1233
h
P312
h
h
h
h
h
C2211
2C2212
C2222
C2233
P322
h
h
h
h
h
C3311
2C3312
C3322
C3333
P333
h
P311
h
2P312
h
P322
h
P333
h33
h
11
(uh , h )
h h h
12 (u , )
h (uh , h )
22
h h h
33 (u , )
D3h (uh , h )
in h ,
h
h
Eh = 2P3
3
+ h Dh ,
in h ,
h
h
h
h
h
eh = C33
33
+ C
+ P3
D3h , in h ,
(3.8)
h
h
h
h
eh33 = C3333
33
+ C33
+ P333
D3h , in h ,
h
h
h
h
E3h = P3
P333
33
+ h33 D3h ,
in h ,
h
h
h
h
C3131
C3132
P131
P231
(3.9)
h
h
h
h
2C3131
2C3132
P131
P231
h
h
h
h
h
h
h
h
2C3232
P132
P232
C3231 C3232 P132 P232 2C3231
=
h
h
h
h
2P132
h11
h12
P131
P132
h11
h12 2P131
h
h
h
h
2P231
2P232
h21
h22
P231
P232
h21
h22
h
C1111
h
C1112
h
C1122
h
C1133
h
h
h
h
C1211 C1212
C1222
C1233
C h
h
h
h
2211 C2212 C2222 C2233
h
h
h
h
C3322
C3333
C3311 C3312
h
h
h
h
P311
P312
P322
P333
P311
h
P312
h
P322
=
h
P333
h33
h
C1111
h
2C1112
h
C1122
h
C1133
h
P311
h
h
h
h
h
C1211
2C1212
C1222
C1233
P312
h
h
h
h
h
C2211 2C2212 C2222 C2233 P322
h
h
h
h
h
C3311
2C3312
C3322
C3333
P333
h
h
h
h
P311
2P312
P322
P333
h33
To guarantee the regularity to the data we increase one condition to the Hypothesis 1
42
3.1.2
h
V0,w
Z
1 h 3
h
h
h
v H ( ) : h {0} vi d = 0,
h
= V0,w
(h ) =
Z
h {0}
h0 = h0 (h ) = { h H 1 (h ) : h = 0 on heD },
,
(3.10)
h2 = h2 (h ) = { h H 1 (h ) : h = h0 on heD }.
h
The spaces V0,w
(h ) and h0 (h ) equipped with the norms
3
X
||v ||V0,w
h (h ) =
i,j=1
||ehij (v h )||0,h
|| h ||h0 (h ) = || h ||0,h ,
!1/2
h
v h V0,w
(h ),
h h0 (h ),
ah ((uh , h ), (vh , h )) = lh (v h , h ),
h
(v h , h ) V0,w
h0 ,
(3.11)
where
h
a ((u , ), (v , )) =
h
h
h
[Cijkl
ehkl (uh ) Pmij
Em
(h )]ehij (v h ) dxh
Z
h
+
[Pmij
ehij (uh ) + hmi Eih (h )]Em ( h ) dxh ,
(3.12)
43
and
h
l (v , ) =
fih vih
dx +
h
N
gih vih
d +
h
L
phi vih dh .
(3.13)
Next, we prove that the problem (3.11) has an unique solution. We remark that the
BVP associated with (3.11) is not simple to write. In fact, we obtain:
jh ijh uh , h
ijh uh , h nhj
huhi i
h
h
h
u
,
i3
h
h
h
h
u
,
vi
i3
= fih in h ,
= gih on hN ,
=
on hdD ,
(3.14)
= phi on hL ,
=
h
0
h
v h = (vih ) V0,w
,
h h
h
h
D
u
,
= 0 in h ,
k
k
Dkh uh , h nhk = 0 on heN ,
h
= h0 on heD .
(3.15)
h
The condition (3.14)5 imposes a restriction on the form of 3i
on h0 that cannot be
expressed in a strong way.
h (xh3 ) =
1 h h,L
1
(L xh3 )h,0
x .
0 +
L
L 3 0
(3.16)
To achieve the limit problem we will suppose that the electric potential applied in both
ends of the beam is constant and independent on xh1 and xh2 . Therefore,
h,0
and
0
h,L
are constants,
0
(3.17)
44
To summarize, we have
h2 = h2 () = h + h0 ,
(3.18)
and
h = h + h ,
h h0 .
(3.19)
a ((u , ), (v , )) =
(3.20)
l2h (v h , h ),
(v , )
h
V0,w
h0 ,
where
l2h (v h , h ) = lh (v h , h ) ah ((0, h ), (vh , h ))
Z
Z
Z
h h
h
h h
h
=
fi vi dx +
gi vi d +
h
3.1.2.1
h
N
h
L
phi vih dh
h
P333
E3h (h )eh33 (v h ) dxh
h
P3
E3h (h )eh (v h ) dxh .
(3.21)
h
Theorem 2 The problem (3.20) has an unique solution (uh , h ) V0,w
h0 .
h
Proof. Firstly, we need to prove that lh is a continuous linear form on V0,w
h0 , which
f h [L2 (h )]3 ,
(3.22)
h
Next, we need to prove that ah is a V0,w
h0 -elliptic. This is a consequence of Korns
and Poincares inequalities because of Korns inequality is still valid in V0,w , space that
45
Mixed formulation
Following the same steps done in Section 2.3.2.4 we obtain the mixed formulation of
problem (3.20)-(3.21). Defining
X h1 = [L2 (h )]9s [L2 (h )]3 ,
h
h
X h0,w = V0,w
h0 and X h2,w = V0,w
h2 ,
for all h , dh X h
1
h
(v h , h )
bhH h , D h , v h , h = lH
ahH
Z
h
h
h
(
, d ), ( , d ) =
h
h h
Cijkl
klh + Pkij
dk ijh dxh
+
bhH
(3.24)
for all v h , h X h
0,w
where
Z
h h
h
h
(
, d ), v ,
=
h
lH
(v h , h )
(3.23)
h h
Pkij
ij + hkl dhl dhk dxh ,
ijh
ehij (v h )dxh
fih
vih
dx
h
N
gih
(3.25)
vih dh
h
L
phi vih dh .
(3.26)
(3.27)
3.2
The major geometric feature of a three-dimensional beam is the fact that the largest
cross sectional dimension is very small compared to its length (h L), causing ill-
46
uh , h with respect to h. The technique of change of variable to a fixed domain and
subsequent rescalling of the displacement and electric potential will allow us to derive a
variational problem equivalent to (3.11)-(3.13) or (3.20)-(3.21) where h shows up in an
explicit way in the rescaled equations.
To start, we perform a change of variable to the reference domain = (0, L)
through the following transformation (Figure 3.2)
h : x = (x1 , x2 , x3 )
h.
xh = h (x) = (xh1 , xh2 , xh3 ) = (hx1 , hx2 , x3 )
(3.28)
All notations refereed to domain are obtained from h for h = 1 and this index is
dropped. For example:
= ,
0 = 10 = {0},
N = [0, L].
x d =
x1 x2 d = 0,
(3.29)
that is, the system Ox1 x2 x3 is a principal system of inertia for . In the view of (3.2)
we will represent the boundary condition
Z
vi d = 0,
0
(xj vi xi vj ) d = 0,
vi d = 0,
v d = 0,
x v3 d = 0,
h
h
xh
(3.30)
47
(3.31)
2 := 2 () = + 0 () = { H 1 () : 0 },
(3.32)
kk0 = ||0, ,
and
e(v) = (eij (v)) ,
1/2
k(v, )kX0 ,w = kvk2V0,w + kk20
1
eij (v) = (i vj + j vi ).
2
Remark 5 In what follows we shall make use of the following decomposition of V0,w ()
(cf. Trabucho & Via
no [1996])
V0,w () = W1 () W2 (),
Z
Z
1
W1 () = H () :
=
x = 0 ,
{0}
{0}
Z
Z
1
2
W2 () = b = ( ) H () :
=
(x1 2 x2 1 ) = 0 .
{0}
(3.33)
(3.34)
(3.35)
{0}
all xh = h (x), x :
(3.36)
(3.37)
Similarly, the electric potential h and the test function h in h0 are associated to the
scaled potential (h) and the scaled (test) function (h) in 0 using the following scaling
48
(h)(x) = h (xh ).
(3.38)
h
h
h
(xh ) = h2 (h)(x), 3
(xh ) = h3 (h)(x), 33
(xh ) = 33 (h)(x),
D h (xh ) = hD (h)(x),
D3h (xh )
(3.39)
= D3 (h)(x),
(3.40)
gh (xh ) = h2 g (x), g3h (xh ) = hg3 (x), for all xh = h (x) hN ,
3
L
2. There exists a function H 1 (0, L), independent of h, such that:
(x
3 ) = h (xh3 ),
(3.41)
h
Pkij
(xh ) = Pkij (x),
x ,
(3.43)
where Cijkl , Pkij and ij are independent of the size of the cross section and satisfy
Hypothesis 1.
h
Combining (3.28) with the notations (3.36)-(3.38) we have for all v h V0,w
, h h0 and
49
xh = h (x), x :
e (v(h))(x) = h2 eh (v h )(xh ),
e (v(h))(x) = eh (v h )(xh ),
33
33
(3.44)
(3.45)
Using the scalings defined previously for the displacement vector and for the electric
potential together with the above assumptions, we can reformulate the variational problem
(3.11)-(3.13) into another variational problem posed in the domain independent of h.
We have the following result:
Proposition 2 The scaled pair (u(h), (h)) is the unique solution of the following problem
+a0 ((u(h), (h)), (v, )) = l(v, ), for all (v, ) X 0,w = V0,w 0 ,
(3.46)
(3.47)
a2 ((u, ), (v, )) =
Z
P3 [E3 ()e (v) e (u)E3 ()] dx
2
a0 ((u, ), (v, )) =
(3.48)
33 E3 ()E3 ()dx
(3.49)
50
fi vi dx +
gi vi d +
pi vi d.
(3.50)
The existence and uniqueness of solution of the problem (3.46)-(3.50) is due to LaxMilgrams Lemma since both a and l are X 0,w -continuous and a is X 0,w -elliptic as the
reader can easily check having in mind properties (3.22), conditions (Hc21 )-(Hc22 ) and
(2.39)-(2.40) for C, P , , and that 0 < h 1.
Remark 6 We note that the restrictions (3.4) about the material coefficients Cijkl, Pkij
and ij allow us to avoid the odd powers for h in (3.46). This is a fundamental fact because
it eliminates some coupled effects in the beam and is the basis to order to complete the
following asymptotic analysis. This analysis would be far more complicated if (3.4) was
not satisfied.
Next we write problem (3.46)-(3.50) in an equivalent form that exhibits the following
scaled principle of virtual work and scaled constitutive law.
Proposition 3 Problem (3.46)-(3.50) is formally equivalent to
(u(h), (h)) X 2,w = V0,w 2 ,
Z
Z
Z
Z
ij (h)eij (v) dx + Dk (h)Ek () dx = fi vi dx +
pi vi d,
gi vi d
(3.51)
where the scaled stress tensor (h) = (ij (h)) and the scaled electrical displacement
D(h) = (Dk (h)) are related to (u(h), (h)) by the following scaled constitutive piezoelectric law (compatible with (3.39), (3.44) and (3.45))
(h) := h4 C e (u(h)) + h2 C33 e33 (u(h)) h2 P3 E3 ((h)),
3 (h) := 2h2 C33 e3 (u(h)) h2 P3 E ((h)),
33 (h) := h2 C33 e (u(h)) + C3333 e33 (u(h)) P333 E3 ((h)),
D (h) := 2h2 P3 e3 (u(h)) + h2 E ((h)),
D3 (h) := h2 P3 e (u(h)) + P333 e33 (u(h)) + 33 E3 ((h)).
(3.52)
51
The following mixed formulation of scaled problem is immediately obtained from (3.38)(3.39), (3.44) and (3.51).
The scaled unknown (((h), D(h)) , (u(h), (h))) satisfies the following variational
problem: Find (((h), D(h)) , (u(h), (h))) X 1 X 2,w such that
for all ( , d) X ,
1
(3.53)
(3.54)
C dx,
(3.56)
( , d)) =
aH,2 ((
, d),
C33 33 + P3 d3 dx +
+2
+
( , d)) =
aH,0 ((
, d),
+
P3 d3 dx
2P3 3 + d d dx
(3.57)
C3333 33 + P333 d3 33 dx
P333 33 + 33 d3 d3 dx,
C33 33 dx
2C33 3 + P3 d 3 dx
fi vi dx
dN
gi vi d
(3.58)
pi vi d.
(3.59)
The pair (((h), D(h)) , (u(h), (h))) can also characterized as the unique solution of this
problem thanks to the Babuska-Brezzi condition.
52
3.3
3.3.1
Weak convergence
In this section we prove that the family (u(h), (h))h>0 weakly converge to (u, ) in
[H 1 ()]3 H 1 (), as h 0, and identify the limit variational problem solved by
(u, ). We follow the method introduced by Trabucho & Via
no [1996].
Then the following weak convergence are guaranteed.
Proposition 4 There exists C > 0, independent of h, such that for all 0 < h 1 the
solution (((h), D(h)), (u(h), (h))) of problem (3.53)-(3.59) verifies
|33 (h)|0, C,
h |3 (h)|0, C,
h |D (h)|0, C,
|D3 (h)|0, C,
k(u(h), (h))k
0,w
C,
k(u(h), (h))kX
h2 | (h)|0, C,
(3.60)
(3.61)
2,w
C.
(3.62)
Proof. Let S(h) [L2 ()]9s and T (h) [L2 ()]3 be the following elements:
S33 (h) = 33 (h),
T (h) = hD (h),
S3 (h) = h3 (h),
S (h) = h2 (h),
T3 (h) = D3 (h).
We have
aT ((S(h), T (h)), (S(h), T (h))) := aH,0 ((S(h), T (h)), (S(h), T (h)))
+ aH,2 ((S(h), T (h)), (S(h), T (h))) + aH,4 ((S(h), T (h)), (S(h), T (h)))
= aH,0 (((h), D(h)), ((h), d(h))) + h2 aH,2 (((h), D(h)), ((h), D(h)))
+ h4 aH,4 (((h), D(h)), ((h), D(h))) .
Hence letting ( , d) = ((h), D(h)) in equation (3.53) we obtain
Z
aT ((S(h), T (h)), (S(h), T (h))) = [C3333 S33 (h)S33 (h) + 33 T3 (h)T3 (h)]dx
Z
+ 2 [C33 S (h)S33 (h) + 4C33 S3 (h)S3 (h)]dx
Z
Z
+ T (h)T (h)dx + C S (h)S (h)dx
(3.63)
(3.64)
53
= (h) )
one has
Z
Z
+
pi ui (h)d + D3 (h)E3 ()dx.
gi ui(h)d
(3.65)
(3.66)
on the one hand. On the other hand, from (3.65), we deduce that
aT ((S(h), T (h)), (S(h), T (h)))
C (kf k0, + kgk0,N + kpk0,L ) ku(h)k1, + Ckk
1,(0,L) kD3 (h)k0,
C ku(h)k1, + CkT3 (h)k0, C k(u(h), (h))k
a2
m
(3.67)
X0,w + C.
(3.68)
bH (e(h), E(h)),
(u(h), (h))
= bH (e(h), E(h)),
(u(h), (h) (h))
,
(3.69)
and taking into account the fact that 0 < h 1 we guarantee the existence of a positive
constant C satisfying
| bH ((e(h), E(h)),
(u(h), (h)))|
X 0,w .
(3.70)
Expliciting in (3.70) the bilinear form b(, ) and using the inequalities of Korn and
54
Poincare we have
bH ((e(h), E(h)),
(u(h),(h)))
Ek ((h))E
dx
k ((h))
2
= ke(u(h))k20, + k(h))k
0,
Cku(h)k1, + Ck(h))k
1, Ck(u(h), (h))k
X0,w .
(3.71)
(3.72)
k(u(h), (h))k
X0,w C.
(3.73)
Finally, we have:
k(u(h), (h))kX2,w = k(u(h), (h))
+ (0, (h))k
X2,w
Ck(u(h), (h))k
H 1 (0,L) .
X0,w + Ckk
(3.74)
h3 (h) 3 ,
hD (h) D ,
h2 (h) ,
D3 (h) D3 ,
u(h) u,
in L2 ()
(3.75)
in L2 ()
(3.76)
in V0,w (),
(3.77)
(h) ,
in 2 (),
(3.78)
(h)
= ,
in 0 ().
(3.79)
(3.80)
e3 (u) = e (u) = 0,
(3.81)
55
E3 () = P3 P333 33 + 33 D3 ,
(3.82)
E () = 0,
(3.83)
C + C33 33 + P3 D3 = 0,
(3.84)
2C33 3 + P3 D = 0,
(3.85)
D 2P3 3 = 0.
Z
e (v)d = 0, v = (v1 , v2 , 0) V0,w (),
Z
Z
d = x d = 0,
Z
D3 E3 ()d = 0, H01 (0, L),
(3.86)
(3.87)
(3.88)
(3.89)
Proof. We immediately see from the previous theorem, the existence of a subsequence,
still indexed by h, and there exists an element denoted by (, D, u, ) such that convergences (3.75)-(3.79) hold. Passing equation (3.53) to the limit as h goes to zero, we obtain
the relations (3.80)-(3.83) in L2 (). To derive (3.84) we take 33 = d3 = 0 in the first
equation of the mixed problem (3.53), multiplying by h2 and passing to the limit taking
into account that e3 (u) = e (u) = E () = 0. Now, we choose 33 = = d3 = 0
in (3.53). Multiplying by h1 and passing to the limit we deduce (3.85) and (3.86). To
prove (3.87) we multiply equation (3.51) by h2 and pass to the limit as h goes to zero.
Choosing appropriate test functions vi V0,w in (3.87) we obtain immediately (3.88) (see
Trabucho & Via
no [1996]). The properties (3.89) are obtained from (3.51) choosing v = 0
and passing to the limit for appropriate 0 .
(3.90)
v3 (x1 , x2 , x3 ) = 3 (x3 ) x (x3 ), 3 V01 (0, L) ,
(3.91)
56
where
V01 (0, L) = H 1 (0, L) : (0) = 0 ,
(3.92)
V02 (0, L) = H 2 (0, L) : (0) = (0) = 0 .
(3.93)
Consequently, we have
u(x1 , x2 , x3 ) = (x3 ),
u (x , x , x ) = (x ) x (x ), V 1 (0, L).
3 1
2
3
3 3
3
3
0
(3.94)
Now, we define 3 to be the space of the electric potential satisfying the condition
(3.83), i.e.,
3 := 3 () = H 1 () : E () = 0 .
(3.95)
z3 H 1(0, L).
(3.96)
= 01,0 , (x1 , x2 ) ,
(3.97)
= 1,L
0 , (x1 , x2 ) .
(3.98)
1,0
and 1,L
constants.
0
0
(3.99)
with
z3 H 1 (0, L),
z3 (0) = 01,0 ,
z3 (L) = 1,L
0 .
(3.100)
(3.101)
57
Consequently, we have
+ 30 ()
(3.102)
(x
1 , x2 , x3 ) = (x1 , x2 , x3 ) (x
3 ) = z3 (x3 )
E3 () = 3 = z3 ,
1
1
(L x3 )1,0
x3 1,L
0
0 ,
L
L
(3.103)
(3.104)
E3 ()
= 3 = z3 +
1 1,L
(0 1,0
0 ).
