Académique Documents
Professionnel Documents
Culture Documents
THESE
Par
HASSANE BOUZAHIR
(Url : www.geocities.com/hbouzahir)
Examinateurs M. ADIMY : Professeur, Université de Pau et des Pays de l’Adour, Pau (France)
E. AIT DADS : Professeur, Université Cadi Ayyad, Marrakech
Rapporteur
H. BOUSLOUS : Professeur, Université Cadi Ayyad, Marrakech
K. EZZINBI : Professeur Habilité, Université Cadi Ayyad, Marrakech
Encadrant
My L. HBID : Professeur, Université Cadi Ayyad, Marrakech
A. RHANDI : Professeur, Université Cadi Ayyad, Marrakech
A la mémoire de mon père
(rahimah ALLAH)
(Blaise Pascal)
REMERCIEMENTS
Au terme de cette thèse, il m’est très agréable de m’acquitter d’une dette de reconnaissance
auprès de toutes les personnes dont l’intervention a favorisé son aboutissement.
J’adresse mes sincères remerciements aux Professeurs Elhadi Ait Dads de l’Université Cadi
Ayyad, Ovide ARINO de l’Université de Pau et des Pays de l ‘Adour, Jack Kenneth
HALE de Georgia Institute of Technology et Jianhong WU de York University de me faire le
grand honneur d’avoir accepter de rapporter mon travail. Je leurs exprime aussi ma grande
gratitude pour leurs précieux conseils et discussions sur mon sujet de recherche.
Je suis très sensible à l’amitié et la gentillesse que j’ai trouvé auprès des Professeurs Mostafa
Adimy de l’Université de Pau et Pays de l’Adour et Rachid Benkhalti de Pacific Lutheran
University. Leurs aides m’ont fait profiter des discussions et collaborations fructueuses qu’on a
eu ensemble.
Les Professeurs Hammadi Bouslous et Moulay Lhassan Hbid ont toujours des conseils
judicieux à me prodiger. Qu’ils trouvent ici l’expression de ma profonde reconnaissance pour
leurs aides et leurs soutiens.
Mes remerciements vont aussi aux Professeurs Abdelilah Dahlane , Mohammed Elalaoui
Talibi, Mohammed Erraoui, Mohammed Khaladi, Salah Labhala, Lahcen Maniar,
Brahim Sadik, Youssef Ouknine pour leurs sens humain et leur soutien.
Je saisis l’occasion pour remercier les collègues avec qui j’ai partagé des moments de joie
comme ceux de peine tout au long de nos années de troisième cycle. Chacun de nous a été une
source d’idées et informations pour les autres. Je remercie également mes autres camarades
étudiants et enseignants du département de mathématiques.
Je témoigne aussi la gentillesse des personnels du département de mathématiques, service de
troisième cycle, service d’intendance, service des bourses, service de relations extérieures,
buvette et veilleurs. Je les remercie beaucoup.
Je n’oublierais jamais de prier la miséricorde pour mon père Elarbi Bouzahir qui a contribué
puissamment à la formation de mon esprit et n’a épargné aucun effort pour que je mène à bien
mes études. Qu’Allah aille son âme.
En fin, j’adresse mes remerciements à mes proches et amis pour leurs aides morales et
matérielles durant la préparation de cette thèse. Mes remerciements les plus vifs vont à mon
frère Abdelhay Bouzahir qui a été le sponsor principal de mes études de troisième cycle.
Avant Propos
Publications
1) Global Attractor for A Class of Partial Functional Differential Equations with
Infinite Delay. ''Topics in Functional Differential and Difference
Equations'', Edited by P. Freitas and T. Faria, Fields Institute
Communications, Vol. 29 , American Mathematical Society, Providence, RI,
(March, 2001). (En collaboration avec K. Ezzinbi).
2) Existence for a Class of Partial Functional Differential Equations with Infinite
Delay. A paraître au «Journal of Nonlinear Analysis, Theory, Methods and
Applications». (En collaboration avec M. Adimy et K. Ezzinbi).
3) Local Existence and Stability for Some Partial Functional Differential
Equations with Infinite Delay. A paraître au «Journal of Nonlinear Analysis,
Theory, Methods and Applications». (En collaboration avec M. Adimy et K.
Ezzinbi).
4) Existence of Periodic Solutions for Some Partial Functional Differential
Equations with Infinite Delay. «Journal of Mathematical Analysis and
Applications», Vol. 256, N. 1, 257-280, (April, 2001). (En collaboration avec
R. Benkhalti et K. Ezzinbi).
Communications
1) Sur les équations différentielles semi-linéaires de type retard infini. A «CIMPA
Summer School on Evolution Equations and Applications», Ouagadougou,
Burkina Faso, 13-31 Juillet, 1998.
2) Local Existence and Stability for a Class of Partial Functional Differential
Equations with Infinite Delay. Au 4ièmes JAUCA, Marrakech, 27-29 Avril,
1999. La partie: «Local Existence and Semiflow for Some Partial Functional
Differential Equations with Infinite Delay» Soumise au numéro spécial du
Journal de la Société de Mathématiques du Maroc consacré au 4ièmes
JAUCA. (En collaboration avec K. Ezzinbi).
3) Existence de Solutions pour une Classe d’Equations aux Dérivées Partielles à
Retard. Au Colloque International sur les équations aux dérivées partielles,
Faculté des Sciences Dhar Mahraz, Fès, 4-8 Mai, 1999.
4) Semiflow Generated by a Class of Partial Functional Differential Equations
with Infinite Delay. A «International Conference on Functional Differential
and Difference Equations» Instituto Superior Técnico, Lisboa-Portugal, 26-
30 Juillet, 1999.
5) A Massera Type Criterion for Linear Partial Functional Differential Equations
with Infinite Delay. Au 6ièmes Journées d'Analyse Numérique et
Optimisation (JANO6) Facultés de Casablanca, 8-10 Mars, 2000. Paru aux
actes de JANO6.
6) Critical Spectrum and Asymptotic Behavior for Linear Partial Functional
Differential Equations with Infinite Delay. A «2nd European-Maghreb
Workshop on Semigroups Theory, Evolution Equations and Applications»,
L'Aquila-Italie, 25-30 Juin, 2000.
7) On Existence of Periodic Solutions for Some Partial Functional Differential
Equations with Infinite Delay. A CIMASI’2000, Ecole Hassania des Travaux
Publics, Casablanca, 23-25 Octobre, 2000.
Content
0.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
0.2 Historique et modèles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
0.2.1 Modèle de prolifération cellulaire . . . . . . . . . . . . . . . . . . . . 6
0.2.2 Modèle de proie-prédateur avec di¤usion dans l’espace . . . . . . . . 6
0.2.3 Modèle de dynamique d’une population distribuée . . . . . . . . . . 7
0.3 Equations di¤érentielles à retard …ni en dimension in…nie . . . . . . . . . . 7
0.4 Equations di¤érentielles à retard in…ni en dimension in…nie . . . . . . . . . 8
0.5 Equations di¤érentielles à retard de type neutre en dimension in…nie . . . . 9
0.6 Description de la thèse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1
3 Local Existence, Stability and Attractiveness for Some Partial Functional
Di¤erential Equations with In…nite Delay2 45
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.2 Local existence and global continuation of integral solutions . . . . . . . . . 46
3.3 Existence of strict solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.4 The solution semigroup and linearized stability . . . . . . . . . . . . . . . . 54
3.5 Attractiveness of solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.6 An application to a reaction di¤usion equation with in…nite delay . . . . . . 65
2
This chapter is based on two papers. The …rst one is in collaboration with M. Adimy and K. Ezzinbi
and it will appear in Journal of Nonlinear Analysis, Theory, Methods and Applications, (2001). The second
one is in collaboration with K. Ezzinbi and it will appear in Fields Institue Communications Series, (2001).
3
This chapter is based on an ongoing work in collaboration with M. Adimy and K. Ezzinbi.
4
This chapter is based on a paper in collaboration with R. Benkhalti and K. Ezzinbi. The paper will
appear in Journal of Mathematical Analysis and Applications, (2001).
2
PRESENTATION GENERALE
0.1 Introduction
Le travail présenté dans cette thèse se situe dans le cadre des équations di¤érentielles à
retard in…ni en dimension in…nie. Nous nous intéressons plus exactement aux équations de
type: 8
< d x(t) = Ax(t) + F (t; x ); t ¸ 0;
t
dt (0.1)
: x = ' 2 B;
0
où A : D(A) ½ E ! E est un opérateur linéaire fermé sur un espace de Banach (E; j:j) ;
pour tout t ¸ 0; la fonction xt 2 B est dé…nie par:
B est l’espace de phase constitué de fonctions dé…nies de ]¡1; 0] à valeurs dans E; véri…ant
certains axiomes (voir Chapitre 1). F : IR+ £ B ! E est une fonction continue.
Le long de cette thèse, nous supposons que D(A) 6= E et que la résolvante de A véri…e la
condition de Hille-Yosida: il existe ! 2 IR tel que l’ensemble résolvant ½(A) de A contient
]!; +1[ et que
© ° °ª
sup (¸ ¡ !)n °(¸I ¡ A)¡n ° < +1; (0.2)
n¸0; ¸>!
Notons que dans le cas où D(A) = E; le théorème de Hille-Yosida a¢rme que A engendre
un C0 semi-groupe (T (t))t¸0 sur l’espace E: Une solution (faible) de l’équation (0.1) est
donnée par la formule de variation de la constante suivante:
8 Z t
< T (t)'(0) + T (t ¡ s)F (s; xs )ds; pour t ¸ 0;
x(t; ') = 0 (0.3)
:
'(t); pour ¡ 1 < t · 0:
L’étude de l’équation (0.1) a fait l’objet de plusieurs travaux. Nous citons essentiellement
Hale et Kato [55]; Arino, Burton et Haddock [20]; Murakami [84]; Hino, Murakami et Naito
[66] et plus récemment, Henriquez ([58]-[60]); Hino, Murakami et Yoshizawa [67]; Liu et
Xu [77]; Naito, Murakami et Shin [91]; Shin [113]; Shin et Naito [114], et Shin, Naito et
Minh [115]. On cite aussi Milota et Petzeltova [82], [83], [98]-[100] et [101] dans le cas où A
3
engendre un semi groupe analytique dans E et Ruan et Wu [103]; Ruess [104]-[106]; Ruess
et Summers [107], [108]; Ruess, Summers et William [109], et Kartsatos et Parrott [69] dans
des cas plus généraus. La liste est loin d’être exhaustive.
Dans ce travail, nous traitons l’existence de solutions faibles ou intégrales et leurs régu-
larités. Nous utilisons la théorie des semi-groupes intégrés introduite par Arendt et al..
Nous entendons par solution faible ou intégrale toute fonction x : (¡1; T ] ! E; T > 0,
satisfaisant l’équation suivante:
8 Z t
< S 0 (t)'(0) + d S(t ¡ s)F (s; xs )ds; pour t 2 [0; T ] ;
x(t; ') = dt 0 (0.4)
:
'(t); pour ¡ 1 < t · 0:
4
d’équations. Au début, Malthus a présenté le modèle suivant:
d
N(t) = bN(t); (0.5)
dt
où N(t) est le nombre d’individus à l’instant t et b est le taux de fécondité.
On s’aperçoit que N évolue exponentiellement par rapport à t, et par conséquent, ce
modèle ne re‡ète pas l’évolution exacte de l’espèce, d’où la nécessité d’introduire d’autres
modèles plus réalistes. Les modèles à retard ont pour objet de résoudre ce problème. Le
modèle “proie-prédateur” de Volterra est constitué de deux populations l’une prédateur se
nourrissant de l’autre, la proie. Volterra a supposé que la croissance des prédateurs en
contact avec la proie n’est pas instantannée, dûe par exemple à une période de gestation.
Parmi les modèles les plus classiques, on trouve le système à retard suivant:
8
> d
< N1 (t) = (b1 ¡ a1 N2 (t))N1 (t);
dt · Z t ¸ (0.6)
> d
: N2 (t) = ¡b2 + a2 N1 (t) + k(µ ¡ t)N1 (µ)dµ N2 (t);
dt t¡r
où k est une fonction qui décrit la manière dont le gain du prédateur à chaque instant t
dépend de l’abondance de la population des proies N1 dans un intervalle de temps passé
[t ¡ r; t] : On distingue deux cas: modèle avec retard …ni (0 < r < +1) et modèle avec
retard in…ni (r = +1): Le modèle peut s’écrire alors sous la forme:
d
N(t) = F (t; Nt ); t ¸ 0; (0.7)
dt
avec la notation Nt (µ) = N(t + µ) pour chaque ¡1 < µ · 0 si r = +1 ou pour chaque
¡r · µ · 0 si 0 < r < +1: La fonction F : IR+ £ X ! E; où X est un espace vectoriel
approprié de fonctions allant de ]¡1; 0] à un espace de Banach E si r = +1 ou de [¡r; 0]
à E si 0 < r < +1: Dans l’exemple (0.6), la fonction F est dé…ne par
0 1
(b1 ¡ a1 Á2 (0))Á1 (0)
B · Z 0 ¸ C
F (t; Á) = @ A; (0.8)
¡b2 + a2 Á1 (0) + k(µ)Á1 (µ)dµ Á2 (0)
¡r
5
quelques modèles qui ont été obtenus dans la littérature et qui peuvent être transformés
en des équations de type (0.1) avec retard …ni ou in…ni. On se contente de donner trois
exemples de modèles de dynamique de populations structurées. Pour d’autres exemples,
nous renvoyons le lecteur aux livres de Wu [124] et Webb [122].
C’est un modèle de production du sang proposé par Rey et Mackey en 1993 et étudié
par Dyson, Villella-Bressan et Webb en 1995. Ce modèle décrit la production des souches
prolifératives et le précurseur des cellules dans la moelle oseuse:
8
> @ @
< @t u(x; t) + @x (xu(x; t)) = ¹u(®x; t ¡ r)(1 ¡ u(®x; t ¡ r));
>
0 < x < 1; t > 0; (0.9)
>
>
:
u(x; µ) = Á(x; t); 0 · x · 1; ¡ r · µ · 0;
Pour l’étude d’une population proie-prédateur, Cohen, Hagan et Simpson [34] ont proposé
le modèle suivant:
µZ +1 ¶
@
u(x; t) = ®¢u(x; t) + h K(s)u(x; t ¡ s)ds
@t µZ0 +1 ¶ (0.10)
¡m H(s)u(x; t ¡ s)ds ;
0
6
µZ +1 ¶
processus de croissance et de décroissance de la proie et m H(s)u(x; t ¡ s)ds est
0
la consommation de la population de la proie par le prédateur.
qui décrit les dynamiques d’une population distribuée d’une seule espèce. N(x; t) représente
la taille de la population totale à la position x et à l’instant t, les deux constantes d et c sont
positives mesurant, respectivement, le taux de di¤usion
Z 0 et la croissance intrinsèque et G :
+
]¡1; 0] ! IR est une fonction intégrable véri…ant G(µ)dµ = 1: La fonction Á est un
¡1
élément d’un espace vectoriel approprié de fonctions allant de ]¡1; 0] vers E := L2 ([0; ¼]):
Le modèle a été étudié par plusieurs auteurs, à savoir: Ruan et Wu [103], Bonilla et Linan
[27], et Britton [30].
où xt est dé…nie par xt (µ) = x(t + µ), pour ¡r · µ · 0; r > 0; et C est l’espace des fonctions
continues de [¡r; 0] à valeurs dans E muni de la topologie de la convergence uniforme.
Lorsque A est générateur in…nitésimal d’un C0 -semi-groupe (T (t))t¸0 dans E, l’équation
(0.12) a fait l’objet de plusieurs travaux. Dans [118], Travis et Webb ont formulé tous les
aspects d’existence et stabilité pour l’équation (0.12). Dans le cas où F est autonome et
globalement Lipschitzienne, ils ont démontré que la solution:
8 Z t
< T (t)'(0) + T (t ¡ s)F (xs )ds; pour t ¸ 0;
x(t; ') = 0 (0.13)
:
'(t); pour ¡ r · t · 0;
7
dé…nit un semi-groupe non linéaire (U(t))t¸0 sur C. Lorsque le semi-groupe (T (t))t¸0 est
compact, ils ont démontré que le semi-groupe solution U(t) est compact pour t > r. Cette
dernière propriété a permis dans le cas linéaire d’étudier la stabilité en terme d’une équation
caractéristique qui caractérise le spectre ponctuel du générateur in…nitésimal de (U(t))t¸0 :
Cette hypothèse de compacité a permis aussi de formuler le théorème de variété centre
pour l’équation (0.12), à savoir que l’espace C s’écrit comme somme directe de trois parties
stable, centre et instable. Dans [23] et [24], Arino et Sanchez ont traité l’équation (0.12)
lorsque A = 0; et ont donné une formule de variation de la constante. Dans [80] et [81],
Memory a étudié le problème d’existence d’une variété stable et instable pour l’équation
(0.12). Il a aussi donné une formule de variation de la constante. Dans [76], Lin, So et Wu
ont étudié l’existence d’une variété centre pour l’équation (0.12). Dans le cas où l’espace
instable est réduit à f0g, ils ont montré que la variété centre obtenue est exponentiellement
attractive. D’autres résultats de base se trouvent dans [119], [120] et [121]. Par la suite,
d’autres travaux ont été faits par plusieurs auteurs: Kunisch et Schappacher ([72], [73]),
Grabosch et Moustakas [46] et Parrott [95] et [96]. Pour plus de détails sur ce sujet, nous
référons le lecteur au livre de Hale et Lunel [56] et à celui de Wu [124]. Récemment, dans
[1], [3], [7] et [45], Adimy et Ezzinbi ont repris l’étude de l’équation (0.12) dans le cas où
l’opérateur A est à domaine non dense et véri…e la condition de Hille-Yosida (0.2).
