Integrodifferential Equations With Parameter-Dependent Operators

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Differential and Integral Equations Volume 15, Number 1, January 2002, Pages 33–45

INTEGRODIFFERENTIAL EQUATIONS WITH


PARAMETER-DEPENDENT OPERATORS

Ronald Grimmer
Department of Mathematics, Southern Illinois University, Carbondale, IL 62901

Min He
Department of Mathematics and Computer Science
Kent State University-Trumbull Campus, Warren, OH 44483

(Submitted by: Glenn Webb)

1. Introduction
Let X be a Banach space with norm  · . Let P be an open subset of
a finite-dimensional normed linear space P with norm | · |. Consider the
abstract integrodifferential equation
 t

 
u (t) = A(ε)u(t) + F (t − s)A(ε) + K(t − s) u(s) ds + f (t), t ≥ 0,
0
u(0) = u0 , (1.1)
where ε is a multiparameter. For each ε ∈ P , A(ε) is a closed non-densely de-
fined linear operator and F (t) and K(t) are bounded operators on a Banach
space X with F (t) ≤ M1 , F  (t) ≤ M2 , K(t) ≤ M3 , t ≥ 0.
One way to consider this problem is via an ODE reformulation of (1.1)
determined by examining the “flow”:
   
u(0) u(t)

−→ t ;
0 F (t − τ + ·)A(ε)u(τ ) dτ + f (t + ·)
f (·)
equation (1.1) can then be replaced by an ordinary differential equation in
a product space of the following form:
          
u(t) A(ε) δ u(t) f (t) u(0) u
= + , = 0 (1.2)
v(t) F (·)A(ε) + K(·) Ds v(t) 0 v(0) 0
on Z = X × C with the usual norm, where C is the space of bounded
uniformly continuous functions on [0, ∞) into X with the usual sup norm
Accepted for publication: September 2000
AMS Subject Classifications: 35, 47.
33
34 Ronald Grimmer and Min He

and v(0) = 0 if v ∈ C. Ds is the generator of the shift semigroup (that


is, the semigroup T (·) generated by Ds satisfies T (t)f )(·) = f (· + t) and
δf = f (0). [F (·)x](t) = F (t)x and [K(·)x](t) = K(t)x. Let
       
A(ε) δ f (t) u(t) u
A(ε) = , F(t) = , z(t) = , z0 = 0 ;
F (·)A(ε) + K(·) Ds 0 v(t) 0
then equations (1.2) may be written as the following abstract Cauchy prob-
lem:
z  (t) = A(ε)z(t) + F(t), z(0) = z0 , for t ≥ 0. (1.3)
Such an approach may be seen in e.g. [5]. Also, we note that a different
approach is taken to a delay problem with fixed delay r (but no parameters)
in [7].
In Section 2, we will discuss continuity in the parameter ε of the resolvent
of the operator A(ε) concerning the cases (a) when the operator A(ε) in (1.1)
has a fixed domain (i.e., D(A(ε)) = D) and (b) when the operator A(ε) in
(1.1) has a variable domain (i.e., D(A(ε))is dependent on ε). In Section 3,
we will discuss continuity in the parameter ε of the integrated semigroup
S(t, ε) generated by A(ε). In consequence, the existence of a unique integral
or classical solution of equation (1.1) and continuity in the parameter ε of
such a solution are accomplished by applying the result on S(t, ε). In the
last section, we will conclude by giving an example of the applicability of
the obtained results.

2. The resolvent of the operator A(ε)


We first consider the case when operators A(ε) have fixed domains.
Lemma 2.1. Assume that
(2.1) For each ε ∈ P, A(ε) is densely defined; i.e., D(A(ε)) = X.
(2.2) D(A(ε)) = D, for all ε ∈ P .
(2.3) There are constants M ≥ 1 and ω ∈ R such that
M
(λI − A(ε))−n  ≤ for λ > ω, n ∈ N, and all ε ∈ P.
(λ − ω)n
(2.4) For each x ∈ D, A(ε)x is continuous in ε ∈ P .
Then (λI − A(ε))−1 x is continuous in ε for each x ∈ X. Thus the C0 -
semigroup T (t, ε) generated by A(ε) is strongly continuous in ε ∈ P , and
the continuity is uniform on bounded t-intervals. In particular, for any ε ∈
P, h ∈ P with ε + h ∈ P , and any t0 ∈ [0, ∞),
sup T (t, ε + h)x − T (t, ε)x = ◦(1) as |h| → 0, for each x ∈ X.
0≤t≤t0
integrodifferential equations with parameter-dependent operators 35