L
(3.105)
Theorem 3 Let us that the beam is made of a class 2 piezoelectric material whose coeffi
cients Ac33 Ac and 133 P3 Ac33 P333 Ac do not depend on x . The limit (u, ) is the
solution of the following problem:
Z
Ac33 33 e33 (u) P333 E3 () e33 (v)dx
Z
+ Ac33 P333 e33 (u) + C3333 E3 () E3 ()dx
Z
Z
Z
=
fi vi dx +
gi vi d +
pi vi d,
(3.106)
(v, ) VBN 30
where
Ac33 =
1
.
Proof. Taking now v VBN () and 30 () in second equation of the mixed problem
(3.53) we obtain
Z
D3 (h) E3 ()dx =
fi vi dx +
33 e33 (v)dx +
D3 E3 ()dx =
fi vi dx +
gi vi d +
N
gi vi dd +
pi vi d,
L
pi vi d,
(3.107)
for all (v, ) VBN 30 . Using (3.80) and (3.82), one has
33 = Ac33 33 e33 (u) P333 E3 () Ac ,
D3 = Ac33 P333 e33 (u) + Ac33 C3333 E3 ()
Ac33 =
1
,
1
P3 Ac33 P333 Ac ,
33
Ac = 33 C33 + P333 P3 .
(3.108)
(3.109)
(3.110)
58
Ac33 (
33 e33 (u) P333 E3 ())e33 (v)dx
Z L Z
Z
c
c
A33 A d 3 dx3 +
0
+
+
=
fi vi dx +
L
0
Z
Ac33 Ac
x d dx3
Z L
Then, if we use conditions (3.88) and the fact that the coefficients are independent of x ,
we get (3.106).
Corollary 3 Any function (u, ) solves (3.106) if and only if it has the form
u (x1 , x2 , x3 ) = (x3 ),
(x1 , x2 , x3 ) = z3 (x3 ),
(3.111)
(3.112)
with V02 (0, L), 3 V01 (0, L) and z3 H 1 (0, L), z3 (0) = 01,0 , z3 (L) = 1,L
0 , and i
and z3 solving the following variational problem (no sum on ):
Z
Z L Z
Z
L
I A33 33 dx3 =
( f d +
g d) dx3
0
0
Z L Z
Z
Z
( x f3 d +
x g3 d)dx3 + ( p d)(L)
( x p3 d) (L),
for all V02 (0, L)
Z L
A()Ac33 [(
33 3 + P333 z3 )3 (P333 3 C3333 z3 )q3 ]dx3
Z Z
Z
Z
L
=
( f3 d +
g3 d)3dx3 + ( p3 d)3(L),
(3.113)
(3.114)
We remark that the limit model, expressed by equations (3.113) and (3.114), is now
written over (0, L) as was considered initially. We note that the flexion model (3.113) is
59
independent of the electric potential, and therefore can be compared with the classical
bending model for an elastic beam. By other hand, the axial displacement is coupled with
electrical potential as can be see in variational problem (3.114).
Our study could stop at this stage, however we intend to prove, in the next section, the
strong convergence and, for that, we need to obtain additional information about limits
mentioned in Corollary 2, using the asymptotic expansions method.
3.4
In order to complete this study, and be able to establish the strong convergence of sequence
(u(h), (h))h>0 , we next assume that the solution of the problem (3.46)-(3.50) can be
expressed in the form
(u(h), (h)) = (u0 , 0 ) + h2 (u2 , 2 ) + h.o.t.,
(3.115)
where
u0 , 0 V0,w 0 , (up , p ) V0,w 0 , p 1,
(3.116)
and the successive coefficients of the powers of h are independent of h. We note that only
the leading term is required to satisfy the electric boundary condition found in definition
of the space V0,w 0 .
(3.117)
virtual work (3.51)-(3.52), we try to characterize the terms u0 , u2 and 0 in the formal
expansions in order to compute (u0 , 0 ).
3.4.0.1
In this section we show that the factors of hq , 4 q 0 disappear, and therefore the
formal expansions (3.117) do not contain any negative terms. For the sake of clarity, the
proof is divided in five parts.
60
D (h) = h2 D2 + D0 + h.o.t.,
D3 (h) = h2 D32 + D30 + h.o.t.,
with
4
= C e (u0 ),
(3.118)
(3.119)
(3.120)
2p
p 1,
(3.121)
2
3
= 2C33 e3 (u0 ) P3 E (0 ),
(3.122)
0
3
= 2C33 e3 (u2 ) P3 E (2 ),
(3.123)
2p
3
= 2C33 e3 (u2p+2 ) P3 E (2p+2 ), p 1,
(3.124)
2
33
= C33 e (u0 ),
(3.125)
0
33
= C33 e (u2 ) + C3333 e33 (u0 ) P333 E3 (0 ),
(3.126)
2p
33
= C33 e (u2p+2 ) + C3333 e33 (u2p ) P333 E3 (2p ), p 1,
(3.127)
D32 = P3 e (u0 ),
(3.128)
(3.129)
p 1,
(3.130)
D2 = 2P3 e3 (u0 ) + E (0 ),
(3.131)
D0 = 2P3 e3 (u2 ) + E (2 )
(3.132)
p 1,
(3.133)
61
e (v) dx
+
=
Z
+h
fi vi dx +
Z
gi vi d +
Dk2 Ek () dx
0
Dk Ek () dx + h.o.t.
(3.134)
(v, ) V0,w 0 .
pi vi d,
L
It follows that
Z
Z
4
e (v) dx =
C e (u0 )e (v) dx,
(3.135)
(3.136)
2
4
= 33
= D32 = 0.
(3.137)
and therefore
2
3
e3 (v) dx
2
e (v) dx
D2 E () dx = 0,
(3.138)
(3.139)
2C33 e3 (u0 ) P3 E (0 ) e3 (u0 ) dx
+
dx = 0.
2P3 e3 (u0 ) + E (0 ) E (0 )
Since
(x
3) =
(3.140)
1
1
(L x3 )1,0
x3 1,L
0 +
0 ,
L
L
1,L
with 1,0
constants, then the previous equation becomes
0 and 0
C33 e3 (u )e3 (u ) dx +
E (0 )E (0 ) dx = 0.
(3.141)
62
(3.142)
0 = 0.
(3.143)
Consequently, we deduce
2
3
= D2 = 0,
(3.144)
e (v) dx = 0,
(3.145)
with
2
(3.146)
Condition (3.116) will allow us to prove that it is not possible to obtain a solution u2 of
(3.145) in the space V0 except for some particular cases. This fact is at the origin of a
boundary layer phenomenon as seen in Trabucho & Via
no [1996], Irago [1999] and Irago
& Via
no [2002].
V0,w such that
Next, we prove that there exists u
+ C33 e33 (u0 ) P3 E3 (0 ) = 0.
C e (u)
(3.147)
= C e (u
(3.148)
e ) e (v) dx = 0.
C e (u2 u
= 0.
(3.149)
(3.150)
We remark that from (3.119) and (3.150) we deduce that u2 is also solution of equation
2
(3.147). In other words,
= 0 if and only if the following problem has at least one
63
solution in V0,w :
= C33 e33 (u0 ) + P3 E3 (0 ),
C e (u)
(3.151)
that is,
e11 (u2 )
(3.152)
(3.153)
2 =
0
0 .
C1211 2C1212 C1222 e12 (u ) C1233 e33 (u ) + P312 E3 ( )
2
0
0
C2211 2C2212 C2222
e22 (u )
C2233 e33 (u ) + P322 E3 ( )
By a simple inversion of the matrix we have
e11 (u2 )
e12 (u2 )
e22 (u2 )
where
= C1211 C1212 C1222
C2211 C2212 C2222
C1211 C1212 C1222
= C1211 2C1212 C1222
C2211 2C2212 C2222
e (u2 ) = C P3 E3 (0 ) C33 e33 (u0 )
+ C12 P312 E3 (0 ) C1233 e33 (u0 ) .
(3.154)
(3.155)
u (x1 , x2 , x3 ) = (x3 ),
V02 (0, L),
(3.156)
u (x , x , x ) = (x ) x (x ), V 1 (0, L).
3 1
2
3
3 3
3
3
0
By consequence:
e33 (u0 ) = 3 x .
(3.157)
64
On the other hand, from (3.143) we deduce that 0 depends only on variable x3 , that is,
0 (x1 , x2 , x3 ) = z3 (x3 ),
z3 H 1 (0, L).
(3.158)
= 1,0
0 ,
z3 (L) = (x1 , x2 , L) = (x
1 , x2 , L) = 1,L
0 . (3.159)
z3 H 1 (0, L),
(3.160)
and
0 (x1 , x2 , x3 ) = 0 (x1 , x2 , x3 ) (x
3 ) = z3 (x3 )
E3 () = 3 = z3 ,
E3 ()
= 3 = z3 +
1
1
(L x3 )1,0
x3 1,L
0
0 ,
L
L
(3.161)
(3.162)
1 1,L
( 1,0
0 ).
L 0
(3.163)
+ C12 P312 z3 C1233 (3 x )
(3.164)
x2 d.
65
z11 =
x2
x21
(C11 C33 + C1112 C1233 ) 2 (C22 P3 + C2212 P312 ),
2
2
z12 = 2x1 x2 (C12 C33 + C1212 C1233 ) + x1 x2 (C11 C33 + C1112 C1233 ),
z21 = 2x1 x2 (C12 C33 + C1212 C1233 ) + x1 x2 (C22 C33 + C2212 C1233 ),
z22 =
x22
x2
(C22 C33 + C2212 C1233 ) 1 (C11 P3 + C1112 P312 ).
2
2
X =
z d,
Z
Z=
d,
Y =
z d,
Z
d.
L=
(3.165)
1
X (0) ,
A()
1
s(0) =
Y (0) Z3 (0) Lz3 (0) .
I1 + I2
s (0) =
(3.166)
66
x21
2 1
u
e2 = k2 (x1 , x3 ) (x2 3 x1 x2 1
x22
)(C22 C33
2 2
+ C2212 C1233 )
(3.167)
Substituting these expressions in the third equation, we show that there exists a function
s, depending only on x3 , such that
2 k1 (x2 , x3 ) = s x2 1 (C22 P3 + C2212 P312 ) + 2x1 2 (C12 C33 + C1212 C1233 )
z3 (C12 P3 + C1212 P312 ) 3 (C12 C33 + C1212 C1233 ),
1 k2 (x1 , x3 ) = s x1 2 (C11 P3 + C1112 P312 ) + 2x2 1 (C12 C33 + C1212 C1233 )
z3 (C12 P3 + C1212 P312 ) 3 (C12 C33 + C1212 C1233 ).
Thus,
h
i
u
e1 = s1 + sx2 x1 (C22 P3 + C2212 P312 ) + x2 (C12 P3 + C1212 P312 ) z3
2
2
x1
x
2
+
(C11 C33 + C1112 C1233 ) (C22 P3 + C2212 P312 ) 1
2
2
h
i
h
i
+ 2x1 x2 (C12 C33 + C1212 C1233 ) + x1 x2 (C22 C33 + C2212 C1233 ) 1
2
x2
x21
+
(C22 C33 + C2212 C1233 ) (C11 P3 + C1112 P312 ) 2
2
2
h
i
Let u
e L2 (). Then
u
e d,
u2
67
3 , z3 H 2 (0, L).
u2 d
x2 u21 x1 u22 d = 0,
1
X (0) ,
A()
s(0) =
1
Y (0) Z3 (0) Lz3 (0) .
I1 + I2
2
Corollary 4 (u2 ) W2 are of the form (3.165) and consequently
= 0 if and only if
2
The analysis of Section 3.4.0.1 culminated in Corollary 4, where it was shown that
= 0,
and consequently the variational problem (3.134) can be written as follows
fi vi dx +
Dk0 Ek () dx + h.o.t.
gi vi d +
pi vi d,
L
(v, ) V0,w 0 .
(3.168)
68
Letting (v, ) VBN 30 in the variational equations of problem (3.168) shows that
Z
0
33
e33 (v) dx
D30 E3 () dx
fi vi dx +
gi vi d +
pi vi d
(3.169)
where
0
33
= C33 e (u2 ) + C3333 e33 (u0 ) P333 E3 (0 ),
g L2 (N )
f3 H 1 [L2 ()],
g3 H 1 [L2 (N )],
pi L2 (L ) L2 (),
(3.170)
then, in the displacement-electric-potential approach, the leading term (u0 , 0 ) of the formal expansion of the scaled displacement and electric potential (u(h), (h)) solves the
following problem
u0 VBN = {v : e (v) = e3 (v) = 0}
(3.171)
0 + 30 = + H 1 () : (x1 , x2 , x3 ) = z(x3 ), z3 H01 (0, L)
Z
Z
Z
Z
Z
0
0
33 e33 (v) dx + D3 E3 ()dx =
fi vi dx +
gi vi d +
pi vi d
(3.172)
(3.173)
(3.174)
0
where 33
and D30 are given by
0
33
= C33
e33 (u0 ) P3 E3 (0 ),
C33
= C3333 C33 [C C33 + C12 C1233 ],
(3.175)
69
where C33
and 3 are positive and bounded.
Any function (u0 , 0 ) solves this problem if and only if it has the form
u0 (x1 , x2 , x3 ) = (x3 ),
0 (x1 , x2 , x3 ) = z3 (x3 ),
(3.176)
(3.177)
with V02 (0, L), 3 V01 (0, L) and z3 H 1 (0, L), z3 (0) = 01,0 , z3 (L) = 1,L
0 , and i
and z3 solving the following variational problem (no sum on ):
Z L
Z L Z
Z
I C33 dx3 =
( f d +
g d) dx3
0
0
Z L Z
Z
Z
( x f3 d +
x g3 d)dx3 + ( p d)(L)
( x p3 d) (L),
for all V02 (0, L)
Z L
A()[(C33
3 + P3z3 )3 (P3 3 3 z3 )q3 ]dx3
Z Z
Z
Z
L
=
( f3 d +
g3 d)3dx3 + (
p3 d)3(L),
(3.178)
(3.179)
where the zeroth order bending moment and axial force components are given by
n3 =
m =
d3 =
0
33
d = A() (C33
3 + P3 z3 ) ,
0
x 33
d = I C33
,
(no sum on )
70
We obtain
Z
0
33
3
dx =
f3 3 dx +
g3 3 d +
p3 3 d.
(3.180)
Next, let the function v = (vi ) VBN and 30 in problem (3.173) be of the particular
form
(3.181)
q3 H 1 (0, L),
= q3 (x3 ),
(3.182)
which gives
Z
0
x 33
dx
x f
dx
x D30 q3
dx =
f dx +
x g
g d +
p d
(3.183)
p x d,
(3.184)
m dx3
(
0
f d +
x f3 d +
g d) dx3
N
(3.185)
x g3 d) dx3
Z
Z
+( p d)(L) ( x p3 d) (L)
Z L
Z L
Z L Z
Z
Z
n3 3 dx3
d3 q3 dx3 =
( f3 d +
g3 d)dx3 + ( p3 d)3(L)
0
(3.186)
Conversely, if functions solve the variational problem (3.178))-(3.179), one sees that
(u0 , 0 ) VBN 30 , with u0 = , u03 = 3 x and 0 = z3 , solves problem (3.173).
We note that on the other hand, using regularity results for variational problem (3.178)
and (3.179) together with (3.170) guarantees the regularity to H 3 (0, L) and 3 , z3
H 2 (0, L).
Theorem 6 The limit problem (3.178)-(3.179) admits a unique solution in the space
2
[V02 (0, L)] V01 (0, L) ( + H 1 (0, L)). Moreover, it is equivalent to the following differ-
71
ential problems:
A()(C33
3 + P3 z3 ) = F3 ,
A()(3z3 P3 3 ) = 0,
in (0, L),
in (0, L),
(3.187)
(4)
C33
I = F + M ,
(0) = 0, (0) = 0,
C I (L) = M L ,
33
in (0, L),
(3.188)
C33
I (L) = FL M (L),
where
I =
(x ) d,
M =
x f3 d +
FiL
pi d,
Fi =
Z
fi d +
gi d,
x g3 d,
N
ML
x p3 d.
C
P
33
3
3
A()
(3 , z3 )
dx3 =
(F3 + F3L , 0)
P3 3
q3
0
0
(3 , q3 ) V01 (0, L) H01 (0, L),
3
q3
(3.189)
dx3 ,
(3.190)
(3.191)
C33
P3
matrix
. By the Lax-Milgram Lemma, there exists a unique solution
P3 3
(3 , z3 ) V01 (0, L) ( + H 1 (0, L)) which satisfies the variational equations (3.187).
72
C33
= Ac33 33 ,
3 = Ac33 C3333 .
3.4.1
u(h) u0 , in Vw,0,
(3.192)
(h) 0 , in L2 ().
(3.193)
Strong convergence
One of the objectives of this section is to analyze the strong convergence of the solution
(u(h), (h)) of the scaled problem (3.46)-(3.50) as h 0. A brief summary of results,
0
which will help us to prove the convergence, is showed here. The pair ((33
, D30 ), (u0 , 0 ))
solves
0
((33
, D30 ), (u0 , 0 )) [L2 ()]2 (VBN ( + 30 ))
Z
Z
0
0
0
C3333 33 + P333 D3 33 dx +
P333 33
+ 33 D30 d3 dx
Z
Z
=
33 e33 (u0 )dx + d3 E3 (0 )dx,
(3.194)
gi vi d +
pi vi d,
(3.195)
X 2,w
0
33 (h) 33
0,
0,
|hD (h)|0, 0,
0,
|h3 (h)|0, 0,
D3 (h) D 0 0.
3 0,
(3.196)
2
h (h) 0,
0,
(3.197)
(3.198)
73
0
T (h) = D(h) D
0,
S(h) = (h)
0
0
33
= 33
,
3
= 0,
= 0,
3 = D0,
D
3
= 0,
D
and
S(h)
= Sij ,
S33 = S33 ,
T (h) = Tij ,
S3 = hS3 ,
S = h2 S ,
T3 = T3 ,
T = hT .
Let
aT (S(h),
T (h)), (S(h),
T (h)) := aH,0 ((S(h), T (h)), (S(h), T (h)))
+ h2 aH,2 ((S(h), T (h)), (S(h), T (h))) + h4 aH,4 ((S(h), T (h)), (S(h), T (h))) ,
(3.199)
= S(h),
n
0 2
+h
| (h)|20,
+h
|D (h)|20,
+ |D3 (h)
2
D30 |0,
(3.200)
Let us examine the behavior of the left-hand side of this inequality as h 0. First, a
simple computation based on variational equation (3.53) shows that
(h) = aT (S(h),
T (h)), (S(h),
T (h))
0 ), (u(h), (h)
0, D
= bH (((h), D(h)), (u(h), (h)
+ (h)))
+ bH ((
+ ))
), (u(h), (h))
), (0, )),
0, D
0, D
+ bH ((
+ bH ((
(3.201)
74
and
0 ), (S(h), T (h))) h2 aH,2 ((
0 ), (S(h), T (h)))
0, D
0, D
2 (h) = aH,0 ((
0 ), (S(h), T (h))).
0, D
h4 aH,4 ((
Using the variational equation in problem (3.54) and the definition (3.55), we have
1 (h) =
fi ui(h)dx +
gi ui(h)d +
0
33
e33 (u(h))dx
pi ui(h)d +
L
D30 E3 ((h))dx
D3 (h)E3 ()dx
D30 E3 ()dx,
and using the weak convergences established in Section 3.3.1 as h 0, we next have
1 (h) 1 =
fi ui dx +
0
33
e33 (u0 )dx
gi uid +
D3 E3 ()dx
pi ui d +
L
D30 E3 (0 )dx
D3 E3 ()dx
D30 E3 ()dx
D30 E3 ()dx
(3.202)
Applying expressions (3.56)-(3.58), the function 2 (h) can be written in following expansive way
0
0 ), (S(h), T (h)))
0, D
h4 aH,4 ((
0
C3333 33
+ P333 D30
0
33 (h) 33
dx
0
P333 33
+ 33 D30
D3 (h) D30 dx
0
C33 33
+ P3 D30 (h)dx,
0
C3333 33
+ P333 D30
0
P333 33
+ 33 D30
0
33 33
dx
D3
D30
dx
0
C33 33
+ P3 D30 dx.