8
moyennant le théorème du point …xe de Horn, l’existence de solutions périodiques lorsqu’il
y a existence de solutions ultimement bornées.
Dans [58]-[60], Henriquez a traité le problème d’existence de solutions pour l’équation (0.1)
et leur régularité. Dans [58] et [59], l’auteur a étudié aussi l’existence de solutions péri-
odiques dans le cas où F est périodique en t.
Dans [103], Ruan et Wu ont montré que sous certaines conditions sur la fonction F et sur
l’espace de phase, l’opérateur solution est une ®-contraction:
Dans [114], Shin et Naito ont étudié le problème d’existence de solutions périodiques dans
le cas où F est linéaire par rapport à la deuxième variable. Moyennant des hypothèses
de compacité, ils ont montré que l’opérateur de Poincaré associé a l’équation est une ®-
contraction. Ce qui a permis grâce à un théorème du point …xe de Chow et Hale [33], de
démontrer que l’opérateur de Poincaré admet un point …xe qui est la donnée initiale d’une
solution périodique.
Dans [91], [92] et [115], Shin, Naito, Minh et Murakami ont donné des propriétés de stabil-
ité pour le semi-groupe solution d’une équation di¤érentielle à retard in…ni en dimension
in…nie. Ils ont aussi caractérisé le générateur in…nitésimal du semi-groupe solution.
Dans [67], Hino, Murakami et Yoshizawa ont traité le problème d’existence de solutions
presque périodiques dans le cas où F est linéaire par rapport à la deuxième variable.
9
croissante ± : [0; r] ! [0; +1) telle que ±(0) = 0 et
¯0 ¯
¯Z ¯
¯ ¯
¯ [d´(µ)] Ã(µ)¯ · ±(s) sup jÃ(µ)j ; Ã 2 C:
¯ ¯
¯ ¯ ¡r·µ·0
¡s
@ @2
[x(»; t) ¡ qx(»; t ¡ h)] = k 2 [x(»; t) ¡ qx(»; t ¡ h)] + f (xt (»; :)) ; » 2 S 1 ; t ¸ 0;
@t @»
(0.15)
où xt (»; µ) = x(»; t + µ), ¡h · µ · 0, t ¸ 0; » 2 S 1 ; k est une constante positive, f est une
fonction continue et 0 · q < 1.
Le livre de Wu [124] contient une analyse détaillée des résultats obtenus dans les articles
[53], [54], [125] et [126]. Puisque H 1 (S 1 ) ½ C(S 1 ); dans [124], l’auteur a également considéré
@2
l’opérateur de Laplace A = k 2 avec domaine C 2 (S 1 ); qui est générateur in…nitésimal d’un
@x
C0 semi-groupe sur E = C(S 1 ): Ce qui a permis d’obtenir des résultats analogues dans un
¡ ¢
cadre plus général : C := C [¡r; 0] ; C(S 1 ) .
Récemment, dans [10], [11] et [12], Adimy et Ezzinbi ont considéré la forme générale
suivante: 8
< d [x(t) ¡ G(t; x )] = A [x(t) ¡ G(t; x )] + F (t; x ); t ¸ 0;
t t t
dt (0.16)
: x = ' 2 C;
0
avec G et F deux fonctions continues sur [0; +1[ £ C ! E. Les auteurs ont développé
des résultats sur l’existence, la régularité et la stabilité dans le cas où l’opérateur A est à
domaine non dense et véri…e la condition de Hille-Yosida (0.2).
Notre contribution à l’étude des équations de type neutre à retard in…ni concerne la
reproduction de quelques résultats concernant l’étude de l’équation (0.16) dans le cas où le
retard est in…ni. Plus précisément, nous nous intéressons à l’équation suivante:
8
< d [x(t) ¡ G(t; x )] = A [x(t) ¡ G(t; x )] + F (t; x ); t ¸ 0;
t t t
dt (0.17)
: x(µ) = '(µ); ¡1 < µ · 0;
10
avec ' 2 B, la fonction G : [0; +1[ £ B ! E est continue et les mêmes notations que
l’équation (0.1).
Signalons que Hernandez et Henriquez [61], [62] et [63] ont étudié l’équation (0.17) dans
le cas où A est générateur d’un semigroupe analytique dans E et G est à valeurs dans le
domaine d’une puissance fractionnaire de ¡A; (D((¡A)¡® ); 0 < ® < 1):
11
Content
0.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
0.2 Historique et modèles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
0.2.1 Modèle de prolifération cellulaire . . . . . . . . . . . . . . . . . . . . 6
0.2.2 Modèle de proie-prédateur avec di¤usion dans l’espace . . . . . . . . 6
0.2.3 Modèle de dynamique d’une population distribuée . . . . . . . . . . 7
0.3 Equations di¤érentielles à retard …ni en dimension in…nie . . . . . . . . . . 7
0.4 Equations di¤érentielles à retard in…ni en dimension in…nie . . . . . . . . . 8
0.5 Equations di¤érentielles à retard de type neutre en dimension in…nie . . . . 9
0.6 Description de la thèse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1
3 Local Existence, Stability and Attractiveness for Some Partial Functional
Di¤erential Equations with In…nite Delay2 45
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.2 Local existence and global continuation of integral solutions . . . . . . . . . 46
3.3 Existence of strict solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.4 The solution semigroup and linearized stability . . . . . . . . . . . . . . . . 54
3.5 Attractiveness of solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.6 An application to a reaction di¤usion equation with in…nite delay . . . . . . 65
2
This chapter is based on two papers. The …rst one is in collaboration with M. Adimy and K. Ezzinbi
and it will appear in Journal of Nonlinear Analysis, Theory, Methods and Applications, (2001). The second
one is in collaboration with K. Ezzinbi and it will appear in Fields Institue Communications Series, (2001).
3
This chapter is based on an ongoing work in collaboration with M. Adimy and K. Ezzinbi.
4
This chapter is based on a paper in collaboration with R. Benkhalti and K. Ezzinbi. The paper will
appear in Journal of Mathematical Analysis and Applications, (2001).
2
PRESENTATION GENERALE
0.1 Introduction
Le travail présenté dans cette thèse se situe dans le cadre des équations di¤érentielles à
retard in…ni en dimension in…nie. Nous nous intéressons plus exactement aux équations de
type: 8
< d x(t) = Ax(t) + F (t; x ); t ¸ 0;
t
dt (0.1)
: x = ' 2 B;
0
où A : D(A) ½ E ! E est un opérateur linéaire fermé sur un espace de Banach (E; j:j) ;
pour tout t ¸ 0; la fonction xt 2 B est dé…nie par:
B est l’espace de phase constitué de fonctions dé…nies de ]¡1; 0] à valeurs dans E; véri…ant
certains axiomes (voir Chapitre 1). F : IR+ £ B ! E est une fonction continue.
Le long de cette thèse, nous supposons que D(A) 6= E et que la résolvante de A véri…e la
condition de Hille-Yosida: il existe ! 2 IR tel que l’ensemble résolvant ½(A) de A contient
]!; +1[ et que
© ° °ª
sup (¸ ¡ !)n °(¸I ¡ A)¡n ° < +1; (0.2)
n¸0; ¸>!
Notons que dans le cas où D(A) = E; le théorème de Hille-Yosida a¢rme que A engendre
un C0 semi-groupe (T (t))t¸0 sur l’espace E: Une solution (faible) de l’équation (0.1) est
donnée par la formule de variation de la constante suivante:
8 Z t
< T (t)'(0) + T (t ¡ s)F (s; xs )ds; pour t ¸ 0;
x(t; ') = 0 (0.3)
:
'(t); pour ¡ 1 < t · 0:
L’étude de l’équation (0.1) a fait l’objet de plusieurs travaux. Nous citons essentiellement
Hale et Kato [55]; Arino, Burton et Haddock [20]; Murakami [84]; Hino, Murakami et Naito
[66] et plus récemment, Henriquez ([58]-[60]); Hino, Murakami et Yoshizawa [67]; Liu et
Xu [77]; Naito, Murakami et Shin [91]; Shin [113]; Shin et Naito [114], et Shin, Naito et
Minh [115]. On cite aussi Milota et Petzeltova [82], [83], [98]-[100] et [101] dans le cas où A
3
engendre un semi groupe analytique dans E et Ruan et Wu [103]; Ruess [104]-[106]; Ruess
et Summers [107], [108]; Ruess, Summers et William [109], et Kartsatos et Parrott [69] dans
des cas plus généraus. La liste est loin d’être exhaustive.
Dans ce travail, nous traitons l’existence de solutions faibles ou intégrales et leurs régu-
larités. Nous utilisons la théorie des semi-groupes intégrés introduite par Arendt et al..
Nous entendons par solution faible ou intégrale toute fonction x : (¡1; T ] ! E; T > 0,
satisfaisant l’équation suivante:
8 Z t
< S 0 (t)'(0) + d S(t ¡ s)F (s; xs )ds; pour t 2 [0; T ] ;
x(t; ') = dt 0 (0.4)
:
'(t); pour ¡ 1 < t · 0:
4
d’équations. Au début, Malthus a présenté le modèle suivant:
d
N(t) = bN(t); (0.5)
dt
où N(t) est le nombre d’individus à l’instant t et b est le taux de fécondité.
On s’aperçoit que N évolue exponentiellement par rapport à t, et par conséquent, ce
modèle ne re‡ète pas l’évolution exacte de l’espèce, d’où la nécessité d’introduire d’autres
modèles plus réalistes. Les modèles à retard ont pour objet de résoudre ce problème. Le
modèle “proie-prédateur” de Volterra est constitué de deux populations l’une prédateur se
nourrissant de l’autre, la proie. Volterra a supposé que la croissance des prédateurs en
contact avec la proie n’est pas instantannée, dûe par exemple à une période de gestation.
Parmi les modèles les plus classiques, on trouve le système à retard suivant:
8
> d
< N1 (t) = (b1 ¡ a1 N2 (t))N1 (t);
dt · Z t ¸ (0.6)
> d
: N2 (t) = ¡b2 + a2 N1 (t) + k(µ ¡ t)N1 (µ)dµ N2 (t);
dt t¡r
où k est une fonction qui décrit la manière dont le gain du prédateur à chaque instant t
dépend de l’abondance de la population des proies N1 dans un intervalle de temps passé
[t ¡ r; t] : On distingue deux cas: modèle avec retard …ni (0 < r < +1) et modèle avec
retard in…ni (r = +1): Le modèle peut s’écrire alors sous la forme:
d
N(t) = F (t; Nt ); t ¸ 0; (0.7)
dt
avec la notation Nt (µ) = N(t + µ) pour chaque ¡1 < µ · 0 si r = +1 ou pour chaque
¡r · µ · 0 si 0 < r < +1: La fonction F : IR+ £ X ! E; où X est un espace vectoriel
approprié de fonctions allant de ]¡1; 0] à un espace de Banach E si r = +1 ou de [¡r; 0]
à E si 0 < r < +1: Dans l’exemple (0.6), la fonction F est dé…ne par
0 1
(b1 ¡ a1 Á2 (0))Á1 (0)
B · Z 0 ¸ C
F (t; Á) = @ A; (0.8)
¡b2 + a2 Á1 (0) + k(µ)Á1 (µ)dµ Á2 (0)
¡r
5
quelques modèles qui ont été obtenus dans la littérature et qui peuvent être transformés
en des équations de type (0.1) avec retard …ni ou in…ni. On se contente de donner trois
exemples de modèles de dynamique de populations structurées. Pour d’autres exemples,
nous renvoyons le lecteur aux livres de Wu [124] et Webb [122].
C’est un modèle de production du sang proposé par Rey et Mackey en 1993 et étudié
par Dyson, Villella-Bressan et Webb en 1995. Ce modèle décrit la production des souches
prolifératives et le précurseur des cellules dans la moelle oseuse:
8
> @ @
< @t u(x; t) + @x (xu(x; t)) = ¹u(®x; t ¡ r)(1 ¡ u(®x; t ¡ r));
>
0 < x < 1; t > 0; (0.9)
>
>
:
u(x; µ) = Á(x; t); 0 · x · 1; ¡ r · µ · 0;
Pour l’étude d’une population proie-prédateur, Cohen, Hagan et Simpson [34] ont proposé
le modèle suivant:
µZ +1 ¶
@
u(x; t) = ®¢u(x; t) + h K(s)u(x; t ¡ s)ds
@t µZ0 +1 ¶ (0.10)
¡m H(s)u(x; t ¡ s)ds ;
0
6
µZ +1 ¶
processus de croissance et de décroissance de la proie et m H(s)u(x; t ¡ s)ds est
0
la consommation de la population de la proie par le prédateur.
qui décrit les dynamiques d’une population distribuée d’une seule espèce. N(x; t) représente
la taille de la population totale à la position x et à l’instant t, les deux constantes d et c sont
positives mesurant, respectivement, le taux de di¤usion
Z 0 et la croissance intrinsèque et G :
+
]¡1; 0] ! IR est une fonction intégrable véri…ant G(µ)dµ = 1: La fonction Á est un
¡1
élément d’un espace vectoriel approprié de fonctions allant de ]¡1; 0] vers E := L2 ([0; ¼]):
Le modèle a été étudié par plusieurs auteurs, à savoir: Ruan et Wu [103], Bonilla et Linan
[27], et Britton [30].
où xt est dé…nie par xt (µ) = x(t + µ), pour ¡r · µ · 0; r > 0; et C est l’espace des fonctions
continues de [¡r; 0] à valeurs dans E muni de la topologie de la convergence uniforme.
Lorsque A est générateur in…nitésimal d’un C0 -semi-groupe (T (t))t¸0 dans E, l’équation
(0.12) a fait l’objet de plusieurs travaux. Dans [118], Travis et Webb ont formulé tous les
aspects d’existence et stabilité pour l’équation (0.12). Dans le cas où F est autonome et
globalement Lipschitzienne, ils ont démontré que la solution:
8 Z t
< T (t)'(0) + T (t ¡ s)F (xs )ds; pour t ¸ 0;
x(t; ') = 0 (0.13)
:
'(t); pour ¡ r · t · 0;
7
dé…nit un semi-groupe non linéaire (U(t))t¸0 sur C. Lorsque le semi-groupe (T (t))t¸0 est
compact, ils ont démontré que le semi-groupe solution U(t) est compact pour t > r. Cette
dernière propriété a permis dans le cas linéaire d’étudier la stabilité en terme d’une équation
caractéristique qui caractérise le spectre ponctuel du générateur in…nitésimal de (U(t))t¸0 :
Cette hypothèse de compacité a permis aussi de formuler le théorème de variété centre
pour l’équation (0.12), à savoir que l’espace C s’écrit comme somme directe de trois parties
stable, centre et instable. Dans [23] et [24], Arino et Sanchez ont traité l’équation (0.12)
lorsque A = 0; et ont donné une formule de variation de la constante. Dans [80] et [81],
Memory a étudié le problème d’existence d’une variété stable et instable pour l’équation
(0.12). Il a aussi donné une formule de variation de la constante. Dans [76], Lin, So et Wu
ont étudié l’existence d’une variété centre pour l’équation (0.12). Dans le cas où l’espace
instable est réduit à f0g, ils ont montré que la variété centre obtenue est exponentiellement
attractive. D’autres résultats de base se trouvent dans [119], [120] et [121]. Par la suite,
d’autres travaux ont été faits par plusieurs auteurs: Kunisch et Schappacher ([72], [73]),
Grabosch et Moustakas [46] et Parrott [95] et [96]. Pour plus de détails sur ce sujet, nous
référons le lecteur au livre de Hale et Lunel [56] et à celui de Wu [124]. Récemment, dans
[1], [3], [7] et [45], Adimy et Ezzinbi ont repris l’étude de l’équation (0.12) dans le cas où
l’opérateur A est à domaine non dense et véri…e la condition de Hille-Yosida (0.2).
