DYNAMIC CONTACT PROBLEMS WITH VELOCITY CONDITIONS
O
ANHCHAU
∗, V
IORICAV
ENERAMOTREANU
∗∗
Laboratoire de Théorie des Systèmes, Université de Perpignan, 52 Avenue de Villeneuve, 66860 Perpignan Cedex, France,
e-mail:
{chau,viorica}@univ-perp.frWe consider dynamic problems which describe frictional contact between a body and a foundation. The constitutive law is viscoelastic or elastic and the frictional contact is modelled by a general subdifferential condition on the velocity, including the normal damped responses. We derive weak formulations for the models and prove existence and uniqueness results.
The proofs are based on the theory of second-order evolution variational inequalities. We show that the solutions of the viscoelastic problems converge to the solution of the corresponding elastic problem as the viscosity tensor tends to zero and when the frictional potential function converges to the corresponding function in the elastic problem.
Keywords: viscoelastic, elastic, subdifferential boundary condition, dynamic process, nonlinear hyperbolic variational
inequality, maximal monotone operator, weak solution
1. Introduction
Contact problems arise in many situations, for instance, in crack and impact mechanics, or in earthquake phenom- ena. Despite the importance of their practical applications and the considerable literature devoted to these topics, many problems involving contact phenomena still re- main open.
A number of papers investigating quasistatic fric- tional contact problems with viscoelastic materials have recently been published (see, e.g., Awbi et al., 2000; Chau et al.; 2001a; 2001b; Han and Sofonea, 2000; 2001).
In (Chau et al., 2001b) frictional contact was modelled by a general velocity-dependent dissipation functional, in (Chau et al., 2001a) a bilateral contact with Tresca’s fric- tion law was analysed, while in (Han and Sofonea, 2001) frictional contact with normal compliance was studied, and in (Awbi et al., 2000; Han and Sofonea, 2000) fric- tional contact with normal damped response was consid- ered. Dynamic contact problems with normal compliance were considered in (Andrews et al., 1997a; 1997b; Kuttler and Shillor, 1999; Martins and Oden, 1987).
This paper constitutes a contribution to the study of second-order evolution contact problems. Our aim is to give versions of the results obtained in (Chau et al., 2001b) to a dynamic process. We investigate models for dynamic frictional contact between a body and an ob- stacle, in which Kelvin-Voigt viscoelastic or elastic con- stitutive laws are considered. The frictional contact is modelled by a general subdifferential boundary condition.
Further examples and detailed explanations concerning the boundary conditions of this form can be found in the monograph by Panagiotopoulos (1985) and more recently in (Chau et al., 2001b). Here, the study of viscoelastic or elastic materials in dynamic processes with subdifferential boundary conditions leads to a non-standard new mathe- matical model, implying nonlinear second-order evolution equations.
We prove the existence and uniqueness of weak solu- tions to the mechanical problems. We also show the con- tinuous dependence of these solutions on the viscosity and frictional potential function, both of which may vary be- cause of simultaneous changes in the viscosity of the body and in the roughness of the surface.
The outline of the paper is as follows. In Section 2
we introduce the notation and a preliminary material. In
Section 3 we formulate the dynamic mechanical problems
with a subdifferential frictional contact condition. Then,
after specifying the assumptions on the data, we derive
variational formulations for the problems, and we prove
an existence and uniqueness result. The proof is based
on second-order evolutionary inequalities with maximal
monotone operators. In Section 4 we prove a convergence
result which shows that the solutions to the viscoelastic
problems converge to the solution to the elastic problem
when the viscosity tends to zero and when the frictional
potential function converges to the corresponding one in
the elastic problem. Finally, in Section 5 we provide some
examples of specific subdifferential conditions to which
our results apply.
2. Notation and Preliminaries
Let S
dbe the space of second-order symmetric tensors on R
d(d = 2, 3), and denote the inner product and the Eu- clidean norm on R
dand S
dby ‘ · ’ and | · |, respectively.
Thus,
u · v = u
iv
i, |v| = (v · v)
1/2, ∀ u, v ∈ R
d, σ · τ = σ
ijτ
ij, |τ | = (τ · τ )
1/2, ∀ σ, τ ∈ S
d. Here and below, the indices i and j run between 1 and d, the summation convention over repeated indices is adopted and the index that follows a comma indicates a partial derivative with respect to the corresponding com- ponent of the independent variable.
Let Ω ⊂ R
dbe a bounded domain with a Lipschitz boundary Γ. We shall use the notation
H = L
2(Ω)
d= {u = (u
i) | u
i∈ L
2(Ω)}, H = {σ = (σ
ij) | σ
ij= σ
ji∈ L
2(Ω)}, H
1= {u = (u
i) | u
i∈ H
1(Ω)}, H
1= {σ ∈ H | σ
ij,j∈ L
2(Ω)}.