75
0
C33 33
+ P3 D30 dx = 0
under assumptions (3.88) and (3.175). Furthermore, we are supposing that the test func0
tions appearing in (3.194) are of the form 33 = 33 33
and d3 = D3 D30 , then the
previous limit reads
2 =
=
(33
0
33
)e33 (u0 )dx
0
33 e33 (u )dx +
D3 E3 ( )dx
0
33
e33 (u0 )dx
D30 E3 (0 )dx,
where 0 = 0 + .
Choosing now v = u0 VBN and = 0 30 as test functions
33 e33 (u )dx +
and
Z
D3 E3 ( )dx
D30 E3 (0
0
33
e33 (u0 )dx
)dx
fi u0i dx
fi u0i dx
gi u0i d
gi u0i d
pi u0i d (3.203)
pi u0i d, (3.204)
We thus infer from these relations that the remaining terms the left-hand side of the
inequality (3.200) converge to zero, i.e.,
= 1 + 2 =
D3 E3 ()dx
D30 E3 ()dx
= 0.
(3.205)
76
2
0 2
(3.206)
1/2
k( , d)kX1 ,
(3.207)
2
0 2
|0, + h2 |D (h)|20,
|D3 (h) D30 |0, + |33 (h) 33
C
u(h) u0 , (h) 0
X 2,w
2
0 2
|0, + h2 |D (h)|20,
|D3 (h) D30 |0, + |33 (h) 33
C
sup
(3.208)
when h 0. This way, we finished to prove that (u(h))h>0 and ((h))h>0 converge
strongly.
3.5
The limit model on the original domain; formulation as a boundary value problem
Since u0 , 0 , 0 and D0 are approximations of u(h), (h), (h) and D(h), respectively,
as h tends to zero, undoing the change of variables (3.28) and the scaling in the different
field, we get the fields
uh,0(xh ) = h1 u0 (x), u03 (x) ,
0
0
0
h,0(xh ) = h2
(x), h3
(x), 33
(x) ,
h,0(xh ) = 0 (x),
Dh,0 (xh ) = hD0 (x), D30 (x) ,
77
h,0
Corollary 6 The approximations (uh,0 , h,0, 33
, D3h,0) are uniquely characterized as follows:
h
uh,0
= (x3 ),
h
h h
uh,0
3 = x 3 ,
h,0 = z3h ,
(3.209)
(3.210)
h,0
h,0
= 3
= 0,
(3.212)
h,0
33
= C33
e33 (uh,0 ) P3 E3 h,0 ,
where
(3.213)
D3h,0 = P3 e33 (uh,0) + 3 E3 h,0 ,
h (xh3 ) = h1 (x3 ),
(3.211)
(3.214)
3h (xh3 ) = 3 (x3 ),
(4)
C33
Ih h
= Fh + Mh
h (0) = 0,
h (0) = 0,
h
C33
Ih h (L) = ML ,
C I h h (L) = F L h M h (L),
33
i
h
A(
)
C
+
P
z
= F3h ,
33
3
3
3
i
= 0,
A( ) 3 z3 P3 3
in (0, L),
h
i
A( h ) C h (L) + P z h (L) = F L h ,
33
3
3
3
3
(3.215)
in (0, L),
in (0, L),
(3.216)
78
=
=
2
xh
Fih
d ,
xh f3h
d +
hhi
d ,
fih
d +
gih d h ,
xh g3h d h ,
h
ML
x hh3 d h .
The normal force and bending moments are given, respectively, by the expressions
h
q =
mh =
h,0
33
d h = C33
Ah ( h ) (3h) + P3 A( h ) (z3h ) ,
h,0
h
xh 33
d h = C33
I (h )
(no sum on ).
It is important to observe that the limit model is described by a system of three partial
differential equations, posed over the one-dimensional set (0, L), two of the fourth order
with respect to the unknowns h describing bending in an elastic beam, and two of the
second order with respect to the axial displacement 3h and electric potential z3h .
Now we shall examine the particular case of a transversely isotropic piezoelectric beam
load on one of its end and verify that our asymptotic piezoelectric model is consistent
with asymptotic elastic beam theory. As in the work of Trabucho & Via
no [1996], we
assume that the resultant of the applied loads P is parallel to axis Ox3 , so that, we have
fih = 0,
gih = 0,
hh1 = 0,
hh3 = 0,
F2h,L = P,
then, we obtain
h h
C33
I1 (1 ) = 0
in (0, L),
h h
C33
I1 (1 ) (L) = 0,
C I h ( h ) (L) = 0
33 1 1
h h
C33
I2 (2 ) = 0
in (0, L),
h h
C33
I2 (2 ) (L) = 0,
C I h ( h ) (L) = P
33 2 2
79
h
h
h
A
C
(
)
P
(z
)
=0
33 3
3
3
Ah P3 (3h ) + 3 (z3h ) = 0
in (0, L),
in (0, L),
3h (0) = 0,
z3h (0) = h,0
z3h (L) = h,L
0 ,
0 ,
Ah C ( h ) (L) P (z h ) (L) = 0.
33 3
3
3
u02 = 2h =
uh,0
3
3h
h 2
P
h
x
3L
(x3 ) ,
3
6I2h C33
xh2 (2h )
h = z3h = h,0
+
0
P3
C33
h,0
0
h,L
0
h,0
0
xh3
P
+ h
I2 C33
xh3
L x3 xh2 ,
2
h,0
h,L
0 0
xh3 ,
L
and consequently the stress tensor component and the electric displacement give
h,0
33
=
D3h,0
P h
(x L)xh2 ,
I2h 3
P
= h P3 (xh3 L)xh2 +
I2 C33
P3 P3
+ 3
C33
h,0
h,L
0 0
.
L
If we ignore the electric field, we can verify that these equations coincide with the asymptotic model established by Trabucho & Via
no [1996]. We can conclude, from this example,
that the asymptotic model for a transversely isotropic beam yields the corresponding classical models of engineering literature.
Chapter
4.1
In this section we describe the boundary conditions and the mechanical problem that will
be the focus of this chapter.
4.1.1
Let us consider the following assumptions with the notations already introduced:
(Hd41 ) The beam is weakly clamped at the extremities h0 hL that is to say hdD = h0 hL
with meas(hdD ) > 0.
(Hd42 ) The surface traction g h = (gih ) act on hdN = hN .
h
h
(Hd43 ) We assume heD = eD
(0, L) with meas(eD
) > 0, and heN = h h \heD .
h
h
h
P
= P33
= P33
= 0,
h3 = 0.
(4.1)
h
a
h0
xh2
h0
heD
h
xh3
hL
xh1
Figure 4.1: Schematic representation of the solid domain detailing the electrical
boundary conditions.
83
In the following we will represent this boundary condition by uh = 0 on hdD .
Remark 8 Condition (Hd45 ) means that the beam is made of a piezoelectric material
having crystal symmetry corresponding to one of the following classes (see e.g. Refs. Royer
& Dieulesaint [2000] and Nye [1985]): monoclinic system (except class m), orthorhombic
system, tetragonal system, hexagonal system (except classes 6 and 6m2), and cubic system.
Let us now define the following closed subspaces of the Sobolev space of order one
1 h m
H ( ) , m 1:
3
h
h
= V0,w
(h ) = {v h H 1 (h ) : v h = 0 on hdD },
V0,w
h = h (h ) = { h H 1 (h ) : h = 0 on h }.
0
0
eD
(4.3)
Since meas(hdD ) > 0 and meas(heD ) > 0, it follows from Korns and Poincare-Friedrichs
inequalities that the following norms are equivalent to the standard Sobolev norms on
h
V0,w
and h0 , respectively,
h
v
h = eh (v h ) h ,
V
0,w
h
h = h h h .
acting on heD , we obtain that the solution (uh , h ) is derived from h = h + h with
(uh , h ) satisfying:
h
Find (uh , h ) V0,w
h0 such that
a((uh , h ), (v h , h )) = l2 (v h , h )
(4.4)
h
(v h , h ) V0,w
h0
where the bilinear form a(, ) and l2 () are defined in Chapter 2 as follow
h
a( u ,
+
, v , ) =
h
Cijkl ehkl (uh ) Pmij Em
h
h h h
Pmij ehij (uh ) + mi Eih h Em
dx ,
l2 (v , ) =
Z
h
dN
ehij (v h )dxh
(4.5)
h h
fi vi + Pkij Ekh (h )ehij (v h ) kl Elh (h )Ekh ( h ) dxh
gih vih dh .
(4.6)
We now consider that the displacement field uh , the electric potential h , the stress
tensor field h and the electric displacement field Dh satisfy the following problem, consisting of the principle of virtual work (cf. (3.13)):
h
(uh , h ) X h2 = V0,w
h2
( h , Dh ) X h1 = ( , d) = (ij , dk ) [L2 (h )]3 L2 (h ) : ij = ji ,
Z
Z
Z
Z
h h
h
h
h h
h
h
h h
h
ij eij (v ) dx +
Dk Ek ( ) dx =
fi vi dx +
gih vih dh
h
(4.7)
h
dN
h
for all (v h , h ) X h0,w = V0,w
h0 .
4.2
As in Section 3.2, we need to transform problem (4.7) into a problem posed over a set
that does not depend on h. Accordingly, we let
:= (0, L),
h
and with each point x , we associate the point xh through the bijection h ,
defined in Chapter 3 by (3.28). Then, one has
hdN = h (dN ),
h0 = h (0 ),
n(x) = nh (xh ),
heN = h (eN ),
hL = h (L ),
h
eD
= h (eD ),
heD = h (eD ),
(4.8)
meas(eD ) > 0.
(4.9)
85
(4.10)
equipped with the following norms equivalent to the usual Sobolev norms:
kvkV = |e(v)|0, ,
where
1
(i vj + j vi ) .
2
= 0 stands for the boundary condition (cf. (3.30))
e(v) = (eij (v)) ,
Here, hvi|dD
kk = ||0, ,
ui d = 0,
{a}
eij (v) =
{a}
(xi uj xj ui ) d = 0,
a = 0, L.
(4.11)
h
With the unknown and test displacement fields uh , v h in V0,w
, we associate the (unknown and test) scaled displacement fields u(h) = (ui (h)) and v(h) = (vi (h)) in V0,w
defined in (3.36)-(3.37). In this chapter, the electric potential h and the test function
= h1 h (xh ),
(h)(x) = h1 h (xh ).
(4.12)
We make the following assumptions on the data: We assume that the constants of the
material, the applied body force density, the applied surface force density and the applied
electric potential are the following form:
1. There exist functions fi L2 () and gi L2 (dN ), independent of h, such that:
fh (xh ) = hf (x),
g h (xh ) = h2 g (x),
= hg3 (x),
(4.13)
hdN .
(4.14)
+ .
(4.15)
h
Motivated by these identities we define for any (v, ) V0,w
H 1 (h ) the tensor
(h; v) and the vector (h; ) as follows:
(4.16)
(4.17)
For simplicity we abbreviate the notation for the special cases v = u(h), = (h),
= 0 , writing
(h) := (u(h)),
We remark that
b := ().
(h) := ((h)),
(h) = eh (uh ),
(4.18)
(h)
= h h ,
and define
b
(h) = (h)((h)
+ 0 ) = (h)
+ (h).
Using the scalings of the unknown and the assumptions on the data, we reformulate,
in the next two theorems, the problems (4.4) and (4.7) now posed over the set .
The variational problem (4.4) becomes
Find (u(h), (h))
(v, )) = l2 (v, ),
(4.19)
(v, ) V0,w 0 ,
where the bilinear form a and the linear form l2 are defined by
Z
a((u, ),
(v, )) =
Cijkl kl (h; u) ij (h; v) dx +
ij i (h; ) j (h; ) dx
Z
m (h; )) dx
+ Pmij (m (h; )
ij (h; v) ij (h; u)
(4.20)
fi vi dx +
dN
87
gi vi d
Pkij bk ij (h; v) dx
ij bi j (h; ) dx
(4.21)
h
Proposition 5 Assume that (uh , h ) V0,w
H 1 (h ) is solution of problem (4.7).
(a) The scaled unknowns (u(h), (h)) V0,w H 1 () satisfy the following variational
gi vi d,
dN
(4.22)
(4.23)
(b) The functions (ij (h)) L2 () and (Di (h)) L2 () defined in (a) are also related to
h
h
h
(xh ) = h2 (h)(x), 3
(xh ) = h3 (h)(x), 33
(xh ) = 33 (h)(x),
(4.24)
Proof. The proof of (a) reduces to simple computations, based on the displacement, the
assumptions on the data, and the formulas h = h1 , 3h = 3 , and
Z
dx = h
dx,
h
dN
d = h
dN
h h d.
the formulas in part (b) follow from the definitions of the functions ij (h) and Di (h) given
in part (a).
(4.25)
(4.27)
where the bilinear forms a2 , a1 , a0 , a1 , a2 , a4 and bH , and the linear form lH are defined
by
a2 ((, D), ( , d)) =
a1 ((, D), ( , d)) =
a0 ((, D), ( , d)) =
33 D3 d3 dx,
Z
Z
P333 D3 33 dx
(4.28)
Z
C3333 33 33 dx +
P33 D3 3 dx +
P333 33 d3 dx,
(4.29)
D d dx,
(4.30)
P3 D3 dx
P3 d3 dx 2 P3 3 d dx,
Z
a2 ((, D), ( , d)) = C33 33 dx + 4 C33 3 3 dx
Z
+
C33 33 dx,
Z
a4 ((, D), ( , d)) = C dx,
(4.31)
(4.32)
(4.33)
fi vi dx
89
gi vi d,
Z
Z
bH ((, D), (v, )) =
ij eij (v)dx
Dk Ek () dx.
4.3
(4.34)
dN
(4.35)
In this section we intend to study the behavior of the solution to the problem (4.22)-(4.24)
when h 0. This study is based on formal asymptotic expansions method.
We now assume that the solution of the problem (4.19)-(4.21) can be expressed in the
form
u(h) = u0 + hu1 + h2 u2 + ...,
ui V0,w ,
(4.36)
(h)
= 0 + h1 + h2 2 + ...,
i 0 ,
(4.37)
where the successive coefficients of the power of h are independent of h. Since (h)
=
(h) ,
then
(h) = 0 + h1 + h2 2 + ... ,
(4.38)
with 0 = + 0 .
The asymptotic developments (4.36) and (4.38) induce the following formal expansions
for tensors (h) and D(h) (cf. (4.23))
(h) = h4 4 + h3 3 + h2 2 + ...,
D(h) = h1 D1 + D 0 + hD 1 + ...,
(4.39)
with q = qij and Dq = (Diq ) independent of h. Inserting developments (4.36)-(4.39)
into problem (4.22) results in a set of variational equations that must be satisfied for
all h > 0 and consequently the terms at the successive powers of h must be zero. This
procedure yields the following problems at the successive powers of h for all (v, )
V0,w 0 :
(P
(P
):
):
e (v) dx = 0,
(4.40)
e (v) dx = 0,
(4.41)
(P
):
(P ) :
Dk1 Ek () dx = 0,
Dk0 Ek () dx
(4.43)
fi vi dx +
gi vi d,
(4.44)
dN
where
4
= C e (u0 ),
(4.45)
= C e (u1 ),
(4.46)
(4.47)
(4.48)
p 0,
2
3
= 2C33 e3 (u0 ),
(4.49)
1
3
= 2C33 e3 (u1 ) P3 E (0 ),
(4.50)
p
3
= 2C33 e3 (up+2 ) P3 E (p+1 ),
p 0,
2
33
= C33 e (u0 ),
(4.51)
1
33
= C33 e (u1 ),
(4.52)
0
33
= C33 e (u2 ) + C3333 e33 (u0 ),
(4.53)
1
33
= C33 e (u3 ) + C3333 e33 (u1 ) P333 E3 (0 ),
(4.54)
p
33
= C33 e (up+2 ) + C3333 e33 (up ) P333 E3 (p1),
D1 = 2P3 e3 (u0 ),
(4.55)
D0 = 2P3 e3 (u1 ) + E (0 ),
Dp = 2P3 e3 (up+1) + E (p ),
p 2,
(4.56)
p 1,
D31 = P3 e (u0 ),
(4.57)
D30 = P3 e (u1 ),
(4.58)
(4.59)
91
4.3.0.1
p 3.
(4.60)
equipped with the norm kvkV0,w = |e(v)| . The space VBN can be equivalently defined
by (see e.g. Ref. Trabucho & Via
no [1996])
(4.61)
Now, in order to obtain some properties of u0 and u2 , we will show that some terms of
4
= 0.
the developments (4.39) are null. We start with problem (P 4 ) and prove that
0
For such, we consider (4.40) and (4.45), and take v = u yielding
C e (u0 )e (u0 ) dx = 0.
(4.62)
= 33
= 0,
D31 = 0.
(4.63)
We turn now to problem (P 2 ). Taking v = u0 in (4.42) and bearing in mind (4.62) and
(4.63) we obtain
Z
C33 e3 (u0 )e3 (u0 ) dx = 0,
which implies
e3 (u0 ) = 0,
(4.64)
D1 = 0.
(4.65)
u0 (x1 , x2 , x3 ) = (x3 ),
H 2 (0, L),
(4.66)
2
In the following we will establish necessary and sufficient conditions for
= 0. This
will allow to characterize the transverse components of the second order displacement u2 .
As a result, we will be able to obtain variational problems that have and 3 as unique
solutions.
2
e V0,w
Theorem 8 A necessary and sufficient condition for
= 0 is that there exists u
such that
C e (e
u) + C33 e33 (u0 ) = 0.
(4.67)
2
e (v) dx = 0,
C e (u2 ) + C33 e33 (u0 ) e (v) dx = 0,
equivalent to
Z
C e (u2 ) C e (e
u) + C e (e
u) + C33 e33 (u0 ) e (v) dx = 0.
e ) e (v) dx = 0,
C e (u2 u
e = u2 by the coerciveness
and therefore taking v = u2 e
u V0,w in this equation leads to u
2
2
e = u2 .
of C, implying
= 0. Inversely, if
= 0 then (4.67) holds with u
Remark 10 From Theorem 8 and (4.47) we conclude that if equation (4.67) has solutions
then u2 is a solution of (4.67). Next we charaterize the solutions of (4.67)
93
Characterization of u0 and u2
C3131 C3132
N=
C3132 C3232
C
0
0
C2233
1122 C1222 C2222
C=
0
0
0
C3131 C3132
0
0
0
0
C3132 C3232
0
C3333
The hypothesis (Hc21 ) and (Hc22 ), introduced in Section 2.1.3, imply that these matrices
are definite positive and therefore det C > 0, det M > 0 and det N > 0.
Before proceeding, we now define some functions that will be useful in the characterization of the elements of V0,w that verify (4.67).
We define the geometry functions (x1 , x2 ) and = ( ) defined by
(x1 , x2 ) =
det M
x ,
det M
1
(det M 11 x21 det M 22 x22 ),
2 det M
det M 1
det M 2
12 =
x x2 ,
21 =
x x1 ,
det M
det M
1
(det M 11 x21 det M 22 x22 )
22
2 det M
11 =
where
M 11
M 12 = M 21 =
M 22
We also introduce the following constants that depend only on the geometry of the cross
x2
d,
X =
d,
Y =
d,
Z=
d.