Proof. The proof is similar to the proof of the C0 -semigroup approximation


theorem in [3, Theorem 3.17], and it is omitted. 
Lemma 2.2. Let B ∈ B(X, Y ). If B ≤ 1/2 , then (I − B)−1 exists, and


(I − B)−1 = Bk .
k=0

Moreover, (I − B)−1  ≤ 2.


Proof. The proof is standard, and it is omitted here. 
Theorem 2.3. Let R(ε) ∈ B(X, Y ) for each ε ∈ P . Assume that R(ε) ≤
1/2 for all ε ∈ P and R(ε)x is continuous in ε for each fixed x ∈ X; then
(I − R(ε))−1 x is continuous in ε for each fixed x ∈ X.

Proof. From Lemma 2.2, we see that the series ∞ k
k=0 R (ε)x is uniformly
convergent to (I − R(ε))−1 x. Hence, the continuity of R(ε)x in ε implies
that (I − R(ε))−1 x is continuous in ε for each fixed x ∈ X. 
Theorem 2.4. Let λ ∈ ∩{ρ(A(ε)) ∩ ρ(Ds )}. Assume that (2.2)–(2.4) are
satisfied; then (λI − A(ε))−1 z is continuous in ε for each z ∈ Z.
Proof. Note that operator A(ε) may be written as
     
I 0 A(ε) 0 I 0 0 δ .
A(ε) = + = A1 (ε)+B,
F (·) I 0 Ds −F (·) I Ds F (·) + K(·) 0
where
     
I 0 A(ε) 0 I 0 0 δ
A1 (ε) = , B= .
F (·) I 0 Ds −F (·) I Ds F (·) + K(·) 0
From Lemma 2.1, we have that (λI − A(ε))−1 u is continuous in ε for each
u ∈ X. Since
    
−1 I 0 (λI − A(ε))−1 0 I 0 u
(λI−A1 (ε)) z = ,
F (·) I 0 (λI − Ds )−1 −F (·) I v
(2.5)
−1
it is clear that (λI − A1 (ε)) z is continuous in ε because F is bounded.
Furthermore, since B is bounded, for λ large enough,
1
B(λI − A1 (ε))−1  ≤
2
(for details, see the argument in [6, p. 76]). Also, B(λI − A1 (ε))−1 z is
continuous in ε for each z ∈ Z because B is bounded. Then from Theorem
36 Ronald Grimmer and Min He

2.3 it follows that (I − B(λI − A1 (ε))−1 )−1 z is continuous in ε. Therefore


we see that
(λI − A(ε))−1 z = (λI − A1 (ε))−1 (I − B(λI − A1 (ε))−1 )−1 z
is continuous in ε for each z ∈ Z. 
We then consider the case when the A(ε) have variable domains.
Assumption Q. Let ε0 ∈ P be given. Then for each ε ∈ P there exist
bounded operators Q1 (ε), Q2 (ε) : X → X with bounded inverses Q−1
1 (ε) and
Q−1
2 (ε), such that A(ε) = Q 1 (ε)A(ε 0 )Q2 (ε).
Note that if A(ε1 ) = Q1 (ε1 )A(ε0 )Q2 (ε1 ), then
A(ε) =Q1 (ε)A(ε0 )Q2 (ε) = Q1 (ε)Q−1 −1
1 (ε1 )Q1 (ε1 )A(ε0 )Q2 (ε1 )Q2 (ε1 )Q2 (ε)
=Q̃1 (ε)A(ε1 )Q̃2 (ε).
Thus, having such a relationship for some ε0 implies a similar relationship
at any other ε1 ∈ P . Without loss of generality then, we may just consider
strong continuity in ε of the resolvent (λI − A(ε))−1 at ε = ε0 ∈ P .
Theorem 2.5. Assume that Assumption Q and (2.3) are satisfied, and
suppose that
(2.6) Qi (ε) (i = 1, 2) and Q−1
2 (ε) are continuous in ε for x ∈ X.
Then (λI − A(ε))−1 z is continuous in ε for each z ∈ Z.
Proof. Let Ã(ε) = Q2 (ε)A(ε)Q−1 2 (ε). Then Ã(ε) = Q2 (ε)Q1 (ε)A(ε0 ).
Clearly, the Ã(ε) have fixed domains; that is, D(Ã(ε)) = D(A(ε0 )) for all
ε ∈ P . Also, (2.6) implies that Ã(ε) satisfies (2.4). Since
(λI − Ã(ε))n = Q2 (ε)(λI − A(ε))n Q−1
2 (ε).