The spaces H, H, H
1and H
1are real Hilbert spaces endowed with the inner products given by
(u, v)
H= Z
Ω
u
iv
idx,
(σ, τ )
H= Z
Ω
σ
ijτ
ijdx,
(u, v)
H1= (u, v)
H+ (ε(u), ε(v))
H, (σ, τ )
H1= (σ, τ )
H+ (Div σ, Div τ )
H, respectively, where ε : H
1→ H and Div : H
1→ H are the deformation and the divergence operators, respec- tively, defined by
ε(u) = (ε
ij(u)), ε
ij(u) = 1
2 (u
i,j+ u
j,i), Div σ = (σ
ij,j).
We denote the norms on the spaces H, H, H
1and H
1by k · k
H, k · k
H, k · k
H1and k · k
H1, respectively.
Let H
Γ= H
1/2(Γ)
d, γ : H
1→ H
Γbe the trace map and ν be the outer unit normal on Γ. For every element v ∈ H
1we still write v for the trace γv of v on Γ, and we denote by v
νand v
τthe normal and tangential components of v on the boundary Γ given by
v
ν= v · ν, v
τ= v − v
νν.
Let H
Γ0be the dual of H
Γand let h·, ·i stand for the pairing between H
Γ0and H
Γ. For every σ ∈ H
1, there exists an element, denoted by σν ∈ H
Γ0, such that
hσν, γvi = (σ, ε(v))
H+(Div σ, v)
H, ∀v ∈ H
1. (1) In addition, if σ is regular enough (e.g. of class C
1), we have
hσν, γvi = Z
Γ
σν · v da, ∀v ∈ H
1. (2) Relations (1) and (2) imply the following Green formula:
(σ, ε(v))
H+ (Div σ, v)
H= Z
Γ
σν · v da, ∀v ∈ H
1. (3) In a similar manner, the normal and tangential compo- nents of σ are defined by
σ
ν= (σν) · ν, σ
τ= σν − σ
νν.
Finally, for every real Hilbert space X we use the classical notation for the spaces L
p(0, T ; X) and W
k,p(0, T ; X), 1 ≤ p ≤ +∞, k = 1, 2, . . . , and we denote by C([0, T ]; X) and C
1([0, T ]; X) the spaces of continuous and continuously differentiable functions from [0, T ] to X, respectively. We recall now an existence and uniqueness result concerning evolution problems, taken from (Barbu, 1976, p. 268).
Theorem 1. Let V and H be two real Hilbert spaces such that V ⊂ H and the inclusion mapping of V into H is continuous and densely defined. We suppose that V is endowed with the norm k · k induced by the inner product (·, ·) and H is endowed with the norm | · |. We denote by V
0the dual space of V , by h·, ·i
V0×Vthe du- ality pairing between an element of V and an element of V
0, and H is identified with its own dual H
0. We assume that M is a maximal monotone set in V × V
0and A is a linear, continuous and symmetric operator from V to V
0satisfying the following coerciveness condition:
hAu, ui
V0×V+ α|u|
2≥ ωkuk
2, ∀ u ∈ V, (4) where α ∈ R and ω > 0. Let g be given in W
1,1(0, T ; H) and u
0, v
0be given with
u
0∈ V, v
0∈ D(M ), {Au
0+ M v
0} ∩ H 6= ∅. (5) Then there exists a unique solution u to the following problem:
d
2u
dt
2+ Au + M du dt
3 g(t) a.e. on (0, T ) u(0) = u
0, du
dt (0) = v
0,
which satisfies
u ∈ W
1,∞(0, T ; V ) ∩ W
2,∞(0, T ; H).
We use Theorem 1 in Section 3 to prove the exis- tence and the uniqueness of the solution to the variational problem associated with our mechanical model.
3. Problem Statement. Existence and Uniqueness Result
In this section we describe the mechanical contact prob- lem, derive its variational formulation and prove an exis- tence and uniqueness result.
The physical setting is the following: We consider a body that occupies a bounded domain Ω ⊂ R
dwith a Lipschitz continuous boundary divided into three disjoint measurable parts Γ
1, Γ
2and Γ
3such that the measure of Γ
1, denoted by |Γ
1|, is positive. Let T > 0 and [0, T ] be the time interval of interest. Let ρ : Ω → R
+be the mass density of the body and f
0: Ω × (0, T ) → R
dthe volume force density acting in Ω × (0, T ). The body is clamped on Γ
1× (0, T ) and therefore the displacement field vanishes there. A surface traction of density f
2: Γ
2→ R
dassumed to be time-independent acts on Γ
2. On Γ
3× (0, T ) the body may come in contact with an obstacle, the so-called foundation, and we suppose that the contact condition may be described by a subdifferential- type inequality.