(4.68)
det M 11
1 u
e1 =
(x 3 ) ,
det
M
det M 22
2 u
e2 =
(x 3 ) ,
det
M
det M 12 + det M 21
1 u
(x 3 ) .
e2 + 2 u
e1 =
det M
(4.69)
det M 11 2
e1 =
(x + 2x1 x2 2 2x1 3 ) + k1 (x2 , x3 ),
u
2 det M 1 1
det M 22 2
u
e2 =
(x + 2x1 x2 1 2x2 3 ) + k2 (x1 , x3 ).
2 det M 2 2
(4.70)
95
Thus,
det M 22 2 det M 12 2
(4.71)
where s are arbitrary functions depending only on x3 . Combining (4.70) with (4.71)
leads to (4.68).
R
R
Let u
e L2 (). Then u
e d, u
e d, , 3 L2 (0, L) and (3.29) together with
(4.68) lead to s , s L2 (0, L).
Corollary 7 Let (e
u ) [L2 ()]2 a solution of (4.67) using the form (4.68). Then, a
e V0,w is that
necessary and sufficient condition for problem (4.67) to have solutions u
u0 VBN be such that H 3 (0, L), 3 H 2 (0, L), and that s, s H 1 (0, L) verify
s(a) =
with a = 0, L.
1
Y (a) + Z3 (a) ,
I1 + I2
(4.72)
Z
det M
x u
e d =
I
(no sum on ).
det
M
R
R
Since u
e H 1 (), then u
e d, x u
e d H 1 (0, L) (no sum on ) and consequently H 3 (0, L), 3 H 2 (0, L). On the other hand, identity (4.68) allows to
write
u
e d = A()s + X H 1 (0, L),
u
e d = (I1 + I2 ) s + Y + Z3 H 1 (0, L).
Hence, the previous regularity results lead to s, s H 1 (0, L). Furthermore, the boundary
condition for u
e (cf. (4.11)) leads to (4.72).
H 1 (0, L) of the form (4.72), then (4.68) implies that u2 H 1 () verifies the boundary
condition (4.11).
Corollary 8 The components u2 of the displacement u2 V0,w are of the form (4.68)
2
with s, s H 1 (0, L) satisfying (4.72) (and consequently
= 0) if and only if u0 VBN
is such that H 3 (0, L), 3 H 2 (0, L).
Remark 11 From (4.68) it follows that in general u2 is not null at the beam ends, for that
would require that both and 3 vanish there. Therefore, if we had chosen strong boundary
conditions in the definition of V0,w , then we would not have been able to ensure that u2 V
2
(the well-known boundary layer phenomenon) and consequently that
= 0.
For now we can only ensure that H02 (0, L), 3 H01 (0, L). The additional
regularity required by the previous corollary will follow from the variational problems to
which , 3 obey. We have the following results.
Theorem 10 Let u0 VBN given by (4.66) and fi L2 (), gi L2 (dN ). Then u0 is
the unique solution of the following variational problem:
Z
0
33
e33 (v) dx
fi vi dx +
dN
(4.73)
0
where 33
is given by
0
33
= Y e33 (u0 ) = Y (3 x ),
and
Y =
det C
det M det N
(4.74)
(4.75)
= 0. From Corollary 8 we get that u2 is of the form (4.68), and consequently (4.53)
becomes (4.74).
Corollary 9 The transverse components and the stretch component 3 of the zeroth
order displacement field are respectively the unique solutions of the following variational
problems (no sum on ):
H02 (0, L),
Z L
Z
Y I dx3 =
0
F dx3
L
0
M dx3 ,
(4.76)
97
and
3 H01 (0, L),
Z L
Z
Y A()3 3 dx3 =
0
F3 3 dx3 ,
(4.77)
with
Fi (x3 ) =
fi d +
M (x3 ) =
dN
x f3 d +
gi d L2 (0, L),
dN
x g3 d L2 (0, L).
(4.78)
(4.79)
Proof. We consider (4.73) with the test function v VBN of the form (4.61), implying
e33 (v) = 3 (x3 ) x (x3 ). Then, taking successively = 0 and 3 = 0 in the resulting
equation we obtain (4.76) and (4.77), respectively.
The following regularity result (see e.g. Ref. Brezis [1983]) ensures the regularity
requirements for and 3 in Theorem 10.
Corollary 10 The solutions of (4.76) and (4.77) are unique and H 3 (0, L), 3
H 2 (0, L).
2
Corollary 11 The components u2 of u2 verify (4.68) and consequently
= 0.
det M
P3 .
det M
(4.80)
Proof. This result follows straightforwardly from (4.59) taking into account that u2
verifies (4.69) and that (4.74) holds.
4.3.0.3
Characterization of u1 and 0
(4.81)
q d = 0}
(4.82)
S = S() = { H 1 () : = 0 on eD }.
(4.83)
1/2
kkS
= ||2 + |1 |2 + |2 |2
1/2
1/2
kkS = |1 |2 + |2 |2
1/2
T := T () = Q() S()
(4.84)
(4.85)
(4.86)
We now define the warping function w(x1 , x2 ) as the unique solution of the variational
problem:
w Q such that
Z
Z
C33 w d = C33 d,
(4.87)
99
(4.88)
(4.89)
for all Q.
(4.90)
C e (u1 )e (u1 ) dx = 0,
(4.91)
= 33
= 0,
D30 = 0.
Equation (4.43) evaluated in v V1 and equation (4.44) for v = 0 in (4.44), and considering (4.50) and (4.56) are given by
Z
2C33 e3 (u1 ) + P3 0 e3 (v) dx = 0,
2P3 e3 (u1 ) 0 dx = 0,
for all v V1 ,
for all 0 ,
(4.92)
(4.93)
which show that u1 and 0 are coupled. In fact, we have the following result.
3
u
b3 = r z3 x z wz ,
(4.94)
C33 u
b3 + z + z + P3 0 e3 (v) dx = 0,
for all v V1 .
(4.95)
We restrict (4.95) to v = (0, 0, (x3)(x1 , x2 )), H01 (0, L), Q and combining it
with (4.87), we get a.e. x3 (0, L)
Z
C33 u
b3 (x3 ) d =
P3 0 (x3 ) d,
for all Q.
Therefore,
U =u
b3 + x z + wz
C33 U(x3 ) d =
P3 0 (x3 ) d,
(4.96)
Q.
(4.97)
(4.98)
(4.99)
101
In view of the previous regularity result we conclude that although (4.91) implies
R
that u1 H 1 (), we can only ensure that u13 L2 (0, L; H 1 ()). Furthermore, u13 d
vanishes at the beam ends as long as z3 H01 (0, L), but in general this is not the case
R
for x u13 d. Therefore, we cannot ensure that u1 V0,w , traducing a boundary layer
phenomenon. On the other hand, the regularity result obtained for u13 implies that
U L2 (0, L; H 1 ()) - cf. (4.96) - and therefore from (4.97) we can only guarantee that
0 is in l not in 0 . Hence, 0 = 0 0 l0 .
(4.91) and (4.99) are concerned. This will allow to use equations (4.92) and (4.93) to
derive a variational problem for 0 and r posed in each cross section {x3 }. In this
context it will be useful to consider for a generic function : R its average along the
x3 -axis, that will be denoted
1
(x1 , x2 ) =
L
(x1 , x2 , s) ds,
(4.100)
(4.101)
+
=
P3 ( r(x3 ) 0 (x3 ) ) d
P3 0 (x3 ) d
0 (x3 ) d,
(4.102)
for all (, ) T.
(4.103)
P3 ( r + z ( w )) + 0 e dx = 0,
for all e .
P3 ( r(x3 ) + z (x3 ) ( w )) + 0 (x3 ) d = 0,
for all S,
Corollary 14 Let us consider (r, 0 ) T the element obtained from (r, 0 ) T using
the definition(4.100). If the material the beam is such that the coefficients C33 , P3 and
do not depend on x3 , then (r, 0 ) is the unique solution of the following 2D variational
problem,
(r, 0 ) T such that
Z
Z
Z
0
C33 r d + d + P3 ( r 0 ) d
P3 0 d
(4.104)
0 d,
for all (, ) T.
C33 d +
for all (, ) T.
We are now in conditions to express z as a function of the electric potential.
Theorem 13 Let u1 be such that (4.91) and (4.99) - and therefore (4.103) - hold. Then
z H01 (0, L) is such that
1
z (x3 ) =
J
P3 ( w ) (0D (x3 )) d,
(4.105)
or
1
z (x3 ) =
J
P3 ( w ) (0D (x3 ) + D
0 (x3 )) d,
(4.106)
103
C33 (z ( w) r) + P3 0 e3 (v) dx = 0,
for all v V1 .
Taking v = ((x3 )x2 , (x3 )x1 , 0), H01 (0, L) in the previous equation we get
Z
C33 (z ( w) r) + P3 0 dx = 0.
Since the coefficients C33 do not depend on x3 , we can reduce previous equation to
C33 ( w) d
Z
d
=
C33 r(x3 ) P3 0 (x3 ) d,
dx3
z (x3 )
d
Jz (x3 ) =
dx3
P3 ( w ) 0 (x3 ) d,
(4.107)
Integrating both sides of (4.107) and taking into account that the coefficients P3 not
depend on x3 we obtain
Jz (x3 ) =
P3 ( w ) 0 (x3 ) d + cz ,
The expression for the constant cz is obtained integrating both sides of the previous
equation over (0, L) and taking into account that z vanishes at the beam ends. This
leads to (4.105). The alternative expression (4.106) is obtained from (4.105) taking into
account that 0 = 0 + 0 implies (cf. (4.101)) 0D = 0D + D
0 .
Remark 12 From the previous result we conclude that in general it is not possible to
Theorem 14 Let
(r, 0 ) = (r, 0 ) + (r D , 0D ) R l0 ,
(4.108)
where (r, 0 ) is the unique solution of the variational problem (4.104). Then element
(4.109)
for all (, ) T,
e
a((e
r , ),
e (, )) =
C33 re d +
e d
Z
Z
1
+ ( P3 ( w )
e d) ( P3 ( w ) d)
J
Z
+ P3 ( re
e ) d
(4.110)
and
Z
Z
D
e
l (, ) =
P3 (0 (x3 )) d (D
0 (x3 )) d
Z
Z
1
D
( P3 ( w ) (0 (x3 )) d) ( P3 ( w ) d)
J
(4.111)
P3 ( w ) d)2 0.
Hence, the uniqueness of solution of the variational problem (4.109)-(4.111) follows from
Lax-Milgram Lemma.
Remark 13 It is worth noting that some of the terms appearing in the bilinear form
and linear functional that define problem (4.109)-(4.111) coincide with the corresponding
ones appearing in (4.104). On the other hand, the set of problems (4.109)-(4.111) has
two important particularities: cross sections having the same electric potential applied
on the boundary have the same solution 0M . These properties can be rather useful when
solving the whole problem numerically (e.g. finite elements). Finally, the decomposition
105
4.3.0.4
In this section we will to derive a model for a special case in which the beam is made
of an anisotropic homogeneous material of class 6mm. The crystal class 6mm represents
an important class of piezo-materials (Ceramic PZT-4, Zinc oxide (ZnO) and Cadmium
sulphide (CdS) are examples of piezoelectric materials belonging to the 6mm class) and
it is contained in crystal class 2. The interest in obtaining this model, is due to the fact
that this type of material (class 6mm) is widely used in engineering applications.
Theorem 15 Let us consider that the beam is made of an anisotropic homogeneous material of class 6mm where the piezo-elastic-dielectric coefficients satisfy (2.21)-(2.23). Then,
equations (4.44) determine in a unique way (u0 , 0 ) V0,w 0 , through the following
expressions:
1. The displacement components ui are of the form:
u0 (x1 , x2 , x3 ) = (x3 ),
(4.112)
(4.113)
F dx3
L
0
M dx3 ,
(b) The additional stretching 3 depends only on x3 , and is the unique solution to
the following problem
3 H01 (0, L),
Z L
Z
Y A()33 dx3 =
0
F3 3 dx3 ,
(4.115)
M (x3 ) =
(4.116)
(4.117)
dN
x f3 d +
dN
x g3 d L2 (0, L).
(4.118)
(4.119)
P14 0
(x3 )(x1 , x2 ),
C44
(4.120)
where z3 H01 (0, L) and the electric potential 0 (x1 , x2 , x3 ) = 0 (x3 )(x1 , x2 ) is the
only solution of
(4.121)
C e (u1 )e (u1 ) dx = 0,
(4.122)
and, therefore,
u1 (x1 , x2 , x3 ) = z (x3 ) + z(x3 ),
C33 u13 + z + z + P3 0 e3 (v) dx = 0.
(4.123)
107
Taking v = (0, 0, (x3 )(x1 , x2 )), H01 (0, L), Q() in the previous equation we
get
C33 u13
d =
C33 (z + z ) + P3 0 d,
for all Q,
(4.124)
valid a.e. x3 (0, L). For a homogeneous anisotropic piezoelectric beam whose constants
of the materials satisfy (2.21)-(2.23), the equation (4.124) takes the following form
Z
u13 d
(z
P14
+ z ) d
C44
0 d,
Therefore, (4.124) and the fact that u13 L2 (0, L; H 1()) lead to
u13 (x1 , x2 , x3 ) = z3 (x3 ) x z (x3 ) w (x1 , x2 ) z (x3 )
P14 0
(x3 )(x1 , x2 ),
C44
(4.126)
where z3 H01 (0, L) and the warping function w(x3 ) (x1 , x2 ) of as the unique solution
of the problem
w H 1 (),
Z
Z
w d =
d,
Z
w d = 0.
(4.127)
where 1 = x2 , 2 = x1 .
Taking v = ((x3 )x2 , (x3 )x1 , 0), H01 (0, L) in (4.123) we get
Z
Z
P14
0
C44
z ( w)
dx = P14 0 dx,
C
44
that is
( w) z dx = 0, H01 (0, L)
or, equivalently,
Z
L
0
Jz dx = 0, H01 (0, L)
(4.128)
where
J =
( w) d =
(x2 2 + x1 1 ) d = 2
d.
P14
0 .
C44
(4.129)
for all e 0 .
Taking e = (x1 , x2 ) (x3 ), S(), H01 (0, L) in the previous equation we have
Z
P14
+ 11 0 (x3 ) d = 0,
P14
C44
4.4
In this section we are going to establish the convergence results justifying the formal
asymptotic expansions.
4.4.1
Following Sene [2001], we establish that the scaled displacement and the scaled electric potential are bounded. Its first published proof for piezoelectric beam appeared in
Figueiredo & Leal [2006].
Taking (v, ) = (u(h), (h))
(u(h), (h)))
=
=
fi ui (h) dx +
dN
gi ui (h) d
ij i (h) j (h) dx
ij bi (h) j (h) dx
(4.130)
2
2
a((u(h), (h)),
(u(h), (h)))
c |(h)| + (h) ,
(4.131)
109
Proof. From condition (Hc22 ), i.e., the ellipticity properties of C and , we get
a((u(h), (h)),
(u(h), (h)))
c1
Z X
3
(ij (h)) dx + c2
i,j=1
ij i (h) j (h) dx
Z X
3
(i (h))2 dx
i=1
2
c |(h)|2 + (h) ,
1 2
b,
m
dN
(4.132)
Proof. Here we invoke the continuity of the trace operator, the fact that f and g
belong to [L2 ()]3 and [L2 (dN )]3 , respectively, and the fact that kkV0,w and kk[H 1 ()]3
are equivalent norms, to obtain
Z
Z
fi ui(h) dx +
dN
Z
Z
gi ui(h) d fi ui(h) dx +
dN
m
1
m
1
|f |2 +
|u(h)|2 + |g|2dN +
|u(h)|2dN
2
2m
2
2m
c1
m
|f |2 + |g|2dN + ku(h)k2[H 1 ()]3
2
m
c
m |f |2 + |g|2dN + ku(h)k2V0,w ,
m
gi ui (h) d
2m
(4.133)
i,j=1
3
=
=
3 X
3m 2 X b
r
|i (h)|2 +
|j (h)|2
2
2m
j=1
j=1
2
2
3m 2 b
3
3
(h) cm +
(h) ,
r |(h)|2 +
2
2m
2m
2m
(4.134)
b
Proof. Let p = sup1i,j,l3 |Plij |. Since (h)
is bounded in [L2 ()]3 , one has
Z
3
X
m 2 b
1
2
2
Plij bl (h) ij (h) dx
p |l (h)| +
|ij (h)|
2
2m
i,j,l=1
=
=
for all h, m > 0.
3
3
3 X
9m 2 X b
p
|l (h)|2 +
|ij (h)|2
2
2m
i,j=1
l=1
9m 2 b
3
3
p |(h)|2 +
|(h)|2 cm +
|(h)|2 ,
2
2m
2m
Proof. Korns inequality and the fact that 0 < h < 1 allows to write
c1 ku(h)k2[H 1 ()]3 |e(h)|2 |(h)|2 ,
or, since kkV0,w and kk[H 1 ()]3 are equivalent norms,
c2 ku(h)k2V0,w |e(h)|2 |(h)|2 .
(4.135)
111
(u(h), (h))).
(4.136)
On the other hand, (4.130) and (4.132) - (4.134) can be combined to yield
2
c7
3
2
2
a((u(h), (h)),
(u(h), (h)))
c6 m + ku(h)kV0,w +
|(h)| + (h) ,
m
2m
2
3
3
c7
2
2
c3
ku(h)kV0,w + c4
|(h)| + c5
(h) cm.
m
2m
2m
2
ku(h)k2V0,w + |(h)|2 + (h) c,
Proof. The previous result follows from (4.135) and (cf. 4.18)
= (h) + (1 ,
(h) = (h) +
2 ,
h 3 ),
c,
(4.137)
Proof. The PoincareFriedrichs inequality allows to write, for any s fixed in (0, L)
Z
[(h)(,
, s)]2 d = k(h)(,
, s)k2L2 ()
h
i
2
2
C() k1 (h)(,
, s)kL2 () + k2 (h)(,
, s)kL2 () .
Z
[(h)(x
[(h)(,
, s)] d ds
1 , x2 , x3 )] dx =
0
Z Lh
i
C()
k1 (h)(,
, s)k2L2 () + k2 (h)(,
, s)k2L2 () ds
0
2
2
= C() |1 (h)|
+ |2 (h)|
.
(h)
in L2 (),
3
(h) in L2 () .
Proof.
(4.138)
(4.139)
(4.140)
(4.141)
,
(h)
and functions u, , and satisfying (4.138)-(4.141).
h>0
h>0
Lemma 9 There exists a subsequence of ((h), (h))0<h<1 , still parameterized by h, and
3
(h) in L2 (),
3
(h) in L2 () ,
+ ,
where
= (1 ,
2 ,
0).
Proof. These results follow straightforwardly from (4.140)-(4.141) taking into account
e
D(h)
= (D (h), h1 D3 (h)),
(4.142)
(4.143)
113
(4.144)
(4.145)
(4.148)
(4.146)
(4.147)
From equalities (4.142)-(4.148) and Lemmas 8 and 9, we conclude that the sequences
3
e
(D(h))
(h))0<h<1 are bounded in [L2 ()] and [L2 ()] , respectively. There0<h<1 and (e
fore, one has the following result.
e
e (h))0<h<1 , still parameterized by h,
Corollary 17 There exists a subsequence of (D(h),
and there exist D [L2 ()] and [L2 ()]s , such that the following weak convergence
hold in L2 () when h tends to zero:
D3 (h) = h1 D3 (h) D3 = P3 + P333 33 33 3 .
