POLONICI MATHEMATICI LVII.2 (1992)
Generalized solutions to boundary value problems for quasilinear hyperbolic systems
of partial differential-functional equations
by Tomasz Cz lapi´ nski (Gda´ nsk)
Abstract. Generalized solutions to quasilinear hyperbolic systems in the second canonical form are investigated. A theorem on existence, uniqueness and continuous dependence upon the boundary data is given. The proof is based on the methods due to L. Cesari and P. Bassanini for systems which are not functional.
1. Introduction. We denote by C(X, Y ) the set of all continuous functions from X to Y , where X, Y are any metric spaces. Let a
0> 0, B = [−b
0, 0] × [−b, b], where b
0∈ R
+, b = (b
1, . . . , b
r) ∈ R
r+, R
+= [0, ∞).
For z : [−b
0, a
0] × R
r→ R
mand (x, y) = (x, y
1, . . . , y
r) ∈ [0, a
0] × R
r, define z
(x,y): B → R
mby z
(x,y)(t, s) = z(x + t, y + s), (t, s) ∈ B. Put Ω = [0, a
0] × R
r× C(B, R
m), and denote by M (m, r) the set of all real m × r matrices.
Let
A : Ω → M (m, m) , A = [A
ij], i, j = 1, . . . , m,
% : Ω → M (m, r), % = [%
ij], i = 1, . . . , m, j = 1, . . . , r, f : Ω → M (m, 1), f = [f
1, . . . , f
m]
T,
be given functions of the variables (x, y, w), y=(y
1, . . . , y
r), w=(w
1, . . . , w
r), and T the transpose symbol. Note that if (x, y) ∈ [0, a
0] × R
ris fixed then A(x, y, ·), %(x, y, ·), f (x, y, ·) are operators on the function space C(B, R
m).
Furthermore, let
B
l: R
r→ M (m, m), B
l= [B
lij], i, j = 1, . . . , m, ψ : R
r→ R
r, ψ = (ψ
1, . . . , ψ
r) ,
where l=1, . . . , N , m≤N , be given functions of the variable y=(y
1, . . . , y
r).
1991 Mathematics Subject Classification: 35L50, 35D05.
Key words and phrases: differential-functional system, second canonical form, gener- alized solutions.
We consider the following hyperbolic differential-functional system in the second canonical form:
(1)
m
X
j=1
A
ij(x, y, z
(x,y)) h
D
xz
j(x, y) +
r
X
k=1
%
ik(x, y, z
(x,y))D
ykz
j(x, y) i
= f
i(x, y, z
(x,y)), i = 1, . . . , m , with the boundary data on N arbitrary (not necessarily distinct) hyper- planes x = a
l, a
l∈ [0, a
0], l = 1, . . . , N ,
(2)
N
X
l=1
B
l(y)z(a
l, y) = ψ(y), y ∈ R
r.
A function z ∈ C(I
a, R
m), I
a= [0, a]×R
r, where a ∈ (0, a
0], is a solution of (1), (2) if z has partial derivatives D
xz, D
ykz, k = 1, . . . , r, a.e. on I
a, satisfies (1) a.e. on I
aand (2) for all y ∈ R
r.
If b
0= 0 and b = 0 then the differential-functional system (1) reduces to a differential system in the second canonical form. Generalized (a.e.) solutions of systems of that form have been investigated in a large number of papers. We mention here the papers of P. Bassanini [1]–[3] and L. Ce- sari [5], [6]. Quasilinear hyperbolic systems with a retarded argument have been studied by Z. Kamont and J. Turo [12]–[15]. In [19]–[22] J. Turo has also considered the existence and uniqueness of generalized solutions for differential-functional hyperbolic systems with operators of Volterra type.
In that case the given functions are superpositions of functions defined on subsets of a finite-dimensional Euclidean space with operators of Volterra type. The main assumptions for the operators have been formulated in the form of linear inequalities given on some function space. Classical solutions to nonlinear equations with the same model of functional dependence have been investigated in [11], [18].
In this paper we consider a new model of differential-functional systems
which has been used in [9] for initial value problems. For each (x, y) ∈
[−b
0, a
0] × R
rand for any z : [−b
0, a
0] × R
r→ R
mwe denote by z
(x,y)the
translation of the restriction of z to B. In other words, the graph of z
(x,y)is
the graph of z : [x − b
0, x] × [y − b, y + b] → R
mshifted to B. Since the given
functions are now operators on z
(x,y)we no longer need the assumption that
the right-hand side of the system is the superposition of some function and
an operator of Volterra type. This simple model is well known for ordi-
nary differential-functional equations (see [10], [16]). It is also very general
since systems of differential equations with retarded argument ([12]–[15]),
differential-integral systems ([17]) and differential-functional systems with
operators of Volterra type ([19]–[22]) can be obtained from (1) by special-
izing the given functions. Our formulation of the problem enables us to get
new examples of differential-functional systems which cannot be obtained from the results cited above. More detailed comparison between the model of functional dependence proposed in this paper and that used in [19]–[22]
is presented in [7], [8]. The method which we use is based on the Banach fixed point theorem in a product space and it is close to that used in [14].
2. Assumptions and notations. Let kηk
m= max
1≤i≤m|η
i| in R
mand kU k
m,k= max
1≤i≤mP
mj=1
|U
ij| in M (m, k). For short we write kηk, kU k.
For w ∈ C(B, R
m) put kwk
0= sup{kw(t, s)k : (t, s) ∈ B}. Let C
0+L(B, R
m) denote the set of all functions w ∈ C(B, R
m) such that
kwk
L= sup{[|t − t| + kw − wk]
−1kw(t, s) − w(t, s)k : (t, s), (t, s) ∈ B} < ∞ . We equip this set with the norm kwk
0+L= kwk
0+ kwk
L, w ∈ C
0+L(B, R
m).