We denote by u = (u
i) : Ω × [0, T ] → R
dthe displacement field, by σ = (σ
ij) : Ω × [0, T ] → S
dthe stress field, and ε(u) = (ε
ij(u)) represents the lin- earized strain tensor. Moreover, dots above a function will represent the derivative with respect to the time variable, i.e. ˙ u = du/dt or ¨ u = d
2u/dt
2.
We now describe the mechanical model for the pro- cess of frictional contact between the body and the obsta- cle. We use a Kelvin-Voigt constitutive law of the form
σ = cAε( ˙ u) + Gε(u),
where A is the viscosity operator, G ≡ (g
ijkh), the elas- ticity tensor, c ≥ 0 is the viscosity coefficient. When c is positive, the body exhibits a viscoelastic behavior, while for c = 0 the body is elastic. We model the frictional contact with a subdifferential boundary condition on Γ
3of the form
u ∈ U, ϕ(v) − ϕ( ˙ u) ≥ −σν · (v − ˙ u), ∀v ∈ U, where U ⊂ H
1represents the set of contact admissible test functions, σν denotes the Cauchy stress vector on the contact boundary and ϕ : Γ
3× R
d→ R is a given
function. The initial displacement field u
0and the initial velocity field v
0are given.
To summarize, the frictional mechanical problem can be formulated as follows.
Problem P
ccc: Find a displacement field u : Ω×[0, T ] → R
dand a stress field σ : Ω × [0, T ] → S
dsuch that
ρ ¨ u = Div σ + f
0in Ω × (0, T ), (6) σ = cAε( ˙ u) + Gε(u) in Ω × (0, T ), (7) u = 0 on Γ
1× (0, T ), (8) σν = f
2on Γ
2× (0, T ), (9) u ∈ U, ϕ(v) − ϕ( ˙ u) ≥ −σν · (v − ˙ u),
∀ v ∈ U on Γ
3× (0, T ), (10) u(0) = u
0, u(0) = v ˙
0in Ω. (11)
To obtain the variational formulation of Problem P
c, we consider the set
V = {v ∈ H
1| v = 0 on Γ
1} ∩ U. (12) Let us define the functional j : V → R ∪ {+∞} by
j(v) =
Z
Γ3
ϕ(v) da if ϕ(v) ∈ L
1(Γ
3),
+∞ otherwise.
(13)
In the sequel, we suppose that:
V is a closed linear subspace in H
1,
is dense in H and contains D(Ω)
d; (14)
j is a proper, convex and lower
semicontinuous functional on V. (15) Since |Γ
1| > 0, Korn’s inequality implies that there exists a constant C
K> 0, depending only on Ω and Γ
1, such that
kε(v)k
H≥ C
Kkvk
H1, ∀ v ∈ V. (16) A proof of Korn’s inequality can be found in (Neˇcas and Hlavaˇcek, 1981, p. 79). We consider the inner product on V given by
(u, v)
V= (ε(u), ε(v))
H, ∀u, v ∈ V, (17) and let k · k
Vbe the norm associated with the inner prod- uct (17), i.e.
kvk
V= kε(v)k
H, ∀v ∈ V. (18)
From (16) it follows that k · k
H1and k · k
Vare equivalent norms on V . Therefore, by (14), (V, k · k
V) is a real Hilbert space. Moreover, by combining Sobolev’s trace theorem and (16), there exists a constant C
0, depending only on Ω, Γ
1and Γ
3, such that
kvk
L2(Γ3)d≤ C
0kvk
V, ∀v ∈ V. (19) We suppose that the viscosity operator A : Ω×S
d→ S
dsatisfies the following conditions:
(i
1) A(x, ·) is monotone on S
d, i.e.
(A(x, τ
1) − A(x, τ
2)) · (τ
1− τ
2) ≥ 0,
∀ τ
1, τ
2∈ S
d, a.e. x ∈ Ω;
(i
2) there exist r ∈ L
∞(Ω) and s ∈ L
2(Ω) such that
|A(x, τ )| ≤ r(x)|τ | + s(x),
∀ τ ∈ S
d, a.e. x ∈ Ω;
(i
3) A(x, ·) is continuous on S
d, a.e. x ∈ Ω;
(i
4) A(·, τ ) is Lebesgue measurable on Ω for all τ ∈ S
d. The elasticity tensor G : Ω × S
d→ S
dis assumed to satisfy the usual properties of ellipticity and symmetry, i.e.