Thus, (λI−Ã(ε))−n = Q2 (ε)(λI−A(ε))−n Q−1 −1


2 (ε). Because Q2 (ε) and Q2 (ε)
are bounded, (2.3) implies that Ã(ε) is also satisfies (2.3). Now applying
Theorem 2.4, we have that (λI − Ã(ε))−1 z is continuous in ε for each z ∈ Z,
where  
Ã(ε) δ
Ã(ε) = .
F (·)Ã(ε) + K(·) Ds
Let


−1
Q2 (ε) 0 −1 Q2 (ε) 0
Q2 (ε) = , Q2 (ε) = .
0 Q2 (ε) 0 Q−1
2 (ε)

Then it is easy to show that Ã(ε) = Q2 (ε)A(ε)Q−1


2 (ε). Because

(λI − A(ε))−1 = Q−1 −1


2 (ε)(λI − Ã(ε)) Q2 (ε)
integrodifferential equations with parameter-dependent operators 37

and (2.6) holds, we have that (λI − A(ε))−1 z is continuous in ε for each
z ∈ Z. 

3. Integrated semigroups and solutions of (1.1)


We start by proving a general theorem concerning continuity in the param-
eter ε of the integrated semigroup Ŝ(t, ε) which is generated by the operator
Â(ε) for the case that the Â(ε) have fixed domains.
In the sequel we use “A is a Hille-Yosida operator” to mean that there
exist M ≥ 1 and ω ∈ R such that λ > ω implies λ ∈ ρ(A) (the resolvent set
of A) and
M
(λI − A)−n  ≤ for λ > ω and n = 1, 2, . . . .
(λ − ω)n

For each ε ∈ P , define the part Â0 (ε) of Â(ε) as

Â0 (ε) = Â(ε) on D(Â0 (ε)) = {u ∈ D(Â(ε)) : Â(ε)u ∈ X̂0 (ε) = D(Â(ε))}.

For each ε ∈ P , let Ŝ(t, ε) be the integrated semigroup generated by the Hille-
Yosida operator Â(ε) on X, and let T̂ (t, ε) be the C0 -semigroup generated
by Â0 (ε) on X̂0 (ε). The results in [2] and [8] show that for each ε ∈ P , the
Hille-Yosida operator Â(ε) generates an integrated semigroup Ŝ(t, ε) on X,
and the part Â0 (ε) of Â(ε) generates a C0 -semigroup T̂ (t, ε) on X̂0 (ε). Also
D(Â0 (ε)) = D(Â(ε)).
Lemma 3.1. Assume that Â(ε) satisfies (2.2) and (2.3), and suppose that
(3.1) (λI − Â(ε))−1 x is continuous in ε for each ε ∈ P and x ∈ X.
Then the integrated semigroup Ŝ(t, ε) generated by Â(ε) is strongly contin-
uous in ε ∈ P , and the continuity is uniform on bounded t-intervals. In
particular, for any ε ∈ P, h ∈ P with ε + h ∈ P , and any t0 ∈ [0, ∞),

sup Ŝ(t, ε + h)x − Ŝ(t, ε)x = ◦(1) as |h| → 0, for each x ∈ X. (3.2)
0≤t≤t0

Proof. First note that the relation between a C0 -semigroup T̂ (t, ε) and
the related integrated semigroup Ŝ(t, ε) is as follows: for λ > ω and x ∈ X,

