Nonlinear Analysis: Real World Applications 7 (2006) 510 – 519

www.elsevier.com/locate/na

A note on partial functional differential equations

with state-dependent delay
Eduardo Hernández∗ , Andréa Prokopczyk, Luiz Ladeira
ICMC, Universidade de São Paulo, São Carlos, Cx 668, 13560-970 SP, Brazil
Received 7 February 2005; accepted 30 March 2005

Abstract
In this paper we study the existence of mild solutions for a class of abstract partial functional
differential equation with state-dependent delay.
䉷 2005 Elsevier Ltd. All rights reserved.

MSC: 34K30; 35R10; 47D06

Keywords: Functional differential equations; Mild solutions; State-dependent delay; Unbounded delay

1. Introduction

In this note we establish the existence of mild solutions for a abstract Cauchy problem
described in the form

x  (t) = Ax(t) + f (t, x
(t,xt ) ), t ∈ I = [0, a], (1)

x0 =  ∈ B, (2)

where A is the infinitesimal generator of a compact C0 -semigroup of bounded linear oper-

ators (T (t))t
0 on a Banach space X; the function xs : (−∞, 0] → X, xs () = x(s + ),
belongs to some abstract phase space B described axiomatically and f : I × B → X,

: I × B → (−∞, a] are appropriate functions.

∗ Corresponding author. Tel.: +55 16 3373 9754; fax: +55 16 3371 2238.
E-mail addresses: lalohm@icmc.sc.usp.br (E. Hernández), andreia@icmc.sc.usp.br (A. Prokopczyk),
ladeira@icmc.sc.usp.br (L. Ladeira).

1468-1218/$ - see front matter 䉷 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.nonrwa.2005.03.014
 E. Hernández et al. / Nonlinear Analysis: Real World Applications 7 (2006) 510 – 519 511

The literature related to ordinary and partial functional differential equations with delay
for which
(t, ) = t is very extensive and we refer the reader to Hale and Lunel [7] and
Wu [17] concerning this matter.
Functional differential equations with state-dependent delay appear frequently in appli-
cations as model of equations and for this reason the study of this type of equations has
received great attention in the last years, see for instance [1–5,8–11,13,14,16,18] and the
references therein. The literature devoted to this subject is concerned with functional differ-
ential equations which the state belong to some finite-dimensional space. Obviously, this
type of systems does not permit the study of partial functional differential equations with
state-dependent delay. This fact is the main motivation of our paper.
Throughout this paper, A : D(A) ⊂ X → X is the infinitesimal generator of a com-
pact semigroup of linear operators (T (t))t
0 on a Banach space X and M
is a constant


such that T (t)  M for every t ∈ I . Related semigroup theory, we suggest the Pazy
book [15].
In this work we will employ an axiomatic definition for the phase space B which is
similar to those introduced in [12]. Specifically, B will be a linear space of functions
mapping (−∞, 0] into X endowed with a seminorm  · B and satisfying the following
axioms:

(A) If x : (−∞, b] → X, b > 0, is continuous on [0, b] and x0 ∈ B, then for every

t ∈ [0, b] the following conditions hold:
(a) xt is in B.
(b) x(t) H xt B .
(c) xt B M(t)x0 B + K(t) sup{x(s) : 0 s t},
where H > 0 is a constant; K, M : [0, ∞) → [1, ∞), K(·) is continuous, M(·) is
locally bounded and H, K(·), M(·) are independent of x(·).
(A1) For the function x(·) in (A), xt is a B-valued continuous function on [0, b].
(B) The space B is complete.