(j
1) there exists a constant m
G> 0 such that G(x, τ ) · τ ≥ m
G|τ |
2, ∀ τ ∈ S
d, a.e. x ∈ Ω;
(j
2) G(x, τ )·σ = τ ·G(x, σ), ∀ τ , σ ∈ S
d, a.e. x ∈ Ω;
(j
3) g
ijkl∈ L
∞(Ω) for all i, j, k, l.
We suppose that the mass density satisfies ρ ∈ L
∞(Ω) and there exists ρ
∗> 0
such that ρ(x) ≥ ρ
∗a.e. x ∈ Ω. (20) In the sequel, we define a new inner product on H given by
((u, v))
H= (ρu, v)
H, ∀u, v ∈ H, (21) and let ||| · |||
Hbe the associated norm, i.e.
|||v|||
H= (ρ v, v)
1/2H, ∀ v ∈ H. (22) Using assumption (20), from (22) it follows that ||| · |||
Hand k·k
Hare equivalent norms on H. Moreover, by (14), the inclusion mapping of (V, k · k
V) into (H, ||| · |||
H) is continuous and dense. We denote by V
0the dual space of V . Identifying H with its own dual, we can write V ⊂ H ⊂ V
0. We use the notation h·, ·i
V0×Vto represent the duality pairing between V
0and V . We have
hu, vi
V0×V= ((u, v))
H, ∀ u ∈ H, ∀ v ∈ V. (23)
We assume that the volume forces and tractions satisfy f
0∈ W
1,1(0, T ; H) and f
2∈ L
2(Γ
2)
d. (24) Let us define the functional J : V → R ∪ {+∞} by
J (v) = j(v) − Z
Γ2
f
2· v da, ∀ v ∈ V. (25) We note that by (24) the integral in (25) is well defined.
We suppose that the initial data of Problem P
csat- isfy
u
0∈ V, v
0∈ D(∂J ), (26) where ∂J denotes the subdifferential of J and D(∂J ) represents its domain.
We also assume that there exists h ∈ H such that (Gε(u
0) + cAε(v
0), ε(v) − ε(v
0))
H+ J (v)
−J (v
0) ≥ ((h, v − v
0))
H, ∀ v ∈ V. (27) For instance, in the case when we have
(σ
c0, ε(v) − ε(v
0))
H+ J (v) − J (v
0)
≥ ((f
0(0), v − v
0))
H, ∀ v ∈ V, with σ
c0:= Gε(u
0) + cAε(v
0), the condition (27) is sat- isfied.
We turn now to derive a variational formulation for the mechanical problem P
c. To this end, let us fix c ≥ 0.
We suppose in the following that {u, σ} are regular func- tions satisfying (6)–(11) and such that ϕ( ˙ u) ∈ L
1(Γ
3).
Let w ∈ V with ϕ(w) ∈ L
1(Γ
3) and t ∈ [0, T ]. Ap- plying (3) to σ for v = w − ˙ u(t) and using (6), we get
(ρ ¨ u(t) − f
0(t), w − ˙ u(t))
H+ (σ(t), ε(w)
− ε( ˙ u(t)))
H= Z
Γ
σ(t)ν · (w − ˙ u(t)) da.
Using (8), (9), (21) and (23), we obtain
h¨ u(t), w − ˙ u(t)i
V0×V+ (σ(t), ε(w) − ε( ˙ u(t)))
H= (f
0(t), w − ˙ u(t))
H+ (f
2, w − ˙ u(t))
L2(Γ2)d+ Z
Γ3
σ(t)ν · (w − ˙ u(t)) da. (28)
Combining (28), (10) and (13), we conclude that h¨ u(t), w − ˙ u(t)i
V0×V+ (σ(t), ε(w) − ε( ˙ u(t)))
H+j(w) − j( ˙ u(t))
≥ (f
0(t), w− ˙ u(t))
H+(f
2, w− ˙ u(t))
L2(Γ2)d. (29)
Taking into account (13), we observe that (29) remains true for all w ∈ V . Consequently, combining (29) and (25), we deduce that
h¨ u(t), w − ˙ u(t)i
V0×V+ (σ(t), ε(w) − ε( ˙ u(t)))
H+ J (w) − J ( ˙ u(t)) ≥ (f
0(t), w − ˙ u(t))
H,
∀ w ∈ V, a.e. t ∈ (0, T ).
Therefore, keeping in mind (7) and (11), we are led to the following variational formulation of the mechanical problem P
c, for each c ≥ 0:
Problem P
VVVccc: Find a displacement field u : [0, T ] → V and a stress field σ : [0, T ] → H
1such that
σ(t) = cAε( ˙ u(t)) + Gε(u(t)) a.e. t ∈ (0, T ), (30) h¨ u(t), w − ˙ u(t)i
V0×V+ (σ(t), ε(w) − ε( ˙ u(t)))
H+J (w) − J ( ˙ u(t)) ≥ (f
0(t), w − ˙ u(t))
H,
∀ w ∈ V, a.e. t ∈ (0, T ), (31) u(0) = u
0, u(0) = v ˙
0in Ω. (32)
We state now our existence and uniqueness result.