8
moyennant le théorème du point …xe de Horn, l’existence de solutions périodiques lorsqu’il
y a existence de solutions ultimement bornées.
Dans [58]-[60], Henriquez a traité le problème d’existence de solutions pour l’équation (0.1)
et leur régularité. Dans [58] et [59], l’auteur a étudié aussi l’existence de solutions péri-
odiques dans le cas où F est périodique en t.
Dans [103], Ruan et Wu ont montré que sous certaines conditions sur la fonction F et sur
l’espace de phase, l’opérateur solution est une ®-contraction:
Dans [114], Shin et Naito ont étudié le problème d’existence de solutions périodiques dans
le cas où F est linéaire par rapport à la deuxième variable. Moyennant des hypothèses
de compacité, ils ont montré que l’opérateur de Poincaré associé a l’équation est une ®-
contraction. Ce qui a permis grâce à un théorème du point …xe de Chow et Hale [33], de
démontrer que l’opérateur de Poincaré admet un point …xe qui est la donnée initiale d’une
solution périodique.
Dans [91], [92] et [115], Shin, Naito, Minh et Murakami ont donné des propriétés de stabil-
ité pour le semi-groupe solution d’une équation di¤érentielle à retard in…ni en dimension
in…nie. Ils ont aussi caractérisé le générateur in…nitésimal du semi-groupe solution.
Dans [67], Hino, Murakami et Yoshizawa ont traité le problème d’existence de solutions
presque périodiques dans le cas où F est linéaire par rapport à la deuxième variable.
9
croissante ± : [0; r] ! [0; +1) telle que ±(0) = 0 et
¯0 ¯
¯Z ¯
¯ ¯
¯ [d´(µ)] Ã(µ)¯ · ±(s) sup jÃ(µ)j ; Ã 2 C:
¯ ¯
¯ ¯ ¡r·µ·0
¡s
@ @2
[x(»; t) ¡ qx(»; t ¡ h)] = k 2 [x(»; t) ¡ qx(»; t ¡ h)] + f (xt (»; :)) ; » 2 S 1 ; t ¸ 0;
@t @»
(0.15)
où xt (»; µ) = x(»; t + µ), ¡h · µ · 0, t ¸ 0; » 2 S 1 ; k est une constante positive, f est une
fonction continue et 0 · q < 1.
Le livre de Wu [124] contient une analyse détaillée des résultats obtenus dans les articles
[53], [54], [125] et [126]. Puisque H 1 (S 1 ) ½ C(S 1 ); dans [124], l’auteur a également considéré
@2
l’opérateur de Laplace A = k 2 avec domaine C 2 (S 1 ); qui est générateur in…nitésimal d’un
@x
C0 semi-groupe sur E = C(S 1 ): Ce qui a permis d’obtenir des résultats analogues dans un
¡ ¢
cadre plus général : C := C [¡r; 0] ; C(S 1 ) .
Récemment, dans [10], [11] et [12], Adimy et Ezzinbi ont considéré la forme générale
suivante: 8
< d [x(t) ¡ G(t; x )] = A [x(t) ¡ G(t; x )] + F (t; x ); t ¸ 0;
t t t
dt (0.16)
: x = ' 2 C;
0
avec G et F deux fonctions continues sur [0; +1[ £ C ! E. Les auteurs ont développé
des résultats sur l’existence, la régularité et la stabilité dans le cas où l’opérateur A est à
domaine non dense et véri…e la condition de Hille-Yosida (0.2).
Notre contribution à l’étude des équations de type neutre à retard in…ni concerne la
reproduction de quelques résultats concernant l’étude de l’équation (0.16) dans le cas où le
retard est in…ni. Plus précisément, nous nous intéressons à l’équation suivante:
8
< d [x(t) ¡ G(t; x )] = A [x(t) ¡ G(t; x )] + F (t; x ); t ¸ 0;
t t t
dt (0.17)
: x(µ) = '(µ); ¡1 < µ · 0;
10
avec ' 2 B, la fonction G : [0; +1[ £ B ! E est continue et les mêmes notations que
l’équation (0.1).
Signalons que Hernandez et Henriquez [61], [62] et [63] ont étudié l’équation (0.17) dans
le cas où A est générateur d’un semigroupe analytique dans E et G est à valeurs dans le
domaine d’une puissance fractionnaire de ¡A; (D((¡A)¡® ); 0 < ® < 1):
11
Chapter 1
This chapter serves as an introduction to the other chapters. It contains the essential
background materials required throughout this thesis. Its organization is as follows. In
Section 1.1, we discuss the axiomatic phase space to functional di¤erential equations with
in…nite delay. In Section 1.2, we collect some useful results on integrated semigroups’ theory
and di¤erential operators with nondense domain. We only state results and leave the details
to references.
We use (E; j:j) or simply E; to denote a Banach space with norm j:j.
12
have been published by many authors until 1978 when J. K. Hale and J. Kato organized
the study of functional di¤erential equations with in…nite delay in [55]. Which leads to
many investigations on this theory. Our results are obtained as far as possible on the phase
space introduced by those latter. That is, a seminormed space satisfying suitable axioms
considered by Kappel and Schappacher [68], and Schumacher [111]. The book by Y. Hino
and al. [66] contains an exhaustive bibliography and a detailed discussion on this subject.
Following this book, we will assume that the phase space B is a linear space of functions
mapping (¡1; 0] into (E; j:j), endowed with a seminorm k:kB and satisfying the following
fundamental axioms :
(A) There exist a positive constant H and functions K(¢); M(¢) : [0; +1) ! [0; +1),
with K continuous and M locally bounded, such that for any ¾ 2 IR and a > 0, if
x : (¡1; ¾ + a] ! E, x¾ 2 B, and x(¢) is continuous on [¾; ¾ + a]; then for all t in
[¾; ¾ + a] the following conditions hold :
(i) xt 2 B,
(ii) kx (t)k · H kxt kB ;
(iii) kxt kB · K(t ¡ ¾) sup kx (s)k + M (t ¡ ¾) kx¾ kB .
¾·s·t
(A1) For the function x(¢) in (A), t 7! xt is a B-valued continuous function for t in
[¾; ¾ + a].
Remark 1.1.2 n[66] Axiomo (B) is equivalent to saying that the space of equivalence classes
b : Á 2 B is a Banach space.
Bb := B = k:k = Á
B
Let C00 be the set of continuous functions Á : (¡1; 0] ! E with compact support
supp(Á).
13
Theorem 1.1.1 [66] Any Á 2 C00 belongs to B . If supp(Á) is contained in [¡r; ¡s] ;
0 · s · r < 1; then there exists a constant ±(r; s) such that
Let us give some examples of concrete functional spaces that veri…es Axioms (A), (A1)
and (B).
Example 1.1.1 For any continuous function g : (¡1; 0] ! (0; +1), let
jÁ(µ)j
Cg0 := fÁ 2 C((¡1; 0] ; E) : lim = 0g;
µ!¡1 g(µ)
jÁ(µ)j
kÁkg := sup :
¡1<µ·0 g(µ)
³ ´
It was proved in Theorems 1.3.2 and 1.3.6 of [66] that if g is nonincreasing, then Cg0 ; k¢kg
satis…es Axioms (A), (A1) and (B).
Example 1.1.2 Consider the space Cr £L1 (g) of all functions Á : (¡1; 0] ! E such that Á
is continuous on [¡r; 0] ; for some r ¸ 0; Lebesgue measurable and g(:) jÁ(:)j is Lebesgue in-
tegrable on (¡1; ¡r) ; where g : (¡1; ¡r) ! IR is a positive Lebesgue measurable function.
A seminorm in Cr £ L1 (g) is de…ned by
Z
¡r
14
To study the existence of periodic solutions in Chapter 5, we need to introduce an
additional axiom which is concerned with the realization of some elements in B. Recall
that Ã
a sequence of functions
! (Án )n2IN : (¡1; 0] ! E; is said uniformly bounded if
sup sup jÁn (µ)j < +1: We say that a sequence of functions (Án )n¸0 2 B converges
n2IN ¡1<µ·0
compactly on (¡1; 0] to Á if the sequence converges uniformly on compact subsets of
(¡1; 0] :
(C) If a uniformly bounded sequence (Án )n in C00 converges to a function Á compactly on
(¡1; 0] ; then Á is in B and kÁn ¡ ÁkB ! 0 as n ! +1:
Section 1.3 of the book [66] contains an exhaustive discussion on the standard spaces
that satisfy most axioms used in this thesis. For example, it is known in Theorem 1.3.2
of this book that the space Cg0 introduced in Example 1.1.1 satis…es Axiom (C) if g is
nonincreasing and g(s) ! 1 as s ! ¡1:
More recently, Ruess and Summers in [107] have considered another axiom that we use
in this thesis to deal with existence of strict solutions and characterization of the generator
of the linear solution semigroup.
(D1) For a sequence (Án )n¸0 in B, if kÁn kB ! 0 then jÁn (s)j ! 0 for each s 2 (¡1; 0] :
Example 1.1.3 As an example, we can verify without di¢culties that the above axiom is
satis…ed by the space Cg0 in the special case g(µ) = e¡°µ ; ° > 0 :
½ ¾
C°0 = Á 2 C((¡1; 0] ; E) : lim e°µ Á(µ) = 0 ; ° > 0;
µ!¡1
with the norm kÁk° =sup e°µ jÁ(µ)j ; Á 2 C°0 : It is satis…ed, in general, by the space
µ·0
½ ¾
C° = Á 2 C((¡1; 0] ; E) : lim e°µ Á(µ) exists in E ; ° > 0:
µ!¡1
B is called a fading memory space if it satis…es the extra axiom (C) and
15
(D2) kW0 (t)ÁkB ! 0 as t ! +1 for all Á 2 B0 :
Let BC be the space of bounded continuous functions mapping (¡1; 0] into E; provided
with the uniform norm topology, then one has.
Proposition 1.1.2 [66] Assuming that B is a fading memory space, then BC ½ B and
there exists a positive constant J such that kÁkB · J kÁkBC : Moreover
As a consequence of this proposition, the functions K(:) and M(:) can be chosen as
K(t) = J and M(t) = (1 + JH) kW0 (t)k : Note that (D2) implies sup kW0 (t)ÁkB < +1
t¸0
by the Banach-Steinhaus theorem. Therefore , whenever B is a fading memory space, the
functions K(:) and M(:) can be assumed bounded on [0; +1) :
B is called a uniform fading memory space if it satis…es the extra axioms (C) and
(D3) kW0 (t)k ! 0 as t ! +1:
One can see that Axiom (D3) holds for B = Cg0 if and only if
½ ¾
g(s + t)
sup : s · ¡t ! 0 as t ! 1: Therefore, C°0 is a uniform fading memory space.
g(s)
On the other hand, if g(s) = 1 + jsjk for some k > 0, then the space Cg0 is a fading memory
space, but not a uniform fading memory space.
We will use Axioms (C), (D2) and (D3) in studying existence of periodic solutions
(Chapter 5). One can see Chapter 7 of [66] for an earlier use in discussing stability and
existence of periodic solutions or almost periodic solutions to FDEs with in…nite delay.
(C1) If (Án )n¸0 is a Cauchy sequence in B with respect to the seminorm and if (Án )n¸0
converges compactly to Á on (¡1; 0], then Á is in B and kÁn ¡ ÁkB ! 0:
The above axiom is also concerned by the realization of elements in B. It is sometimes
considered in stability investigation additionally to the following axiom :
(C2) There exists a constant ° 0 such that the function ("¸ - u) : (¡1; 0] ! E de…ned by
e¸µ u for µ · 0, belongs to B for Re(¸) > ° 0 and u 2 E, and that
is …nite for each ¸ with Re(¸) > ° 0 ; and bounded for Re(¸) > ° 1 for some ° 1 ¸ ° 0 :
One of the important consequence of Axiom (C1) is that it allows computing the integral
in B from the integral in E ( see Lemma 2.2.6 in Chapter 2 or [91]). Here, we use Axiom
16
(C1) in studying existence of strict solutions.
Some applications of Axioms (C); (C1) and (C2) can be found in [56] and [66] and most
recently [91], [115], [92] and [114] where they were essentially used in studying stability and
periodicity of solutions to FDEs with in…nite delay.
17
restrictions, the simplest of which is that f takes its values in D(A). It is the integrated
semigroups theory that allows the range of the operator f to be a subset of E not necessarily
contained in D(A).
(Once) integrated semigroups are motivated by formally de…ning
Zt
S(t) = T (s)ds;
0
De…nition 1.2.1 [18] A family (S(t))t¸0 ½ L(E) is called an integrated semigroup if the
following conditions are satis…ed :
(i) S(0) = 0;
(ii) for any x 2 E; S(t)x is a continuous
Z s function of t ¸ 0 with values in E;
(iii) for any t; s ¸ 0 S(s)S(t) = (S(t + ¿ ) ¡ S(¿ ))d¿ :
0
It’s well known from [97] that if R(¸) is injective, then there exists a unique operator A
satisfying (!; +1) ½ ½(A) ( the resolvent set of A) such that
18
De…nition 1.2.3 [18] A linear operator A : D(A) ½ E ! E is called a generator of an
integrated semigroup, if there exists ! 2 IR such that (!; +1) ½ ½(A), and there exists a
strongly continuous exponentially bounded family (S(t))t¸0 of linear bounded operators such
Z +1
¡1
that S(0) = 0 and (¸I ¡ A) = ¸ e¡¸t S(t)dt for all ¸ > !:
0
Proposition 1.2.1 [18] Let A be the generator of an integrated semigroup (S(t))t¸0 : Then
for all x 2 E and t ¸ 0;
0t 1
Zt Z
S(s)xds 2 D(A) and S(t)x = A @ S(s)xdsA + tx:
0 0
and
Zt
S(t)x = S(s)Ax ds + tx:
0
Corollary 1.2.2 [18] Let A be the generator of an integrated semigroup (S(t))t¸0 : Then
for all x 2 E and t ¸ 0 one has S(t)x 2 D(A).
Moreover, let x 2 E. Then S(¢)x is right-sided di¤erentiable in t ¸ 0 if and only if
S(t)x 2 D(A). In that case
S 0 (t)x = AS(t)x + x:
Proposition 1.2.3 [64] Let A : D(A) ½ E ! E be a linear operator and (S(t))t¸0 ½ L(E)
an exponentially
Z t bounded family. The following
µZ t assertions
¶ are equivalent
(i) S(s)xds 2 D(A) and S(t)x = A S(s)xds + tx, ( t ¸ 0; x 2 E);
0 0
(ii) (S(t))t¸0 is an integrated semigroup on E generated by A.
An important special case is when the integrated semigroup is locally Lipschitz contin-
uous (with respect to time).
19
De…nition 1.2.4 [71] An integrated semigroup (S(t))t¸0 is called locally Lipschitz contin-
uous, if for all a > 0 there exists a constant l(a) > 0 such that
De…nition 1.2.5 [71] We say that a linear operator A : D(A) ½ E ! E satis…es the
¹ ¸ 0 and ! 2 IR such that (!; +1) ½ ½(A)
Hille-Yosida condition (HY ) if there exist M
and
¹
sup f(¸ ¡ !)n kR(¸; A)n k ; n 2 IN; ¸ > !g · M: (HY )
The following theorem shows that the Hille-Yosida condition (HY ) characterizes gen-
erators of locally Lipschitz continuous integrated semigroups.
Theorem 1.2.6 Let A be a Hille-Yosida operator and (S(t))t¸0 be the locally Lipschitz
continuous integrated semigroup generated by A: It is well known from [117], that the deriv-
ative (S 0 (t))t¸0 on D (A) is a strongly continuous semigroup generated by the part A0 of
the operator A in D(A), which is de…ned by
8 n o
< D(A ) = x 2 D(A) : Ax 2 D (A) ;
0
: A0 x = Ax; for x 2 D(A0 ):
20
Throughout this text, a linear operator that satis…es the Hille-Yosida condition (HY )
without being necessarily densely de…ned is called a Hille-Yosida operator.
Next, we give some results for the existence of solutions to the following Cauchy problem
8
< d u(t) = Au(t) + f(t); t ¸ 0;
dt (1.2)
: u(0) = x 2 E;
De…nition 1.2.6 [40] We say that a function u : [0; a] ! E; a > 0, is a strict solution of
Equation (1.2) in [0; a] if the following conditions hold :
(i) u 2 C 1 ([0; a] ; E) \ C([0; a] ; D(A));
(ii) u satis…es Equation (1.2) on [0; a].