Ŝ(t, ε)x = λŜ0 (t, ε)(λI − Â(ε))−1 x − T̂ (t, ε)(λI − Â(ε))−1 x + (λI − Â(ε))−1 x
(3.3)
38 Ronald Grimmer and Min He
t
where Ŝ0 (t, ε) = 0 T̂ (s, ε) ds. Since for each ε, Â(ε) satisfies (2.2), X̂0 (ε) =
.
D = X̂0 which is independent of ε. From (3.1), it follows that
(λI − Â0 (ε))−1 x is continuous in ε for each x ∈ X̂0 and λ > ω
because
(λI − Â0 (ε))−1 = (λI − Â(ε))−1 on X̂0 .
Also, it follows from Lemma 2.1 that
sup T̂ (t, ε + h)x − T̂ (t, ε)x = ◦(1) as |h| → 0, for each x ∈ X̂0 . (3.4)
0≤t≤t0

Next we proceed to prove three claims.


Claim 1. For each x ∈ X̂0 , Ŝ0 (t, ε)x is uniformly continuous in ε on [0, t0 ]
for any t0 > 0.
In fact, for each x ∈ X̂0 , and for any fixed ε ∈ P and h ∈ P with ε+h ∈ P ,
 t
Ŝ0 (t, ε + h)x − Ŝ0 (t, ε)x ≤ T̂ (s, ε + h)x − T̂ (s, ε)x ds
0
 t0
≤ T̂ (s, ε + h)x − T̂ (s, ε)x ds (since t ≤ t0 ).
0
It follows from (3.4) that Claim 1 holds.
Claim 2. For each x ∈ X, Ŝ0 (t, ε)(λI − Â(ε))−1 x is uniformly continuous
in ε on [0, t0 ] for any t0 > 0.
In fact, for each x ∈ X, and for any fixed ε ∈ P and h ∈ P with ε + h ∈ P ,
Ŝ0 (t, ε + h)(λI − Â(ε + h))−1 x − Ŝ0 (t, ε)(λI − Â(ε))−1 x
≤ Ŝ0 (t, ε + h)(λI − Â(ε + h))−1 x − Ŝ0 (t, ε + h)(λI − Â(ε))−1 x
+ Ŝ0 (t, ε + h)(λI − Â(ε))−1 x − Ŝ0 (t, ε)(λI − Â(ε))−1 x
≤ S0 (t, ε + h) · (λI − Â(ε + h))−1 x − (λI − Â(ε))−1 x
+ [Ŝ0 (t, ε + h) − Ŝ0 (t, ε)](λI − Â(ε))−1 x
 t0

≤ M e dτ (λI − Â(ε + h))−1 x − (λI − Â(ε))−1 x
ωτ
0
+ [Ŝ0 (t, ε + h) − Ŝ0 (t, ε)](λI − Â(ε))−1 x
= ◦ (1) as |h| → 0, uniformly on [0, t0 ],
by continuity in ε of (λI − Â(ε))−1 x and Claim 1. Thus, Claim 2 is true.
Claim 3. For each x ∈ X, T̂ (t, ε)(λI − Â(ε))−1 x is uniformly continuous
in ε on [0, t0 ] for any t0 > 0.
integrodifferential equations with parameter-dependent operators 39

Using a discussion similar to that of Claim 2, we can show that Claim 3


holds by applying (3.1) and (3.4).
Finally, the statement (3.3) together with Claim 1–Claim 3 imply that
(3.2) holds.
Theorem 3.2. Assume that (2.2)–(2.4) are satisfied; then the integrated
semigroup S(t, ε) generated by A(ε) is strongly continuous in ε ∈ P , and
the continuity is uniform on bounded t-intervals. In particular, for any ε ∈
P, h ∈ P with ε + h ∈ P , and any t0 ∈ [0, ∞),
sup S(t, ε + h)z − S(t, ε)z = ◦(1) as |h| → 0, for each z ∈ Z.
0≤t≤t0