Furthermore, for any q ∈ R
+we set C(B, R
m, q) = {w ∈ C(B, R
m) : kwk
0≤ q}, C
0+L(B, R
m, q) = {w ∈ C
0+L(B, R
m) : kwk
0+L≤ q}.
Assumption H
1. 1
oA ∈ C(Ω, M (m, m)) and there is a constant ν > 0 such that det A(x, y, w) ≥ ν for any (x, y, w) ∈ Ω.
2
oThere are nondecreasing functions n, l : R
+→ R
+such that for any q ∈ R
+we have
(i) kA(x, y, w)k ≤ n(q), (x, y, w) ∈ [0, a
0] × R
r× C(B, R
m, q),
(ii) kA(x, y, w) − A(x, y, w)k ≤ l(q)[|x − x| + ky − yk + kw − wk
0], (x, y, w), (x, y, w) ∈ [0, a
0] × R
r× C
0+L(B, R
m, q).
R e m a r k 1. Assumption H
1yields that A
−1(x, y, w) exists for any (x, y, w) ∈ Ω. Furthermore, there are nondecreasing n
0, l
0: R
+→ R
+such that condition 2
oof Assumption H
1holds with A, n, l replaced by A
−1, n
0, l
0, respectively.
Let θ denote the set of all functions l : [0, a
0] × R
+→ R
+such that l(·, q) : [0, a
0] → R
+is measurable for any q ∈ R
+, and l(x, ·) : R
+→ R
+is nondecreasing for a.e. x ∈ [0, a
0].
Assumption H
2. 1
o%(·, y, w) : [0, a
0] → M (m, r) is measurable for all (y, w) ∈ R
r× C(B, R
m).
2
o%(x, ·) : R
r× C(B, R
m) → M (m, r) is continuous for a.e. x ∈ [0, a
0].
3
oThere are a nondecreasing function n
1: R
+→ R
+and l
1∈ θ such that for any q ∈ R
+we have
(i) k%(x, y, w)k ≤ n
1(q), (x, y, w) ∈ [0, a
0] × R
r× C(B, R
m, q),
(ii) k%(x, y, w)−%(x, y, w)k≤l
1(x, q)[ky − yk+kw − wk
0] for all (y, w), (y, w) ∈ R
r× C
0+L(B, R
m, q) and for a.e. x ∈ [0, a
0].
Assumption H
3. 1
of (·, y, w) : [0, a
0] → M (m, 1) is measurable for all (y, w) ∈ R
r× C(B, R
m).
2
of (x, ·) : R
r× C(B, R
m) → M (m, 1) is continuous for a.e. x ∈ [0, a
0].
3
oThere are a nondecreasing function n
2: R
+→ R
+and l
2∈ θ such that for any q ∈ R
+we have
(i) kf (x, y, w)k ≤ n
2(q), (x, y, w) ∈ [0, a
0] × R
r× C(B, R
m, q),
(ii) kf (x, y, w) − f (x, y, w)k ≤ l
2(x, q)[ky − yk + kw − wk
0] for all (y, w), (y, w) ∈ R
r× C
0+L(B, R
m, q) and for a.e. x ∈ [0, a
0].
For (x, y, w) ∈ Ω, l = 1, . . . , N , we put
(3) A(x, y, w) = E + e A(x, y, w), A
−1(x, y, w) = E + A(x, y, w), B
l(y) = E
l+ e B
l(y), E = [δ
ij], E
l= [δ
liδ
ij], i, j = 1, . . . , m , and
σ
0= sup n X
Nl=1
k e B
l(y)k : y ∈ R
ro , σ
1= sup{k e A(x, y, w)k : (x, y, w) ∈ Ω} , σ
2= sup{kA(x, y, w)k : (x, y, w) ∈ Ω} .
Assumption H
4. 1
oψ ∈ C(R
r, R
m) and there are constants Γ, Λ ∈ R
+such that for all y, y ∈ R
rwe have
kψ(y)k ≤ Γ, kψ(y) − ψ(y)k ≤ Λky − yk .
2
oB
l∈ C(R
r, M (m, m)), det B
l(y) 6= 0 for all l = 1, . . . , N , y ∈ R
r, and there is a constant G ∈ R
+such that
N
X
l=1
kB
l(y) − B
l(y)k ≤ Gky − yk, y, y ∈ R
r. 3
o(σ
0+ σ
1)(1 + σ
2) < 1.
We denote by B
1(a, Q, Q
1, Q
2), where a ∈ (0, a
0], Q, Q
1, Q
2∈ R
+, the set of all functions z ∈ C(I
a, R
m) such that for all (x, y), (x, y) ∈ I
awe have
(i) kz(x, y)k ≤ Q,
(ii) kz(x, y) − z(x, y)k ≤ Q
1|x − x| + Q
2ky − yk.
Note that B
1(a, Q, Q
1, Q
2) is a closed subset of the Banach space of all continuous and bounded vector functions z : I
a→ R
mwith the norm kzk
Ia= sup{kz(x, y)k : (x, y) ∈ I
a}.
For a ∈ (0, a
0] define the constants L
ia(Q, Q
1, Q
2) = R
a0
l
i(t, Q + Q
1+ Q
2) dt, i = 1, 2. In the sequel we write L
iafor short.