In the case where x is not su¢ciently regular (that is, x is just in D(A)) there may
not exist a strict solution u(t) 2 E but, following the work of Da Prato and Sinestrari [40],
Equation (1.2) may still have a so-called integral solution. This motivates the following
de…nition :
De…nition 1.2.7 [40] Given f 2 L1loc (0; +1; E) and x 2 E, we say that
u : [0; +1) ! E is an integral solution of Equation (1.2) if the following assertions are
true
(i) Zu 2 C([0; +1) ; E),
t
(ii) u(s)ds 2 D(A); for t ¸ 0;
0 µZ t ¶ Z t
(iii) u(t) = x + A u(s)ds + f(s)ds; for t ¸ 0:
0 0
21
From this de…nition, we deduce that for an integral solution u, we have u(t) 2 D(A);
Z Z t+h
1 t+h
for all t > 0, because u(t) = lim u(s)ds and u(s)ds 2 D(A). In particular,
h!0 h t t
x 2 D(A) is a necessary condition for the existence of an integral solution to Equation
(1.2).
Theorem 1.2.8 ([32], [40]) Suppose that A is a Hille-Yosida operator, x 2 D(A) and
f : [0; +1) ! E is a continuous function. Then the problem (1.2) has a unique integral
solution which is given by the following variation of constants formula
0t 1
Z
d
u(t) = S 0 (t)x + @ S(t ¡ s)f(s)dsA ; for t ¸ 0; (1.4)
dt
0
Proposition 1.2.9 ([1], [71] and [116]) Let A : D(A) µ E ! E be a Hille-Yosida operator,
(S(t))t¸0 be the integrated semigroup generated by A and G : [0; a] ! E ; a > 0, be a
Bochner-integrable function. Then, the function B : [0; a] ! E de…ned by
Zt
B(t) = S(t ¡ s)G(s) ds
0
is continuously di¤erentiable
Z t on [0; a] and satis…es for t 2 [0; a] ;
1
(i) B 0 (t) =lim S 0 (t ¡ s)S(h)G(s) ds;
h&0 h 0
(ii) B(t) 2 D(A); Z t
(iii) B 0 (t) = AB(t) + G(s) ds:
Z t 0
22
Let us end with some classical examples of Hille-Yosida operators with nondense domain
( for more details, see [40]).
1)
8
>
< E = C([0;
>
©
a] ; IR);
ª
D(A) = u 2 C 1 ([0; a] ; IR); u(0) = 0 ;
>
>
: Au = ¡u0 :
2)
8
> ®
< E = fu 2© C ([0; a] ; IR); u(0) = 0g ; 0 < ® < 1;ª
>
D(A) = u 2 C 1+® ([0; a] ; IR); u(0) = u0 (0) = 0 ;
>
>
: Au = ¡u0 ;
where
½ ¾
® ju(t) ¡ u(s)j
C ([0; a] ; IR) = u : [0; a] ! IR; [u]C ® ([0;a];IR) = sup <1 ;
0·t·s·a jt ¡ sj®
and
© ª
C 1+® ([0; a] ; IR) = u : [0; a] ! IR; u0 2 C ® ([0; a] ; IR) ;
with
kukC ® ([0;a];IR) = kukC([0;a];IR) + [u]C ® ([0;a];IR) :
3)
8
>
< E = C([0;
>
©
a] ; IR);
ª
D(A) = u 2 C 2 ([0; a] ; IR); u(0) = u(a) = 0 ;
>
>
: Au = u00 :
23
8
>
> E = W01;p (0; a; C(-; IR)); 1 · p < 1;
>
> ( )
< u 2 C([0; a] ; D(¢)) \ C 1 ([0; a] ; C 1 (-; IR));
D(A) = ;
>
>
> u(0) = 0 and ¢u ¡ u0 2 W01;p (0; a; C(-; IR))
>
:
Au = ¢u ¡ u0 :
where
½ Z t ¾
W01;p (0; a; E) = u : [0; a] ! E; u(t) = u0 (s)ds and u0 2 Lp (0; a; E) :
0
6) Let
8
>
> E = W01;p (0; a; C(-; IR));
>
> ( )
< u 2 C([0; a] ; D(¢)) \ C 1 ([0; a] ; C(-; IR));
D(A) = ;
>
>
> u(0) = 0 and ¢u ¡ u0 2 W01;p (0; a; C(-; IR))
>
:
Au = ¢u ¡ u0 :
24
Chapter 2
2.1 Introduction
Let (E; j:j) be a Banach space. In all this thesis, B is a linear space of functions mapping
(¡1; 0] into E, endowed with a seminorm k:kB and satisfying the fundamental axioms (A);
(A1) and (B) introduced in the previous chapter. Since, in Axiom (A) the functions K(:)
and M(:) are, respectively, continuous and locally bounded, we will use the notations Ka
and Ma to denote, respectively, max K(t) and sup M(t) for a …xed constant a > 0.
0·t·a 0·t·a
The present chapter is devoted to some existence results to the following class of partial
functional di¤erential equations (PFDEs) with in…nite delay
8
< d x(t) = Ax(t) + F (t; x ); t ¸ 0;
t
dt (2.1)
: x = ';
0
25
b > 0; and t 2 [0; b] ; xt represents the mapping de…ned from (¡1; 0] into E by
Surely there are works dealing with existence of solutions to Equation (2.1) with dense
domain D(A) ( cf., for instance, [56], [66], [124], [91] and [58]-[60]). However, most existing
results concern the case where A = 0 and the extension to the case of non-dense domain
looks quite natural. We state that we can solve Equation (2.1) without assuming necessarily
that A is densely de…ned. Some of our results are an extension of Henriquez’s ones ([58]-
[60]) to the situation where A satis…es the resolvent estimates of the well-known Hille and
Yosida theorem while the domain of A is not dense in E. Techniques employed in [1], [7]
and [45], have been generalized to the study of Equation (2.1). That is, under di¤erent
appropriate assumptions, existence and uniqueness of solutions (called integral solutions)
are …rst proved, using a …xed point argument. Then, the integral solutions are shown to be
strict solutions under more restrictive assumptions.
The main tool in the approach followed in our proves is the theory of integrated semi-
groups introduced in Section 1.2 in Chapter 1.
Theorem 2.2.1 [60] Assuming that A is the generator of a strongly continuous semigroup
(T (t))t¸0 on E, and F : [0; a]£B ! E is a continuous mapping in t and uniformly Lipschitz
continuous on B. Then, for given ' 2 B; the Cauchy problem (2.1) has exactly one mild
26
solution which is given by
8 Z t
< T (t)'(0) + T (t ¡ s)F (s; xs )ds; 0 · t · a;
x(t) = 0
:
'(t); ¡1 < t · 0:
De…nition 2.2.2 Let ' 2 B. We say that a function x : (¡1; a] ! E is a strict solution
of Equation (2.1) in (¡1; a] if the following conditions hold :
(i) x 2 C 1 ([0; a] ; E) \ C([0; a] ; D(A));
(ii) x satis…es Equation (2.1) on (¡1; a].
From the closedness property of the operator A, we can prove the following.
Proposition 2.2.2 (i) If x is an integral solution of Equation (2.1) in (¡1; a] ; then for
all t 2 [0; a] ; x(t) 2 D(A): In particular '(0) 2 D(A):
(ii) If x is an integral solution of Equation (2.1) in (¡1; a] ; such that x : [0; a] ! E
belongs to C 1 ([0; a] ; E) or C([0; a] ; D(A)); then x is also a strict solution of Equation (2.1)
in (¡1; a].
Z
t+h Z t+h
1 1 1
A x(s)ds = (x(t + h) ¡ x(t)) ¡ F (s; xs )ds:
h h h t
t
27
Since F is continuous, then the right side of the above equality tends to
d
x(t) ¡ F (t; xt ), as h tends to 0+ ; and
dt
Z
t+h
1
x(s)ds tends to x(t); as h tends to 0+ :
h
t
From the closedness of A; we get x(t) 2 D(A) and
d
Ax(t) = x(t) ¡ F (t; xt ):
dt
From what we deduce that x is a strict solution.
On the other hand, if x belongs to C([0; a] ; D(A)): Again from De…nition 2.2.1, for all
t 2 [0; a] and h > 0;
Z
t+h Z t+h
1 1 1
(x(t + h) ¡ x(t)) = Ax(s)ds + F (s; xs )ds:
h h h t
t
Since Ax(:) and F are continuous, the right side of the above equality tends to
28
Theorem 2.2.3 Let F : [0; a] £ - ! E be a continuous mapping and suppose that the
conditions (H2:1) and (H2:2) hold. If ' 2 - with '(0) 2 D(A); then Equation (2.1) has
at least one integral solution x : (¡1; b] ! E; for some b 2 (0; a] : Moreover,
8 Z t
< S 0 (t)'(0) + d S(t ¡ s)F (s; xs )ds; 0 · t · b;
x(t) = dt 0
:
'(t); ¡1 < t · 0:
Proof. The main idea of the proof is to use the Schauder …xed point theorem.
Let ' 2 - with '(0) 2 D(A): First, there exist constants r1 > 0 and ¹ ¸ 0 such that
Br1 (') = fà 2 B : kà ¡ 'kB · r1 g µ - and jF (s; Ã)j · ¹ for all s 2 [0; r1 ] and à 2
Br1 ('): Consider the function y : (¡1; +1) ! E de…ned by
(
S 0 (t)'(0); for t ¸ 0;
y(t) =
'(t); for t · 0:
By virtue of Axioms (A ¡ i) and (A1), yt 2 B and for r2 2 ]0; r1 [, there exists b1 2 ]0; r1 ]
such that kyt ¡ 'kB · r2 for all t 2 [0; b1 ] :
Set M a := sup kS 0 (s)kD(A) and let b be a constant such that
0·s·a
½ ¾
r1 ¡ r2
0 < b · min b1 ; :
¹
M M a Ka ¹
We introduce the space
29
For any u 2 Fb ('); we have
letting n to +1, yields ku0 ¡ 'kB = 0: In addition, Axiom (A ¡ iii) implies that for any
t 2 [0; b] ;
Then, for any t 2 [0; b] ; unt ! ut in B: From this together with the following inequality
T maps elements of Fb (') into Fb ('): In fact, by (H2:2) and Axiom (A1), for every
u 2 Fb ('); the mapping s 7! F (s; us ) is continuous on [0; b] : So, for every u 2 Fb ('); the
30
Z t
mapping t 7! S(t ¡ s)F (s; us )ds is continuously di¤erentiable on [0; b] : From this, the
0
mapping v := T u is continuous on [0; b] : Then, v 2 Fb : To prove that v 2 Fb ('), we put
w = v ¡ y: Then, we get for any t 2 [0; b] ;
31
Z t
d
¸R(¸; A) S(t ¡ s)F (s; us )ds
dt 0
Z t
= S 0 (t ¡ s)¸R(¸; A)F (s; us )ds;
0
Z t¡"
= S 0 (") S 0 (t ¡ s ¡ ")¸R(¸; A)F (s; us )ds
0
Z t
+ S 0 (t ¡ s)¸R(¸; A)F (s; us )ds:
t¡"
Since S 0 (") is compact, there exists a compact set W" such that
½ Z t¡" ¾
0 0
S (") S (t ¡ s ¡ ")¸R(¸; A)F (s; us )ds : u 2 Fb ('); ¸ > ! µ W" :
0
Furthermore,
¯Z t ¯
¯ ¯ ¸ ¹
¯ S (t ¡ s)¸R(¸; A)F (s; us ) ds¯¯ ·
0
M M a ¹"; for all u 2 Fb ('):
¯ ¸¡!
t¡"
This implies, by letting ¸ to +1; that f(T u)(t) ¡ S 0 (t)'(0); u 2 Fb (')g is totally bounded,
that is uniformly bounded for u 2 Fb ('); and therefore Range(T )(t) is relatively compact.
On the other hand, for every 0 · t0 < t · b and ¸ > !; one has
This leads to
32
Letting ¸ to +1 and using the equicontinuity of (S 0 (¢)x)x2W , we obtain
Using similar argument for 0 · t < t0 · b, we can conclude that Range(T )(t) is equicon-
tinuous.
Finally, the Schauder …xed point theorem implies that T has a …xed point x in Fb ('): The
fact that x is an integral solution of Equation (2.1) is a consequence of the same property
in Theorem 1.2.8. This ends the proof of Theorem 2.2.3.
In order to de…ne the integral solution to Equation (2.1) in its maximal interval of
existence, we add the following condition :
(H2:3) F is continuous from [0; +1) £ B into E and takes bounded sets of [0; +1) £ B
into bounded sets of E.
Theorem 2.2.4 Assume that the conditions (H2:1); (H2:2) and (H2:3) hold and let ' 2
B such that '(0) 2 D(A): Then, Equation (2.1) has at least one integral solution x(:; ') in
a maximal interval (¡1; b' ) : Moreover,
i) b' = +1 or
ii) b' < +1 and lim sup jx(t; ')j = +1:
t!b¡
'
Proof. Let x(:; ') be an integral solution of Equation (2.1) in (¡1; b1 ] : We know that
x(t) 2 D(A); for all t 2 [0; b1 ] : Repeating the procedure used in the local existence result.
This yields existence of b2 > b1 and a function x(:; xb1 (:; ')) : (¡1; b2 ] ! E which satis…es
for t 2 [b1 ; b2 ]
Zt
0 d
x(t; xb1 (:; ')) = S (t ¡ b1 )x(b1 ; ') + S(t ¡ s)F (s; xs (:; xb1 (:; ')))ds:
dt
b1
Proceeding inductively, we obtain the maximal interval of existence (¡1; b' ) of the solution
x(:; '). Assume that b' < 1 and lim sup jx(t; ')j is bounded. Then, from (H2:3)
t!b'
there exists a constant ¹ > 0 such that jF (t; xt (:; ')j · ¹ for all t 2 [0; b' ) : As before,
let M b' = sup kS 0 (s)kD(A) and x : [t0 ; b' ) ! E; t0 2 (0; b' ) ; be the restriction of
0·s·b'
x(:; ') to [t0 ; b' ) : Consider t0 · t · t0 < b' and 0 < " < t0 . Let ´1 > 0 be such that
j(S 0 (t0 ) ¡ S 0 (t)) '(0)j · " for jt ¡ t0 j < ´1 : For ¸ > !; one has
33
j¸R(¸; A) (x(t0 ) ¡ x(t))j · j¸R(¸; A) (S 0 (t0 ) ¡ S 0 (t)) '(0)j
Z t
+ j(S 0 (t0 ¡ s) ¡ S 0 (t ¡ s)) ¸R(¸; A)F (s; xs )j ds
t¡"
Z t0
+ S 0 (t0 ¡ s)¸R(¸; A)F (s; xs )ds
t
Z t¡"
+ j(S 0 (t0 ¡ " ¡ s) ¡ S 0 (t ¡ " ¡ s)) S 0 (")¸R(¸; A)F (s; xs )j ds:
0
Moreover, S 0 (") f¸R(¸; A)F (s; xs ) : ¸ > !g is included in a compact set W . Since, the
family (S 0 (¢)x)x2W is uniformly equicontinuous in [0; b' ¡ "] ; there exists ´ 2 > 0 such that
¯¡ 0 0 ¢ ¯
¯ S (t ) ¡ S 0 (t) u¯ < "; for u 2 W;
Therefore, lim x(t; ') exists. If we de…ne x(b' ; ') := lim x(t; '); we can extend x(:; ')
t!b' t!b'
beyond b' and we contradict the maximality of (¡1; b' ). This …nishes the proof.
34
Theorem 2.2.5 Assume that the conditions (H2:1) and (H2:4) are satis…ed. Then, for
' 2 B such that '(0) 2 D(A), Equation (2.1) has a unique global integral solution x which
is given by 8 Z t
< S 0 (t)'(0) + d S(t ¡ s)F (s; xs )ds; t ¸ 0;
x(t) = dt 0 (2.5)
:
'(t); ¡1 < t · 0:
Moreover, x(t; ') is a continuous function of ', in the sense that if ' 2 B such that '(0)
2 D(A) and t ¸ 0, then there exists a positive constant ¯ such that, for à 2 B and Ã(0)
2 D(A), we have
jx(t; ') ¡ x(t; Ã)j · ¯ k' ¡ ÃkB :
Proof. Let a > 0 be …xed and ' 2 B such that '(0) 2 D(A). Consider the set
Za (') := fy 2 C([0; a] ; E) : y(0) = '(0)g and z 2 Za (').
Let z~ : (¡1; a] ! E be the mapping de…ned by
(
z(t); t 2 [0; a] ;
z~(t) =
'(t); ¡1 < t · 0:
By virtue of condition (H2:4) and AxiomZ (A1) the mapping s 7! F (s; z~s ) is continuous on
t
[0; a]. This implies that the mapping t 7! S(t¡s)F (s; z~s ) ds is continuously di¤erentiable
0
on [0; a].