Example (The phase space Cr × Lp (g; X)). Let g : (−∞, −r) → R be a positive
Lebesgue integrable function and assume that there exists a non-negative and locally
bounded function  on (−∞, 0] such that g( + ) ()g() for all  0 and  ∈
(−∞, −r)\N , where N ⊆ (−∞, −r) is a set with Lebesgue measure zero. The space
Cr × Lp (g; X) consists of all classes of functions  : (−∞, 0] → X such that  is contin-
uous on [−r, 0], Lebesgue measurable and gp is Lebesgue integrable on (−∞, −r).
The seminorm in Cr × Lp (g : X) is defined by

 −r 1/p
p
B := sup{() : −r 0} + g()() d .
−∞

Assume that g(·) verifies the conditions (g-5), (g-6) and (g-7) in the nomenclature of [12].
In this case, B = Cr × Lp (g; X) verifies axioms (A), (A1), (B) see [12, Theorem 1.3.8]
for details. Moreover, when r = 0 and p = 2 we have that H = 1, M(t) = (−t)1/2 and
0
K(t) = 1 + ( −t g() d)1/2 for t
0.
512 E. Hernández et al. / Nonlinear Analysis: Real World Applications 7 (2006) 510 – 519

Example (The phase space Cg0 (X)). Let g : (−∞, 0] → [0, ∞) be a continuous positive
function such that g(0) = 1 and assume that g(·) verifies conditions (g-1) and (g-2) in [12].
Let Cg0 (X) be the space consisting of all continuous functions  : (−∞, 0] → X such that
()/g() → 0 as  → −∞ endowed with the norm

()
g := sup .
−∞<  0 g()

Then, Cg0 satisfies axioms (A), (A1), (B), see [12, Theorem 1.3.2]. Moreover, H = 1, K(t) =
sup{g −1 () : −t 0} and M(t) = G(t) = e e−()t for t
0, see [12, Theorem 1.3.6].

Remark 1. Let  ∈ B and t 0. The notation t represents the function defined by
t () = (t + ). Consequently, if the function x(·) in axiom A is such that x0 = , then
xt = t . We observe that t is well defined for every t < 0 since the domain of (·) is
(−∞, 0].
We also note that, in general, t ∈
/ B. Consider for example the characteristic function
X[−r,0] , r > 0, in the space Cr × Lp (g; X).

The terminology and notations are those generally used in functional analysis. In par-
ticular, for Banach spaces (Z,  · Z ), (W,  · W ), the notation L(Z, W ) stands for the
Banach space of bounded linear operators from Z into W. Moreover, Br (x, Z) denotes the
closed ball with center at x and radius r > 0 in Z. Additionally, for a bounded function
 : I → [0, ∞) and 0 t a we employ the notation t for

t = sup{(s) : s ∈ [0, t]}. (3)

To prove our results we will use the classic Schauder Fixed Point Theorem and of the
following result, which is referred in the literature as Leray Schauder Alternative Theorem,
see [6, Theorem 6.5.4].

Theorem 1.1. Let D be a convex subset of a Banach space X and assume that 0 ∈ D. Let
G : D → D be a completely continuous map. Then the map G has a fixed point in D or the
set {x ∈ D : x = G(x), 0 <  < 1} is unbounded.

The paper has three sections. In Section 2, we establish the existence of mild solutions
for the abstract Cauchy problem (1), (2). Section 3 is reserved for examples.

2. Existence results

In this section, we establish the existence of mild solutions for the abstract Cauchy
problem (1), (2). To prove our results we always assume that
: I × B → (−∞, a] is
continuous and that  ∈ B and f (·) verifies the next conditions.
 E. Hernández et al. / Nonlinear Analysis: Real World Applications 7 (2006) 510 – 519 513

H1 . The function f : I × B → X satisfies the following properties.

(a) The function f (·, ) : I → X is strongly measurable for every  ∈ B.
(b) The function f (t, ·) : B → X is continuous for each t ∈ I .
(c) There exist an integrable function m : I → [0, ∞) and a continuous non-decreasing
function W : [0, ∞) → (0, ∞) such that
f (t, ) m(t)W (B ), (t, ) ∈ I × B.
H
. The function t → t is well defined and continuous from the set R(
− ) = {
(s, ) :
(s, ) ∈ I ×B,
(s, )0} into B and there exists a continuous and bounded function
J  : R(
− ) → (0, ∞) such that t B J  (t)B for every t ∈ R(
− ).
Remark 2. Related to condition H
, we point here that this condition is frequently veri-
fied by functions continuous and bounded. In fact, if the space B verifies axiom C2 in the
nomenclature of [12], then there exists a constant L > 0 such that B L sup  0 ()
for every  ∈ B continuous and bounded, see [12, Proposition 7.1.1] for details. Conse-
quently, t B Lsup  0 ()/B B for every continuous and bounded function
 ∈ B\{0} and every t 0. We also observe that the spaces Cr × Lp (g; X), Cg0 (X) verify
axiom C2 , see [12, p. 10,16] for details.