Theorem 2. Assume that (14), (15), (i
1)–(i
4), (j
1)–(j
3), (20), (24), (26) and (27) hold. Then for each c ≥ 0 there exists a unique solution {u, σ} to Problem P
Vcsuch that
u ∈ W
1,∞(0, T ; V ) ∩ W
2,∞(0, T ; H), (33) σ ∈ L
2(0, T ; H), Div σ ∈ L
∞(0, T ; H). (34)
We conclude that, under the assumptions of Theo- rem 2, Problem P
chas a unique weak solution {u, σ}
having the regularity (33), (34).
Proof. Let us fix c ≥ 0. We consider the Hilbert spaces H = L
2(Ω)
dand V given by (12). We introduce the operator A : V → V
0defined by
hAu, vi
V0×V= (Gε(u), ε(v))
H, ∀ u, v ∈ V. (35) Using (18), (j
2) and (j
3), we see that A ∈ L(V, V
0), and (j
1) implies that A satisfies the condition (4) with α = 0 and ω = m
G.
Define now the set-valued operator M
c: V → V
0by
M
c= B
c+ ∂J, (36)
where B
c: V → V
0is given by
hB
cu, vi
V0×V= c(Aε(u), ε(v))
H, ∀ u, v ∈ V. (37)
From (37) and (i
1), we have hB
cu − B
cv, u − vi
V0×V= c(Aε(u)−Aε(v), ε(u)−ε(v))
H≥ 0, ∀ u, v ∈ V, so the operator B
cis monotone. Using (37) and (18), we have
kB
cu−B
cvk
V0≤ ckAε(u)−Aε(v)k
H, ∀ u, v ∈ V, and, keeping in mind (i
2), (i
3), (i
4) and Krasnoselski’s theorem (see Kavian, 1993, p. 60), we find that B
c: V → V
0is a continuous operator. Using again (37) and (i
2), we find that B
cis bounded.
From (15) and (25) we deduce that J is proper, con- vex and lower semicontinuous, which implies that ∂J is maximal monotone. Consequently, since B
cis mono- tone, bounded and hemicontinuous from V to V
0, we conclude (Barbu, 1976, p. 39) that M
c= B
c+ ∂J is maximal monotone.
Moreover, the initial data u
0, v
0satisfy (5) due to (26) and (27). Thus, all the requirements of Theorem 1, with A defined by (35), M = M
cgiven in (36) and g = f
0, are satisfied. By defining σ by (30), it follows that there exists a unique solution {u, σ} to Problem P
Vcsatisfying (33).
It remains to show (34) for σ. From (30), (i
2), (j
2), (j
3) and (33) it follows that σ ∈ L
2(0, T ; H). Let t ∈ [0, T ] and ψ ∈ D(Ω)
d. Since J ( ˙ u(t) ± ψ) = J ( ˙ u(t)) <
+∞, choosing w = ˙ u(t) ± ψ ∈ V (see (14)) in (31), using (3), (21) and (23), we obtain
ρ ¨ u = Div σ + f
0in H.
Now, taking into account (33) and (24), we arrive at Div σ ∈ L
∞(0, T ; H), and thus (34) is satisfied. The uniqueness of the solution follows from Theorem 1. The proof of Theorem 2 is now complete.
4. Convergence as Viscosity Vanishes
In this section we investigate the behaviour of the solu- tion to the viscoelastic problem P
Vcwhen the viscosity operator converges to zero and when the frictional poten- tial function tends to the potential of the corresponding elastic problem. We suppose in the sequel that (14), (i
2)–
(i
4), (j
1)–(j
3), (20), (24) hold and the following addi- tional property is satisfied:
(i
5) A(x, ·) is strongly monotone on S
d, i.e. there exists m
A> 0 such that
(A(x, τ
1) − A(x, τ
2)) · (τ
1− τ
2)
≥ m
A|τ
1− τ
2|
2, ∀ τ
1, τ
2∈ S
d, a.e. x ∈ Ω.