Let a ∈ (0, a
0] be so small that
(4) L
1a(1 + Q
2) < 1 ,
and let Q, Q
1, Q
2∈ R
+, p ∈ (0, 1). Then we denote by B
2(a, Q, Q
1, Q
2, p)
the set of all functions h ∈ C(∆
a, M (m, r)), ∆
a= [0, a] × [0, a] × R
r, such
that for all (x, x, y), (ξ, x, y), (ξ, x, y) ∈ ∆
a, i = 1, . . . , m, we have
(i) h
i(x, x, y) = 0,
(ii) kh
i(ξ, x, y) − h
i(ξ, x, y)k ≤ n
1(Q)|ξ − ξ| + γ
an
1(Q)|x − x| + pky − yk, where γ
a= [1 − L
1a(1 + Q
2)]
−1, h
i= (h
i1, . . . , h
ir). It is easily seen that B
2(a, Q, Q
1, Q
2, p) is a closed subset of the Banach space of all continuous and bounded matrix functions h : ∆
a→ M (m, r) with the norm khk
∆a= max
1≤i≤msup{kh
i(ξ, x, y)k : (ξ, x, y) ∈ ∆
a}. Indeed, for all (ξ, x, y) ∈ ∆
a, i = 1, . . . , m, we have
kh
i(ξ, x, y)k = kh
i(ξ, x, y) − h
i(x, x, y)k ≤ n
1(Q)|ξ − x| ≤ n
1(Q)a , which yields khk
∆a≤ n
1(Q)a.
For any h ∈ B
2(a, Q, Q
1, Q
2, p) let g : ∆
a→ M (m, r) be defined by (5) g
i(ξ, x, y) = y + h
i(ξ, x, y), (ξ, x, y) ∈ ∆
a, i = 1, . . . , m ,
where g
i= (g
i1, . . . , g
ir), h
i= (h
i1, . . . , h
ir). Then for all (x, x, y), (ξ, x, y), (ξ, x, y) ∈ ∆
a, i = 1, . . . , m, we have
(i) g
i(x, x, y) = y,
(ii) kg
i(ξ, x, y)−g
i(ξ, x, y)k ≤ n
1(Q)|ξ−ξ|+γ
an
1(Q)|x−x|+(1+p)ky−yk.
3. The operators V
(1), V
(2)and their properties. Let h·, ·i be a scalar product in R
m, let g
i(t, x, y) be defined by (5) and set
z
i∗(t, x, y) = (z
1(t, g
i(t, x, y)), . . . , z
m(t, g
i(t, x, y))) , A
∗i(t, x, y)
= (A
i1(t, g
i(t, x, y), z
(t,gi(t,x,y))), . . . , A
im(t, g
i(t, x, y), z
(t,gi(t,x,y)))) . We consider the operators Z = V
(1)[z, h], H = V
(2)[z, h] defined for z ∈ B
1(a, Q, Q
1, Q
2), h ∈ B
2(a, Q, Q
1, Q
2, p) as follows:
Z(x, y) = A
−1(x, y, z
(x,y))(∆
1(x, y) + ∆
2(x, y) + ∆
3(x, y)) ,
∆
k(x, y) = (∆
k1(x, y), . . . , ∆
km(x, y)), k = 1, 2, 3,
∆
1i(x, y) = ψ
i(g
i(a
i, x, y)) +
x
R
ai
f
i(t, g
i(t, x, y), z
(t,gi(t,x,y))) dt,
∆
2i(x, y) =
x
R
ai
hD
tA
∗i(t, x, y), z
i∗(t, x, y)i dt (6)
+ hA
∗i(a
i, x, y), z
i∗(a
i, x, y)i ,
∆
3i(x, y) = −
N
X
l=1
hB
li(g
i(a
i, x, y)), z(a
l, g
i(a
i, x, y))i,
i = 1, . . . , m, (x, y) ∈ I
a;
and
(7)
H(ξ, x, y) = [H
ij(ξ, x, y)], i = 1, . . . , m, j = 1, . . . , r, H
i(ξ, x, y) =
ξ
R
x
%
i(t, g
i(t, x, y), z
(t,gi(t,x,y))) dt,
i=1, . . . , m, (ξ, x, y) ∈ ∆
a. R e m a r k 2. By (3) we may simultaneously replace D
tA
∗i, A
i, B
liin the above equalities by D
tA e
∗i, e A
i, e B
li, respectively.
Lemma 1. Suppose that Assumptions H
1–H
4hold and that (8) (1 + σ
2)Γ + (1 + σ
2)(σ
0+ σ
1)Q < Q,
(9) l
0(Q)(1 + Q
1)[Γ + (σ
0+ σ
1)Q] + (1 + σ
2)[n
1(Q)(Λ + GQ) + l(Q)Q(2 + n
1(Q) + 2Q
1+ Q
2n
1(Q))]
+ n
1(Q)(1 + σ
2)(σ
0+ σ
1)Q
2< Q
1, (10) l
0(Q)(1+Q
2)[Γ +(σ
0+σ
1)Q]+(1+σ
2)[(1+p)(Λ+GQ)+l(Q)Q(1+Q
2)]
+ (1 + p)(1 + σ
2)(σ
0+ σ
1)Q
2< Q
2. Then for a∈(0, a
0] sufficiently small the operator V
(1)maps B
1(a, Q, Q
1, Q
2)
× B
2(a, Q, Q
1, Q
2, p) into B
1(a, Q, Q
1, Q
2).
P r o o f. Let Z=V
(1)[z, h], z∈B
1(a, Q, Q
1, Q
2), h ∈ B
2(a, Q, Q
1, Q
2, p).