Consider the operator P : Za (') ¡! Za (') de…ned by
Z
0 d t
(P z)(t) := S (t)'(0) + S(t ¡ s)F (s; z~s )ds:
dt 0
Thanks to Proposition 1.2.5, we can assume without loss of generality that ! > 0. By
virtue of Proposition 1.2.9, we have for every y; z 2 Za ('); t 2 [0; a] and ¸ > !;
Z t
¸R(¸; A) ((P y)(t) ¡ (P z)(t)) = S 0 (t ¡ s)¸R(¸; A) (F (s; y~s ) ¡ F (s; z~s )) ds:
0
Then,
Zt
¸ ¹ 2 !a
j¸R(¸; A) ((P y)(t) ¡ (P z)(t))j · M e jF (s; y~s ) ¡ F (s; z~s )j ds:
¸¡!
0
35
This yields
Z t
¹ 2 e!a
j(P y)(t) ¡ (P z)(t)j · LM K(s) sup jy(») ¡ z(»)j ds;
0 0·»·s
Z t
¹ 2 Ka e!a
· LM sup jy(») ¡ z(»)j ds:
0 0·»·s
¹ !a k' ¡ Ãk
· H Me B
Z t
+LM¹ 2 Ka e!a ¹ 2 Ma e!a k' ¡ Ãk :
max jx(») ¡ y(»)j ds + aLM B
0 0·»·s
36
By Gronwall’s lemma, it follows that
where
¹ !a (H + aLMM
¯ = Me ¹ a ) exp(aLM
¹ 2 Ka e!a ):
Hence the continuity of x(t; ') in term of ' and the uniqueness of the solution of Equation
(2.5) are guaranteed. This completes the proof .
Under more restrictive conditions, we obtain existence of strict solutions to Equation (2.1).
In order to compute the integral in B from the integral in E, we suppose that B is normed
and satis…es one of the following two extra axioms.
(C1) If (Án )n¸0 is a Cauchy sequence in B with respect to the seminorm and if (Án )n¸0
converges compactly to Á on (¡1; 0], then Á is in B and kÁn ¡ ÁkB ! 0; as n ! 1:
(D1) For a sequence ('n )n¸0 in B, if k'n kB ! 0; as n ! 1; then j'n (µ)j ! 0; as n ! 1;
for each µ 2 (¡1; 0] :
Lemma 2.2.6 [92] Let B be a normed space which satis…es Axiom (C1) and f : [0; a] ! B;
a > 0; be a continuous function such that f(t)(µ) is continuous for (t; µ) 2 [0; a] £ (¡1; 0].
·Z a ¸ Z a
Then, f(t)dt (µ) = f(t)(µ)dt; µ 2 (¡1; 0] :
0 0
Lemma 2.2.7 Let B satis…es Axiom (D1) and f : [0; a] ! B be a continuous function.
Then for all µ 2 (¡1; 0] ; the function f(:)(µ) is continuous and
·Z a ¸ Z a
f(t)dt (µ) = f(t)(µ)dt; µ 2 (¡1; 0] :
0 0
Z a
a n ka
Proof. We have f(t)dt = lim § f( ) in B: By use of Axiom (D1), we get
0 n!+1 n k=1 n
·Z a ¸
a n kn
f (t)dt (µ) = lim § f( )(µ); µ 2 (¡1; 0] :
0 n!+1 n k=1 a
On the other hand, the same axiom implies that the function f(:)(µ) is continuous on [0; a].
From what we infer that for all µ 2 (¡1; 0] ; the function f(:)(µ) is integrable on [0; a] and
Z a
a n kn
f(t)(µ)dt = lim § f( )(µ); µ 2 (¡1; 0] :
0 n!+1 n k=1 a
37
This ends the proof of the lemma.
To prove the strictness of solutions, we need also to make the following assumption.
(H2:5) F : [0; 1) £ B ! E is continuously di¤erentiable and the derivatives D1 F , D2 F
satisfy the following locally Lipschitz condition :
for any compact set Q ½ [0; 1) £ B, there exists a constant L > 0 such that
(
kD1 F (t; ') ¡ D1 F (t; Ã)k · L k' ¡ ÃkB ;
kD2 F (t; ') ¡ D2 F (t; Ã)k · L k' ¡ ÃkB ;
for all (t; '), (t; Ã) 2 Q:
Theorem 2.2.8 Assume that B is a normed space which satis…es Axiom (C1) or Axiom
(D1) and the conditions (H2:1); (H2:4) and (H2:5) hold. Let ' 2 B continuously di¤er-
entiable such that
'0 2 B; '(0) 2 D(A); '0 (0) 2 D(A) and '0 (0) = A'(0) + F (0; '):
Proof. By Theorem 2.2.5, we know that Equation (2.1) has a unique global integral
solution x; which is given by
Z
d t
x(t) = S 0 (t)'(0) +
S(t ¡ s)F (s; xs )ds; for t ¸ 0: (2.6)
dt 0
The assumption '(0) 2 D(A) implies that S 0 (t)'(0) = S(t)A'(0) + '(0); and x can be
written as Z
d t
x(t) = S(t)A'(0) + '(0) + S(t ¡ s)F (s; xs )ds: (2.7)
dt 0
Consider the following equation
8
< d y(t) = Ay(t) + D F (t; x ) + D F (t; x ) y ; t ¸ 0;
1 t 2 t t
dt (2.8)
: y(t) = '0 (t); ¡1 < t · 0:
Using Axiom (A1), we can see that the mapping G : (¡1; 0] £ B ! E de…ned by
G(t; Ã) = D1 F (t; xt ) + D2 F (t; xt )Ã, satis…es (H2:4). By Theorem 2.2.5, Equation (2.8)
has a unique global integral solution y which is given by
8
>
>
> y(t) = S 0 (t)(A'(0) + F (0; '))
>
> Z
>
< d t
+ S(t ¡ s) (D1 F (s; xs ) + D2 F (s; xs ) ys ) ds; t ¸ 0;
dt 0 (2.9)
>
>
>
>
>
>
:
y(t) = '0 (t); ¡1 < t · 0:
38
Consider the mapping w : (¡1; +1) ! E; de…ned by
8 Z t
< '(0) + y(s) ds; t ¸ 0;
w(t) = 0 (2.10)
:
'(t); ¡1 < t · 0:
and t 7! S(t ¡ s)F (s; ws )ds are continuously di¤erentiable and satisfy
0
Z t
d
S(t ¡ s)F (s; ws )ds = S(t)F (0; ')
dt 0
Z t (2.12)
+ S(t ¡ s) (D1 F (s; ws ) + D2 F (s; ws ) ys ) ds:
0
Let 8 Z
>
> d t
>
> I1 = S(t ¡ s) (F (s; xs ) ¡ F (s; ws )) ds;
>
> dt 0
>
< Z t
I2 = ¡ S(t ¡ s) (D1 F (s; xs ) ¡ D1 F (s; ws )) ds;
>
> 0
>
> Z t
>
>
>
: I3 = ¡ S(t ¡ s) (D2 F (s; xs ) ¡ D2 F (s; ws )) ys ds;
0
and, for a > 0 …xed
½ ¾
° 0 ° ¡ ° 0° ¢
° ° ° °
b0 = max sup kS(s)k ; sup S (s) ; sup K(s) jy(s)j + M(s) ' B :
0·s·a 0·s·a 0·s·a
39
Let t 2 [0; a]. It’s easy to see that
Z t
kI2 k · b0 L kxs ¡ ws kB ds;
0
Let
¡ ¢
¹ :
b1 = b0 L 1 + b0 + M
Then, we obtain Z t
sup jx(s) ¡ w(s)j · b1 kxs ¡ ws kB ds
0·s·t 0
Then, Z t
kxt ¡ wt kB · Ka b1 kxs ¡ ws kB ds:
0
Using Gronwall’s lemma, we conclude that
Consequently, x(t) = w(t) for all t 2 (¡1; a]. Hence, x is continuously di¤erentiable on
(¡1; a]. This ends the proof.
40
8 Z t
>
> @
>
> w(t; ») = d¢w(t; ») + k(t; s; w(s; »))ds + f(t; w(t; »));
>
> @t ¡1
>
>
>
> t ¸ 0; » 2 -;
>
<
(2.13)
>
>
>
> w(t; ») = 0; t ¸ 0; » 2 @-;
>
>
>
>
>
>
>
: w(µ; ») = w (µ; »); ¡ 1 < µ · 0; » 2 - ;
0
@2 n
where d > 0, - is a bounded open set in IRn with regular boundary @-, ¢ = § 2 and
i=1 @³ i
k : IR+ £ IR £ IRn ! IRn ; f : IR+ £ IRn ! IRn and w0 : (¡1; 0] £ - ! IRn are functions.
¡ ¢
We choose E = C -; IRn ; and we consider the operator A : D(A) ½ E ! E de…ned
by ( © ª
D(A) = y 2 E : ¢y 2 E and y=@- = 0
Ay = d¢y:
© ª
It is known that D(A) = y 2 E : y=@- = 0 6= E and
8
< ½(A) = (0; +1) ;
° ° 1
: ° ¡1 °
°(¸I ¡ A) ° · ; for ¸ > 0:
¸
This implies that A satis…es (H2:1) on E.
As a concrete phase space, we choose the space
jÁ(µ)j
B = Cg0 := fÁ 2 C((¡1; 0] ; E) : lim = 0g;
µ!¡1 g(µ)
jÁ(µ)j
kÁkB := sup ;
¡1<µ·0 g(µ)
g(t + µ)
G(t) = sup ;
¡1<µ·¡t g(µ)
is locally bounded.
41
From Theorems 1.3.2 and 1.3.6 of [66], this space satis…es Axioms (A), (A1), (B) and
(C1).
In addition, we set for t ¸ 0 and » 2 -
8
>
> x(t)(») = w(t; »);
>
>
<
'(µ)(») = w0 (µ; »);
>
> Z 0
>
>
: F (t; Á)(») = f(t; Á(0)(»)) + k(t; t + µ; Á(µ)(»)) dµ; for Á 2 B:
¡1
Now it will be easy to adapt our previous results to solving Equation (2.13). The
properties of the mapping F depend on k, f and on the choice of the space B .
We assume that k : IR+ £ IR £ IRn ! IRn and f : IR+ £ IRn ! IRn are continuous
functions that satisfy :
where ´ is a …xed positive constant, C; N : (¡1; 0] ! [0; +1) are two measurable functions
such that, C(¢)g(¢) and N(¢)g(¢) are integrable on (¡1; 0] :
Under the above conditions, F : [0; +1) £ B ! E satis…es condition (H2:4). In fact,
given t ¸ 0; Á 2 B and a sequence (tn )n¸0 of [0; +1) such that tn ! t ; we have
and
kk(tn ; tn + µ; Á(µ)(¢))k · C(µ) jÁ(µ)j ;
jÁ(µ)j
· C(µ)g(µ) sup ;
¡1<µ·0 g(µ)
= C(µ)g(µ) kÁkB :
By continuity of k, we have
42
Then, the Lebesgue convergence theorem allows us to assert that
µZ 0 ¶
sup jk(tn ; tn + µ; Á(µ)(»)) ¡ k(t; t + µ; Á(µ)(»))j dµ ! 0;
¡1 n
»2-
and
Z 0 Z 0
kk(t; t + µ; Á(µ)(¢)) ¡ k(t; t + µ; Ã(µ)(¢))k dµ · kÁ ¡ ÃkB N(µ)g(µ)dµ:
¡1 ¡1
where Z 0
L := ´H + N(µ)g(µ)dµ:
¡1
43
vi) f and k are continuously di¤erentiable, with the following conditions :
(b1) kD1 k(t; s; u(¢)) ¡ D1 k(t; s; v(¢))k + kD2 k(t; s; u(¢)) ¡ D2 k(t; s; v(¢))k
· N1 (s ¡ t) ju ¡ vj ; for s · t; u; v 2 E;
where D1 ; D2 and D3 are the derivatives, ´ 1 and ´2 are two …xed positive constants, C1 ,
N1 , C2 , N2 : (¡1; 0] ! [0; +1) are measurable functions such that C1 (¢)g(¢); N1 (¢)g(¢);
C2 (¢)g(¢) and N2 (¢)g(¢) are integrable on (¡1; 0] : We can verify without di¢culties, that
all conditions of Theorem 2.2.8 hold. Then the integral solution w(t; ») = x(t)(») is a strict
one.
44
Chapter 3
3.1 Introduction
This chapter is concerned with local existence of solutions and stability to the autonomous
case of the PFDE with in…nite delay considered in the previous chapter
8
< d x(t) = Ax(t) + F (x ); t ¸ 0;
t
dt (3.1)
: x = ' 2 B;
0
45
semigroup around an equilibrium is exponentially stable, then the equilibrium of Equation
(3.2) is also exponentially stable. After that, we deal with existence of a global attractor.
Finally, we give an application to a reaction di¤usion equation with in…nite delay
We still assume that the phase space (B; k:kB ) is a seminormed linear space of func-
tions mapping (¡1; 0] into E, that satis…es the fundamental Axioms (A); (A1) and (B)
introduced in Chapter 1. We also keep the assumption (H2:1) on the operator A.
Contrarily to the results obtained in Chapter 2, to prove the local existence of solutions to
Equation (3.2), we suppose a local Lipschitz condition on the nonlinear part F; that is,
(H3:1) F is Lipschitz continuous on the seminorm-balls of B, i.e., for each r > 0 there
exists a constant c0 (r) > 0 such that if '1 ; '2 2 B and k'1 kB ; k'2 kB · r then,
Theorem 3.2.1 Assume that (H2:1) and (H3:1) hold. Let ' 2 B such that '(0) 2 D(A).
Then, there exists a maximal interval of existence (¡1; b' ) ; b' > 0, and a unique integral
solution x(:; ') of Equation (3.2), de…ned on (¡1; b' ) and either
Moreover, x(t; ') is a continuous function of ', in the sense that if ' 2 B; '(0) 2 D(A) and
t 2 [0; b' ) ; then there exist positive constants ® and " such that, for à 2 B; Ã(0) 2 D(A)
and k' ¡ ÃkB < "; we have t 2 [0; b' ) and
Proof. Note that (H3:1) implies that, for each r > 0 there exists c0 (r) > 0 such that for
' 2 B and k'kB · r, we have
46
Let ' 2 B, '(0) 2 D(A), r = k'kB + 1 and c1 = c0 (r)r + jF (0)j : De…ne the function
(
S 0 (t)'(0); for t ¸ 0,
y(t) =
'(t); for t · 0.
By virtue of the axioms (A ¡ i) and (A1), we deduce that yt 2 B and the map t ! yt is
continuous. Then, for b1 2 (0; 1) ; there exists b2 2 (0; 1) such that
Without loss of generality, we can assume that ! > 0: Let 0 < b · b2 such that
¹ !b c1 b < 1 ¡ b1 :
Kb Me
47
Since kus ¡ 'kB · 1, for s 2 [0; b] and r = k'kB + 1, we deduce that kus kB · r, for
s 2 [0; b] : Then,
jF (us )j · c0 (r) kus kB + jF (0)j · c1 :
On the other hand, by Axiom (A ¡ iii) we see that for any t 2 [0; b]
¹ and !
and for suitable constants M
¯ ¯
¯ Zt ¯
¯d ¯
¯
jw(t)j = ¯ S(t ¡ s)F (us )ds¯¯ ;
¯ dt ¯
0
Zt
¹ !t c1
· Me e¡!s ds;
0
¹ !t c1 b;
· Me
1
< (1 ¡ b1 ):
Kb
Then, we deduce that
kvt ¡ 'kB < 1, for any t 2 [0; b] :
we obtain
kT u ¡ T vkFb = sup j(T u)(t) ¡ (T v)(t)j ,
t2[0;b]
and ¯ Z t ¯
¯d ¯
j(T u)(t) ¡ (T v)(t)j = ¯¯ S(t ¡ s)(F (us ) ¡ F (vs ))ds¯¯ ;
dt Z 0
t
·M ¹e !t jF (us ) ¡ F (vs )j ds;
0 Z
t
·M ¹ e!t c0 (r) kus ¡ vs kB ds;
0
¹ e!b c0 (r) Kb b ku ¡ vk :
·M Fb
Then,
¹ !b c0 (r) Kb b ku ¡ vk :
kT u ¡ T vkFb · Me Fb
48
Note that r ¸ 1, then
¹ !b c0 (r)Kb b · Me
Me ¹ !b c1 Kb b < 1:
³ ´
This means that T is a strict contraction in F'b ; k:k '
Fb . So, T has a …xed point x in Fb .
Let u be another …xed point of T . Then,
(
ku ¡ xkFb = 0;
u0 = x0 = ':
This means that T has one and only one …xed point x in F'
b . We conclude that Equa-
tion (3.2) has one and only one integral solution which is de…ned on the interval (¡1; b].