(4.149)
Proof. Properties (4.149) are obtained from (4.142)-(4.148) taking the limit as h tends
to zero.
4.4.2
Identification of the limit problem and the weak convergence of the scaled displacement and electric potential to
the leading terms
The aim of this section is to derive the limit models for a homogeneous anisotropic material
of class 2 and class 6mm and to compare with the equations found in Section 4.3 .
In order to identify the limit model, we consider all the notation defined in Section 4.3,
and denote by V0,w () = W1 () W2 () the space of scaled displacements as follow
Z
Z
1
W1 () = H () :
=
x = 0, a = 0, L
{a}
{0}
Z
Z
1
2
W2 () = b = ( ) H () :
=
= 0, a = 0, L
{a}
(4.150)
(4.151)
{a}
(4.152)
v 0 W1 (),
(4.153)
v dx = 0, v W2 (),
Z
Z
x = = 0,
(4.154)
u VBN :
(4.156)
3 v 0 dx = 0,
(4.155)
e (u) = e3 (u) = 0,
(4.157)
= (1 ,
2 ,
0)T ,
(4.158)
= (1 , 2 , 0)T ,
(4.159)
33 = Ac33 33 e33 (u) Ac ,
(4.160)
E () = P3 3 + D ,
Ac33 33 =
33
,
33 C3333 + P333 P333
(4.161)
Ac = 33 C33 + P333 P3
(4.162)
fi vi dx +
dN
gi vi d,
v VBN .
(4.163)
115
3
Proof. From the weak convergence u(h) u in [H 1 ()] , since u(h) V0,w , we deduce
that u V0,w . Moreover,
(4.164)
In addition, from identities (4.16) and from weak convergence of Lemma 8 we have
(4.165)
in L2 ().
The uniqueness of the limits appearing in (4.164)-(4.165) implies (4.157) besides e (u) =
e3 (u) = 0. Now, from (4.140) we conclude that
i (h)
i = i in H 1 ().
(4.166)
h3 (h))
( , 3 ) in [L2 ()]3 .
Therefore, given (4.166) and the uniqueness of the limit appearing in the right-hand side
of (4.141) we conclude that
( , 3 ) = ( ,
0),
that is (4.158), which implies (4.159).
In order to establish (4.156)-(4.157), we consider the following equation obtained from
(4.26) by setting di = 0 and passing to the limit when h 0:
Z
C3333 33 + C33 + P333 D3 33 dx =
(4.167)
which implies
e33 (u) = C3333 33 + C33 + P333 D3 ,
(4.168)
e3 (u) = e (u) = 0.
(4.169)
P3 P333 33 + 33 D3 d3 dx = 0,
1
P3 + P333 33 .
33
P333
e33 (u) = C3333 33 + C33 +
P3 + P333 33 ,
33
from which we arrive at (4.160) with
1
,
Ac33 =
Ac = 33 C33 + P333 P3 .
(h) e (v)dx + 2h
3 (h) e3 (v)dx = h
Z
fi vi dx +
gi vi d .
dN
for all v W2
v dx = 0,
(4.170)
x d =
d = 0.
(4.171)
In a similar way, taking v = (0, 0, v3) and di = 0 in (4.27) and multiplying by h one has
h
3 (h) e3 (v)dx = h
Z
fi vi dx +
dN
gi vi d .
P3 3 + D d dx =
E ()d dx
(4.172)
117
Let us now consider the restriction of equation (4.26) to v VBN . Then since e3 (u) =
e (u) = 0 we obtain
Z
Dk (h)Ek () dx =
33 e33 (v)dx +
fi vi dx +
D E () dx =
fi vi dx +
gi vi d,
dN
gi vi d,
(4.173)
dN
for all (u, ) VBN 0 . Taking di = 0 and substituting (4.160) in the previous equation
we deduce
Z
Z
Z
c
c
A33 33 e33 (u) A e33 (v)dx = fi vi dx +
gi vi d,
(4.174)
dN
for all u VBN . For homogeneous anisotropic material, the coefficient Ac33 Ac does not
depend either on x1 or x2 and therefore equation (4.174) becomes, by property (4.171),
Z
fi vi dx +
dN
gi vi d, v VBN .
(4.175)
det C
= Ac33 33 ,
det M det N
D E () dx = 0, for all 0 .
(4.176)
in [H 1 ()]3 ,
(4.177)
Proof. From uniqueness of solution of (4.114) and comparing with (4.163) we conclude
that u = u0 and u3 = u03 and therefore u = u0 . In section 4.4.1 it was showed that
(u(h)) is weakly convergent in V0,w (). By unicity of the limit this implies that the whole
sequence (u(h))h>0 converges to u0 .
Remark 14 We note that the variational problem (4.176) cannot be expressed in order
We shall now establish the weak convergence of sequence (u(h), (h))h>0 to the first
term of the asymptotic expansion (u0 , 0 ) V0,w 0 , for the anisotropic homogeneous
material of class 6mm.
Theorem 16 Let us consider that the beam is made of a material anisotropic of class
6mm. Sequences (u(h))h>0 and ((h))h>0 satisfy as h 0
u(h) u0
in [H 1 ()]3 ,
(4.178)
(h) 0
in L2 ()
(4.179)
Proof. From (2.21)-(2.23) and using some algebraic software tools, we show that
C11 = C1111 = C2222 ,
33
e33 (u)e33 (v)dx =
fi vi dx +
gi vi d.
(4.180)
dN
E2 () = P232 32 + 22 D2
(4.181)
1
(E2 () + P14 32 ).
11
(4.182)
or equivalently,
D1 =
1
(E1 () + P14 31 ),
11
D2 =
119
1
[ (E1 () + P131 31 )]E1 () dx +
11
1
(E2 () + P232 32 )]E2 () dx = 0,
22
(4.183)
for all 0 . If, in the previous situation, we consider the 6mm material case, or,
equivalently
Z
1
E ()E () dx +
11
P14 3 E () dx = 0,
for all 0 .
(4.184)
1
Choosing in (4.184) test functions of the form = (x1 , x2 )(x3 ) with H0,
()
eD
1
and H0 (0, L), we have
L
0
Z
Z
Z L
1
E () d dx3 +
P14
3 d dx3 = 0,
11
0
(4.185)
and consequently we find that is a solution of the following problem a.e. in [0,L]
Z
1
E () d + P14
11
3 d = 0,
(4.186)
for all H0,eD (). For the homogeneous case, the coefficient P14 is independent on x1
and x2 , then from property (4.153) satisfied by function 3 we obtain the second term
of (4.186) is identically zero and the equation reads
Z
1
d = 0,
11
(4.187)
for all H0,eD (). We prove, using algebraic manipulation software tools, that
1
P14 P14
= 11 +
,
11
C66
and therefore the previous equation coincides with (4.121).
From the unicity of the limits u = u0 , u3 = u03 and = 0 we conclude that the weak
convergence (4.178) and (4.178) hold for the whole sequence u(h) and (h), respectively.
4.5
This section is subdivided in two section in which models for materials of class 2 and 6mm
will be written for the original beam h .
4.5.1
We now return to the actual beam h and define the following spaces (cf. (4.82)-(4.85)):
h
Q = Q ( ) = { H ( ) :
d h = 0},
h
S h = S h ( h ) = { H 1 ( h ) : = 0 on eD
},
T h = T h ( h ) = Qh S h ,
hl = hl (h ) = L2 (0, L; H 1 ( h )),
Rh = Rh (h ) = L2 (0, L; Qh ( h )),
hl0 = hl0 (h ) = L2 (0, L; S h ( h )).
Given the scalings (3.36), (4.12) and (4.24), the developments (4.36), (4.38 and (4.39)
induce formal developments on uh , h , h and Dh , respectively, whose leading terms
we will identify and characterize in the following. For that we will undo the change of
variable xh = h (x) and accordingly define the de-scaled quantities:
h (xh3 ) = h1 (x3 ),
3h (xh3 ) = 3 (x3 ),
0h (xh ) = h0 (x) ,
r h (xh ) = hr (x) ,
C33 h h h d h ,
for all h H 1 ( h ),
(4.188)
J =h J =
C33 (h h w h )(h h w h ) d h
C33 (h h w h )h d h , (4.189)
121
and the second moment of area of the cross section with respect to axis Oxh ( 6= )
Ih
(xh )2 d h ,
where 1h (xh1 , xh2 ) = xh2 , 2h (xh1 , xh2 ) = xh1 . We also define the resultant of the applied loads
and moments in each cross section (cf. (4.78)-(4.79)):
Fh (xh3 )
F3h (xh3 )
Mh (xh3 )
= h F (x3 ) =
2
= h F3 (x3 ) =
3
= h M (x3 ) =
+
+
h
dN
h
dN
h
dN
For a generic function h : R we define its average along the xh3 -axis (cf. (4.100)),
(xh1 , xh2 )
1
=
L
L
0
(z )
(xh3 )
1
= h
J
P3 (h w h h )h (h 0h (xh3 ) + h h0 (xh3 )) d h ,
Z
Z
h h h
h
=
P3 0 d
h h0 h d h ,
h
for all (, ) T h .
(4.190)
4.5.2
(4.191)
(4.192)
where
C16
x1 ,
C11 + C13
C16
2 (x1 , x2 ) =
x2
C11 + C13
1 (x1 , x2 ) =
C16
(x1 , x2 ) = ( )(x1 , x2 ) =
(C11 + C13 )
(x21 x22 )
2
x1 x2
x1 x2
(x22 x21 )
2
The warping function warping function w h (xh1 , xh2 ) as the unique solution of the variational problem:
Z
w H ( ),
wh = 0
h
Z
Z
w h h d h =
(xh2 1h h xh1 2h h ) d h ,
h
(4.193)
h
for all H ( ).
h,0
Corollary 20 The approximations (uh,0 , h,0, 33
, Dh,0) are uniquely characterized as
follows:
h
uh,0
= (x3 ),
h
h
uh,0
3 = x ( ) (x3 ),
h
uh,1
= z ,
(4.195)
(4.196)
P14 h,0
,
C14
(4.197)
h,0 hl ,
(4.198)
h
h h
uh,1
3 = z3 x (z )
h,0 = h,0 h ,
(4.194)
123
h,0
h,0
= 3
= 0,
(4.199)
h,0
33
= Y e33 (uh,0) = Y [(h ) x (h ) (x3 )],
(4.200)
(4.201)
h
h
Y
A(
)
= F3h
3h (0) = 3h (L) = 0,
in (0, L),
(4)
Y Ih h
= Fh + Mh
in (0, L),
h,0 = 0 in h ,
R11
R11 h h,0 nh = 0 on eN
,
h,0 = h on h ,
0
eD
(4.202)
(4.203)
C11 + C13
33 C66 + P36 P36
P14
1
= P14
+ 11 =
.
C44
11
Y =
(4.204)
R11
(4.205)
We note that for the homogeneous transversely isotropic beam model, the electrical
and mechanical phenomena are decoupled. The boundary problems found in (4.194)(4.202) of Corollary 20 are respectively called the one-dimensional bending equations and
the one-dimensional stretching equations of a linearly elastic beam. Together with the
property that u(0) is a Bernoulli-Navier displacement field (Theorem 15), they constitute
the linear Bernoulli-Navier model of a linearly elastic beam. Furthermore, the electric
potential can be obtained through a two-dimensional Laplaces equation, expressed by
the boundary value problem (4.203).
Chapter
In many applications weakly curved beams are used instead of straight beams. In this
chapter, a weakly curved beams is a beam which the length of the centreline much
greater than the diameter of the cross section and the curvature is of the order of the
diameter of the cross section. This type of curved beams are also known as shallow arches.
We present a zeroth-order model for a transversely isotropic - 6mm symmetric class piezoelectric shallow arch under the influence of an applied electric potential on both end
faces, obtained by asymptotic methods.
Let us briefly outline the content of this chapter, which closely follows Alvarez-Dios
& Via
no [1996]. In the next two sections we introduce the notation and present the
piezoelectricity problem in its Hellinger-Reissner variational principle. In Section 5.3 we
rescale the three-dimensional problem posed in a straight reference rod. In Sections 5.5
5.1
As before, let h and L be two positive scalars and let h denote an open bounded, simply
connected subset of R2 , with Lipschitz continuous boundary bounded h having area
A( h ) = h2 . We suppose that system Oxh1 xh2 x3 is a principal system of inertia associated
to h , therefore axis Ox3 passes through the mass center of h {x3 }, which means that
Z
h {x3 }
xh
d =
(5.1)
h {x3 }
parameterized by its arc length sh (x3 ), x3 [0, L]. The Frenet trihedron (t,h , n,h , b,h )
1
h
t,h = (t,h
i ) = h (i ),
A
,h
h h h
n1
Ah h
1 1
1
h h h
n,h = n,h
= h h Ah h
2
2 2 ,
A B
h
n,h
h
3
,h
b1
h
2
1
b,h = b,h
h
= h
,
2
1
B
h
h h
b,h
h
3
1 2 2 1
(5.2)
(5.3)
(5.4)
where
h
Ah = |h | = h
+ 1,
(5.5)
h
h h
h h 2
Bh = |h h |2 = h
+ (1 2 1 2 ) .
(5.6)
(5.7)
(5.8)
(5.9)
127
where k h (sh (x3 )) h (x3 ) and h (sh (x3 )) h (x3 ) are the centrelines curvature and
torsion. The curvature and torsion are extracted from the curve parametrization as follows
|h h |
=
h (x3 ) =
|h |3
Bh
,
(Ah )3
h (h h )
1
h
h h
(x3 ) =
= h (h
1 2 2 1 ).
2
h
h
B
h
As assumed in Alvarez-Dios
& Via
no [1998], the family of curves C h , 0 < h < 1,
satisfies the following hypothesis:
(HC1) h C 3 [0, L].
(HC2) For all x3 [0, L], (t,h , n,h , b,h ) is a positive oriented orthonormal basis of R3 .
(HC3) Frenet equations (5.7)-(5.9) hold for curvature h and h belonging to C[0, L].
h h (
h ) R3 in the following manner (Figure 5.1)
Now we define the map h :
h xh = (h1 (x3 ), h2 (x3 ), x3 ) + xh1 n,h (x3 ) + xh2 b,h (x3 )
(5.10)
ij
for all xh h ,
oh (xh ) := det h h (xh )
for all xh h ,
where the scalar oh (xh ) and the vectors bhij (xh ), xh h are of the form (see Alvarez-Dios
& Via
no [1998])
h
Ah (1
xh1 h )
h Ah h ,h
,h
h
h
b1j (x ) = nj +
x2 tj ,
h
o
h Ah h ,h
,h
h
h
b2j (x ) = bj
x1 tj ,
oh
1
bh3j (xh ) = h t,h
,
o j
o (x ) =
Ah
xh1
Bh
,
Ah
(5.11)
(5.12)
(5.13)
(5.14)
0
b
xh
h0
b
hL
h
h
0
b
Figure 5.1: The reference configuration of the shallow arch described in curvilinear and
Cartesian coordinates, and in a fixed domain.
h, h
h , is the union of the end faces
h = h ( h {0}),
h =
The boundary of
0
L
h
h h
h
h
h = h (h ),
dD
0
eN
h = h (h ),
dN
dN
h = h (h ),
eD
dN
hdN = hN hL ,
heD = h0 hL .
h = h (h ),
eN
N
with
The reference configuration of the shallow arch can thus be described in terms of
the three-dimensional curvilinear coordinates, or in terms of the Cartesian coordinates
h } . Let h := /xh (hence h := /xh )
h = (xh ) {
xh1 , x
h2 , x
3 , of the same point x
i
i
129
and ih will stand for the differential operator / xhi . In an analogous way, we denote by
h , h a function defined in h and a function defined in ,
h a function defined in
related by
= (xh ); xh = h (x).
h (
xh ) = h (xh ) = (x),
x
5.2
h } .
We now review the formulation of the linearly piezoelectric problem in the set {
h
We assume that the shallow arch, whose reference configuration is
, is clamped on the
h
h
h
h , to a
portion = and submitted to a mechanical volume force of density f in
dD
The left end of the beam is weakly clamped, and therefore the boundary condition (3.2),
introduced in Chapter 3, is also assumed.
h
Then, the body undergoes a mechanical displacement field u
h = (
uhi ) :
R3 and
h
an electrical potential h :
R satisfying the following equilibrium equations:
h , h ) = fi in
ih
ijh (u
h h h h
, ) n
ij (u
j = gih on
h , h )
i3
(u
= phi on
h
h (u
kh D
h ) = 0 in
k ,
h
h (u
D
h ) n
k = 0 on
k ,
h
= h0 on
N ,
(5.15)
L
h
h ,
eN
(5.16)
eD
h,
in
h
h = 2P h
h
h22 D
E
2
232 32 +
2
h,
in
(5.17)
+ P3
D
h
h
eh12 = 2C1212
12
h,
in
h
h
h h
eh31 = 2C3131
31
+ P131
D1
h,
in
h
h
h h
eh32 = 2C3232
32
+ P232
D2
h,
in
h
h
h
h h
eh33 = C3333
33
+ C33
+ P333
D3
h.
in
(5.18)
These constitutive equations describe piezoelectric materials of crystal class 6mm in which
the components of the elastic, piezoelectric and dielectric material satisfy (cf. (3.8) in
Chapter 3)
h
h
Ciijj
6= 0, Cklkl
6= 0
h
P3
6= 0
for k 6= l
h
h
P3
6= 0, P333
6= 0, hii 6= 0,
(5.19)
h
Pkij
and hij
vanish.
(5.20)
this chapter.
Now, we define the following spaces for the admissible displacements and admissible
electric potentials
n
o
h :=
h (
h ) = h H 1 (
h ) : h = 0 on
h =
h
h ,
0
0
eD
0
L
Z
Z
h
h
h
h
h
1
h
3
h
h
h
h
)= v
)] :
=
=0 .
V0,w := V0,w (
[H (
v
d
xi vj xj vi d
h {0}
h {0}
131
h := V h
h and X
h := [L2 (
h )]9 L2 (
h)
We have already defined the spaces X
0,w
0,w
0
1
s
equipped with the norms,
1/2
h 2
h 2
k(
v h , h )kX
h
=
k
v
k
+
k
k
,
h
h)
V
H 1 (
0,w
0,w
h
h
k(
h , d )kX
h |2L2 ( h ) + |d |2L2 ( h )
h = |
1
1/2
respectively, and k
v h kV h = |
eh (
v h )|0, h .
0,w
h ) such that h = h on
h .
H 1 (
0
eD
h ) is defined by
In addition, the closed convex set on H 1 (
n
o
h :=
h (
h ) = h H 1 (
h ) : h h
h ,
2
2
0
h = V h
h.
and consequently we have X
2,w
0,w
2
hX
h ,
h ), (
((
h, D
uh , h )) X
1
2,w
h
h
h
h,
h ), (
a
hH ((
h, D
h , d )) + bhH ((
h , d ), (
uh , h )) = 0,
(
h , d ) X
1
h
h
h ,
h ), (
bH ((
h , D
v h , h )) = lH
(
v h , h ),
(
v h , h ) X
0,w
(5.21)
h
h
h h
h h
bh ((
v h , h )) =
H , d ), (
h h
lh (
H v , ) =
h
h
h dh d
Pkij
ijh + hkl D
l
k x ,
ehij (
v h ) d
xh
ijh
fih vih d
xh
gih
vih
h
d
(5.22)
dhk Ekh (h ) d
xh ,
(5.23)
h.
phi vih d
(5.24)
h X
h is the unique
h ), (
As we mentioned in Chapter 3, the pair ((
h, D
uh , h )) X
1
2,w
solution of the problem (5.21)-(5.24).