From Pazy [15], we adopt the following concept of mild solution for (1), (2).

Definition 2.1. A function x : (−∞, a] → X is called a mild solution of the abstract

Cauchy problem (1), (2) if x0 = , x
(s,xs ) ∈ B for every s ∈ I and
 t
x(t) = T (t)(0) + T (t − s)f (s, x
(s,xs ) ) ds, t ∈ I .
0

To prove our first theorem we need the next preliminary result.

Lemma 2.1. Let  ∈ B and J = ( , 0] be such that t ∈ B for every t ∈ J . Assume that
there exists a locally bounded function J  : J → [0, ∞) such that t B J  (t)B
for every t ∈ J . If x : (−∞, a] → X is continuous on I and x0 = , then
xs B (Ma + J  (max{ , −|s|})B + Ka xmax{0,s} , s ∈ ( , a].

Proof. The assertion follows from the inequalities

xs B = s B J  (s)B , s ∈ J ,
xs B M(s)B + K(s)xs , s ∈ I. 

Now, we can prove our first existence result.

Theorem 2.2. Assume conditions H1 , H
be satisfied. If

 a

a lim inf W ()
1 > MK m(s) ds,
→∞+  0

then there exists a mild solution of (1), (2).

514 E. Hernández et al. / Nonlinear Analysis: Real World Applications 7 (2006) 510 – 519

Proof. Let Y = {u ∈ C(I : X) : u(0) = (0)} endowed with the uniform convergence
topology and
: Y → Y be the operator defined by
 t

x(t) = T (t)(0) + T (t − s)f (s, x̄
(s,x̄s ) ) ds, t ∈ I, (4)
0

where x̄ : (−∞, a] → X is such that x̄0 =  and x̄ = x on I. From axiom (A), the strong
continuity of (T (t))t
0 and our assumptions on , we infer that
x(·) is well defined and
continuous.
Let 
¯ : (−∞, a] → X be the extension of  to (−∞, a] such that () ¯ = (0) on I and
J
 = sup{J  (s) : s ∈ R(
− )}. We affirm that there exists r > 0 such that
(Br (|¯ I , Y )) ⊂
¯ I , Y ). If we assume that this property is false, then for every r > 0 there exist x r ∈
Br (|
¯ I , Y ) and t r ∈ I such that r < 
x r (t r ) − (0). Then, from Lemma 2.1 we find that
Br (|

r < 
x r (t r ) − (0)
 tr
r
T (t )(0) − (0) + T (t r − s)f (s, x r
(s,x r ) ) ds
s
0
 tr

+ 1)H B + M
(M
m(s)W (x r
(s,x r ) B ) ds
s
0
 tr

+ 1)H B + M
(M
m(s)W ((Ma + J
 )B + Ka x r a ) ds
0  a

+ 1)H B + MW
(M
((Ma + J
 )B + Ka (r + (0))) m(s) ds
0

and hence
 a

a lim inf W ()
1 MK m(s) ds,
→∞  0

which is contrary to our assumption.

Let r > 0 be such that
(Br (| ¯ I , Y )) ⊂ Br (|¯ I , Y ). Next, we will prove that
(·) is
completely continuous from Br (| ¯ I , Y ) into Br (|
¯ I , Y ).
Step 1. The set
(Br (|
¯ I , Y ))(t) = {
x(t) : x ∈ Br (| ¯ I , Y )} is relatively compact in X
for every t ∈ I .
The case t = 0 is obvious. Let 0 <  < t a. If x ∈ Br (| ¯ I , Y ), from Lemma 2.1 follows
that

x̄
(t,x̄t ) B r ∗ := (Ma + J
 )B + Ka (r + (0))

and so that
   
  a
  ∗∗
∗
T ( − s)f (s, x̄
(s,x̄s ) ) ds  r := MW (r ) m(s) ds,  ∈ I. (5)