Proof. From Theorem 2.4, we see that A(ε) satisfies (3.1). Because of (2.3)
and the fact that Ds is a shift semigroup, it follows from (2.5) that A1 (ε)
satisfies (2.3). Since B is bounded, we have that A(ε) satisfies (2.3). Now
the desired result follows from Lemma 3.1. 
Now consider the case when the A(ε) have variable domains.
Proposition 3.3. Assume Assumption Q. Let Ã(ε) = Q2 (ε)A(ε)Q−1 2 (ε).
Then Ã0 (ε) = Q2 (ε)A0 (ε)Q−1
2 (ε).
Proof. Note that for each ε ∈ P ,
(λI − A0 (ε))−1 = (λI − A(ε))−1 on X0 (ε) = D(A(ε))
(λI − Ã0 (ε))−1 = (λI − Ã(ε))−1 on X̃0 (ε) = D(Ã(ε))
and for x ∈ D(Ã(ε)), Q−1
2 (ε)x ∈ D(A(ε)).
Thus, for each x ∈ D(Ã0 (ε)),
(λI − Ã0 (ε))−1 x = (λI − Ã(ε))−1 x = Q2 (ε)(λI − A(ε))−1 Q−1
2 (ε)x
(3.5)
= Q2 (ε)(λI − A0 (ε))−1 Q−1 −1 −1
2 (ε)x = (λI − Q2 (ε)A0 (ε)Q2 (ε)) x.

Since X̃0 (ε) = D(Ã0 (ε)) and all operators in (3.5) are bounded, we have that
(3.5) holds for every x ∈ X̃0 (ε). Hence, Ã0 (ε) = Q2 (ε)A0 (ε)Q−1
2 (ε). 
Let S(t, ε) be the integrated semigroups which are generated by the A(ε)
and let Ã(ε) = Q2 (ε)A(ε)Q−1 2 (ε) where


−1
Q2 (ε) 0 −1 Q2 (ε) 0
Q2 (ε) = , Q2 (ε) = .
0 Q2 (ε) 0 Q−1
2 (ε)

Proposition 3.4. Assume that Assumption Q, (2.3), and (2.6) are satisfied.
Then Ã(ε) generates an integrated semigroup S̃(t, ε) and
S(t, ε) = Q−1
2 (ε)S̃(t, ε)Q2 (ε).
40 Ronald Grimmer and Min He

Proof. First note that D(A(ε)) = D(A(ε)) × C, and thus D(A(ε)) =


D(A(ε)) × C.
Claim 1. A(ε)(u, v)t ∈ D(A(ε)) if and only if u ∈ D(A(ε)) and A(ε)u ∈
D(A(ε)).
In fact, A(ε)(u(t), v(t))t = (A(ε)u(t) + δv(t), F (t)A(ε)u + K(t)u + Ds v)t ,
which implies that u ∈ D(A(ε)) and A(ε)u ∈ D(A(ε)) because (A(ε)u(t) +
δv(t)) ∈ D(A(ε)) and δv(t) = v(0) = 0. The other way is clear.
Claim 2. D(A0 (ε)) = D(A0 (ε)) × C.
In fact, “(u, v)t ∈ D(A0 (ε))” ⇐⇒ “(u, v)t ∈ D(A(ε)) and A(ε)(u, v)t ∈
D(A(ε))” ⇐⇒ “u ∈ D(A(ε)), A(ε)u ∈ D(A(ε)) and v ∈ C” ⇐⇒ “u ∈
D(A0 (ε)) and v ∈ C” ⇐⇒ “(u, v)t ∈ D(A0 (ε)) × C.”
For (u, v)t ∈ D(A0 (ε)), it follows from Claim 2 that u ∈ D(A0 (ε)). Hence,
by Proposition 3.3,



u u
A0 (ε) = A(ε)
v v





A(ε) δ u A0 (ε) δ u
= =
F (·)A(ε) + K(·) Ds v F (·)A0 (ε) + K(·) Ds v



Q−1
2 (ε)Ã0 (ε)Q2 (ε) δ u
=
F (·)Q−1
2 (ε) Ã0 (ε)Q2 (ε) + K(·) D s v



−1

Q2 (ε) 0 Ã0 (ε) δ Q2 (ε) 0 u
=
0 Q2 (ε) F (·)Ã0 (ε) + K(·) Ds 0 Q−1
2 (ε) v


u
=Q−12 (ε)Ã0 (ε)Q2 (ε) v .