Let [0; b' ) ; b' > 0; be the maximal interval of existence of the integral solution x(:; ')
of Equation (3.2). Assume that b' < +1 and lim sup jx(t; ')j < +1. Then, from Axiom
t!b¡
'
(A ¡ iii), we can see that there exists a constant r > 0 such that kxs (:; ')kB · r, for all
s 2 [0; b' ). Consider t; t + h 2 [0; b' ) and h > 0;
Using the same reasoning, one can show a similar result for h < 0: By applying the well-
known Gronwall lemma, we can deduce that x is uniformly continuous on [0; b' ). Conse-
quently, x(:; ') can be extended to b' , which contradicts the maximality of [0; b' ) :
49
We prove now that the solution depends continuously on the initial data. Let ' 2 B such
that '(0) 2 D(A) and t 2 [0; b' ) be …xed. Set
It is clear that
kÃkB · k'kB + " < r:
We put (
b0 = sup fs > 0 : kx¾ (:; Ã)kB · r, for all ¾ 2 [0; s]g ;
¡ ¢
c(t) = (Mb0 + HKb0 M ¹ e!t ) exp Kb Me¹ !t c0 (r)t :
0
c(t)" < 1.
kxs (:; ') ¡ xs (:; Ã)kB · Mb0 k' ¡ ÃkB + Kb0 sup jx(»; ') ¡ x(»; Ã)j ;
0·»·s
©¯ ¯
· Mb0 k' ¡ ÃkB + Kb0 sup ¯S 0 (») ('(0) ¡ Ã(0))¯
0·»·s
¯ Z ¯¾
¯d » ¯
+ ¯¯ S(» ¡ ¾) (F (x¾ (:; ')) ¡ F (x¾ (:; Ã))) d¾¯¯ ;
d» 0
¡ ¢
· ¹ !t + Mb k' ¡ Ãk
HKb0 Me 0 B
Z s
¹ !t c0 (r)
+Kb0 Me kx¾ (:; ') ¡ x¾ (:; Ã)kB d¾:
0
50
It follows that b0 cannot be the largest number s > 0 such that kx¾ (:; Ã)kB · r; for
¾ 2 [0; s] : Thus, b0 ¸ t and t < bà . Furthermore, kxs (:; Ã)kB · r; for s 2 [0; t]. Then,
using the inequality (3.4), we deduce the continuous dependence on the initial data. This
completes the proof.
Theorem 3.3.1 Assume that the phase space B is a normed space which sati…es Axiom
(C1) or Axiom (D1) and the conditions (H2:1); (H3:1) and (H3:2) hold. Let ' 2 B be a
continuously di¤erentiable function with '0 2 B and
'(0) 2 D(A); '0 (0) 2 D(A) and '0 (0) = A'(0) + F ('):
If x(:; ') : (¡1; b' ) ! E is the integral solution of Equation (3.2) given by Theorem 3.2.1,
then x(:; ') is a strict solution of Equation (3.2).
Proof. Let ' 2 B such that '(0) 2 D(A); '0 (0) 2 D(A) and '0 (0) = A'(0) + F ('): Let
x := x(:; ') be the unique integral solution of Equation (3.2) on (¡1; b' ) and b1 2 (0; b' ).
Consider the following equation
8
< dy (t) = Ay(t) + F 0 (x )y ; t ¸ 0;
t t
dt
: y = '0 2 B.
0
Using a similar argument as in the prove of Theorem 3.2.1, we can show that the above
equation has a unique integral solution y : [0; b1 ] ! E, which is given by
8 Z t
< S 0 (t)'0 (0) + d S(t ¡ s)F 0 (xs )ys ds; for t 2 [0; b1 ] ;
y(t) = dt 0
: 0
' (t); for t 2 (¡1; 0] :
51
We will prove that x = z: Using the expression of y, we obtain
Zt
0
z(t) = '(0) + S(t)' (0) + S(t ¡ s)F 0 (xs )ys ds; for t 2 [0; b1 ] :
0
From '(0) 2 D(A); '0 (0) 2 D(A) and '0 (0) = A'(0) + F ('), we deduce
On the other hand, by use of Lemma 2.2.6 or Lemma 2.2.7, we can infer from (3.5) that
Zt
zt = ' + ys ds, for t 2 [0; b1 ] :
0
Z t
Consequently, the maps t 7! zt and t 7! S (t ¡ s) F (zs ) ds are continuously di¤eren-
0
tiable and we have
Zt Zt
d
S (t ¡ s) F (zs ) ds = S (t) F (') + S (t ¡ s) F 0 (zs ) ys ds:
dt
0 0
Zt
0
z(t) = S (t)'(0) + S(t)F (') + S(t ¡ s)F 0 (zs )ys ds:
0
52
Therefore, Z t
d
x (t) ¡ z (t) = S (t ¡ s) (F (xs ) ¡ F (zs )) ds
dtZ 0
t
¡ S (t ¡ s) (F 0 (xs ) ¡ F 0 (zs )) ys ds:
0
Then, we deduce
µZ t
jx (t) ¡ z (t)j · ¯e!b1 jF (xs ) ¡ F (zs )j ds
Z t 0 ¶
+ 0 0
kF (xs ) ¡ F (zs )kL(B;E) kys kB ds :
0
Let à !
r = max sup kxs kB ; sup kys kB ; sup kzs kB :
0·s·b1 0·s·b1 0·s·b1
There exist c0 (r); v0 (r) ¸ 0 such that if '1 ; '2 2 B and k'1 kB ; k'2 kB · r, then
(
jF ('1 ) ¡ F ('2 )j · c0 (r) k'1 ¡ '2 kB ;
kF 0 ('1 ) ¡ F 0 ('2 )kL(B;E) · v0 (r) k'1 ¡ '2 kB :
Zt
!b1
sup jx(s) ¡ z(s)j · ¯e (c0 (r) + rv0 (r)) kxs ¡ zs kB ds:
0·s·t
0
Thus, we obtain
Zt
!b1
kxt ¡ zt kB · Kb1 ¯e (c0 (r) + r v0 (r)) kxs ¡ zs kB ds:
0
By Gronwall’s lemma, kxt ¡ zt kB = 0 for any t 2 [0; b1 ] : Using Axiom (A1 ¡ ii); we
conclude that x(t) = z(t), for all t 2 (¡1; b1 ] : Consequently, the integral solution x is
continuously di¤erentiable. Then, from Proposition 2.2.2 we conclude that x is a strict
solution of Equation (3.2). This completes the proof of the theorem.
53
3.4 The solution semigroup and linearized stability
In this section, we give some properties of the solution map associated to Equation (3.2).
We make the following autonomous version of (H2:4).
(H3:3) F is globally Lipschitz continuous on B :
Using this assumption and a …xed point theorem, we have shown in Theorem 2.2.5 that for
all ' 2 B such that '(0) 2 D(A), Equation (3.2) has a unique integral solution which is
de…ned on IR.
Let n o
X := ' 2 B : '(0) 2 D(A) :
where x(:; ') is the unique integral solution of Equation (3.2). We can prove the following
result.
(v) there exist two positive locally bounded functions m(:); n(:) : IR+ ! IR+ such that, for
all '1 ; '2 2 X and t ¸ 0;
Proof. (i) and (ii) are consequence of the uniqueness of the integral solution of Equation
(3.2).
54
(iii) and (iv) can be proved by using the integral Equation (2.5) and some properties of B.
To prove (v); we can see that
Proposition 3.4.2 Assume that (H2:1) ; (H2:2) ; and (H3:3) hold. Then, the semigroup
(U(t))t¸0 is decomposed on X as follows
where (U2 (t))t¸0 is the semigroup on X given, for ' 2 X , t ¸ 0 and µ 2 (¡1; 0], by
(
S 0 (t + µ)'(0); t + µ ¸ 0;
(U2 (t)') (µ) =
'(t + µ); t + µ · 0;
Proof. Let (Ãn )n¸0 be a bounded sequence in X : First, we prove that the sequence
((U1 (t)Ã n )(µ))n¸0 ; t > 0; µ 2 (¡1; 0] ; is totally bounded in E and the sequence (U1 (t)Ã n )n¸0 ; t >
0; is equicontinuous in (¡1; 0] : Let t > 0 and µ 2 (¡1; 0] be …xed. We have
(
(U1 (t + µ)Ã n ) (0); t + µ ¸ 0,
(U1 (t)Ã n )(µ) =
0; t + µ · 0:
Let 0 < " < t + µ: Then, for ¸ > !; we have
Z
t+µ
55
Z
t+µ¡"
where x(:; Ãn ) is the unique integral solution of Equation (3.2) for the initial condition Ãn .
Since S 0 (") is compact, there exists a compact set W" such that
8t+µ¡" 9
< Z =
0 0
S (") S (t + µ ¡ " ¡ s)¸R(¸; A)F (xs (:; Ãn ))ds : n ¸ 0; ¸ > ! ½ W" :
: ;
0
where
° °
®t = sup fjF (xs (:; Ã n )j : n ¸ 0g and M t = sup °S 0 (s)°D(A) :
s2[0;t] s2[0;t]
This shows, by letting ¸ to +1, that ((U1 (t)Ã n )(µ))n¸0 ; t > 0; µ 2 (¡1; 0] ; is totally
bounded. To establish the equicontinuity, let t > 0 and µ 0 2 (¡1; 0] be …xed. For
µ 2 (¡1; 0] such that jµ ¡ µ0 j small enough and µ0 < µ; we obtain
56
Which leads to
j¸R(¸; A) ((U1 (t)Ãn )(µ) ¡ (U1 (t)Ãn )(µ0 ))j
¯ Z t+µ0 ¯
¯ 0 ¯
¯
· ¯(S (µ ¡ µ0 ) ¡ I) S (t + µ 0 ¡ s)¸R(¸; A)F (xs (:; à n ))ds¯¯
0
0
¸M¹
+ M t ®t (µ ¡ µ0 ):
¸¡!
Besides, there exists a compact set W such that
8t+µ 9
< Z 0¯ ¯ =
¯S 0 (t + µ0 ¡ s)¸R(¸; A)F (xs (:; Ãn ))¯ ds : n ¸ 0; ¸ > ! µ W:
: ;
0
Letting ¸ to +1 and using the fact that (S 0 (:)x)x2W is equicontinuous at the right in 0;
gives
lim j(U1 (t)Ã n )(µ) ¡ (U1 (t)Ã n )(µ 0 )j = 0:
µ!µ0
µ>µ 0
where L := F 0 (0) and (T (t))t¸0 is the solution semigroup on X associated to Equation (3.6).
Proposition 3.4.3 Assume that (H2:1); (H3:3) and (H3:4) hold. Then, the derivative
at zero of the nonlinear semigroup (U(t))t¸0 , associated to Equation (3.2) is the linear
semigroup (T (t))t¸0 associated to Equation (3.6).
57
Proof. It su¢ces to show that for each t ¸ 0 and " > 0; there exists ± > 0 such that
Note that
kU(t)' ¡ T (t)'kB = kxt (:; ') ¡ yt (:; ')kB ;
where x(:; ') ( respectively, y(:; ')) is the unique integral solution of Equation (3.2) ( respectively,
Equation (3.6)).
From Axiom (A ¡ iii), we see that for all t ¸ 0
Consequently,
0 1
Zt
¹ !t @" k'k + kLk
kU(t)' ¡ T (t)'kB · K(t)Me e¡!s kU(s)' ¡ T (s)'kB dsA :
B L(B;E)
0
Hence, we conclude that U(t) is di¤erentiable at 0 and (D' U(t))(0) = T (t); for each t ¸ 0:
58
De…nition 3.4.1 ( Desh and Schappacher [41]) Let (V (t))t¸0 be a strongly continuous
semigroup on a Banach space (Y; j:jY ). A point x0 2 Y is called an equilibrium of (V (t))t¸0
if V (t)x0 = x0 for all t ¸ 0: An equilibrium x0 2 Y is called locally exponentially stable if
there exist ± > 0, ¹ > 0 and k ¸ 1 such that
Theorem 3.4.4 [41] Let (V (t))t¸0 be a nonlinear strongly continuous semigroup in a Ba-
nach space Y . Assume that x0 2 Y is an equilibrium of (V (t))t¸0 such that V (t) is Fréchet-
di¤erentiable at x0 for each t ¸ 0, with W (t) the Fréchet-derivative at x0 of V (t): Then,
(W (t))t¸0 is a strongly continuous semigroup of bounded linear operators on Y: Moreover,
if (W (t))t¸0 is exponentially stable, then x0 is a locally exponentially stable equilibrium of
(V (t))t¸0 .
By using the above theorem in the Banach space of equivalence classes B = k:kB , we get
our main result concerning the stability of an equilibrium of (U(t))t¸0 .
Theorem 3.4.5 Assume that (H2:1); (H3:3) and (H3:4) hold. If (T (t))t¸0 is exponen-
¹ ¸ 1 and ! > 0 such that kT (t)k · Me
tially stable on X ; i.e., there exist constants M ¹ ¡!t ,
for t ¸ 0, then zero is a locally exponentially stable equilibrium of (U(t))t¸0 on X ; that is,
there exist ± > 0, ¹ > 0 and k ¸ 1 such that
To study the stability of solutions to the linearized equation (3.6), we need to charac-
terize the in…nitesimal generator of the linear semigroup (T (t))t¸0 on X : To do this, we
assume that B satis…es the additional axiom (D1) used in strictness of solutions.
59
Let x(:; ') be the integral solution of Equation (3.2). In the proof of Theorems 2.2.8 and
3.3.1, we showed that there exists a continuous mapping y : IR ! E with y0 = '0 such
that
Zt
xt (:; ') = ' + ys ds; for t ¸ 0:
0
1
Which implies that lim (xt (:; ') ¡ ') exists and is equal to '0 : Then, ' 2 D(AT ).
t
t!0+
Conversely, let ' 2 X such that
1
lim (xt (:; ') ¡ ') = Ã = AT '; exists in X :
t!0+ t
Axiom (D1) implies that
1
lim (xt (µ; ') ¡ '(µ)) exists for all µ · 0 and is equal to Ã(µ):
t!0+ t
Then, for µ 2 (¡1; 0), we have
1
Ã(µ) = lim ('(t + µ) ¡ '(µ)) = D+ '(µ);
t!0+ t
where D+ ' is the right derivative of '; and D+ ' = à in (¡1; 0) : Since à is continuous,
D+ ' is also continuous in (¡1; 0) : Let us introduce the following result.
Lemma 3.4.7 [97] Let ' be continuous and di¤erentiable on the right on [a; b). If D+ ' is
continuous on [a; b) ; then ' is continuously di¤erentiable on [a; b) :
From the above lemma, we deduce that the function ' is continuously di¤erentiable in
(¡1; 0) and '0 = Ã: On the other hand, for µ = 0; one has lim '0 (µ) exists and equals to
µ!0¡
Ã(0): From this, we infer that the function ' is continuously di¤erentiable in (¡1; 0] and
'0 = Ã 2 X : Since t ! xt (:; ') is continuously di¤erentiable in [0; +1) and using Axiom
(D1), we can state that the map t ! x(t; ') is continuously di¤erentiable in [0; +1) : Then,
from Proposition 2.2.2, x is a strict solution of Equation (3.6), and for µ = 0; one has
1
'0 (0) = Ã(0) = lim (x(t) ¡ '(0)) = x0 (0) and x0 (0) = A'(0) + L('):
t!0 t
Form this we conclude that
'0 (0) = A'(0) + L('):
60
In the sequel, we give a necessary and su¢cient condition to ensure the stability of the
linear semigroup (T (t))t¸0 on X . We need the following results.
Let H : Z ! Z be a closed linear operator with a dense domain D(H) in a Banach space
Z. We denote by ¾(H) the spectrum of H; and by ¾p (H) the point spectrum of H.
We call essential spectrum of H (denotation by ¾e (H)) the set of all ¸ in ¾(H) for which
at least one of the following holds :
(i) Im(¸I ¡ H) = f(¸I ¡ H)z : z 2 Zg is not closed,
(ii) the generalized eigenspace M¸ (H) = [ ker(¸I ¡ H)k of ¸ is in…nite dimensional;
k¸1
(iii) ¸ is a limit point of ¾(H); that is, ¸ 2 ¾(H)= f¸g:
where the Kuratowski’s measure of noncompactness of the bounded subset B of the Banach
space Z is de…ned by :
From the above formula, we deduce that re (H + K) = re (H) for any compact operator
K : Z ! Z; that is, re is invariant under compact perturbations.
Since ®(H) · kHk ; by the formula (3.7) one has
Let ¸ 2 ¾(H) ¡ ¾e (H); it follows from [122] that ¸ is an isolate point, dim M¸ (H) < 1
and ¸ 2 ¾p (H).
De…ne the growth bound of the semigroup (T (t))t¸0 by
½ ¾
¡!t
!((T (t))t¸0 ) := inf ! 2 IR : sup e kT (t)k < 1 :
t¸0
61
Set
s1 (AT ) = sup fRe ¸ : ¸ 2 ¾(AT ) ¡ ¾e (AT )g :
It is noted in Chapter 1 that C° together with the norm kÁk° = sup e°µ jÁ(µ)j ; Á 2 C° , is
µ·0
a standard example of phase space B that satis…es axioms (A), (A1), (B) and (D1).