0 0
 E. Hernández et al. / Nonlinear Analysis: Real World Applications 7 (2006) 510 – 519 515

Using (5) we find that

 t−

x(t) = T (t)(0) + T () T (t −  − s)f (s, x̄
(s,x̄s ) ) ds
0
 t
+ T (t − s)f (s, x̄
(s,x̄s ) ) ds
t−
∈ {T (t)(0)} + T ()Br ∗∗ (0, X) + C ,

where diam(C )2MW
(r ∗ ) t m(s) ds, which proves that
(Br (| ¯ I , Y ))(t) is totally
t−
bounded and hence relatively compact in X.
¯ I , Y )) is equicontinuous on I.
Step 2. The set of functions
(Br (|
Let 0 < t < a and  > 0. From Step 1 and the strong continuity of (T (t))t
0 , we can
choose 0 <  a − t such that
T (h)x − x < , x ∈
(Br (|
¯ I , Y ))(t),
when 0 < h < . Under these conditions, for x ∈ Br (|
¯ I , Y ) and 0 < h <  we get

x(t + h) −
x(t) T (t + h)(0) − T (t)(0) + (T (h) − I )
x(t)
 t+h
+M
m(s)W (r ∗ ) ds
t
 t+h


 M(T
(r ∗ )
(h) − I )(0) +  + MW m(s) ds,
t

where r ∗ := (Ma + J
 )B + Ka (r + (0)). It proves that

(Br (| ¯ I , Y )) is right
equicontinuous at t ∈ (0, a). An analogous procedure shows that
(Br (| ¯ I , Y )) is right
equicontinuous at zero and left equicontinuous at t ∈ (0, a]. Thus, the set of functions

(Br (|
¯ I , Y )) is equicontinuous on I.
Step 3. The map
(·) is continuous on Br (| ¯ I , Y ).
Let (x n )n∈N be a sequence in Br (| ¯ I , Y ) such that x n → x in Y.
¯ I , Y ) and x ∈ Br (|
At first, we study the convergence of the sequences (x n
(s,x n ) )n∈N , s ∈ I . If s ∈ I is such
s
that
(s, x s ) > 0, we can fix N ∈ N such that
(s, xsn ) > 0 for every n > N. In this case, for
n > N we see that
x n
(s,x n ) − x
(s,x s ) B x n
(s,x n ) − x
(s,x n ) B + x
(s,x n ) − x
(s,x s ) B
s s s s
Ka x n () − x()
(s,x n ) + x
(s,x n ) − x
(s,x s ) B
s s
Ka x n − xa + x
(s,x n ) − x
(s,x s ) B ,
s

xn
which proves that
(s,x n ) → x
(s,x s ) in B as n → ∞ for every s ∈ I such that
s

(s, x s ) > 0. Similarly, if
(s, x s ) < 0 and N ∈ N is such that
(s, xsn ) < 0 for every n > N,
we get
x n
(s,x n ) − x
(s,x s ) B = 
(s,x n ) − 
(s,x s ) B ,
s s

which also shows that xn
(s,xsn ) → x
(s,x s ) in B as n → ∞ for every s ∈ I such that

(s, x s ) < 0.
516 E. Hernández et al. / Nonlinear Analysis: Real World Applications 7 (2006) 510 – 519

Combining the previous arguments, we can prove that x n
(s,x n ) →  for every s ∈ I
s
such that
(s, x s ) = 0.
From the previous remarks, f (s, x n
(s,x n ) ) → f (s, x
(s,x s ) ) for every s ∈ [0, a]. Now,
s
condition H1 and the Lebesgue Dominated Convergence Theorem permit to assert that
n

x →
x in Y. Thus,
(·) is continuous.
The existence of a mild solution for (1), (2) is now a consequence of the Schauder Fixed
Point Theorem. The proof is completed. 

Theorem 2.3. Let H
, H1 be satisfied. If
(t, )t for every (t, ) ∈ I × B and
 a  ∞

ds
Ka M m(s) ds < ,
0 C W (s)

where C = (Ka MH
+ Ma + J
 )B and J
 = sup{J  (s) : s ∈ R(
− )}, then there exists
a mild solution of (1), (2).