Using (2.3), (2.5), and the fact “Ds generates a shift semigroup,” we have
that A(ε) satisfies (2.3). Thus A0 (ε) generates a C0 -semigroup T (t, ε). Be-
cause of (2.6), the fact that Ã(ε) = Q2 (ε)A(ε)Q−1
2 (ε), and the fact that

(λI − A(ε))−n = Q−1


2 (ε)(λI − Ã(ε))
−n
Q2 (ε),

we have that Ã(ε) satisfies (2.3) and Ã0 (ε) generates a C0 -semigroup T̃ (t, ε).
It is clear that T (t, ε) = Q−1
2 (ε)T̃ (t, ε)Q2 (ε).
Now, using (3.3), one can easily show that S(t, ε) = Q−1 2 (ε)S̃(t, ε)Q2 (ε).
Lemma 3.5. Assume that Assumption Q, (2.3), and (2.6) are satisfied.
Then the integrated semigroup S̃(t, ε) generated by Ã(ε) is strongly continu-
ous at ε0 , and the continuity is uniform on bounded t-intervals. In particular,
integrodifferential equations with parameter-dependent operators 41

for any h ∈ P with ε0 + h ∈ P , and any t0 ∈ [0, ∞),


sup S̃(t, ε0 + h)z − S̃(t, ε0 )z = ◦(1) as |h| → 0, for each z ∈ Z.
0≤t≤t0

Proof. First note that Ã(ε) = Q2 (ε)Q1 (ε)A(ε0 ), and hence D(Ã(ε)) has
a fixed domain because D(Ã(ε)) = D(A(ε0 )) for all ε ∈ P . Secondly, we
see that (2.6) implies that there exists δ > 0 so that Q2 (ε) and Q−1
2 (ε) are
uniformly bounded on Bδ (ε0 ) = {ε : |ε − ε0 | ≤ δ}. Since
(λI − Ã(ε))−n = Q2 (ε)(λI − A(ε))−n Q−1
2 (ε)

and assumption (2.3) holds, we see that Ã(ε) satisfies (2.3). Also, Ã(ε)
satisfies (2.4) because Qi (ε) (i = 1, 2) is continuous in ε. Now the desired
result directly follows from Theorem 3.2.
Theorem 3.6. Assume that Assumption Q, (2.3), and (2.6) are satisfied;
then the integrated semigroup S(t, ε) generated by A(ε) is strongly continuous
at ε0 ∈ P , and the continuity is uniform on bounded t-intervals. In particu-
lar, for an arbitrary ε0 ∈ P, h ∈ P with ε0 + h ∈ P, and any t0 ∈ [0, ∞),
sup S(t, ε0 + h)z − S(t, ε0 )z = ◦(1) as |h| → 0, for each z ∈ Z. (3.6)
0≤t≤t0

Proof. From Proposition 3.4, S(t, ε) = Q−12 (ε)S̃(t, ε)Q2 (ε). Note that for
z ∈ Z = X × C and ε0 + h ∈ Bδ (ε0 ) = {ε : |ε − ε0 | ≤ δ},
S(t, ε0 + h)z − S(t, ε0 )z
= Q−1 −1
2 (ε0 + h)S̃(t, ε0 + h)Q2 (ε0 + h)z − Q2 (ε0 )S̃(t, ε0 )Q2 (ε0 )z
≤ Q−1
2 (ε0 + h)S̃(t, ε0 + h) Q2 (ε0 + h)x − Q2 (ε0 )z
+ Q−1
2 (ε0 + h) (S̃(t, ε0 + h) − S̃(t, ε0 ))Q(ε0 )z
+ (Q−1 −1
2 (ε0 + h) − Q2 (ε0 ))S̃(t, ε0 )Q2 (ε0 )z .