By the estimates k∆
1(x, y)k ≤ max
1≤i≤m
n
|ψ
i(g
i(a
i, x, y))|
+
x
R
ai
f
i(t, g
i(t, x, y), z
(t,gi(t,x,y))) dt o
≤ Γ +
x
R
ai
n
2(Q) dt ≤ Γ + n
2(Q)a,
k∆
2(x, y)k ≤ max
1≤i≤m
n
x
R
ai
hD
tA e
∗i(t, x, y), z
i∗(t, x, y)i dt + |h e A
∗i(a
i, x, y), z
i∗(a
i, x, y)i|
o
≤
x
R
ai
l(Q)(1 + n
1(Q) + Q
1+ Q
2n
1(Q))Q dt + σ
1Q
≤ l(Q)(1 + n
1(Q) + Q
1+ Q
2n
1(Q))Qa + σ
1Q, k∆
3(x, y)k ≤ max
1≤i≤m
−
N
X
l=1
h e B
li(g
i(a
i, x, y)), z(a
l, g
i(a
i, x, y))i
≤ σ
0Q ,
we obtain kZ(x, y)k
≤ (1 + σ
2)[Γ + n
2(Q)a + l(Q)(1 + n
1(Q) + Q
1+ Q
2n
1(Q))Qa + σ
1Q + σ
0Q] , (x, y) ∈ I
a. By (8) we can choose a so small that
(11) (1 + σ
2)[Γ + (σ
1+ σ
0)Q]
+(1 + σ
2)[n
2(Q) + l(Q)(1 + n
1(Q) + Q
1+ Q
2n
1(Q))Q]a ≤ Q , and thus for (x, y) ∈ I
awe obtain
(12) kZ(x, y)k ≤ Q .
For any (x, y), (x, y) ∈ I
a, we have
Z(x, y) − Z(x, y) = α
0+ α
1+ α
2+ α
3, where
α
0= (A
−1(x, y, z
(x,y)) − A
−1(x, y, z
(x,y)))[∆
1(x, y) + ∆
2(x, y) + ∆
3(x, y)], α
i= A
−1(x, y, z
(x,y))[∆
i(x, y) − ∆
i(x, y)], i = 1, 2, 3 .
We can estimate the above terms as follows:
kα
0k ≤ [l
0(Q)(1 + Q
2)|x − x| + l
0(Q)(1 + Q
2)ky − yk]
× [k∆
1(x, y)k + k∆
2(x, y)k + k∆
3(x, y)k]
≤ [Γ + n
2(Q)a + l(Q)(1 + n
1(Q) + Q
1+ Q
2n
1(Q))Qa + σ
1Q + σ
0Q]l
0(Q)(1 + Q
1)|x − x|
+ [Γ + n
2(Q)a + l(Q)(1 + n
1(Q) + Q
1+ Q
2n
1(Q))Qa + σ
1Q + σ
0Q]l
0(Q)(1 + Q
1)ky − yk,
kα
1k ≤ (1 + σ
2) max
1≤i≤m
n
|ψ
i(g
i(a
i, x, y)) − ψ
i(g
i(a
i, x, y))|
+
x
R
ai
[f
i(t, g
i(t, x, y), z
(t,gi(t,x,y)))
− f
i(t, g
i(t, x, y), z
(t,gi(t,x,y)))] dt +
x
R
x
f
i(t, g
i(t, x, y), z
(t,gi(t,x,y))) dt o
≤ (1 + σ
2)[Λγ
a+ L
2a(1 + Q
2)γ
a+ n
2(Q)a]|x − x|
+ (1 + σ
2)[Λ(1 + p) + L
2a(1 + Q
2)(1 + p)]ky − yk, kα
2k ≤ (1 + σ
2) max
1≤i≤m
n
x
R
ai
h e A
∗i(t, x, y) − e A
∗i(t, x, y), D
tz
∗i(t, x, y)i dt
+
x
R
ai
hD
tA e
∗i(t, x, y), z
∗i(t, x, y) − z
∗i(t, x, y)i dt +
x
R
x
hD
tA e
∗i(t, x, y), z
∗i(t, x, y)i dt + |h e A
∗i(x, x, y) − e A
∗i(x, x, y), z
i∗(x, x, y)i|
+ |h e A
∗i(a
i, x, y), z
i∗(a
i, x, y) − z
∗i(a
i, x, y)i| o
≤ (1 + σ
2){l(Q)(1 + Q
2)γ
a(Q
1+ Q
2n
1(Q))a + l(Q)(1 + n
1(Q) + Q
1+ Q
2n
1(Q))Q
2γ
aa + l(Q)(1 + n
1(Q) + Q
1+ Q
2n
1(Q))Q + l(Q)(1 + Q
1)Q + σ
1Q
2γ
a}|x − x|
+ (1 + σ
2){l(Q)(1 + Q
2)(1 + p)(Q
1+ Q
2n
1(Q))a + l(Q)(1 + n
1(Q) + Q
1+ Q
2n
1(Q))Q
2(1 + p)a + l(Q)(1 + Q
2)Q + σ
1Q
2(1 + p)}ky − yk, kα
3k ≤ (1 + σ
2)
× max
1≤i≤m
n
N
X
l=1
h e B
li(g
i(a
i, x, y)) − e B
li(g
i(a
i, x, y)), z(a
l, g
i(a
i, x, y))i +
N
X
l=1
h e B
li(g
i(a
i, x, y)), z(a
l, g
i(a
i, x, y)) − z(a
l, g
i(a
i, x, y))i o
≤ (1 + σ
2)γ
a[GQ + σ
0Q
2]|x − x| + (1 + σ
2)(1 + p)[GQ + σ
0Q
2]ky − yk . In estimating α
2we have used integration by parts. From the above in- equalities we obtain
kZ(x, y) − Z(x, y)k ≤ W
1a|x − x| + W
2aky − yk , where
W
1a= l
0(Q)(1 + Q
1)[Γ + (σ
1+ σ
0)]
+ (1 + σ
2)[γ
a(Λ + GQ) + l(Q)Q(2 + n
1(Q) + 2Q
1+ Q
2n
1(Q))]
+ γ
a(1 + σ
2)(σ
1+ σ
0)Q
2+ W
1a0, W
2a= l
0(Q)(1 + Q
2)[Γ + (σ
1+ σ
0)]
+ (1 + σ
2)[(1 + p)(Λ + GQ) + l(Q)Q(1 + Q
2)]
+ (1 + p)(1 + σ
2)(σ
1+ σ
0)Q
2+ W
2a0,
and W
1a0≥ 0, W
2a0≥ 0 are some constants such that lim
a→0+W
1a0= lim
a→0+W
2a0= 0. By (9), (10) we can choose a so small that
(13) W
1a≤ Q
1, W
2a≤ Q
2.