5.2.1
Now our objective is to transform the problem (5.21)-(5.24) expressed in Cartesian coordinates into another problem in curvilinear coordinates. We define the transformation
from the old to the new test functions,
h )]3 v h = v
v
h [H 1 (
h h [H 1 (h )]3 ,
h ) h = h h H 1 (h ),
h H 1 (
h h
h (
h ) and
h (
h ), and
which in turn induce a bijection between the spaces V0,w
( ),
0
2
h
h
We define the spaces X h0,w = V0,w
h0 , X h1 = [L2 (h )]9s L2 (h ) and X h2,w = V0,w
h2 .
d
xh = oh (xh ) dxh .
(5.25)
133
h and h is given by
(b) The area of elements along
h = oh (xh )
d
oh (xh )dh
q
h
h
h
on hN ,
(5.26)
on h0 hL ,
(5.27)
q
o (x ) = bhi (xh )nh (xh )bhi (xh )nh (xh ).
h
(5.28)
on hN ,
(5.29)
on h0 hL .
(5.30)
jh vih (
xh ) = bhkj (xh ) kh vih (xh ),
(5.31)
v h V0,w (h ),
h 0 (h ),
(5.32)
(5.33)
we obtain ehij (
v h ) = ehij (v h ) and Eih (h ) = Eih ( h )
From all these above, the next theorem follows, which gives a new formulation of the
oh (xh ) > 0,
h , D h , (uh , h ) X h1 X h2,w defined by
h (xh ) = h (
xh ),
h (
h (xh ) =
xh ),
h (
D h (xh ) = D
xh ),
ahH h , D h , ( h , dh ) + bhH h , dh , uh , h = 0,
for all h , dh X h ,
1
h
(v h , h ),
bhH h , Dh , v h , h = lH
(5.34)
(5.35)
for all v h , h X h ,
0,w
where
h ), ( h , dh ))
ahH ((
h, d
h
h h
Cijkl
klh + Pkij
dk ijh oh (xh ) dxh
Z
+
h ), (v h , h ))
bhH ((
h, d
h
lH
(v h , h )
fih
=
vih
Z
h
h h
Pkij
ij + hkl dhl dhk oh (xh ) dxh ,
o (x )dx
h
N
h
L
(5.36)
(5.37)
(5.38)
h,
for all x
h = h (xh )
h ,
gih (xh ) = gih (
xh ) for all x
h = h (xh )
N
phi (xh ) = phi (
xh )
h ,
for all x
h = h (xh )
L
h
h
Cklij
(xh ) = Cklij
(
xh ),
h
h
Pkij
(xh ) = Pkij
(
xh ),
h h
ij (x ) = hij (
xh ),
xh = h (xh ), xh h .
5.3
135
As usual, our first task is to define a problem equivalent to problem (5.34)-(5.38), but now
posed over a domain that does not depend on h. For such, we use the transformation (3.28)
introduced in Chapter 3, and consequently the scalings (3.36)-(3.39) to the unknowns and
R
the assumptions (3.40)-(3.43) on the data. The scalings for the functions bij (h) :
R are defined by:
and o(h) :
bij (h)(x) := bhij (xh ),
o(h)(x) := oh (xh ),
o(h)(x) := oh (xh ),
(5.39)
(5.40)
Using the scalings and the assumptions on the data, we can recast the variational
for all ( , d) X ,
1
(5.41)
(5.42)
C o(h)dx,
dk Ek ()o(h)dx,
+2
( , d)) =
aH,0 ((
, d),
+
2C33 3 + P3 d 3 o(h)dx
P3 d3 o(h)dx +
2P3 3 + d d o(h)dx,
P333 33 + 33 d3 d3 o(h)dx,
fi vi o(h) dx
gi vi o(h) o(h) d
pi vi o(h) d,
(5.43)
and
0 () = H 1 () : = 0 on eD ,
2 () = H 1 () : 0 0 () ,
o
n
1
3
V0,w () = v H () : hvi = 0 on dD .
(5.44)
(5.45)
(5.46)
The pair (((h), D(h)) , (u(h), (h))) can also be characterized as the unique solution
of this problem thanks to the Lions-Stampachias theorem and the Lax-Milgrams Lemma.
In the next three theorems there will be established some geometrical, mechanical
and electrical preliminaries needed in the sequel for the asymptotic analysis of linearly
piezoelectric shallow arches. The geometrical preliminaries results are due to Alvarez-Dios
& Via
no [1998] and are summarized in the next two Lemmas.
Lemma 10 (Alvarez-Dios
& Via
no [1998]) Let h C3 (0, L; R3 ) be such that its
Frenet trihedron (t,h , n,h , b,h ) is a positively oriented basis of R3 and satisfies Frenet
equations for curvature h and torsion h . If h satisfies (5.40) then, for all h > 0, we
have:
{Ah }r = 1 + h2 s0 (h, ), r R,
h = h{c + h2 s1 (h, )}, h = d + h2 s2 (h, ),
2
n,h
1 = b1 + h s3 (h, ),
,h
2
2
n,h
2 = b2 + h s4 (h, ), n3 = h{h1 + h s5 (h, )}
137
2
b,h
2 = b1 + h s7 (h, ),
2
b,h
3 = h{h2 + h s8 (h, )},
,h
,h
2
2
2
t,h
1 = h{t1 + h s9 (h, )} t2 = h{t2 + h s10 (h, )}, t3 = t3 + h s11 (h, ),
max
h>0
x3 [0,L]
(5.47)
and c, d, b , h and ti are functions defined in [0, L], independent of h and satisfying the
following properties:
(H1b ) b21 + b22 = 1,
(H2b ) b1 h1 b2 h2 = t1 = 1 ,
(H3b ) b1 h2 + b2 h1 = t2 = 2 ,
(H4b ) b1 = d b2 ,
b2 = d b1 .
Moreover, if 1 (x3 )2 (x3 ) 6= 0 for all x3 [0, L], then we have (x3 ) 6= 0, for all h > 0
and for all x3 [0, L], and also
c=
(1 )2 + (2 )2 ,
(5.48)
b = /c,
h1 =
(5.49)
1 1 + 2 2
,
c
t = ,
d=
h2 =
1 2 2 1
,
c
t3 = 1,
1
2 2 1
.
c2
(5.50)
(5.51)
(5.52)
We note that the limits of these functions for h = 0 are functions of x3 [0, L] only, i.e.,
Lemma 11 (Alvarez-Dios
& Via
no [1998]) There exists h0 = h0 (, c, d, b , h ) such
h
h and for all h h0 ,
that the Jacobian matrix h (xh ) is non-singular for all xh
h and
and h is an orientation-preserving map for all h h0 . We have for all xh
for all h h0 :
o(h) = 1 + h2 o# (h, ),
o(h) = 1 + h2 o# (h, ),
3 #
b(h)(x) = C 0 (x) + hC 1 (x) + h2 C #
2 (h, )(x) + h C 3 (h, )(x),
(5.53)
(5.54)
b1 (x3 )
b2 (x3 ) 0
0
0
1
0
0
h1 (x3 ) + x2 d(x3 )
C 1 (x) = 0
0
h2 (x3 ) x1 d(x3 ) ,
t1 (x3 ) t2 (x3 )
0
#
#
b (h, ) b12 (h, )
0
11
#
#
C#
(h,
)(x)
=
b
(h,
)
b
(h,
)
0
,
2
21
22
#
0
0
b33 (h, )
0
0
b#
(h,
)
13
#
C#
(h,
)(x)
=
0
0
b
(h,
)
,
3
23
#
#
b31 (h, )(x) b32 (h, )
0
(5.55)
(5.56)
(5.57)
(5.58)
sup
0<hh0
sup
0<hh0
max |
o# (h, )(x)| C0 ().
(5.59)
(5.60)
(5.61)
h with
Using all these results we establish a relation for the electric vector field in
respect to the electric vector field in h and complete the Lemma 4.4. introduced by
Alvarez-Dios
& Via
no [1998].
139
h } . Then
h = h (h (x)) {
for all x
h = h1
1 h x
h = h1
2 h x
(b1 1 b2 2 ) + h2 1 3 + h2 b#
1 (h, )
+h4 b#
31 (h, )3
(b2 1 + b1 2 ) + h2 2 3 + h2 b#
(h,
)
+h4 b#
32 (h, )3
(x),
(5.62)
(x),
(5.63)
o
n
h = (h1 + x2 d)1 + (h2 x1 d)2 + 3 + h2 b#
3 h x
(h,
)
(x),
i
i3
(b1 1 v b2 2 v ) + h2 1 3 v + h2 b#
1 (h, ) v
h
h
2
=h
1 v x
(x),
+h4 b#
(h,
)
v
3
31
2
2 #
(b
v
)
+
h
v
+
h
b
(h,
)
v
1 1 3
2 2 3
3
1 3 3
1
h = h1
1 v3h x
(x),
+h4 b#
(h,
)
v
3 3
31
(b2 1 v + b1 2 v ) + h2 2 3 v + h2 b#
(h,
)
v
2
h
h
2
=h
2 v x
(x),
+h4 b#
(h,
)
v
3
32
2
2 #
(b2 1 v3 + b1 2 v3 ) + h 2 3 v3 + h b2 (h, ) v3
h
h
1
=h
2 v3 x
(x),
+h4 b#
32 (h, )3 v3
v
+
(h
+
x
d)
v
+
(h
x
d)
v
3
1
2
1
2
1
2
h = h1
3 vh x
(x),
+h2 b# (h, ) v
i3
(5.64)
(5.65)
(5.66)
(5.67)
(5.68)
(5.69)
o
n
h = 3 v3 + (h1 + x2 d)1 v3 + (h2 x1 d)2 v3 + h2 b#
3 v3h x
(h,
)
v
i 3 (x).
i3
(5.70)
Consequently, we obtain
Eh (h )(
xh ) = h1 {E (h)()}(x),
h (h )(
E
xh ) = E3 (h)()(x),
3
(5.71)
and
where
eh (
v h )(
xh ) = h2 {e (h)(v)}(x),
eh3 (
v h )(
xh ) = h1 {e3 (h)(v)}(x) (v) ,
eh (
h
xh ) = e33 (h)(v)(x),
33 v )(
(5.72)
(5.73)
E2 () = (b2 1 + b1 2 ),
(5.74)
(5.75)
e11 (v) = b1 1 v1 b2 2 v1 ,
(5.76)
e22 (v) = b2 1 v2 + b1 2 v2 ,
1
[b1 (1 v2 + 2 v1 ) + b2 (1 v1 2 v2 )] ,
2
1
e13 (v) = [b1 1 v3 b2 2 v3 + 3 v1 + (h1 + x2 d)1 v1 + (h2 x1 d)2 v1 ] ,
2
1
e23 (v) = [b2 1 v3 + b1 2 v3 + 3 v2 + (h1 + x2 d)1 v2 + (h2 x1 d)2 v2 ] ,
2
e12 (v) =
(5.77)
(5.78)
(5.79)
(5.80)
#
and e#
ij (h, ; v) and Ei (h, ; ) defined by
2 #
E1# (h, ; ) = 1 3 + b#
(h,
)
+
h
b
(h,
)
3
1
31
2 #
E2# (h, ; ) = 2 3 + b#
(h,
)
+
h
b
(h,
)
3
2
32
E3# (h, ; ) = b#
i3 (h, )i ,
(5.81)
(5.82)
(5.83)
2 #
e#
11 (h, ; v) = 1 3 v1 + b1 (h, ) v1 + h b31 (h, )3 v1 ,
#
#
v
+
v
+
b
(h,
)
v
+
b
(h,
)
v
2
1
1
2
2 3 1
1 1 3 2
,
e#
12 (h, ; v) =
2
+h2 b# (h, ) v + h2 b# (h, ) v
31
3 2
32
(5.84)
(5.85)
3 1
2 #
e#
22 (h, ; v) = 2 3 v2 + b2 (h, ) v2 + h b32 (h, )3 v2 ,
1
#
#
2 #
e#
(h,
;
v)
=
v
+
b
(h,
)
v
+
b
(h,
)
v
+
h
b
(h,
)
v
3 3
3
i
3 3 ,
3
3
i3
2
#
e#
33 (h, ; v) = bi3 (h, )i v3 .
(5.86)
(5.87)
(5.88)
sup
0<hh0
max |e#
ij (h, ; v)|0, C1 ()kvk1, ,
(5.89)
(5.90)
i,j
141
1 (
x )=h
(b1 1 b2 2 ) + h r 1 (h, ; ) + h s1 (h, ; ) (x),
(5.91)
n
o
4 #
2 h (
xh ) = h1 (b2 1 + b1 2 ) + h2 r #
(h,
;
)
+
h
s
(h,
;
)
(x),
2
2
n
o
xh ) = (h1 + x2 d)1 + (h2 x1 d)2 + 3 + h2 r#
(h,
;
)
(x),
3 h (
3
max |r #
i (h, ; )|0, +
i
sup
max
0<hh1
|s#
(h, ; )|0,
(5.92)
C1 ()kk1, .
Theorem 18 Let the scalings to the unknown and to the data be as above. Then, for each
h h0 , the scaled unknown (((h), D(h)) , (u(h), (h))) satisfies the following variational
problem (cf. (5.41)-(5.43)):
Find a pair (((h), D(h)) , (u(h), (h))) X 1 X 2,w such that
for all ( , d) X ,
1
(5.93)
#
(5.94)
= lH
(v, ) + h2 lH
(h, )(v, )
bH (( , d), (v, ))
ij eij (v) dx
dk Ek ()dx,
Z
( , d)) =
aH,4 ((
, d),
C dx,
Z
Z
( , d)) =
aH,2 ((
, d),
C33 33 + P3 d3 dx + C33 33 dx
Z
Z
+2
2C33 3 + P3 d 3 dx
P3 d3 dx
Z
+
2P3 3 + d d dx,
and
Z
=
C3333 33 + P333 d3 33 dx +
Z
Z
Z
lH (v, ) = fi vi dx
gi vi d
pi vi d.
( , d))
aH,0 ((
, d),
P333 33 + 33 d3 d3 dx,
#
#
The bounded bilinear forms b#
H (h, )(, ),aH (, ) and the linear form lH (h, ) are defined
by
b#
H (h, )(( , d), (v, ))
( , d))
a#
, d),
H ((
Z
=h
C o dx + h
#
(5.95)
P3 d3 o# dx
C33 33 P3 d3 o# dx + 2h2
2C33 3 P3 d 3 o# dx
Z
Z
+ h2 C33 33 o# dx + h2
2P3 3 + h2 d d o# dx
Z
Z
#
+
C3333 33 P333 d3 33 o dx +
P333 33 + 33 d3 d3 o# dx,
(5.96)
Z
Z
#
#
lH (h, )(v, ) = fi vi o (h, )dx
gi vi [o# (h, ) + o# (h, )]d
N
Z
Z
#
2
pi vi o (h, )d h
gi vi o# (h, )
o# (h, )d.
(5.97)
+ h2
143
Applying inequalities (5.60) and (5.61), it is easy to verify that the last bilinear and linear
forms are bounded.
5.4
by
n
o
VBN
= VBN
() = v V0,w : e (v) = e3 (v) = 0 ,
VBN
= v : R3 : v (x1 , x2 , x3 ) = (x3 ), V02 (0, L),
where
and
v3 (x1 , x2 , x3 ) = 3 (x3 ) b (x1 , x2 , x3 ) (x3 ), 3 V01 (0, L) ,
V01 (0, L) = H 1 (0, L) : (0) = 0 ,
V02 (0, L) = H 2 (0, L) : (0) = (0) = 0 ,
b1 (x1 , x2 , x3 )
b2 (x1 , x2 , x3 )
b1 (x3 ) b2 (x3 )
b2 (x3 ) b1 (x3 )
x1
x2
(5.98)
(5.99)
(5.100)
b = p
,
(1 )2 + (2 )2
(5.101)
(5.102)
b21 + b22 = 1.
(5.103)
In Alvarez-Dios
& Via
no [1998] we can see the proof of the following equivalence
between the norms
|v|V
BN
= e33 (v)
0,
C2 () kvk1, ,
v VBN
.
2
2
:= () = L () : Ei () L () ,
(5.104)
kk =
X 2
||2 +
Ek ()
0,
)1/2
Theorem 19 Let be a domain in R3 and let eD be such that meas (eD ) > 0.
Then, there exists a constant C3 () such that
||
0
:= E ()
0,
(
)1/2
X 2
=
C3 () kk1, ,
Ek ()
k
0 .
0,
(5.105)
b1
b2
2 =
b2
b1
0
3
(h1 + x2 d) (h2 x1 d) 1
1
11
b1 b2 0
0
12 0
0 b2 b2
=
21 b2 b1 0
0
0
0 b2 b1
22
E1 ()
E () H 1 () 3 ,
2
E3 ()
1 E1 ()
2 E1 ()
4
1
.
H
()
1 E2 ()
2 E2 ()
Now, deriving (5.75) with respect to the variables x1 and x2 , respectively, we deduce from
previous conditions that
31 = (h1 + x2 d)11 + (h2 x1 d)22 1 E3 (),
32 = (h1 + x2 d)12 + (h2 x1 d)22 2 E3 (),
145
In work of Duvaut & Lions [1976], it is shown that any H 1 () such that i
H 1() in fact belongs to L2 (), and so (i) is proved.
(ii) It is easy to show that the identity mapping from H 1 () equipped with kk1,
into equipped with kk is continuous, since there clearly exists a constant d such
that kk d1 kk1, for all H 1 (), and surjective, thanks to the step (i). The
inverse mapping H 1 () 7 E () is continuous, that is
kk1, d2 kk
for all ,
z H01 (0, L) .
b1 1 b2 2 = 0,
(5.106)
b2 1 + b1 2 = 0.
(5.107)
(5.108)
b2 (1 2 ) + b1 (1 + 2 ) = 0.
(5.109)
1 + 2 = 0.
(5.110)
11 12 = 0,
11 + 21 = 0,
(5.111)
21 22 = 0,
12 + 22 = 0.
(5.112)
and
147
z2 (x3 ) = z3 (x3 ).
(5.113)
0,
C kk1, ,
3 .
(5.114)
sup
(,d)X1
5.5
bH (( , d) , (v, ))
|( , d)|
n
o
C kvk21, + kk21,
for all (v, ) VBN
3 .
Repeating the argument used in Section 3.3, we can see that the scaled unknowns
(u(h), (h)) given in problem (5.41)-(5.43) converge in [H 1 ()]3 H 1 () as h 0 toward a limit (u, ) and this limit can be identified with the solution of a one-dimensional
variational problem, which will be later identified.
5.5.1
Weak convergence
Proposition 7 There exists C > 0, independent of h, such that for all 0 < h 1 the
solution (((h), D(h)), (u(h), (h))) of problem (5.41)-(5.43) verifies
|33 (h)|0, C(),
h |3 (h)|0, C(),
h |D (h)|0, C(),
k(u(h), (h))k
X0,w C(),
h2 | (h)|0, C(),
(5.115)
(5.116)
(5.117)
S (h) = h2 (h),
S3 (h) = h3 (h),
T (h) = hD (h),
T3 (h) = D3 (h),
(5.118)
(5.119)
(C3333 S33 (h)S33 (h) + 33 T3 (h)T3 (h) o(h)dx
C33 S (h)S33 (h) + 4C33 S3 (h)S3 (h) o(h)dx
T (h)T (h)o(h)dx
C S (h)S (h)o(h)dx.
o# (h, )
#
2o (h, )
h
2
h,
(5.120)
12 C2 h
2
C1 12 C2 h3 12 C2 hC0 () |(S(h), T (h))|20, .
dx
(5.121)
149
=
+
+
gi ui(h)(1 + h2 o# )(1 + h2 o# )d +
fi ui(h)(1 + h2 o# )dx
pi ui (h)(1 + h2 o# )d
(5.122)
D3 (h)E3 ()(1
+ h2 o# )dx.