Since Q−1 −1
2 (ε0 + h)S̃(t, ε0 + h) and Q2 (ε0 + h) are uniformly bounded,
and, from Lemma 3.5 and (2.6), S̃(t, ε), Q2 (ε), and Q−12 (ε) are continuous
in ε, we have (3.6).
Remarks. The results in [2, 8] show that the classical or integral solution
of (1.3) can be expressed in terms of the integrated semigroup S(t, ε) gen-
erated by A(ε). Hence, the existence of the integrated semigroup S(t, ε)
and continuity in the parameter ε of S(t, ε) imply the existence of a unique
classical or integral solution of (1.3) and continuity in the parameter ε of
such a solution, and thus the same results apply to (1.1).
42 Ronald Grimmer and Min He

4. A viscoelastic equation dependent on parameters


We consider a problem from one-dimensional linear viscoelasticity which
can be addressed easily with our theory. In particular, we consider the
problem of longitudinal vibrations in a viscoelastic rod of length one. The
equation can be written
ρutt (t, x) = σx (t, x), t ≥ 0, 0 ≤ x ≤ 1, (4.1)
where σ is the longitudinal stress and ρ is the density. Following Achenbach,
[1, pp. 399-401], and assuming that the material is initially quiescent, we
write σ as follows.
 t
σ(t, x) = G(0)ux (t, x) + G (t − s)ux (s, x) ds
0
 t
= G(0)ux (t, x) + G (s)ux (t − s, x) ds. (4.2)
0
For our purposes here, we assume that the rod is physically restrained on
each end with a dashpot device. This gives rise to boundary conditions of
the form
αut (t, 0) − σ(t, 0) = 0, ut (t, 1) + βσ(t, 1) = 0. (4.3)
Here α and β are positive constants depending on the strength of the dash-
pots.
Let us introduce the new variable v = ut . Writing our equation in terms
of this new variable, we obtain
ρvt = σx
 t
σt = G(0)vx + G (s)vx (t − s) ds + G (t)v(0) (4.4)
0
αv(t, 0) − σ(t, 0), v(t, 1) + βσ(t, 1), α > 0, β > 0.
Now, the constant ρ and function G depend on the material while the α and
β depend on the external dashpots. Thus the α and β can be varied. That
is, they may be considered as parameters in the problem.
We also assume that G(t) > 0, G (t) and G (t) are bounded continuous
functions defined for t ≥ 0 with |G(t)| ≤ M1 , |G (t)| ≤ M2 and |G (t)| ≤ M3 ,
t ≥ 0. The associated abstract equation obtained by taking v = ut , w = ux ,
and z = (v, w)t and is given by
 t
dz(t)
= A(ε)z(t) + F (t − s)A(ε)z(s) ds + f (t), z(0) = z0 (4.5)
dt 0
integrodifferential equations with parameter-dependent operators 43
2
on X = i=1 (C[0, 1], 
· ∞ ), where


0 ∂x
A(ε) = , ε = (α, β) ∈ R+ 2
,
∂x 0


v αv(0) = w(0),
D(A(ε)) = ∈ C 1 [0, 1] × C 1 [0, 1] α, β ≥ 0 ,
w v(1) = −βw(1),




a(t − s) 0 g(t) u1
F (t − s) = , f (t) = , z0 = .
0 0 0 u0
Note that (4.2) is a special case of (1.1). Also, for each ε,

 
v
2
αv(0) = w(0),
D(A(ε)) = X0 (ε) = ∈X= C[0, 1] α, β ≥ 0 .
w v(1) = −βw(1),
i=1
Thus, the operator A(ε) is not densely defined and the domain of A(ε) is
dependent on ε.
To apply Theorem 3.6, we need to show that A(ε) satisfies
Assumption


v̄ v
Q, (2.3) and (2.6). Now make the change of variables z̄ = =U
w̄ w

1 1

where U = 1 2 2
1 . We have
2 −2


−1 ∂x 0
A1 (ε) ≡ U A(ε)U = ,
0 −∂x


v (1 − α)v(0) = (1 + α)w(0), 
D(A1 (ε)) = ∈ C [0, 1] × C [0, 1]
1 1
.
w (1 + β)v(1) = −(1 − β)w(1),
1−β
Defining ᾱ = 1−α
1+α , β̄ = 1+β , A1 (ε) has the domain


v ᾱv(0) = w(0),
D(A1 (ε)) = ∈ C [0, 1] × C [0, 1]
1 1
ᾱ, β̄ ∈ (−1, 1] .
w v(1) = −β̄w(1),
Since A1 (ε) = U A(ε)U −1 and (λI − A(ε))−1 = U −1 (λI − A1 (ε))−1 U and
U, U −1 are bounded and independent of ε, it suffices to show that the oper-
ator A1 (ε) satisfies all hypotheses of Theorem 3.6.
For ε = (ᾱ, β̄), with ᾱ, β̄ ∈ (−1, 1],
1
A1 (ε) = Q(ε)A1 (0)Q(ε),
1 + ᾱβ̄