Hence, for any (x, y), (x, y) ∈ I
awe get
(14) kZ(x, y) − Z(x, y)k ≤ Q
1|x − x| + Q
2ky − yk .
Finally, if a ∈ (0, a
0] is so small that inequalities (4), (11), (13) hold, then by (12), (14) we have Z ∈ B
1(a, Q, Q
1, Q
2). This completes the proof.
Lemma 2. Suppose that Assumption H
2holds, Q, Q
1, Q
2∈ R
+, p ∈ (0, 1) and a ∈ (0, a
0] is so small that
(15) L
1a(1 + p)(1 + Q) ≤ p .
Then the operator V
(2)maps B
1(a, Q, Q
1, Q
2) × B
2(a, Q, Q
1, Q
2, p) into B
2(a, Q, Q
1, Q
2).
P r o o f. Let H=V
(2)[z, h], z∈B
1(a, Q, Q
1, Q
2), h∈B
2(a, Q, Q
1, Q
2, p). It is easy to see that for any (ξ, x, y), (ξ, x, y), (ξ, x, y) ∈ ∆
a, i = 1, . . . , m, we have
H
i(x, x, y) = 0, kH
i(ξ, x, y) − H
i(ξ, x, y)k ≤ n
1(Q)|ξ − ξ|, kH
i(ξ, x, y) − H
i(ξ, x, y)k
≤
ξ
R
x
k%
i(t, g
i(t, x, y), z
(t,gi(t,x,y))) − %
i(t, g
i(t, x, y), z
(t,gi(t,x,y)))k dt
≤
ξ
R
x
l
1(t, P + Q)(1 + Q)kg
i(t, x, y) − g
i(t, x, y)k dt
≤ L
1a(1 + Q)(1 + p)ky − yk ≤ pky − yk
by (15). Note that (15) implies (4), and thus for any (ξ, x, y), (ξ, x, y) ∈ ∆
a, i = 1, . . . , m, we have
kH
i(ξ, x, y) − H
i(ξ, x, y)k ≤
x
R
x
k%
i(t, g
i(t, x, y), z
(t,gi(t,x,y)))k dt +
ξ
R
x
k%
i(t, g
i(t, x, y), z
(t,gi(t,x,y))) − %
i(t, g
i(t, x, y), z
(t,gi(t,x,y)))k dt
≤ [n
1(Q) + L
1a(1 + Q)γ
a]|x − x| = γ
a|x − x| .
From the above inequalities we derive H ∈ B
2(a, Q, Q
1, Q
2, p). This is our claim.
For a ∈ (0, a
0] define constants
E
1a= l
0(Q)[Γ + n
2(Q)a + l(Q)(1 + n
1(Q) + Q
1+ Q
2n
1(Q))Qa + σ
1Q + σ
0Q] + (1 + σ
2)[L
2a+ l(Q)(Q
1+ Q
2n
1(Q))a
+ l(Q)(1 + n
1(Q) + Q
1+ Q
2n
1(Q))a + l(Q)Q + σ
1+ σ
0],
E
2a= (1 + σ
2)[Λ + L
2a(1 + Q
2) + l(Q)(1 + Q
2)(Q
1+ Q
2n
1(Q))a + l(Q)Q
1(1 + n
1(Q) + Q
1+ Q
2n
1(Q))a + σ
1Q
2+ GQ + σ
0Q
2] . Lemma 3. Let Assumptions H
1–H
4hold , Q, Q
1, Q
2∈ R
+, p ∈ (0, 1) and a ∈ (0, a
0] be so small that inequality (4) holds. Then for any z, z
0∈ B
1(a, Q, Q
1, Q
2), h, h
0∈ B
2(a, Q, Q
1, Q
2, p), we have
kV
(1)[z, h] − V
(1)[z
0, h
0]k
Ia≤ E
1akz − z
0k
Ia+ E
2akh − h
0k
∆a, (16)
kV
(2)[z, h] − V
(2)[z
0, h
0]k
Ia≤ L
1akz − z
0k
Ia+ L
1a(1 + Q)kh − h
0k
∆a. (17)
P r o o f. Let z, z
0∈ B
1(a, Q, Q
1, Q
2) and h, h
0∈ B
2(a, Q, Q
1, Q
2, p). For any (x, y) ∈ I
awe have
V
(1)[z, h](x, y) − V
(1)[z
0, h
0](x, y) = β
0+ β
1+ β
2+ β
3, where
β
0= (A
−1(x, y, z
(x,y)) − A
−1(x, y, z
(x,y)0))[∆
1(x, y) + ∆
2(x, y) + ∆
3(x, y)], β
i= A
−1(x, y, z
0(x,y))[∆
i(x, y) − ∆
0i(x, y)], i = 1, 2, 3,
and formulas for ∆
01, ∆
02, ∆
03arise from (6) by replacing z, h by z
0, h
0, re- spectively. Analogously to Lemma 1 we have
kβ
0k ≤ l
0(Q)[Γ + n
2(Q)a + l(Q)(1 + n
1(Q)
+ Q
1+ Q
2n
1(Q))Qa + σ
1Q + σ
0Q]kz − z
0k
Ia,
kβ
1k ≤ (1 + σ
2)L
2akz − z
0k
Ia+ (1 + σ
2)[Λ + L
2a(1 + Q
2)]kh − h
0k
∆a, kβ
2k ≤ (1 + σ
2)[l(Q)(Q
1+ Q
2n
1(Q))a
+ l(Q)(1 + n
1(Q) + Q
1+ Q
2n
1(Q))a + l(Q)Q + σ
1]kz − z
0k
Ia+ (1 + σ
2)[l(Q)(1 + Q
2)(Q
1+ Q
2n(Q))a
+ l(Q)(1 + n
1(Q) + Q
1+ Q
2n
1(Q))a + σ
1Q]kh − h
0k
∆a, kβ
3k ≤ (1 + σ
2)σ
0kz − z
0k
Ia+ (1 + σ
2)[GQ + σ
0Q
2]kh − h
0k
∆a.