We focus our attention on the convergence to fric- tional elasticity and the continuity with respect to the fric- tion potential function. Thus, we consider a sequence of problems P
Vcnobtained from Problem P
Vcin which we set c = c
n, where (c
n) is a sequence of viscosity coef- ficients such that c
n→ 0 as n → ∞; ϕ = ϕ
n, where ϕ
n: Γ
3× R
d→ R are given functions; u
n0and v
n0stand for the initial displacements and velocities, respec- tively. We have the following variational problem, for each n:
Problem P
VVVcccnnn: Find a displacement field u
n: [0, T ] → V and a stress field σ
n: [0, T ] → H
1such that
σ
n(t) = c
nAε( ˙ u
n(t))+Gε(u
n(t)) a.e. t ∈ (0, T ), (38) h¨ u
n(t), w− ˙ u
n(t)i
V0×V+(σ
n(t), ε(w)−ε( ˙ u
n(t)))
H+J
n(w) − J
n( ˙ u
n(t)) ≥ (f
0(t), w− ˙ u
n(t))
H,
∀ w ∈ V, a.e. t ∈ (0, T ), (39) u
n(0) = u
n0, u ˙
n(0) = v
n0in Ω. (40)
Here J
nis defined by (25) for j = j
n, where j
nis given by (13) for ϕ = ϕ
n.
Next we consider the elastic problem P
V0obtained from P
Vcfor c = 0 and the data ϕ : Γ
3× R
d→ R, u
0, v
0given.
Problem P
VVV000: Find a displacement field u : [0, T ] → V and a stress field σ : [0, T ] → H
1such that
σ(t) = Gε(u(t)) a.e. t ∈ (0, T ), (41) h¨ u(t), w − ˙ u(t)i
V0×V+ (σ(t), ε(w) − ε( ˙ u(t)))
H+J (w) − J ( ˙ u(t)) ≥ (f
0(t), w − ˙ u(t))
H,
∀ w ∈ V, a.e. t ∈ (0, T ), (42) u(0) = u
0, u(0) = v ˙
0in Ω. (43)
Here, the functionals J and j are defined by (25) and (13), respectively.
We assume that j satisfies (15);
j
nis a proper, convex and lower
semicontinuous function on V for all n; (44) there exist m ∈ [1, 2) and α > 0 such that, for all n,
|ϕ
n(x, y) − ϕ(x, y)|
≤ αc
n|y|
m, ∀ y ∈ R
d, a.e. x ∈ Γ
3; (45)
there exists h ∈ H such that
(Gε(u
0), ε(v) − ε(v
0))
H+ J (v) − J (v
0)
≥ ((h, v − v
0))
H, ∀ v ∈ V ; (46) for each n there exists h
n∈ H such that
(Gε(u
n0) + c
nAε(v
n0), ε(v) − ε(v
n0))
H+J
n(v) − J
n(v
n0)
≥ ((h
n, v − v
n0))
H, ∀ v ∈ V ; (47) and v
0∈ D(∂J ), v
n0∈ D(∂J
n). Finally,
u
n0→ u
0in V, v
n0→ v
0in H as n → ∞. (48) Let us remark that if we have, for all n, v
n0= v
0and
(σ
n0, ε(v) − ε(v
0))
H+ J
n(v) − J
n(v
0)
≥ ((f
0(0), v − v
0))
H, ∀v ∈ V, (σ
0, ε(v) − ε(v
0))
H+ J (v) − J (v
0)
≥ ((f
0(0), v − v
0))
H, ∀v ∈ V, with σ
n0:= Gε(u
n0) + c
nAε(v
n0), σ
0:= Gε(u
0), then the assumptions (46)–(48) are satisfied.
From Theorem 2 it follows that, for each n, Prob- lem P
Vcnhas a unique solution {u
n, σ
n} with regu- larity u
n∈ W
1,∞(0, T ; V ) ∩ W
2,∞(0, T ; H), σ
n∈ L
2(0, T ; H), Div σ
n∈ L
∞(0, T ; H), and Problem P
V0has a unique solution {u, σ} with regularity u ∈ W
1,∞(0, T ; V ) ∩ W
2,∞(0, T ; H), σ ∈ L
2(0, T ; H), Div σ ∈ L
∞(0, T ; H).
We are now in a position to formulate our conver- gence result.
Theorem 3. Let (c
n) be a sequence in (0, +∞) such that c
n→ 0 as n → ∞. Suppose that (14), (15), (i
2)–(i
5), (j
1)–(j
3), (20), (24), (44)–(47) and denote by {u
n, σ
n}, {u, σ} the unique solutions to Problems P
Vcnand P
V0, respectively. Then there exists a constant C > 0, depending on u and on the data, but independent of n, such that for all n we have
ku
n− uk
C([0,T ];V )+ ku
n− uk
C1([0,T ];H)+ kσ
n− σk
L2(0,T ;H)≤ C(ku
n0− u
0k
V+ kv
n0− v
0k
H+ √ c
n).