1 β̄
where Q(ε) = . It is clear that Assumption Q and (2.6) are satis-
−ᾱ 1
fied for any ε0 = (ᾱ0 , β̄0 ) in (−1, 1].
44 Ronald Grimmer and Min He

Now the remaining thing is to show that A1 (ε) satisfies (2.3).


For any (f, g)t ∈ X, consider



v f λv − v  = f
(λI − A1 (ε)) = , i.e.,
w g λw + w = g
with the boundary conditions ᾱv(0) = w(0), v(1) = −β̄w(1).
We want to estimate (λI − A1 (ε))−1 (f, g)t for a given (f, g)t ∈ X and
λ > 0. Let


v −1 f
= (λI − A1 (ε)) .
w g


 v 
Case 1.   = v∞ . If v∞ = |v(x0 )| where x0 ∈ (0, 1), then
w
v  (x0 ) = 0 and λv(x0 ) = f (x0 ). Thus,


1 1 1 f 
v∞ ≤ |f (x0 )| ≤ f ∞ ≤  
λ λ λ g
implies

 1


 v   f 
  = v∞ ≤  .
w λ g
If v∞ = |v(0)|, we may assume that v(0) > 0. Then v  (0) ≤ 0 and
0 < λv(0) ≤ f (0). Thus,

 1


 v  1  f 
  = v∞ = v(0) ≤ f (0) ≤  .
w λ λ g
If v∞ = |v(1)|, then v(1) = −β̄w(1) implies |v(1)| ≤ |β̄||w(1)| ≤ |w(1)|
(since |β̄| ≤ 1). Also


 v 
w∞ ≤   = v∞ = |v(1)| ≤ |w(1)| ≤ w∞
w
implies


 v 
  = w∞ = |w(1)|.
w
We may assume that w(1) > 0. Then w (1) ≥ 0, and 0 < λw(1) ≤ g(1).
Thus,

 1


 v  1  f 
  = w∞ = |w(1)| ≤ g(1) ≤  .
w ∞ λ λ g


 v 
Case 2.   = w∞ . The proof is very similar to that of Case 1,
w
and thus it is omitted here.
integrodifferential equations with parameter-dependent operators 45

Now Theorem 3.6 and Remarks in Section 3 apply to Equation (4.1).


For related material regarding hyperbolic equations with non-densely de-
fined operators, see [5]. In addition, for a treatment of abstract delay prob-
lems with application to the wave equation with delay, see [4] and [7].

References
[1] J.D. Achenbach, “Wave Propagation in Elastic Solids,” North-Holland, Amsterdam,
1973.
[2] G. Da Prato and E. Sinestrari, Differential operators with non dense domain, Annali
Scuola Norm. Sup. Pisa, 14 (1987), 285–344.
[3] E. D. Davies, “One-Parameter Semigroups,” Academic Press, London, 1980.
[4] R. Grimmer and J. Liu, Integrodifferential equations with non-densely defined oper-
ators, in “Differential Equation with Applications in Biology, Physics and Engineer-
ing,” J. Goldstein, F. Kappel, and W. Schappacher (eds.), Marcel Dekker Inc., 1991,
185–199.
[5] R. Grimmer and E. Sinestrari, Maximum norm in one-dimensional hyperbolic prob-
lems, Diff. Int. Equ., 5 (1992), 421–432.
[6] A. Pazy, “Semigroups of Linear Operators and Applications to Partial Differential
Equations,” Springer-Verlag, New York, 1983.
[7] E. Sinestrari, Wave equation with memory, Differential and Integral Eq., to appear.
[8] H.R. Thieme, Integrated semigroups and integrated solutions to abstract Cauchy prob-
lems, J. Math. Anal. Appl., 152 (1990), 416–447.

You might also like