Using the estimates (5.60) and (5.61) for o# and o# , respectively, the Youngs inequality
2
2ab am + mb2 for m > 0 and the fact that fi L2 (), gi L2 (N ), pi L2 (L ) and
H 1 (0, L), we obtain
bH ((S(h), T (h)), (u(h), (h)))
C3 + h2 C0 () (kf k0, + kgk0, + kpk0, ) ku(h)k1,
+ C4 + h2 C0 () k(h)k
2
(C5 + h2 C5 ()) k(u(h), (h))k
X0,w +
(C6 + h2 C6 ())2
+ mkT3 (h)k20, .
2m
Combining the previous inequality with (5.121), we deduce the existence of h7 small
enough, when m is large enough, and of constant C7 () such that
k(S(h), T (h))k2X1 C7 () k(u(h), (h))k
X0,w ,
if 0 < h h7 .
(5.123)
X1 in equation (5.41) we get
Now, putting ( , d) = e, E
(u(h), (h)))
(u(h), (h))) + bH ((e , E),
(0, (h)))
bH ((e , E),
= bH ((e , E),
+ bH ((e , E),
(0, (h))),
= a ((S(h), T (h)), (e, E))
(5.124)
and taking into account bound (5.60) for o# (h, ) and the fact that 0 < h 1 we
guarantee the existence of h9 small enough and of constant C9 () satisfying
(u(h), (h)))
C8 ()(k(S(h), T (h))k
bH ((e, E),
X1
0,
+ 1)k(e, E)k
X0,w ,
(5.125)
| bH ((e, E),(u(h),
(h)))|
=
eij (h)(u(h))eij (h)(u(h))o(h)dx
Z
+
Ek ((h))E
k ((h))o(h)dx
Z n
o2
=
eij (h)(u(h)) + h2 e#
(h,
;
u)
(1 + h2 o# (h, ))dx
ij
Z
o2
n
2 #
+
Ek ((h))
+ h Ek (h, ; (h))
(1 + h2 o# (h, ))dx
(5.126)
By Alvarez-Dios
& Via
no [1998] we have
|eij (h)(u(h))|20,
2 #
=
eij (h)(u(h)) + h eij (h, ; u) dx C10 () ku(h)k21, .
(5.127)
Using the same argument and applying (5.105), we prove the existence of h11 small enough
and C11 () such that
2
|Ek (h)((h))|
0,
Z n
Ek ((h))
+h
o2
dx
h3 #
h
#
2
2
2
4
2
|E
(
(h))|
|E (h, ; (h))|
|Ek ((h))|
+
h
|E
(h,
;
(h))|
0,
0,
0,
0,
k
2 k
2 k
1
h3 #
2
2
|Ek ((h))|
0,
0,
2
2
)
(
2
1
1
h3
2
2
(5.128)
1,,
1, C11 ()k(h)k
2 C3 ()
2
if 0 < h h11 . Therefore, we prove the existence of h13 small enough and C13 () such
that
Z
Z
2
2
(u(h), (h)))|
| bH ((e, E),
C12 ()
(eij (u(h))) dx + (Ek ((h)))
dx
2
C13 ()k(u(h), (h)k
X 0,w ,
if 0 < h h13 .
(5.129)
Combining (5.125) with (5.129), and applying Babuska-Brezzi condition, we deduce the
existence of h14 small enough and C14 () such that
k(u(h), (h))k
X0,w
(u(h), (h)))
bH ((e, E),
k(u(h), (h))kX0,w
151
if 0 < h h14 . The previous inequality together with (5.123) allows to obtain,
k(u(h), (h))k
X0,w C(),
(5.130)
+ (0, (h))k
X2,w
k(u(h), (h))k
X 0,w + k(h)k
1,(0,L) C()
(5.131)
in L2 ()
(5.132)
in L2 ()
(5.133)
u(h) u,
in V0,w (),
(5.134)
(h) ,
in 2,w (),
(5.135)
(h)
in 0,w ().
(5.136)
33 (h) 33 ,
h3 (h) 3 ,
hD (h) D ,
h2 (h) ,
D3 (h) D3 ,
(5.137)
d =
(5.138)
x d = 0,
(5.139)
e3 (u) = e (u) = 0
(5.140)
E3 () = P3 P333 33 + 33 D3 ,
(5.141)
E () = 0,
(5.142)
e#
(h, ; u) = C + C33 33 + P3 D3 ,
(5.143)
2C33 3 + P3 D = 0,
(5.144)
D 2P3 3 = 0,
h
i
Ad33 P333 e33 (u)E3 () E3 ()e33 (v) dx
Z
dx
e33 (u)33 + E3 ()d3 dx = 0,
(5.145)
33 e33 (v)
D3 E3 ()
dx =
fi vi dx +
gi vi d +
N
pi vi d,
(5.146)
where Ad33 =
1
3333 33 +P333 P333 .
C
e#
(h, ; u) = C + C33 33 + P3 D3 ,
since e3 (u) = e (u) = E () = 0. Setting 33 = = d3 = 0 in equation (5.42) and
multiplying it by h1 and passing to the limit we deduce (5.144).
We multiply equation (5.42) by h2 . Taking the limit when h 0 we obtain (5.137)
when h goes to zero. Choosing now = 0 and the test functions (vi ) V0,w as in
Z
h
i
2 #
2
33 (h) e33 (v) + h e33 (h, ; v) o(h)dx + 2h
3 (h) e#
3 (h, ; v)o(h)dx
Z
Z
#
2
2
+h
(h) e (h, ; v)o(h)dx + h
D (h)E# (h, ; )o(h)dx
Z
h
i
2 #
+ D3 (h) E3 () + h E3 (h, ; ) o(h)dx
Z
Z
Z
=
fi vi o(h)dx +
gi vi o(h)
o(h)d +
pi vi o(h)
o(h)d,
33 e33 (v) dx
+
e#
(h, ; v) dx
gi vi d +
153
D3 E3 () dx
fi vi dx
pi vi d,
33 e33 (v) dx
+
gi vi d +
dx +
D3 E3 () dx
fi vi dx
(5.147)
pi vi d.
From (5.138) we obtain that the second term in the above equation vanishes, and therefore
the previous equation reads.
Z
33 e33 (v) dx
where
33
D3
D3 E3 () dx
C3333 P333
P333 33
fi vi dx +
!1
gi vi d +
pi vi d.
(5.148)
Applying properties (5.138) and taking into account that the coefficients Ad33 (
33 C33 +
P333 P3 ) and Ad (P333 C33 + 33 P3 ) do not depend on x , we obtain mixed variational
33
problem (5.145)-(5.146).
5.5.2
In order to be able to show that the coefficient vanishes in expressions (5.139) and
(5.141), we use now the displacement-electric potential-stress-electric displacement approach instead of the displacement-electric potential approach used in Section 3.4.
We assume that the solution of a problem (5.93)-(5.97) can be expressed as the asymptotic developments
(h) = h4 4 + h2 2 + 0 + h2 2 + . . .
ij L2 (),
D(h) = h4 D4 + h2 D 2 + D 0 + h2 D 2 + . . . ,
(u(h), (h)) = (u0 + h2 u2 + . . . , 0 + h2 2 + . . .),
Dk L2 (),
(5.149)
(5.150)
(5.151)
aH,0 ( 4 , D 4 ), ( , d) = 0
for all ( , d) X ,
1
bH ( 4 , D4 ), (v, ) = 0
for all (v, ) X .
0,w
2
4
2
4
a
(
,
D
),
(
,
d)
=
a
(
,
D
),
(
,
d)
H,0
H,2
#
aH (h, ) ( 4 , D4 ), ( , d) ,
for all ( , d) X 1 ,
bH ( 2 , D 2 ), (v, ) = 0
for all (v, ) X .
0,w
Problem P0 reads:
aH,0 ( 0 , D0 ), ( , D) = aH,2 ( 2 , D 2 ), ( , d)
a#
(h, ) ( 2 , D 2 ), ( , d) ,
for all ( , d) X 1 ,
bH ( 0 , D 0 ), (v, ) = lH
(v, ),
for all (v, ) X .
0,w
Problem P2 reads:
2
0
2
0
a
(
,
D
),
(
,
D)
=
a
(
,
D
),
(
,
d)
H,0
H,2
a#
(h,
)
(
,
D
),
(
,
d)
,
for all ( , d) X 1 ,
n
bH ( 2 , D2 ), (v, ) = lH
(v, ), (v, ) X 0,w .
(5.152)
(5.153)
(5.154)
(5.155)
(5.156)
(5.157)
(5.158)
155
Furthermore, we have
4
= C e (u0 ),
(5.159)
2
0
(5.160)
2p
p 0,
(5.161)
4
3
= 0,
(5.162)
2
3
= 2C33 e3 (u0 ) P3 E (0 ),
(5.163)
0
3
= 2C33 e3 (u2 ) P3 E (2 ),
(5.164)
2p
3
= 2C33 e3 (u2p+2 ) P3 E (2p+2 ),
p 1,
(5.165)
4
33
= 0,
(5.166)
2
33
= C33 e (u0 ),
(5.167)
0
33
= C33 e (u2 ) + C3333 e33 (u0 ) P333 E3 (0 ),
(5.168)
2p
33
= C33 e (u2p+2 ) + C3333 e33 (u2p ) P333 E3 (2p ),
p 1,
(5.169)
D4 = 0,
(5.170)
D2 = 2P3 e3 (u0 ) + E (0 ),
(5.171)
D0 = 2P3 e3 (u2 ) + E (2 ),
(5.172)
p 1,
D34 = 0,
(5.173)
D32 = P3 e (u0 ),
(5.174)
(5.175)
p 1.
(5.176)
In this section we show that the formal expansion of the tensor (5.149) and (5.150) induced
by (5.151) do not contain any negative powers of h.
4
Choosing 33 = 33
and d3 = D34 for the test functions in the first equation of (5.152)
we obtain
4
33
= D34 = 0 in L2 () .
(5.177)
e3 (u0 ) = 0,
(5.178)
2
0
e33 (u0 ) = C33
+ C3333 33
+ P333 D30 ,
(5.179)
E (0 ) = 0,
(5.180)
0
2
P333 33
+ 33 D30 ,
E3 (0 ) = P3
(5.181)
2
2
4
= D32 = 33
= 0, and 3
= D2 = 0.
(5.182)
and consequently
e (u2 ) + e#
(h, ; u ) = C + C33 33 + P3 D3 ,
in L2 () ,
(5.183)
or, equivalently,
0
0
0
2
C
= e (u2 ) + e#
(h, ; u ) C33 33 P3 D3 ,
(5.184)
0
0
0
+ e#
e11 (u)
0
,
+ e#
e12 (u)
12 (h, ; u ) = 0
0
0
0
+ e#
e22 (u)
22 (h, ; u ) C2233 33 + P322 D3 = 0
(5.185)
157
(5.186)
0
0
+ e#
C33 33
+ P3 D30 e (u)
(h,
;
u
)
dx = 0,
0
0
+ e#
e (u2 ) + e#
(h, ; u ) + e (u)
(h, ; u )
for all L2 (). After some algebraic manipulations, the previous equation reads
Z h
i
#
0
0
e (u2 ) + e#
(h,
;
u
)
+
e
(
u)
+
e
(h,
;
u
)
dx = 0,
u2 ) we obtain that e (u
u2 ) = 0, and
for all L2 (). Putting = e (u
= u2 . Substituting u
by u2 in (5.186) we get
therefore u
2
= 0.
2
= 0, which is equivalent to saying that u2 satisfies (5.185). Then,
We assume then that
combining the previous equation with the expressions (5.179) and (5.181), we obtain
1133 33 + P311 P333 e33 (u0 )
C
0
d
e11 (u2 ) = e#
0 ,
11 (h, ; u ) + A33
0
e12 (u2 ) = e#
12 (h, ; u ),
e22 (u2 )
0
e#
22 (h, ; u )
with
Ad33
1
.
Ad33 =
C3333 33 + P333 P333
5.5.2.2
(5.187)
(5.188)
(5.189)
(5.190)
0
VBN
u0 (x1 , x2 , x3 ) = (x3 ),
u0 (x , x , x ) = (x ) b (x ), V 1 (0, L),
2
3
3 3
3
3 1
3
0
(5.191)
0 = z3 , z3 H 1 (0, L),
0 + 3 : 0 (x1 , x2 , 0) = (0)
(5.192)
= 00 ,
0
(x1 , x2 , L) = (L)
= L0 ,
where
b1 (x1 , x2 , x3 )
b2 (x1 , x2 , x3 )
b1 (x3 ) b2 (x3 )
b2 (x3 ) b1 (x3 )
x1
x2
(5.193)
and
b21 + b22 = 1.
(5.194)
Proof. Taking into account relations (5.177)-(5.178) we get expressions (5.191) for the
(5.195)
e33 (u0 ) = 3 b + ,
(5.196)
(5.197)
E3 (0 ) = 3 0 = z3 ,
(5.198)
E3 (0 ) = 3 0 = z3 +
b1 = b1 x1 b2 x2 ,
1 L
( 00 ),
L 0
(5.199)
b2 = b2 x1 + b1 x2 .
(5.200)
u21 =
s1 (x3 ) +
1 b
s(x3 )
b1 2
1
x
b1 1
(Cm 3 Ce z3 ) + x1 r 0 (x3 )
+x1 b1 (2 2 1 1 ) x1 b2 (1 2 + 2 1 ) + 1
(5.201)
u22 =
s2 (x3 )
1 b
s(x3 )
b1 1
1
x
b1 2
(Cm 3 Ce z3 ) + x2 r 0 (x3 )
x1 b2 (2 2 1 1 ) x1 b1 (1 2 + 2 1 ) + 2
159
(5.202)
where s , s, z L2 (0, L) are arbitrary functions depending only on the variable x3 and
r 0 (x3 ) =
1
1
Cm (2 2 + 1 1 ) 1 1 b1 (2 2 1 1 ) + b2 (1 2 + 2 1 ) ,
b1
b1
x21 x22
2
2
= Cm 2b2 b1 x1 x2 +
b1 + b2 ,
2
2
2
x1 x22
2
2
= Cm x1 x2 b2 b1 2b1 b2
,
2
2
2
x2 x21
2
2
= Cm 2b2 b1
+ x2 x1 b1 + b2 ,
2
2
2
x2 x21
2
2
b2 b1 2b1 b2 x2 x1 .
= Cm
2
2
(5.203)
(5.204)
(5.205)
(5.206)
Ce = Ad33 C1133 P333 + P311 C3333 .
(5.207)
0
[b1 (1 u22 + 2 u21 ) + b2 (1 u21 2 u22 )] = 2e#
12 (h, ; u ),
(5.208)
0
b2 1 u22 + b1 2 u22 = Cm e33 (u0 ) + Ce E3 (0 ) e#
22 (h, ; u ).
(5.209)
#
0
0
b1 1 u21 2 u22 b2 2 u21 + 1 u22 = e#
22 (h, ; u ) e11 (h, ; u ).
(5.210)
(5.211)
1 u21
1 u21
2 u22
2 u22
b22
2 u21
b1 b2
2 u21
1 u22
1 u22
#
#
0
0
= b2 e22 (h, ; u ) e11 (h, ; u ) ,
#
#
0
0
= b1 e22 (h, ; u ) e11 (h, ; u ) ,
(5.212)
(5.213)
(5.214)
0
b21 (1 u22 + 2 u21 ) + b1 b2 (1 u21 2 u22 ) = 2b1 e#
12 (h, ; u ),
(5.215)
respectively. Adding up equation (5.213) with (5.214) and subtracting equation (5.215)
to (5.212) we obtain
1 u21
2 u22
h
i
#
#
0
0
0
= b1 e22 (h, ; u ) e11 (h, ; u ) 2b2 e#
12 (h, ; u ),
(5.216)
and
h
i
#
#
0
0
0
2 u21 + 1 u22 = 2b1 e#
(h,
;
u
)
b
e
(h,
;
u
)
e
(h,
;
u
)
.
2
12
22
11
(5.217)
(5.218)
12 u21 22 u22 = 0,
(5.219)
21 u21 + 11 u22 = 0,
(5.220)
22 u21 + 12 u22 = 0.
(5.221)
b1 b2 12 u21
b1 1
1 1
Cm e33 (u0 )
Ce E3 (0 )
(5.222)
161
b21 + b22 1 2b1 b2 2 .
(5.223)
x21
Cm b21 + b22 1 2b1 b2 2 .
2
(5.224)
b21 22 u22
0
0
= b1 2 2 2 Cm e33 (u ) Ce E3 ( )
= b2 1 b1 1 u21 + b1 Cm 2 e33 (u0 )
= b2 1 b2 2 u21 1 1 Cm e33 (u0 ) Ce E3 (0 ) + b1 Cm 2 e33 (u0 )
= b22 22 u22 + Cm b2 1 e33 (u0 ) + b1 2 e33 (u0 ) .
Using again the relations (5.191), (5.193) and (5.194) we conclude that
22 u22 = Cm 2b2 b1 1 + b22 b21 2 .
Solving this differential equation, we prove that there exist functions s1 and s2 such that
u21 = k2 (x2 , x3 ) + x1 k1 (x2 , x3 ) +
x21
Cm b21 + b22 1 2b1 b2 2 ,
2
x22
Cm 2b2 b1 1 + b22 b21 2 .
2
Putting in (5.218) and (5.219) the components (5.225) and (5.226) we get
u22 = s2 (x1 , x3 ) + x2 s1 (x1 , x3 ) +
1 s1 (x1 , x3 ) = Cm
b21 + b22 1 2b1 b2 2 ,
2 k1 (x2 , x3 ) = Cm 2b2 b1 1 + b22 b21 2 ,
(5.225)
(5.226)
(5.227)
(5.228)
b21 + b22 1 2b1 b2 2 ,
k1 (x2 , x3 ) = k3 (x3 ) + x2 Cm 2b2 b1 1 + b22 b21 2 .
(5.229)
(5.230)
u21 = k2 (x2 , x3 ) + x1 k3 (x3 ) + x1 x2 Cm 2b2 b1 1 + b22 b21 2
+
x21
Cm b21 + b22 1 2b1 b2 2 ,
2
x22
Cm 2b2 b1 1 + b22 b21 2
2
+ x2 x1 Cm b21 + b22 1 2b1 b2 2 .
(5.231)
(5.232)
b21 + b22 1 2b1 b2 2 ,
11 s2 (x1 , x3 ) = Cm 2b2 b1 1 + b22 b21 2 ,
(5.233)
(5.234)
and therefore, we prove the existence of the functions k4 (x3 ), k5 (x3 ), s4 (x3 ), s5 (x3 ) such
that
x22
Cm b21 + b22 1 2b1 b2 2 ,
2
x2
s2 (x1 , x3 ) = s5 (x3 ) + x1 s4 (x3 ) 1 Cm 2b2 b1 1 + b22 b21 2 .