From the above estimates we get (16). In a similar way we derive (17), which completes the proof.
4. The main theorem
Theorem. Suppose that Assumptions H
1–H
4, conditions (8)–(10) and the inequality
(18) l
0(Q)[Γ + Q(σ
1+ σ
0)] + (1 + σ
2)l(Q)Q + (1 + σ
2)(σ
1+ σ
0) < k , are satisfied , where k ∈ (0, 1) is a constant such that k > (σ
1+ σ
0)(1 + σ
2).
Then for a ∈ (0, a
0] sufficiently small and for any system of numbers a
l∈
[0, a], l = 1, . . . , N , there is a function z ∈ B
1(a, Q, Q
1, Q
2) which is a
unique solution of the problem (1), (2) in the class B
1(a, Q, Q
1, Q
2). Fur- thermore, if z, z
0are solutions of (1), (2) with functions ψ, ψ
0respectively then
(19) kz − z
0k
Ia≤ (1 − k)
−1(1 + σ
2)kψ − ψ
0k
0, where kψ − ψ
0k
0= sup{kψ(y) − ψ
0(y)k : y ∈ R
r}.
P r o o f. Let V = (V
(1), V
(2)), where V
(1), V
(2)are defined by (6), (7).
From Lemmas 1–3 it follows that
V : B
1(a, Q, Q
1, Q
2) × B
2(a, Q, Q
1, Q
2, p)
→ B
1(a, Q, Q
1, Q
2) × B
2(a, Q, Q
1, Q
2, p) is continuous provided that a is so small that (11), (13) and (15) hold. By (18) we may additionally assume that a is small enough that
(20) E
1a< k, L
1a(1 + Q
2) < k, E
2aL
1a< (k − E
1a)(k − L
a(1 + Q
2)) . Let α, β > 0 satisfy L
1a(k − E
1a)
−1≤ α/β ≤ (k − L
1a(1 + Q
2))E
2a−1. Analogously to [4] we define in B
1(a, Q, Q
1, Q
2) × B
2(a, Q, Q
1, Q
2, p) the following weighted norm:
(21) kwk
∗= αkzk
Ia+ βkhk
∆a, w = (z, h) ,
where z ∈ B
1(a, Q, Q
1, Q
2), h ∈ B
2(a, Q, Q
1, Q
2, p). It is easy to check (cf. [4]) that V is a contraction with constant k with respect to this norm.
Thus V has a unique fixed point w = V w, w = (z, h) ∈ B
1(a, Q, Q
1, Q
2) × B
2(a, Q, Q
1, Q
2, p). Let us prove that z is a solution of (1), (2). From (6) by integration by parts we obtain
(22) 0 = A
−1(x, y, z
(x,y))[∆
1(x, y) + e ∆
2(x, y) + ∆
3(x, y)], (x, y) ∈ I
a, where ∆
1, ∆
3are defined by (6) with z = z, h = h and e ∆
2(x, y) = ( e ∆
21(x, y), . . . , e ∆
2m(x, y)) is defined by
∆ e
2i(x, y) = −
x
R
ai
hA
∗i(t, x, y), D
tz
∗i(t, x, y)i dt, i = 1, . . . , m . Multiplying (22) by A(x, y, z
(x,y)) we obtain
∆
1i(x, y) + e ∆
2i(x, y) + ∆
3i(x, y) = 0, (x, y) ∈ I
a, i = 1, . . . , m , which for x = a
iyields that z satisfies the boundary condition (2). Thus (22) reduces to
x
R
ai
[f
i(t, g
i(t, x, y), z
(t,gi(t,x,y))) − hA
∗i(t, x, y), D
tz
∗i(t, x, y)i] dt = 0 ,
(x, y) ∈ I
a, i = 1, . . . , m, where g is defined by (5) for h = h. By the same
considerations as in [6] (in particular by using the group property for g) we
see that z satisfies (1) a.e. in I
a. By the reverse considerations and by the uniqueness of the fixed point of V we conclude that z is a unique solution of (1), (2).
It remains to prove (19). Let w = (z, h) = V w, w
0= (z
0, h
0) = V w
0be the fixed points of V = (V
(1), V
(2)) defined by (6), (7) with functions ψ, ψ
0respectively. We then have
kz − z
0k
Ia≤ (1 + σ
2)kψ − ψ
0k
0+ E
1akz − z
0k
Ia+ E
2akh − h
0k
∆a, kh − h
0k
∆a≤ L
1akz − z
0k
Ia+ L
1a(1 + Q
2)kh − h
0k
∆a.