Proof. Let Y = C(I : X) and

: Y → Y be the operator defined by (4). In order to use
Leray Schauder Alternative Theorem, next we will shall a priori estimates for the solutions
of the integral equation z = 
z,  ∈ (0, 1). If x  = 
x  ,  ∈ (0, 1), then from Lemma
2.1 we have that
 t


x (t) MH B +
(s, x 
Mf
(s,x̄s ) ) ds
0
 t

B + M
 MH
m(s)W ((Ma + J
 )B + Ka x  max{0,
(s,x  )} ) ds
s
0 t

B + M
 MH
m(s)W ((Ma + J
 )B + Ka x  s ) ds,
0

since
(s, xs )s for every s ∈ I . If  (t) = (Ma + J
 )B + Ka x  t , we obtain that
 t
 (t)(Ka MH
+ Ma + J
 )B + Ka M
m(s)W ( (s)) ds. (6)
0

Denoting by  (t) the right-hand side of (6), follows that

 (t)Ka Mm(t)W (  (t))

and hence
  (t)  a  ∞
ds
ds
Ka M m(s) ds < ,
 (0)=C W (s) 0 C W (s)

which implies that the set of functions {  (·) :  ∈ (0, 1)} is bounded in C(I : R). Thus,
the set {x  (·) :  ∈ (0, 1)} is also bounded in C(I : X).
Arguing as in the proof of Theorem 2.2 we can prove that
(·) is completely continuous.
These remarks and Theorem 1.1 shows the existence of mild solution for (1), (2). The
proof is now completed. 
 E. Hernández et al. / Nonlinear Analysis: Real World Applications 7 (2006) 510 – 519 517

3. Examples

In this section, we discuss briefly a pair of examples of partial functional differential

equations with state-dependent delay.
Let X = L2 ([0, ]) and let A be the operator given by Af = f  with domain D(A) :=
{f ∈ X : f  ∈ X, f (0) = f () = 0}. It is well known that A is the infinitesimal generator
of a compact C0 -semigroup (T (t))t
0 on X.
Consider the differential system
 t
ju(t, ) j2 u(t, )
= + a1 (s − t)u
jt j2 −∞
    
× s −
1 (t)
2 a2 () | u(t, )| d ,  ds,
2
0
t ∈ I = [0, a],  ∈ [0, ], (7)
submitted to the initial conditions
u(t, 0) = u(t, ) = 0, t
0, (8)
u(, ) = (, ),  0, 0 , (9)
where  ∈ B = C0 × L2 (g; X), the functions ai : R → R,
i : [0, ∞) → [0, ∞), i = 1, 2
∞
are continuous, the function a2 (·) is positive and L1 = ( 0 a12 (s)/g(s) ds)1/2 < 1.
By defining the operators f : I × B → X,
: I × B → R by
 0
f (t, )() = a1 (s)(s, ) ds,
−∞
  

(s, ) = s −
1 (s)
2 a2 () | (0, )|2 d ,
0

we can transform system (7)–(9) into the abstract Cauchy problem (1), (2). Moreover, f (·)
is a continuous linear operator and f L(B,X) L1 .
The next results are consequence of Theorem 2.3 and Remark 2.

Theorem 3.4. Let  ∈ B be such that condition H
holds. Then there exists a mild solution
of (7)–(9).

Corollary 3.1. Let  ∈ B continuous and bounded. Then there exists a mild solution of
(7)–(9) on I.

Now, we consider briefly the existence of mild solution for the differential system
ju(t, ) j2 u(t, )
= + a(t)b(u(t − (u(t, 0)), )), (10)
jt j2
u(t, 0) = u(t, ) = 0, (11)
u(, ) = (, ), 0 (12)
518 E. Hernández et al. / Nonlinear Analysis: Real World Applications 7 (2006) 510 – 519

for t ∈ I = [0, a] and  ∈ [0, ]. Here,  ∈ B = Cg0 (X), the functions a : I → R,
b : R × J → R,  : R → R+ are continuous and we assume the existence of positive
constants b1 , b2 such that | b(t) | b1 | t | +b2 for every t ∈ R.
By defining the maps f (t, )() = a(t)b((0, )) and
(t, ) = t − ((0, 0)), we
represent system (10), (11) by the abstract Cauchy problem (1), (2). Moreover, a simple
estimates shows that f (t, ) a(t)[b1 B + b2 1/2 ] for all (t, ) ∈ I × B.
Under these conditions, the next results are valid.