Consequently, if (48) holds, then as n → ∞, u
n→ u in C([0, T ]; V ) ∩ C
1([0, T ]; H),
σ
n→ σ in L
2(0, T ; H).
Proof. Let t ∈ [0, T ]. Taking w = ˙ u(t) in (39) and using (38), we have
h¨ u
n(t), ˙ u(t) − ˙ u
n(t)i
V0×V+ c
n(Aε( ˙ u
n(t)), ε( ˙ u(t)) − ε( ˙ u
n(t)))
H+ (Gε(u
n(t)), ε( ˙ u(t)) − ε( ˙ u
n(t)))
H+ j
n( ˙ u(t)) − j
n( ˙ u
n(t))
≥ (f
0(t), ˙ u(t) − ˙ u
n(t))
H+ (f
2, ˙ u(t) − ˙ u
n(t))
L2(Γ2)d, and taking w = ˙ u
n(t) in (42), using (41), we obtain
h¨ u(t), ˙ u
n(t) − ˙ u(t)i
V0×V+ (Gε(u(t)), ε( ˙ u
n(t)) − ε( ˙ u(t)))
H+ j( ˙ u
n(t)) − j( ˙ u(t))
≥ (f
0(t), ˙ u
n(t) − ˙ u(t))
H+ (f
2, ˙ u
n(t) − ˙ u(t))
L2(Γ2)d.
Adding the last two inequalities, we deduce that for each t ∈ [0, T ],
h¨ u
n(t) − ¨ u(t), ˙ u(t) − ˙ u
n(t)i
V0×V+ c
n(Aε( ˙ u
n(t)) − Aε( ˙ u(t)), ε( ˙ u(t)) − ε( ˙ u
n(t)))
H+ c
n(Aε( ˙ u(t)), ε( ˙ u(t)) − ε( ˙ u
n(t)))
H+ (Gε(u
n(t)) − Gε(u(t)), ε( ˙ u(t)) − ε( ˙ u
n(t)))
H+ j
n( ˙ u(t)) − j
n( ˙ u
n(t)) + j( ˙ u
n(t)) − j( ˙ u(t)) ≥ 0.
Integrating this inequality on [0, t] and using (23), (j
2), (i
5), (18), (40), (43), we conclude that
1
2 ||| ˙ u
n(t) − ˙ u(t)|||
2H+ c
nm
AZ
t0
k ˙ u
n(s) − ˙ u(s)k
2Vds
+ 1
2 (Gε(u
n(t)) − Gε(u(t)), ε(u
n(t)) − ε(u(t)))
H≤ 1
2 |||v
n0− v
0|||
2H+ c
nZ
t 0kAε( ˙ u(s))k
Hk ˙ u(s) − ˙ u
n(s)k
Vds
+ 1
2 (G(ε(u
n0) − ε(u
0)), ε(u
n0) − ε(u
0))
H+
Z
t 0|j
n( ˙ u(s))−j
n( ˙ u
n(s))+j( ˙ u
n(s))−j( ˙ u(s))| ds.
Using (j
1), (j
3) and (18), and since by (i
2), (i
3) and (i
4) we have Aε( ˙ u) ∈ L
2(0, T ; H), we deduce that
1
2 ||| ˙ u
n(t) − ˙ u(t)|||
2H+c
nm
AZ
t 0k ˙ u
n(s) − ˙ u(s)k
2Vds
+ m
G2 ku
n(t) − u(t)k
2V≤ 1
2 |||v
n0− v
0|||
2H+ c
n2m
AZ
t 0kAε( ˙ u(s))k
2Hds
+ c
nm
A2
Z
t 0k ˙ u
n(s) − ˙ u(s)k
2Vds
+ C
2 ku
n0− u
0k
2V+ Z
t0
(|j
n( ˙ u(s)) − j( ˙ u(s))|
+|j
n( ˙ u
n(s)) − j( ˙ u
n(s))|) ds, (49) where C is a positive constant independent of n and may change from line to line. From the fact that ϕ
n( ˙ u
n(s)) ∈ L
1(Γ
3), ϕ( ˙ u(s)) ∈ L
1(Γ
3), by using (45), it follows that ϕ
n( ˙ u(s)) ∈ L
1(Γ
3), ϕ( ˙ u
n(s)) ∈ L
1(Γ
3). Conse- quently, using (45), Young’s inequality and (19), we can write, for all s ∈ [0, T ],
|j
n( ˙ u(s)) − j( ˙ u(s))| + |j
n( ˙ u
n(s)) − j( ˙ u
n(s))|
≤ αc
nZ
Γ3
| ˙ u(s)|
mda + αc
nZ
Γ3
| ˙ u
n(s)|
mda
≤ α(1 + 2
m−1)c
nZ
Γ3
| ˙ u(s)|
mda
+ α2
m−1c
nZ
Γ3
| ˙ u
n(s) − ˙ u(s)|
mda
≤ α(1 + 2
m−1) c
nZ
Γ3
| ˙ u(s)|
mda
+ c
nZ
Γ3
"
2 − m 2
α2
m−1τ
2−m2+ m
2 τ
m2| ˙ u
n(s) − ˙ u(s)|
2i da
≤ C(c
nk ˙ u(s)k
mL2(Γ3)d+ c
n) + c
nm
2 τ
m2k ˙ u
n(s) − ˙ u(s)k
2L2(Γ3)d≤ C(c
nC
0mk ˙ u(s)k
mV+ c
n) + c
nm
2 τ
m2C
02k ˙ u
n(s) − ˙ u(s)k
2V,
where τ > 0 is a constant that will be chosen below.