From this we obtain (1 − E
1a− E
2aL
1a(1 − L
1a(1 + Q
2))
−1)kz − z
0k
Ia≤ (1 + σ
2)kψ − ψ
0k
0. By (20) we have 1 − k ≤ 1 − E
1a− E
2aL
1a(1 − L
1a(1 + Q
2))
−1, and from the last inequality we derive (19). This ends the proof.
5. Examples. Put b Ω = [0, a
0] × R
r× R
m× R
m.
Assumption H
5. 1
oA ∈ C( b b Ω, M (m, m)) and there is a constant ν > 0 b such that det b A(x, y, u, v) ≥ b ν for any (x, y, u, v) ∈ b Ω.
2
oThere are constants n, l ∈ R
+such that for all (x, y, u, v), (x, y, u, v) ∈ Ω we have b
k b A(x, y, u, v)k ≤ n ,
k b A(x, y, u, v) − b A(x, y, u, v)k ≤ l[|x − x| + ky − yk + ku − uk + kv − vk] . 3
o%(·, y, u, v) : [0, a b
0] → M (m, r), b f (·, y, u, v) : [0, a
0] → M (m, 1) are measurable for any (y, u, v) ∈ R
r× R
m× R
mand %(x, ·) : R b
r× R
m× R
m→ M (m, r), b f (x, ·) : R
r× R
m× R
m→ M (m, 1) are continuous for a.e.
x ∈ [0, a
0].
4
oThere are Lebesgue integrable functions b l
1, b l
2: [0, a
0] → R
+and con- stants n
1, n
2∈ R
+such that for a.e. x ∈ [0, a
0] and for all (y, u, v), (y, u, v) ∈ R
r× R
m× R
mwe have
k %(x, y, u, v)k ≤ n b
1, k b f (x, y, u, v)k ≤ n
2,
k %(x, y, u, v) − b %(x, y, u, v)k ≤ l b
1(x)[ky − yk + ku − uk + kv − vk] , k b f (x, y, u, v) − b f (x, y, u, v)k ≤ l
2(x)[ky − yk + ku − uk + kv − vk] . Let α, β, γ : [0, a
0] × R
r→ R
1+r, where α = (α
0, α
1, . . . , α
r), β = (β
0, β
1, . . . , β
r), γ = (γ
0, γ
1, . . . , γ
r), be given functions such that −b
0≤ α
0(x, y), β
0(x, y), γ
0(x, y) ≤ x for all (x, y) ∈ [0, a
0] × R
r. Suppose that α is Lipschitz continuous on [0, a
0] × R
rand that β, γ are Lipschitz continuous with respect to the second variable on R
r. Then, if we define
A(x, y, w) = b A(x, y, w(0, 0), w(α(x, y) − (x, y))) ,
%(x, y, w) = %(x, y, w(0, 0), w(β(x, y) − (x, y))) , b
f (x, y, w) = b f (x, y, w(0, 0), w(γ(x, y) − (x, y))) ,
and if Assumption H
5holds then the functions A, %, f satisfy Assumptions H
1–H
3respectively. Thus the hyperbolic system with retarded argument (23)
m
X
j=1
A b
ij(x, y, z(x, y), z(α(x, y))) h
D
xz
j(x, y)
+
r
X
k=1
%(x, y, z(x, y), z(β(x, y)))D b
ykz
j(x, y) i
= b f (x, y, z(x, y), z(γ(x, y))), i = 1, . . . ., m, is a particular case of (1).
The differential-integral system (24)
m
X
j=1
A b
ijx, y, z(x, y),
˜ α(x,y)
R
α(x,y)
z(t, s)K(t, s, x, y) dt ds h
D
xz
j(x, y)
+
r
X
k=1
% b
ikx, y, z(x, y),
β(x,y)˜
R
β(x,y)
z(t, s)K
1(t, s, x, y) dt ds
D
ykz
j(x, y) i
= b f
ix, y, z(x, y),
˜ γ(x,y)
R
γ(x,y)
z(t, s)K
2(t, s, x, y) dt ds
, i = 1, . . . , m ,
where α, α, β, e e β, γ, e γ : [0, a
0] × R
r→ R
1+rand K, K
1, K
2: [0, a
0] × R
r× [0, a
0] × R
r→ M (m, m), is also a particular case of (1). In this case if we make suitable assumptions on α, α, β, e e β, γ, e γ, K, K
1, K
2and if Assumption H
5holds then A(x, y, w) defined by
A(x, y, w) = b A
x, y, w(0, 0),
˜ α(x,y)
R
α(x,y)
w(t − x, s − y)K(t, s, x, y) dt ds satisfies Assumption H
1, and %, f similarly defined satisfy Assumptions H
2, H
3respectively.
Our last example is the system (25)
m
X
j=1
A b
ij(x, y, z(x, y), z(α(x, y, z
(x,y)))) h
D
xz
j(x, y)
+
r
X
k=1
% b
ik(x, y, z(x, y), z(β(x, y, z
(x,y))))D
ykz
j(x, y) i
= b f
i(x, y, z(x, y), z(γ(x, y, z
(x,y)))) ,
i = 1, . . . , m, α, β, γ : Ω → R
1+r.