Theorem 3.5. If  ∈ B verifies condition H
, then there exits a mild solution of (10)–(12).

Corollary 3.2. If  is continuous and bounded, then there exists a mild solution of (10)–(12).

Acknowledgements

Andréa Prokopczyk wishes to acknowledge the support of Cnpq, Brazil, for this research.

References

[1] W.G. Aiello, H.I. Freedman, J. Wu, Analysis of a model representing stage-structured population growth with
state-dependent time delay, SIAM J. Appl. Math. 52 (3) (1992) 855–869.
[2] D. Alexander, D. Michael, L. Elena, On equations with delay depending on solution, Nonlinear Anal. TMA
49 (5) (2002) 689–701.
[3] O. Arino, K. Boushaba, A. Boussouar, A mathematical model of the dynamics of the phytoplankton-nutrient
system. Spatial heterogeneity in ecological models (Alcalá de Henares, 1998), Nonlinear Anal. RWA 1 (1)
(2000) 69–87.
[4] Y. Cao, J. Fan, T.C. Gard, The effects of state-dependent time delay on a stage-structured population growth
model, Nonlinear Anal. TMA 19 (2) (1992) 95–105.
[5] C. Fengde, S. Dexian, S. Jinlin, Periodicity in a food-limited population model with toxicants and state
dependent delays, J. Math. Anal. Appl. 288 (1) (2003) 136–146.
[6] A. Granas, J. Dugundji, Fixed Point Theory, Springer, New York, 2003.
[7] J.K. Hale, S.M. Verduyn Lunel, Introduction to Functional–Differential Equations, Applied Mathematical
Sciences, vol. 99, Springer, New York, 1993.
[8] F. Hartung, Parameter estimation by quasilinearization in functional differential equations with state-
dependent delays: A numerical study, Proceedings of the Third World Congress of Nonlinear Analysts,
Part 7, Catania, 2000; F. Hartung, Nonlinear Anal. TMA 47(7) (2001) 4557–4566.
[9] F. Hartung, Linearized stability in periodic functional differential equations with state-dependent delays,
J. Comput. Appl. Math. 174 (2) (2005) 201–211.
[10] F. Hartung, T.L. Herdman, J. Turi, Parameter identification in classes of neutral differential equations with
state-dependent delays, Nonlinear Anal. TMA Ser. A: Theory Methods 39 (3) (2000) 305–325.
[11] F. Hartung, J. Turi, Identification of parameters in delay equations with state-dependent delays, Nonlinear
Anal. TMA 29 (11) (1997) 1303–1318.
[12] Y. Hino, S. Murakami, T. Naito, Functional–differential equations with infinite delay, Lecture Notes in
Mathematics, vol. 1473, Springer, Berlin, 1991.
[13] Y. Kuang, H.L. Smith, Slowly oscillating periodic solutions of autonomous state-dependent delay equations,
Nonlinear Anal. TMA 19 (9) (1992) 855–872.
[14] B. Mária, Periodic solutions for differential equations with state-dependent delay and positive feedback,
Nonlinear Anal. TMA 53 (6) (2003) 839–857.
[15] A. Pazy, Semigroups of Linear Operators and Applications to Partial Differential Equations, Applied
Mathematical Sciences, vol. 44, Springer, New York, Berlin, 1983.
 E. Hernández et al. / Nonlinear Analysis: Real World Applications 7 (2006) 510 – 519 519

[16] R. Torrejón, Positive almost periodic solutions of a state-dependent delay nonlinear integral equation,
Nonlinear Anal. TMA 20 (12) (1993) 1383–1416.
[17] J. Wu, Theory and Applications of Partial Functional–Differential Equations, Applied Mathematical Sciences,
vol. 119, Springer, New York, 1996.
[18] L. Yongkun, Periodic solutions for delay Lotka Volterra competition systems, J. Math. Anal. Appl. 246 (1)
(2000) 230–244.