Assume the following hypothesis.
(H3:5) There exist two positive constants ¹ and ± such that
° 0 °
°S (t)° · ¹e¡±t ; t ¸ 0:
D(A)
Proposition 3.4.8 Suppose that (H2:1), (H2:2), (H3:3), (H3:4) and (H3:5) hold, and
B = C° : If s1 (AT ) < 0; then zero is an exponentially stable equilibrium of (U(t))t¸0 :
where (T2 (t))t¸0 is the semigroup on X given, for ' 2 X , t ¸ 0 and µ 2 (¡1; 0], by
(
S 0 (t + µ)'(0); t + µ ¸ 0;
(T2 (t)') (µ) =
'(t + µ); t + µ · 0;
62
For every Á 2 X ; we have
which implies that kT2 (t)Ák° · ¹e¡ min(±;°)t kÁk° : Using (3.8), we deduce that
Which implies that for t large enough, re (T (t)) < 1; and by (3.9) we get ! e ((T (t))t¸0 ) < 0.
So, we conclude that !((T (t))t¸0 ) < 0 and the semigroup (T (t))t¸0 is exponentially asymp-
totically stable. Consequently, from Theorem 3.4.5, we deduce that zero is an exponentially
stable equilibrium of (U(t))t¸0 : This completes the proof of the proposition.
63
(iv) A compact invariant set A is said to be a maximal compact invariant set if every
compact invariant set of the semigroup belongs to A.
(v) An invariant set A is said to be a global attractor if A is maximal compact invariant
set which attracts each bounded set B ½ Y .
(vi) The semigroup T is said to be point dissipative (compact dissipative) if there is a
bounded set B ½ Y that attracts each point of Y (each compact set of Y ) under T.
Hereafter, we suppose hypothesis (H2:1) and (H3:3): We also suppose that the initial
function ' is an element of
n o
X := Á 2 B : Á(0) 2 D(A) :
Remark 3.5.1 It’s easy to see that the assertion (v) in Proposition 3.4.1 implies that for
all t ¸ 0, U(t) is a bounded operator on X .
Proposition 3.5.1 (Section 3.4) Under the conditions (H2:1) ; (H2:2) and (H3:3); let
U(t) = U1 (t) + U2 (t); t ¸ 0; where (U2 (t))t¸0 is the solution semigroup of the equation
(
x(t) = S 0 (t)'(0); t ¸ 0
x0 = ' 2 X :
¹ ¡´t ; t ¸ 0:
kU2 (t)k · Me
According to Section 3.4, we can show that the abstract assumption (H3:6) is satis…ed
at least in the case where B =C° with ° > 0: In fact, without loss of generality, we can
assume that (H3:5) is veri…ed, otherwise, we consider A ¡ ¸I instead of A such that
S¸0 (t) := e¸t S 0 (t) is exponentially stable for ¸ properly chosen ( refer to Proposition 1.2.5).
As a consequence of Proposition 3.5.1, we obtain the following result.
Proposition 3.5.2 Assume that (H2:1) ; (H2:2); (H3:3) and (H3:6) hold. Then the
solution semigroup (U(t))t¸0 is asymptotically smooth on X .
64
Lemma 3.5.3 [51] Suppose that T is decomposed as T (t) = T1 (t) + T2 (t) : Y ! Y; t ¸ 0;
such that (T1 (t))t¸0 is compact and there is a continuous function u : IR+ £ IR+ ! IR+
such that u(t; r) ! 0 as t ! +1 and kT2 (t)yk · u(t; r) if kyk · r: Then, the semigroup T
is asymptotically smooth on Y .
Theorem 3.5.4 If (H2:1) ; (H2:2); (H3:3) and (H3:6) hold, then each of the following
conditions implies the existence of a global attractor A for Equation (3.2).
(a) Equation (3.2) is compact dissipative.
(b) Equation (3.2) is point dissipative and orbits of bounded sets are bounded.
65
This implies that assumption (H2:1) is satis…ed. On the other hand, we can see that
with the norm kÁk° =sup e°µ jÁ(µ)j ; Á 2 C° : As noted in Examples 1.1.1 and 1.1.3, this
µ·0
space satis…es Axioms (A), (A1), (B) and (D1). Set
8
>
> x(t)(») = w(t; »); t ¸ 0; » 2 [0; ¼] ;
>
<
'(µ)(») = w0 (µ; »); µ · 0; » 2 [0; ¼] ;
> Z 0
>
>
: F (Á)(») = bÁ(0)(») + f(Á(¡¿ )(»)) + c G(µ)Á(µ)(»)dµ; » 2 [0; ¼] ; Á 2 C° :
¡1
We suppose that
(i) f is locally Lipschitz continuous,
(ii) G(:)e¡°: is integrable on (¡1; 0], Ã !
(iii) w0 2 C((¡1; 0] £ [0; ¼] ; IR), with lim e°µ sup jw0 (µ; »)j exists, and w0 (0; 0) =
µ!¡1 0·»·¼
w0 (0; ¼) = 0:
We have, for every Á1 ; Á2 2 C° ,
Z 0
sup G(µ) jÁ1 (µ)(») ¡ Á2 (µ)(»)j dµ
0·»·¼ ¡1
Z 0
¡ ¢
= sup e¡°µ G(µ) e°µ jÁ1 (µ)(») ¡ Á2 (µ)(»)j dµ;
0·»·¼
µZ 0 ¡1 ¶
· e¡°µ G(µ)dµ sup e°µ jÁ1 (µ)(») ¡ Á2 (µ)(»)j :
¡1 ¡1<µ·0
0·»·¼
sup jf(Á1 (¡¿ )(»)) ¡ f(Á2 (¡¿ )(»))j · b(r) sup jÁ1 (¡¿ )(») ¡ Á2 (¡¿ )(»)j ;
0·»·¼ 0·»·¼
· b(r)e°¿ kÁ1 ¡ Á2 k° :
66
We conclude that the assumptions (i), (ii) and (iii) imply that F is locally Lipschitz con-
tinuous and ' 2 C° with '(0) 2 D(A). Consequently, Theorem 3.2.1 ensures the existence
of a maximal interval of existence (¡1; bw0 ) and a unique integral solution w(t; ») on
(¡1; bw0 ) £ [0; ¼] .
To investigate that the integral solution w of Equation (3.10) is a strict one, we add the
following assumptions.
(iv) f is continuously di¤erentiable and f 0 is locally
à Lipschitz continuous,
µ ¶!
@
(v) w0 2 C 2 ((¡1; 0] £ [0; ¼] ; E), with lim e°µ sup w0 (µ; ») exists,
µ!¡1 0·»·¼ @µ
@ @
w0 (0; 0) = w0 (0; ¼) = 0 and
@µ @µ
@ @2
w0 (0; ») = a 2 w0 (0; ») + bw0 (0; »))
@µ Z @» 0
+c G(µ)w0 (µ; »))dµ + f(w0 (¡¿ ; »)); for » 2 [0; ¼] :
¡1
Z0
0
F (Á)(Ã)(») = bÃ(0)(») + c G(µ)Ã(µ)(»)dµ + f 0 (Á(¡¿ )(»))Ã(¡¿ )(»):
¡1
67
It is known from [43] that A0 generates a strongly continuous compact and exponentially
stable semigroup (T0 (t))t¸0 in D(A).
We deduce from Proposition 3.4.8 that the asymptotic behavior of the linearized semi-
group (T (t))t¸0 is completely determined by the point spectrum of its generator AT . In
order to characterize the eigenvalues of AT , let ¾p (AT ) denote the point spectrum of AT :
Then, ¸ 2 ¾p (AT ) if and only if there exists Á 2 D(AT ); Á 6= 0 such that AT Á = ¸Á: It
follows that Á(µ) = e¸µ u with u 6= 0; u 2 D(A) and ¸u = Au + F 0 (0)(e¸: u): On the other
hand, Á 2 C° . This is true only if Re ¸ ¸ ¡°. Then, ¸ 2 ¾p (AT ) if and only if there exists
u 2 D(A) and u 6= 0 such that
Z0
Re ¸ ¸ ¡° and ¸u = Au + bu + c G(µ)e¸µ udµ:
¡1
It follows that the exponential stability of solutions of Equation (3.11) is determined by the
following characteristic equation
8
>
< Re ¸ ¸ ¡°;
Z 0
(3.12)
>
: ¸¡b¡c G(µ)e¸µ dµ = ¡an2 ; n ¸ 1:
¡1
Lemma 3.6.1 [78] All roots of System (3.12) have negative real part if
Z0
G(µ)dµ = 1 and c < a ¡ b: (3.13)
¡1
From this lemma, we conclude that if condition (3.13) is satis…ed, then the solution
semigroup associated to Equation (3.11) is exponentially stable. In that case, by Proposition
3.4.8 we conclude that zero is a locally exponentially stable equilibrium of the solution
semigroup associated to Equation (3.10).
68
Chapter 4
This chapter establishes some extensions to the theory of existence and linearized stability
obtained in Chapters 2 and 3.
We start with a brief description of the equation under investigation. Then, in Section
4.2, we prove global existence and uniqueness of integral solutions by use of a …xed point
argument. We also prove that the integral solutions are strict solutions under more restric-
tive assumptions. Next, in Section 4.3, we state some properties of the solution operator
associated to the autonomous case. Finally, in Section 4.4, we investigate the stability near
an equilibrium. Mainly, we prove that the equilibrium of the solution semigroup associ-
ated to the semilinear autonomous case is locally exponentially stable when its linearized
solution semigroup around an equilibrium is exponentially stable.
1
This chapter is based on an ongoing work in collaboration with M. Adimy and K. Ezzinbi.
69
4.1 Introduction
In [10]-[12], Adimy and Ezzinbi dealed with some general equations of the type (0.14) or
(0.15) with …nite delay. The authors studied equations of the form (0.15) or :
8
< @ Dv = ADv + F (v ) ; t ¸ 0;
t t t
@t (4.1)
: v = ' 2 C;
0
where A is a nondensely de…ned linear operator that satis…es the Hille-Yosida condition on
a Banach space (E; j:j); C is the space of all continuous functions on [¡r; 0] with values in
E, provided with the uniform norm topology. D : C ! E is a bounded linear operator given
by
Z0
DÁ := Á(0) ¡ [d´(µ)] Á(µ);
¡r
where A generates a strongly continuous semigroup on a Banach space (E; j:j) and B is the
space of functions mapping (¡1; 0] into E, which satis…es axioms (A); (A1) and (B): G
and F are continuous functions from [0; +1) £ B into E and for each x : (¡1; b] ! E;
b > 0; and t 2 [0; b], xt represents, as usual, the mapping de…ned from (¡1; 0] into E by
The authors have obtained some results about the existence and regularity by use of the
analytic semigroups theory.
70
In the previous chapters, we had developed many similar results on PFDEs with in…nite
delay under the condition that A only satis…es the usual Hille-Yosida conditions expect the
density of D(A) in E: In this chapter, we prove that the same results can be reproduced
for PNFDEs. We consider the case of in…nite delay for the equation considered by Adimy
and Ezzinbi in [12] :
8
< d [x(t) ¡ G(t; x )] = A [x(t) ¡ G(t; x )] + F (t; x ); t ¸ 0;
t t t
dt (4.3)
: x = ' 2 B:
0
Equation (4.3) is the true generalization to the in…nite delay case of Equation (??). How-
ever, according to [10], Equation (4.2) can be written as Equation (4.3), if we suppose that
Range(G) µ D(A):
In this chapter, we suppose that A is nondensely de…ned and satis…es the Hille-Yosida
condition. As in [10]-[12], we will establish some results about existence of integral and
strict solutions to Equation (4.3). In autonomous case, we prove that the solutions de…ne
a nonlinear strongly continuous semigroup on B. As a consequence, we give a linearization
principle around an equilibrium.
71
(iv) for 0 · t · a;
Zt Zt
x(t) = G(t; xt ) + '(0) ¡ G(0; ') + A (x(s) ¡ G(s; xs ))ds + F (s; xs )ds:
0 0
De…nition 4.2.2 Let ' 2 B. We say that a function x : (¡1; a] ! E is a strict solution
of Equation (4.3) in (¡1; a] if the following conditions hold :
(i) x(t) ¡ G(t; xt ) 2 C 1 ([0; a] ; E) \ C([0; a] ; D(A));
(ii) x satis…es Equation (4.3) on (¡1; a].
Lemma 4.2.1 From the closedness property of the operator A, we have the following state-
ments :
(i) If x is an integral solution of Equation (4.3) in (¡1; a] ; then for all t 2 [0; a] ; x(t) ¡
G(t; xt ) 2 D(A): In particular '(0) ¡ G(0; ') 2 D(A):
(ii) If x is an integral solution of Equation (4.3) in (¡1; a] ; such that x(t) ¡ G(t; xt ) be-
longs to C 1 ([0; a] ; E) or C([0; a] ; D(A)); then x is also a strict solution of Equation (4.3)
in (¡1; a].
Z
t+h Z
t+h
1
x(t) ¡ G(t; xt ) = lim (x(s) ¡ G(s; xs ))ds and (x(s) ¡ G(s; xs ))ds 2 D(A):
h!0 h
t t
If x(s) ¡ G(s; xs ) is di¤erentiable, since F is continuous, then the right side of the above
equality tends to
d
(x(t) ¡ G(t; xt )) ¡ F (t; xt ), as h tends to 0+ ; and
dt
Z
t+h
1
(x(s) ¡ G(s; xs ))ds tends to x(t) ¡ G(t; xt ); as h tends to 0+ :
h
t
72
From the closedness of A; we get x(t) ¡ G(t; xt ) 2 D(A) and
d
A(x(t) ¡ G(t; xt )) = (x(t) ¡ G(t; xt )) ¡ F (t; xt ) for t 2 [0; a] :
dt
From what we deduce that x is a strict solution.
On the other hand, if x(t)¡G(t; xt ) belongs to C([0; a] ; D(A)): Again from De…nition 4.2.1,
for all t 2 [0; a] and h > 0;
1
fx(t + h) ¡ G(t + h; xt+h ) ¡ x(t) + G(t; xt )g =
h
Z
t+h Z t+h
1 1
A(x(s) ¡ G(s; xs ))ds + F (s; xs )ds:
h h t
t
Since A(x(s) ¡ G(s; xs )) and F are continuous, the right side of the above equality tends
to A(x(t) ¡ G(t; xt )) + F (t; xt ); as h tends to 0+ : Which implies that x(t) ¡ G(t; xt ) is
di¤erentiable at the right in t and
d+
(x(t) ¡ G(t; xt )) = A(x(t) ¡ G(t; xt )) + F (t; xt ):
dt
It’s known that when the right derivative is continuous then the C 1 property holds. Then,
since A(x(t) ¡ G(t; xt )) + F (t; xt ) is continuous on [0; a], we conclude that x(t) ¡ G(t; xt ) is
d
di¤erentiable on [0; a] and (x(t) ¡ G(t; xt )) = A(x(t) ¡ G(t; xt )) + F (t; xt ): This …nishes
dt
the proof of the lemma.
To obtain global existence and uniqueness of the integral solution, additionally to
(H2:1), we make the following hypothesis :
(H4:1) G : [0; +1) £ B ! E is continuous and there exists ®0 > 0 such that ®0 K(0) < 1
and
jG(t; '1 ) ¡ G(t; '2 )j · ®0 k'1 ¡ '2 kB for '1 ; '2 2 B and t ¸ 0:
(H4:2) F : [0; +1) £ B ! E is continuous and there exists ¯ 0 > 0 such that
jF (t; '1 ) ¡ F (t; '2 )j · ¯ 0 k'1 ¡ '2 kB for '1 ; '2 2 B and t ¸ 0:
Before stating the results, we …rst rewrite Equation (4.3) in an integrated form, as follows.
For an initial value ' 2 B such that G(0; ') 2 D(A). From the integrated semigroup theory,
73
we know that x : (¡1; +1) ! E is an integral solution of Equation (4.3) if and only if x
solves the following system
8 Z t
< G(t; x ) = S 0 (t)G(0; ') + d S(t ¡ s)F (s; xs )ds; t ¸ 0;
t
dt 0 (4.4)
:
x0 = ':
Theorem 4.2.2 Assume that the conditions (H2:1); (H4:1) and (H4:2) are satis…ed.