Integrating this inequality on [0, t] and using Hölder’s in- equality, we obtain
Z
t 0(|j
n( ˙ u(s)) − j( ˙ u(s))| + |j
n( ˙ u
n(s)) − j( ˙ u
n(s))|) ds
≤ Cc
n+ c
nm 2 τ
m2C
02Z
t 0k ˙ u
n(s) − ˙ u(s)k
2Vds
≤ Cc
n+ c
nm
A3
Z
t 0k ˙ u
n(s) − ˙ u(s)k
2Vds, (50)
for τ chosen such that (m/2)τ
2/mC
02≤ m
A/3. Com- bining (49) with (50) and using the equivalence of the norms ||| · |||
Hand k · k
Hon H, we infer that
k ˙ u
n(t) − ˙ u(t)k
2H+ c
nZ
t 0k ˙ u
n(s) − ˙ u(s)k
2Vds
+ ku
n(t) − u(t)k
2V≤ C(kv
n0− v
0k
2H+ ku
n0− u
0k
2V+ c
n).
From the continuity of the embedding V ⊂ H, it follows that
ku
n− uk
2C([0,T ];V )+ ku
n− uk
2C1([0,T ];H)+c
nk ˙ u
n− ˙ uk
2L2(0,T ;V )≤ C(kv
n0− v
0k
2H+ ku
n0− u
0k
2V+ c
n). (51)
It remains to prove that for all n we have kσ
n− σk
L2(0,T ;H)≤ C(ku
n0− u
0k
V+ kv
n0− v
0k
H+ √
c
n). (52)
From (38) and (41) we deduce that kσ
n(t) − σ(t)k
2H≤ 2c
2nkAε( ˙ u
n(t))k
2H+ 2kGε(u
n(t)) − Gε(u(t))k
2H, a.e. t ∈ (0, T ). Condition (i
2) implies that there exist two positive constants δ
1and δ
2such that kAτ k
2H≤ δ
1kτ k
2H+ δ
2, ∀τ ∈ H. It follows that
kσ
n− σk
2L2(0,T ;H)≤ 2c
2nδ
1k ˙ u
nk
2L2(0,T ;V )+ 2T δ
2c
2n+ 2 Z
T0
kGε(u
n(t)) − Gε(u(t))k
2Hdt.
Using again (j
3) and (18), we deduce that kσ
n− σk
2L2(0,T ;H)≤ 4c
2nδ
1k ˙ u
n− ˙ uk
2L2(0,T ;V )+ 4c
2nδ
1k ˙ uk
2L2(0,T ;V )+ 2T δ
2c
2n+ 2C Z
T0
ku
n(t) − u(t)k
2Vdt.
Keeping in mind (51), we arrive at
kσ
n− σk
2L2(0,T ;H)≤ C(c
n+ 1)(ku
n0− u
0k
2V+ kv
n0− v
0k
2H+ c
n) + Cc
2n, so, since c
n→ 0 as n → ∞, we obtain (52). This completes the proof.
We conclude by Theorem 3 that the weak solution to the frictional elastic problem P
0can be approached by the weak solution to the frictional viscoelastic problem P
cnwhen the coefficient of viscosity is small enough and the corresponding friction potential functions satisfy (45).
In addition to the mathematical interest in the convergence properties proved in Theorem 3, this is of importance from the mechanical point of view, because the frictional elas- ticity appears as a limit case of frictional viscoelasticity.
Remark 1. A similar argument to the one used in the proof of Theorem 3 shows that Theorem 3 remains true when we replace the condition (45) with the following property: there exist an integer p ≥ 1 and numbers m
1, . . . , m
p∈ [1, 2), α
1, . . . , α
p> 0 such that for ev- ery n one has
|ϕ
n(x, y) − ϕ(x, y)| ≤ c
n pX
i=1