2
Hence
u21 = k5 (x3 ) + x2 k4 (x3 ) + x1 k3 (x3 ) + x1 x2 Cm 2b2 b1 1 + b22 b21 2
2
x1 x22
+
(5.235)
(5.236)
(5.237)
(5.238)
163
(5.239)
s4 (x3 ) = k4 (x3 ) b2 (2 2 1 1 ) b1 (1 2 + 2 1 ) ,
(5.240)
so that we have
u21 = k5 (x3 ) + x2 k4 (x3 ) + x1 s3 (x3 ) + x1 b1 (2 2 1 1 )
x1 b2 (1 2 + 2 1 ) + x1 x2 Cm 2b2 b1 1 + b22 b21 2
2
x1 x22
+
(5.241)
(5.242)
b2
1
1
k4 (x3 ) + (Cm 3 Ce z3 ) + Cm (2 2 + 1 1 )
b1
b1
b1
1
1
1
Cm (2 2 + 1 1 )
(2 2 + 1 1 )
b1
2b1
b21 b22 (2 2 1 1 ) + 2b1 b2 (1 2 + 2 1 ) ,
r 0 (x3 ) =
1
1
(x2 b1 + x1 b2 ) k4 (x3 ) + x1 (Cm 3 Ce z3 ) + x1 r 0 (x3 )
b1
b1
+ x1 b1 (2 2 1 1 ) + x1 x2 Cm 2b2 b1 1 + b22 b21 2
2
x1 x22
x1 b2 (1 2 + 2 1 ) +
u21 = k5 (x3 ) +
(5.244)
u22 = s5 (x3 )
(5.245)
We evaluate now the equation (5.156) for test functions in spaces (v, ) VBN
3 .
2
Then, there exists u2 satisfying (5.185), and consequently
= 0. Clearly, (5.179) and
(5.181) become
0
e33 (u0 ) = C3333 33
+ P333 D30 ,
0
E3 (0 ) = P333 33
+ 33 D30 ,
(5.246)
> 0,
(5.247)
0
33
= Ad33 33 e33 (u0 ) P333 E3 (0 ) ,
(5.248)
D30 = Ad33 P333 e33 (u0 ) + C3333 E3 (0 ) ,
(5.249)
where
1
,
Ad33 =
C3333 33 + P333 P333
(5.250)
E3 (0 ) = z3 .
(5.251)
0
C3333 33
+ P333 D30 33 dx +
0
33
e33 (v)dx +
pi vi d,
0
P333 33
+ 33 D30 d3 dx
d3 E3 (0 )dx
D30 E3 ()dx =
= 0,
fi vi dx +
(5.252)
2
(5.253)
VBN
3 ,
5.5.3
165
Strong convergence
As in Section 3.4.1, and using similar techniques, we prove the following result.
Theorem 20 For 0 < h < 1, let (((h), D(h)), (u(h), (h))) X1 X2 be the solution of
(5.41)-(5.43), and Ad33 (
33 C33 + P333 P3 ) and Ad33 (P333 C33 + 33 P3 ) be independent
on x . Then:
u(h) u0 ,
strongly in [H 1 ()]3 ,
(5.254)
(h) 0 ,
strongly in H 1 (),
(5.255)
strongly in L2 (),
(5.256)
strongly in L2 ().
(5.257)
0
33 (h) 33
,
hD (h) 0,
h2 (h) 0,
h3 (h) 0,
D3 (h) D30 ,
X 1 one has
Proof. For any (
, d)
(
+ aH,2 (
(
+ aH,4 (
(
Ck(
2 . (5.258)
aH,0 (
, d),
, d)
, d),
, d)
, d),
, d)
, d)k
X1
Replacing (
, d)
T (h)), where
S3 = hS3 ,
S = h2 S ,
T3 = T3 ,
T = hT ,
with
0
T (h) = D(h) D
0
3 = D0,
= 0,
= 0, D
3
0,
S(h) = (h)
0
0
33
= 33
,
= 0,
D
(5.259)
+h
aH,0
| (h)|20,
+h
|D (h)|20,
o
0 2
+ D3 (h) D3 0,
0 ), ((h)
0)
, D(h) D
0 , T (h) D
((h)
0
(5.260)
0
0
0
0
, D(h) D ), ((h)
, D(h) D )
h aH,2 ((h)
2
0
0
0
0
, D(h) D ), ((h)
, D(h) D ) = (h).
h aH,4 ((h)
4
(5.261)
0 ), ((h)
0)
0, D
0 , D(h) D
aH,0 (
0 ), ((h)
0)
0, D
0 , D(h) D
h2 aH,2 (
0 ), ((h)
0)
0, D
0 , D(h) D
h4 aH,4 (
or equivalently, it reads
0 ), (u(h), (h))
0 , D(h) D
(h) = bH ((h)
0 ), (u0 , 0 ))
0 , D(h) D
+ bH (((h)
Z
2
0
h
C33 33
+ P3 D30 (h)o(h)dx,
(5.262)
0 ), (u(h), (h)
0 , D(h) D
1 (h) = bH ((h)
+ )
+ ))
+ bH
=
fi ui (h)o(h)dx +
gi ui (h)o(h)
o(h)d +
N
0 ), (u(h), (h)
,D
(
+ )
pi ui (h)o(h)
o(h)d
L
Z
Z
0
D3 (h)E3 ((h))o(h)dx
Due to the weak convergence result, and to the assumed estimates for fi , gi , pi and o# ,
one has, as h 0,
1
fi ui dx +
gi ui d +
0
33
e33 (u0 )dx
D3 E3 ()dx
pi ui d +
L
D3 E3 ()dx
D30 E3 (0 + )dx
D30 E3 ()dx.
(5.263)
167
On the other hand, the two last terms of (5.262) can be written in the following expansive
way
Z
0
0
0
2
0
, D(h) D ), (u , ))h
2 (h) = bH (((h)
C33 33
+ P3 D30 (h)o(h)dx
Z
Z
0
= (33 (h) 33
) e33 (u0 )o(h)dx (D3 (h) D30 ) E3 (0 )o(h)dx
Z
0
h2
C33 33
+ P3 D30 (h)o(h)dx.
0
(33
0
33
) e33 (u0 )dx
0
C33 33
+ P3 D30 dx.
2 =
=
D3 E3 (0 )dx
Z
0
33
D30 E3 (0 )dx
fi u0i dx
gi u0i d
pi u0i d
D3 E3 ()dx
N
L
Z
Z
Z
Z
0
0
0
+ fi ui dx +
gi ui d +
pi ui d + D30 E3 ()dx
(D30 D3 ) E3 ()dx,
D3 E3 ()dx
D30 E3 ()dx
(D30 D3 ) E3 ()dx
= 0.
(5.264)
1/2
2
0 2
0,
0,
0,
+ (C2 + h2 C0 ()) k( , d)kX1 ,
which together with the Babuska-Brezzi (inf-sup) condition allows us to prove the existence the existence of D1 () such that
n
2
|bH (( , d), (u(h) u0 , (h) 0 ))|
D1 () D3 (h) D30 0,
k( , d)kX1
(,d)X1
o
2
2
2
0 2
2
2
4
+ 33 (h) 33
+
h
|D
(h)|
+
h
|
(h)|
+
h
|
(h)|
0,
0,
0, + D1 (),
0,
sup
equality and applying Korns and Poincares inequalities, and evoking inequalities (5.127)
and (5.128), we obtain
u(h) u0 , (h) 0
X2,w
n
2
0 2
D1 () D3 (h) D30 0, + 33 (h) 33
0,
o
2
2
2
2
2
4
+h |D (h)|0, + h |3 (h)|0, + h | (h)|0, + D1 (),
when h 0. In this way, we finished to prove that (u(h))h>0 and ((h))h>0 converge
strongly.
5.6
The limit scaled one-dimensional problem: existence and uniqueness of the solution
Substituting equations (5.248) and (5.249) in the variational equation (5.253) we have
Z
Ad33 33 e33 (u0 ) P333 E3 (0 ) e33 (v)dx
Z
+ Ad33 P333 e33 (u0 ) + C3333 E3 (0 ) E3 ()dx
Z
Z
Z
=
fi vi dx +
gi vi d +
pi vi d,
(v, ) VBN
3 ,
(5.265)
169
Lemma 15 The problem (5.265) has one and only one solution.
Proof. In order to prove the ellipticity of the bilinear form of the problem (5.265) we
take into account Korns inequality: there exists a constant c > 0 such that (cf. (5.104))
|e (v)| = |e33 (v)| c()kvk1, ,
(5.266)
The properties of the coefficients 33 and C3333 and the inequality (5.114) together guarantee that
Z
Z
d
2
A33 33 (e33 (v)) dx + Ad33 C3333 (E3 ())2 dx c1 |e33 (v)|2 + c2 |E3 ()|2
c() kvk21, + kk21, ,
Using Stampacchias Theorem and Lax-Milgramm Lemma, the existence and uniqueness of solution of problem (5.265) is guaranteed.
Putting now 3 = 0, = 0 and = 0, = 0 successively in (5.265) and taking into
account the relations (5.195)-(5.200) we get the following result.
Corollary 25 The element (1 , 2 , 3 , z3 ) [V02 (0, L)] V01 (0, L) ( + H01 (0, L)) is the
Z L
Z L
A33 33 I dx3 +
Ad33 33 3 + + P333 z3 A dx3
0Z
0
Z L
L
=
F dx3
M dx3 + FL (L) ML (L),
0
0
Z L
0Z
Z0 L
=
F3 3 dx3 + F3L 3 (L),
(5.267)
(5.268)
fi d +
gi d, M =
Z
Z
FiL =
pi d, ML = b p3 d,
Z
I =
b b d.
b f3 d
b g3 d,
5.7
In this section our goal is to give an approach to the piezoelectric shallow arch occupying
the volume {h } . In view of the scalings
h (x3 ) := h1 (x3 ),
3h (x3 ) := 3 (x3 ),
in ,
h (0)(
u
xh ) := (h1 u0 (x), u03(x)),
h (0)(
xh ) := 0 (x),
h
h
h
0
0
0
(
(0)(
xh ),
3
(0)(
xh ),
33
(0)(xh )) := (h2
(x), h3
(x), 33
(x)),
h (0)(
3h (0)(xh )) := (hD0 (x), D30 (x)),
(D
xh ), D
h and h :
h {
h }1 are
h = (
for all x
xh1 , xh3 , xh3 ), where the mappings h :
those defined in Sections 5.2.1 and 5.3, we obtain the following corollary.
uh (0)(
xh ) = h ,
h,
uh3 (0)(
xh ) = 3h b,h
,
h (0)(
xh ) = z3h ,
(5.269)
(5.270)
z3h (L) = L,h
0
(5.271)
(0) =
3
(0) = 0,
n
o
h
h
h
b,h h
h h
h z h, ,
33
(0) = Ad,h
(
)
(
)
+
(
)
(
)
+
P
33
33
3
333 3
n
o
h (0) = Ad,h P h ( h ) b,h ( h ) + (h ) ( h ) C h z h, ,
D
3
33
333
3
3333 3
171
(5.272)
(5.273)
(5.274)
h = xh .
for every point x
(b) The mechanical and electrical field (ih , z3h,0 ) is the solution of the following onedimensional boundary problems (cf. (5.267)):
h,0
h,0
h
h
=
F
+
M
,
h (0) = (h ) (0) = 0,
where
in (0, L),
h,L
mh,0
(L) = M ,
h, h
h
h
h
C33
I (h ) (L) + nh,0
3 (L)( ) (L) = F M (L)
h,0
h,0
n
+
d
= F3h ,
in (0, L),
3
3
h,0
h,0
n3 (L) = F3h,L ,
dh,0
3 (0) = d3 (L) = 0
nh,0
3
mh,0
=
dh,0
3 =
n
o
h
h,
h
33
(0) d h = Ah C33
(3 ) + (h ) (h ) + P3h, z3h, ,
h, h
h
b,h
33
(0) d h = C33
I (h ) ,
n
o
h (0) d h = Ah P h, ( h ) + (h ) ( h ) h,z h, ,
D
3
3
3
3
3
h
h,
h33
C33
= h h 33 h h = Ad,h
33
C3333 33 + P333 P333
h
C3333
h
h,
=
= Ad,h
3
33 C3333 ,
h
h
h h
C + P P
3333 33
333
333
P h
h
P3h, = h h 333 h h = Ad,h
33 P333 ,
C3333 33 + P333 P333
Fih (xh3 )
b,h
h
b,h
d ,
FiL,h (xh3 ) =
fih d h
h
N
phi d h ,
gih d h ,
Mh (xh3 )
ML,h (xh3 ) =
h
h
b,h
1 = x1 b1 (x3 ) x2 b2 (x3 ),
h f3h
d +
h
N
h g3h d h ,
h ph3 d h ,
h
h
b,h
2 = x1 b2 (x3 ) + x2 b1 (x3 ).
A major conclusion is thus that we have been able to rigorously justify one-dimensional
equations for piezoelectric shallow arches by showing that (up to appropriate scalings)
their solution can be identified (in the sense of Corollary 26(a)) with the [H 1 ()]3 H 1 ()limit of the three-dimensional solution as the diameter of the cross section of the beam
approaches zero.
We observe, in first place, that if we ignore the electric field considering a linearly
elastic beam, whose material satisfies the conditions
h,
h,
= 0,
33 = P3
1
h,
C33
= h ,
C3333
the one-dimensional model found here does indeed coincide with the linearly elastic sha
llow arch model derived by Alvarez-Dios
& Via
no [1998]. On the other hand, we note
that our asymptotic shallow arch model also coincides with asymptotic straight rod model
derived in Chapter 3, taking the subclass 6mm of the anisotropic piezoelectric material.
Chapter
6.0.1
Conclusions
The present research proposes three asymptotic models for anisotropic piezoelectric beams
submitted to an electric potential. In the three cases, the procedure starts with a change
of variable that is used to establish the original 3-D piezoelectricity problem in a fixed
reference domain, which does not depend on the diameter of the beam cross section
Then, the asymptotic method introduced by Lions (Lions [1973]) is used to perform a
mathematically rigorous dimensional reduction.
The main contributions of the present work towards the development of reduced
anisotropic beam models are the following:
An interesting approach to the three-dimensional piezoelectricity problem for an
anisotropic beam of class 2 with one end fixed and subjected to an electric potential
applied on the right and left ends of the beam has been presented in the thesis. In
this approach the piezoelectric beam is modeled as a one-dimensional domain. It is
required that the displacements satisfy a weak clamping boundary condition to avoid
the boundary layer phenomenon. The expression of the displacement second order
term has been used to identify the limit model and to prove the strong convergence
result, see Section 3.4.0.1 and Section 3.4.1.
174
6.0.2
The present work can be considered as a starting point for the derivation of models for
anisotropic piezoelectric beams using the Asymptotic Method as a tool. Moreover, there
are still vast uncharted areas where the asymptotic method would be adequate to justify
lower-dimensional models in a rigorous way. To complete and extend the present research,
a number of future tasks could be performed:
To complete our analysis, the most urgent work is to prove that, for the anisotropic
piezoelectric beam of class 2, the displacement vector field and the electric potential
weakly (and strongly) converge towards the leading terms of the electric potential
expansions.
Study the contribution of higher-order terms (correctors) of the asymptotic displacement - electric potential expansions in order to allow the construction of higher
order models.
References
Alvarez-Dios,
J.A. (1992). Teoria asint
otica de vigas el
asticas lineales anis
otropas e
no homogeneas. Ph.D. thesis, Dpto. Matematica Aplicada, Universidad de Santiago
Compostela.
176
References
Boor, A. & Tabaka, J. (2008). Good vibrations. Applying piezoelectric material to the
world of sports, http://www.engr.wisc.edu/wiscengr/april05/goodvibrations.
shtml.
Brezis, H. (1983). Analyse fonctionnelle. Theorie et applications. RAM, Wiley, Chichester, Masson.
Collard, C. & Miara, B. (2003). Two-dimensional models for geometrically nonlinear
thin shells. Asymptotic Analyis, 31, 113151.
Crawley, E. & Luis, J. (1987). Use of piezoelectric actuators as elements of intelligent
structures. AIAA Journal , 25, 13731385.
Destuynder, P., Legrain, I., Castel, L. & Richard, N. (1992). Theorical, numerical and experimental discussion on the use of piezoelectric devices for control-structure
interaction. European J. Mech. A. Solids, 11, 181213.
Duvaut, G. & Lions, J.L. (1976). Inequalities in mechanics and physics. SpringerVerlag, Berlin.
Figueiredo, I. & Leal, C. (2005). A piezoelectric anisotropic plate model. Asymptot.
Anal., 44, 327346.
Figueiredo, I. & Leal, C. (2006). A generalized piezoeletric Bernoulli-Navier
anisotropic rod model. Journal Elasticity, 31, 85106.
Gantner, A. (2005). Mathematical Modelling and Numerical Simulation of Piezoelectrical Agitated Surface Acoustic Waves. Ph.D. thesis, Naturwissenschaftlichen Fakultt
der Universitt Augsburg.
Girault, V. & Raviart, P.A. (1986). Finit Element Methods for Navier Stokes.
Springer-Verlag, Berlin.
Haenel, C. (2000). Analyse et simulation numerique de coques piezoeletriques. Ph.D.
thesis, University of Paris 6.
Hwu, C., Chang, W.C. & Gai, H. (2004). Vibration suppression of composite sandwich beams. J. Sound Vib., 272, 120.
Ikeda, T. (1990). Fundamentals of Piezoelectricity. Oxford University Press.
Irago, H. (1999). Deduction of Saint-Venants principle for the asymptotic residue of
elastic rods. Journal of Elasticity.
References
177
178
References
Sabu, N. (2002). Vibrations of thin piezoelectric flexural shells: Two-dimensional approximation. Journal of Elasticity, 68, 145165.
Sabu, N. (2003). Vibrations of thin piezoelectric shallow shells: Two-dimensional approximation. Journal of Elasticity, 113, 333352.
Saravanos, D. & Heyliger, P. (1999). Mechanics and computational models for
laminated piezoelectric beams, plates and shells. Appl. Mech. Rev., 52, 305320.
Sene, A. (2001). Modelling of piezoelectric thin plates. Asymptotic Analyis, 25, 120.
Taylor, G., Gagnepain, J., Meeker, T., Nakamura, T. & Shuvalov, L. (1985).
Piezoelectricity, , Vol. 4 . Gordon and Breach, New York.
Tiersten, H. (1969). Linear Piezoelectric Plate Vibrations. Penum Press, New York.
o, J. (1996). Mathematical Modelling of Rods. In Handbool of
Trabucho, L. & Vian
Numerical Analysis IV (P. Ciarlet, J. Lions eds.). North-Holland, Amsterdam.
o, J.M., Ribeiro, C. & Figueiredo, J. (2005a). Asymptotic modelling of a
Vian
piezoelectric beam. In Congreso de Metodos Numericos en Ingeniera, Granada, Spain,
A310, pp. 1-17.
o, J.M., Ribeiro, C. & Figueiredo, J. (2005b). Asymptotic modelling of a
Vian
piezoelectric beam. In II ECCOMAS Thematic Conference on Smart Structures and
Materials, Lisbon, EO26MOD, pp. 1-12.
Vidoli, S. & Batra, C. (2000). Derivation of plate and rod equations for a piezoelectric
body from a mixed three-dimensional variational principle. Journal of Elasticity.
Vidoli, S., Batra, C. & DellIsolaf, F. (2000). Saint - Venants problem for a
second-order piezoelectric prismatic bar. International Journal of engineering science.
Viriyasrisuwattana, P., Licht, C. & Weller, T. (2007). Mathematical modelling
of piezoelectric thin plates and slender beams through functional analysis. Thai Journal ,
31, 2537.
Weller, T. & Licht, C. (2002). Analyse asymptotique de plaques minces lineairement
piezoelectriques. C. R. Acad. Sci. Paris, 335, 309314.
Weller, T. & Licht, C. (2007). Modeling of linearly electromagneto-elastic thin plates.
C. R. Mecanique, 335, 201206.
Weller, T. & Licht, C. (2008). Asymptotic modeling of linearly piezoelectric slender
rods. C. R. Mecanique, 336, 572577.