Lemma 4. Suppose that 1
oAssumption H
1holds;
2
oα(·, y, w), β(·, y, w), γ(·, y, w) : [0, a
0] → R
1+rare measurable for any (y, w) ∈ R
r× C(B, R
m), and α(x, ·), β(x, ·), γ(x, ·) : R
r× C(B, R
m) → R
1+rare continuous for a.e. x ∈ [0, a
0];
3
ofor any (x, y, w) ∈ Ω we have
α(x, y, w) ∈ B, β(x, y, w) ∈ B, γ(x, y, w) ∈ B ;
4
othere is a nondecreasing function D : R
+→ R
+such that for all q ∈ R
+, (x, y, w), (x, y, w) ∈ [0, a
0] × R
r× C
0+L(B, R
m, q) we have
kα(x, y, w) − α(x, y, w)k ≤ D(q)[|x − x| + ky − yk + kw − wk
0] ; 5
othere are nondecreasing functions D
1, D
2: R
+→ R
+such that for all q ∈ R
+and (y, w), (y, w) ∈ R
r× C
0+L(B, R
m, q) and for a.e. x ∈ [0, a
0] we have
kβ(x, y, w) − β(x, y, w)k ≤ D
1(q)[ky − yk + kw − wk
0] , kγ(x, y, w) − γ(x, y, w)k ≤ D
2(q)[ky − yk + kw − wk
0] . Then the functions A, %, f defined by
A(x, y, w) = b A(x, y, w(0, 0), w(α(x, y, w) − (x, y))) ,
%(x, y, w) = %(x, y, w(0, 0), w(β(x, y, w) − (x, y))) , b f (x, y, w) = b f (x, y, w(0, 0), w(γ(x, y, w) − (x, y))) , satisfy Assumptions H
1, H
2, H
3respectively.
R e m a r k 3. Systems (23), (24) can also be obtained from the theory of differential-functional equations with operators of Volterra type [19]–[22].
System (25) cannot be obtained from that theory.
References
[1] P. B a s s a n i n i, On a recent proof concerning a boundary value problem for quasilin- ear hyperbolic systems in the Schauder canonic form, Boll. Un. Mat. Ital. (5) 14-A (1977), 325–332.
[2] —, Iterative methods for quasilinear hyperbolic systems, ibid. (6) 1-B (1982), 225–
250.
[3] —, The problem of Graffi–Cesari , in: Nonlinear Phenomena in Math. Sci., V. Lak- shmikantham (ed.), Proc. Arlington 1980, Academic Press, 1982, 87–101.
[4] P. B a s s a n i n i e E. F i l l i a g g i, Schemi iterativi a accelerazione della convergenza per operatori di contrazione nel prodotto di due spazi di Banach, Atti Sem. Mat.
Fis. Modena 28 (1979), 249–279.
[5] L. C e s a r i, A boundary value problem for quasilinear hyperbolic systems, Riv. Mat.
Univ. Parma 3 (1974), 107–131.
[6] L. C e s a r i, A boundary value problem for quasilinear hyperbolic systems in the Schauder canonic form, Ann. Scuola Norm. Sup. Pisa (4) 1 (1974), 311–358.
[7] T. C z l a p i ´n s k i, On the Cauchy problem for quasilinear hyperbolic systems of partial differential-functional equations of the first order , Z. Anal. Anwendungen 10 (1991), 169–182.
[8] —, A boundary value problem for quasilinear hyperbolic systems of partial differen- tial-functional equations of the first order , Boll. Un. Mat. Ital. (7) 5-B (1991), 619–
637.
[9] T. C z l a p i ´n s k i and Z. K a m o n t, Generalized solutions of quasi-linear hyperbolic systems of partial differential-functional equations, to appear.
[10] J. H a l e, Functional Differential Equations, Springer, New York 1971.
[11] Z. K a m o n t, Existence of solutions of first order partial differential-functional equa- tions, Comment. Math. 25 (1985), 249–263.
[12] Z. K a m o n t and J. T u r o, On the Cauchy problem for quasilinear hyperbolic system of partial differential equations with a retarded argument , Boll. Un. Mat. Ital. (6) 4-B (1985), 901–916.
[13] —, —, On the Cauchy problem for quasilinear hyperbolic systems with a retarded argument , Ann. Mat. Pura Appl. 143 (1986), 235–246.
[14] —, —, A boundary value problem for quasilinear hyperbolic systems with a retarded argument , Ann. Polon. Math. 47 (1987), 347–360.
[15] —, —, Generalized solutions of boundary value problems for quasilinear systems with retarded argument , Radovi Mat. 4 (1988), 239–260.
[16] V. L a k s h m i k a n t h a m and S. L e e l a, Differential and Integral Inequalities, Vol.
2, Academic Press, New York 1969.
[17] N. M a t t i o l i and M. C. S a l v a t o r i, A theorem of existence and uniqueness in nonlinear dispersive optics, Atti Sem. Mat. Fis. Univ. Modena 28 (1979), 405–424.
[18] A. S a l v a d o r i, Sul problema di Cauchy per una struttura ereditaria di tipo iperbol- ico. Esistenza, unicit`a e dipendenza continua, ibid. 32 (1983), 329–356.
[19] J. T u r o, A boundary value problem for quasilinear hyperbolic systems of hereditary partial differential equations, ibid. 34 (1985–86), 15–34.
[20] —, On some class of quasilinear hyperbolic systems of partial differential-functional equations of the first order , Czechoslovak Math. J. 36 (111) (1986), 185–197.
[21] —, Existence and uniqueness of solutions of quasilinear hyperbolic systems of partial differential-functional equations, Math. Slovaca 37 (1987), 375–387.
[22] —, A boundary value problem for hyperbolic systems of differential-functional equa- tions, Nonlinear Anal. 13 (1) (1989), 7–18.
INSTITUTE OF MATHEMATICS UNIVERSITY OF GDA ´NSK WITA STWOSZA 57 80-952 GDA ´NSK, POLAND
Re¸cu par la R´edaction le 4.11.1991 R´evis´e le 15.2.1992