Then, for given ' 2 B such that G(0; ') 2 D(A), Equation (4.3) has a unique integral
solution x(:; ') which is the unique solution of Equation (4.4) on (¡1; +1) :
Proof. Let a > 0 and C([0; a] ; E) be the space of continuous functions from [0; a] to E;
provided with the uniform norm topology. Let ' 2 B such that G(0; ') 2 D(A). Consider
the nonempty closed subset de…ned by
Using the hypothesis and Axiom (A ¡ iii); we can see that for every z1 ; z2 2 Za (') and
t 2 [0; a] ;
¡ ¢
¹ !a a Ka kz1 ¡ z2 k
j(Jz1 )(t) ¡ (Jz2 )(t)j · ®0 + ¯ 0 Me Za (') :
74
(H4:3) G and F are continuously di¤erentiable and their partial derivatives are locally
Lipschitzian with respect to the second argument in the sense that for any compact set
Q ½ [0; 1) £ B, there exists a constant ¯ 1 > 0 such that
8
>
> kD' F (t; ') ¡ D' F (t; Ã)k · ¯ 1 k' ¡ ÃkB ;
>
>
< kD F (t; ') ¡ D F (t; Ã)k · ¯ k' ¡ Ãk ;
t t 1 B
(4.5)
>
> kD ' G(t; ') ¡ D ' G(t; Ã)k · ¯ k' ¡ ÃkB;
>
> 1
:
kDt G(t; ') ¡ Dt G(t; Ã)k · ¯ 1 k' ¡ ÃkB
for all (t; '), (t; Ã) 2 Q and t ¸ 0, where Dt F , D' F; Dt G and D' G denote the derivatives
with respect to t and '.
Theorem 4.2.3 Assume that B is normed and sati…es Axiom (C1) or Axiom (D1) and
the conditions (H2:1); (H4:1); (H4:2) and (H4:3) hold. Then, for each continuously
di¤erentiable function ' 2 B such that
'0 2 B; G(0; ') 2 D(A); D' G(0; ')'0 + Dt G(0; ') 2 D(A)
and D' G(0; ')'0 + Dt G(0; ') = AG(0; ') + F (0; ');
the integral solution asserted by Theorem 4.2.2, is a strict solution of Equation (4.3).
Proof. Let a > 0: Then, by Theorem 4.2.2, we know that Equation (4.3) has a unique
integral solution x := x(:; ') which is given by
Z
d t
G(t; xt ) = S 0 (t)G(0; ') + S(t ¡ s)F (s; xs )ds; for t 2 [0; a] : (4.6)
dt 0
By Lemma 4.2.1, it’s enough to show that x is continuously di¤erentiable on [0; a].
From Corollary 1.2.2, the assumption G(0; ') 2 D(A) implies that
S 0 (t)G(0; ') = S(t)AG(0; ') + G(0; '); and Equation (4.6) can be written as
Z
d t
G(t; xt ) = G(0; ') + S(t)AG(0; ') + S(t ¡ s)F (s; xs )ds: (4.7)
dt 0
Consider the following equation
8
> @
>
> [D' G(t; xt )yt + Dt G(t; xt )] = A [D' G(t; xt )yt + Dt G(t; xt )]
>
< @t
>
+Dt F (t; xt ) + D' F (t; xt ) yt ; t 2 [0; a] ; (4.8)
>
>
>
>
>
: y = '0 :
0
75
Assumptions (H4:1) and (H4:2) imply respectively that,
kD' G(t; Ã)k · ®0 and kD' F (t; Ã)k · ¯ 0 for all à 2 B and t ¸ 0:
Then, using the same reasoning as in the proof of Theorem 4.2.2, one can show that
Equation (4.8) has a unique integral solution y on (¡1; a] given by
8
>
>
> D' G(t; xt )yt = ¡Dt G(t; xt ) + S 0 (t)(D' G(0; ')'0 + Dt G(0; '))
>
> Z
>
< d t
+ S(t ¡ s) (Dt F (s; xs ) + D' F (s; xs ) ys ) ds; t 2 [0; a] ;
dt 0 (4.9)
>
>
>
>
>
>
: y(t) = '0 (t); ¡ 1 < t · 0:
Zt
wt = ' + ys ds; for t 2 [0; a] : (4.10)
0
Next, we show that x = w on (¡1; a] : As in the proof of the preceding theorem, we proceed
by steps, that is, we …rst take a small enough such that ®0 Ka < 1; then we use the same
arguments to see the similar result on [a; 2a] ; :::; [na; (n + 1)a] for any n ¸ 2:
Using (4.9) and the expressions satis…ed by ', we obtain for 0 · t · a
Z t Z t
D' G(s; xs )ys ds = ¡ Dt G(s; xs ) ds + S (t) (AG(0; ') + F (0; '))
0 Z t0 (4.11)
+ S (t ¡ s) (Dt F (s; xs ) + D' F (s; xs ) ys ) ds:
0
On the other hand, from (4.10), the function t 7! wt is continuously di¤erentiable. It follows
that for t 2 [0; a] ; one has
Z
d t
S (t ¡ s) F (s; ws ) ds = S (t) F (0; ')
dt 0 Z t
+ S (t ¡ s) (Dt F (s; ws ) + D' F (s; ws ) ys ) ds:
0
76
Consequently, for t 2 [0; a] ;
Z t
d
S (t) F (0; ') = S (t ¡ s) F (s; ws ) ds
dtZ 0 (4.12)
t
¡ S (t ¡ s) (Dt F (s; ws ) + D' F (s; ws ) ys ) ds:
0
Zt
d
z1 (t) = G(0; ') + S(t)AG(0; ') + S (t ¡ s) F (s; xs ) ds; (4.13)
dt
0
and Z t
d
z2 (t) = (G(s; ws ))ds + G(0; ');
Z 0 ds
t
= (Dt G(s; ws ) + D' G(s; ws )ys ) ds + G(0; '):
0
Then, we infer from expression (4.11)
Z t
z2 (t) = (Dt G(s; ws ) ¡ Dt G(s; xs )) ds
Z0 t
+ (D' G(s; ws ) ¡ D' G(s; xs )) ys ds
0
+G(0; ') + S (t) (AG(0; ') + F (0; '))
Z t
+ S (t ¡ s) (Dt F (s; xs ) + D' F (s; xs ) ys ) ds:
0
Therefore,
Z t
d
z1 (t) ¡ z2 (t) = S (t ¡ s) F (s; xs ) ds ¡ S (t) F (0; ')
dt
Z 0
t
¡ (Dt G(s; ws ) ¡ Dt G(s; xs )) ds
Z 0
t
¡ (D' G(s; ws ) ¡ D' G(s; xs )) ys ds
Z 0
t
¡ S (t ¡ s) (Dt F (s; xs ) + D' F (s; xs ) ys ) ds:
0
77
Expression (4.12) yields
Z t
d
z1 (t) ¡ z2 (t) = S (t ¡ s) (F (s; xs ) ¡ F (s; ws )) ds
dt
Z 0
t
¡ (Dt G(s; ws ) ¡ Dt G(s; xs )) ds
Z 0
t
¡ (D' G(s; ws ) ¡ D' G(s; xs )) ys ds
Z 0
t
+ S (t ¡ s) (Dt F (s; ws ) ¡ Dt F (s; xs )) ds
Z 0
t
+ S (t ¡ s) (D' F (s; ws ) ¡ D' F (s; xs )) ys ds:
0
Consequently, we deduce the existence of a positive constant ¾(a) such that
Zt
jz1 (t) ¡ z2 (t)j · ¾(a) kxs ¡ ws kB ds for t 2 [0; a] :
0
and
Zt
jx(t) ¡ w(t)j · ®0 kxt ¡ wt kB + ¾(a) kxs ¡ ws kB ds
0
Zt
· ®0 Ka sup jx(s) ¡ w(s)j + ¾(a) kxs ¡ ws kB ds:
0·s·t
0
Therefore, one has
Consequently, x(t) = w(t) for all t 2 (¡1; a] ; a = a0 . Repeating the same procedure in
[a; 2a] ; :::; [na; (n + 1)a] ; n ¸ 2, we deduce that x(t) = w(t) for all t 2 (¡1; +1) and x
is continuously di¤erentiable on (¡1; +1). Finally, by Lemma 4.2.1, we get that x is a
strict solution. Hence, the proof of Theorem 4.2.3 is complete.
78
4.3 The solution semigroup in autonomous case
In this section, we suppose that F and G are autonomous in t. In that case, Equation (4.3)
becomes 8
< d [x(t) ¡ G(x )] = A [x(t) ¡ G(x )] + F (x ) ; t ¸ 0;
t t t
dt (4.14)
: x = ' 2 B;
0
where F and G are Lipschitz continuous on B. We verify that the integral solutions of
Equation (4.14) satisfy the properties of a nonlinear strongly continuous semigroup on a
subset Y of B. We also prove that this semigroup satis…es the translation property and a
Lipschitz property.
Let G : B ! E be the mapping de…ned by G' = '(0) ¡ G' and Y be the set de…ned by
n o
Y = ' 2 B : G(') 2 D(A) :
where x(:; ') is the integral solution of Equation (4.14). From Lemma 4.2.1, it follows that
Proposition 4.3.1 Assume that the conditions (H2:1); (H4:1) and (H4:2) hold. Then
(U(t))t¸0 is a nonlinear strongly continuous semigroup on Y, that is,
(i) U(0) = I,
(ii) U(t + s) = U(t)U(s) for all t; s ¸ 0;
(iii) for all ' 2 Y, U(t)(') is a continuous function of t ¸ 0 with values in Y.
Moreover,
(iv) for all t ¸ 0, U(t) is continuous from Y into Y.
(v) (U(t))t¸0 satis…es for t ¸ 0 and µ 2 (¡1; 0] the translation property
(
(U(t + µ)(')) (0) if t + µ ¸ 0;
(U(t)(')) (µ) =
'(t + µ) if t + µ · 0;
(vi) for all T > 0, there are two functions m; n 2 L1 ([0; T ] ; IR+ ) such that
for all '1 ; '2 2 Y;
79
Proof. The proofs of (i) and (v) are straightforward. (iii) follows from Axiom (A1) and
the uniqueness of the integral solutions to Equation (4.14). To prove (vi); without loss of
generality we suppose that ! > 0: Let x1 := x(:; '1 ) and x2 := x(:; '2 ): Choose " > 0. For
t 2 [0; "]
° °
kU(t)'1 ¡ U(t)'2 kB = °x1t ¡ x2t °B ;
¯ ¯
· K(t) sup ¯x1 (s) ¡ x2 (s)¯ + M (t) k'1 ¡ '2 kB ;
0·s·t
©¯ ¯ ª
· K" sup ¯G(x1s ) ¡ G(x2s )¯ + jS 0 (s)(G('1 ) ¡ G('2 ))j
0·s·t ¯ Z s ¯
¯d ¡ ¢ ¯
+K" sup ¯ ¯ S(s ¡ ¾) F (x¾ ) ¡ F (x¾ ) d¾¯¯
1 2
0·s·t ds 0
+M" k'1 ¡ '2 kB ;
Then
° ° £ ¤
kU(t)'1 ¡ U(t)'2 kB · ®0 K" sup °x1s ¡ x2s °B + K" Me
¹ !t (H + ®0 ) + M" k'1 ¡ '2 k
B
0·s·t ¯ Z s ¯
¯d ¡ ¢ ¯
+K" sup ¯ ¯ S(s ¡ ¾) F (x¾ ) ¡ F (x¾ ) d¾ ¯¯
1 2
0·s·t ds 0
Choose " > 0 such that 1 ¡ K" ®0 > 0; then for t 2 [0; "]
° ° ©
sup °x1s ¡ x2s °B · (1 ¡ ®0 K" )¡1 [K" Me
¹ !" (H + ®0 )
0·s·t ¾
Z t ° 1 °
¹ 2
+M" ] k'1 ¡ '2 kB + K" M e ¯ 0!" ° 2 °
sup xs ¡ xs B d¾ :
0 0·s·¾
80
where
© ª
¹ !" (H + ®0 ) + M" ] exp (1 ¡ ®0 K" )¡1 K" M
À 0 (") = (1 ¡ ®0 K" )¡1 [K" Me ¹ 2 e!" ¯ 0 " :
Let L = F 0 (0); then the linearized equation of (4.17) around 0 is of the following form:
8
< d (Dx ) = ADx + Lx ; t ¸ 0;
t t t
dt (4.18)
: x = ' 2 B;
0
associated, respectively, to the equations (4.17) and (4.18). We have the following result :
Theorem 4.4.1 Assume that (H2:1); (H4:2); (H4:4) and (H4:5) hold, then for every
t ¸ 0, the Fréchet-derivative at zero of U(t) is T (t).
81
Lemma 4.4.2 Assume that (H4:4) holds. Let ' 2 B and g : [0; +1) ! E be a continuous
function. Suppose that there exists a function x : (¡1; +1) ! E which is continuous on
[0; +1) and satis…es (
Dxt = g(t); t ¸ 0;
x0 = 0:
Then, for each T > 0; there exists a function b 2 L1 ([0; T ] ; IR+ ) such that
Proof of the lemma. Let " > 0 small enough: Then, for t 2 [0; "]
with
K"
b0 (") = .
1 ¡ K" kD0 k
Similarly, by Axiom (A ¡ iii); for t 2 ["; 2"]
where
K" + M" b0 (")
b1 (") =
1 ¡ K" kD0 k
M" b20 (")
= b0 (") + :
K"
82
Using the same argument, we can see that for t 2 [2"; 3"]
and inductively, for t 2 [n"; (n + 1)"] for any integer n such that (n + 1)" · T;
n bp0 (") p
bn (") = b0 (") § M :
p=0 K"p "
Then, we obtain the inequality (4.19) for any T > 0. This completes the proof of the
lemma.
Proof of Theorem 4.4.1. It su¢ces to show that for each t > 0 and " > 0; there exists
± > 0 such that
kU(t)(') ¡ T (t)'kB · " k'kB for all k'kB · ±:
We have
Z
d t
D [U(t)(') ¡ T (t)'] = S(t ¡ s) (F (U(s)(')) ¡ F (T (s)')) ds
dt Z0
d t
+ S(t ¡ s) [F (T (s)') ¡ L (T (s)')] ds:
dt 0
and Z t
d
g(t) = S(t ¡ s) [F (T (s)') ¡ L (T (s)')] ds
dt 0Z
t
d
+ S(t ¡ s) (F (U(s)(')) ¡ F (T (s)')) ds:
dt 0
Then (
D [vt ¡ wt ] = g(t) for t ¸ 0;
v0 ¡ w0 = 0:
83
Using Lemma 4.4.2, we obtain
By virtue of the di¤erentiability of F at 0 and (vi) of Proposition 4.3.1, we deduce that for
" > 0; there exists ± > 0 such that
Z t
e¡!s jF (T (s)') ¡ L (T (s)') dsj · " k'kB ; for k'kB · ±:
0
Consequently,
µ Z t ¶
¹ !t " k'k + ¯ 0
jg(t)j · Me e¡!s kU(s)(') ¡ T (s)'kB ds
B
0
kU(t)(') ¡ T (t)'kB ·
µ Z t ¶
¹
b(t)Me !t " k'kB + ¯ 0 e¡!s kU(s)(') ¡ T (s)'kB ds :
0
We conclude that U(t) is di¤erentiable at 0 and D' U(t)(0) = T (t) for each t ¸ 0:
Basing on Proposition 4.3.1, Theorem 4.4.1 and Theorem 3.4.4, we obtain a principle
of linearized stability to Equation (4.17) as follows.
Theorem 4.4.3 Under the same assumptions as in Theorem 4.4.1, if the zero equilibrium
of (T (t))t¸0 is exponentially stable, then the zero equilibrium of (U(t))t¸0 is locally expo-
nentially stable in the sense that there exist ± > 0, ¹ > 0 and k ¸ 1 such that
84
Chapter 5
5.1 Introduction
The goal of this chapter is to discuss the existence of periodic solutions to the equation
with in…nite delay considered in Chapters 2 and 3 :
8
< d x(t) = Ax(t) + F (t; x ); t ¸ 0;
t
dt (5.1)
: x = Á 2 B;
0
with the same notations as before. Hereafter, the function F : [0; +1) £ B ! E is contin-
uous and p-periodic in t; for some constant p > 0.
To our knowledge, existence of periodic solutions to Equation (5.1) with A being densely
de…ned has been investigated by Henriquez [58]-[59], Hino, Murakami and Yoshizawa [67]
and recently, Shin and Naito [114]. The authors discussed the problem of existence of
solutions de…ned in the sense of classical semigroups’ theory. In [58], existence of periodic
solutions was obtained by application of the Sadovskii …xed point theorem on the well-
1
This chapter is based on a paper in collaboration with R. Benkhalti and K. Ezzinbi. The paper will
appear in Journal of Mathematical Analysis and Applications, (2001).
85
known Poincaré map. In [59], a method of approximation of Equation (5.1) by a family of
similar equations with …nite delays was introduced. From which some results on existence of
periodic solutions and stability were deduced. In [67], assuming the existence of a bounded
solution which is BC-uniformly stable, the existence of an almost periodic solution was
established under the condition that A generates a compact C0 -semigroup. In [114], Shin
and Naito established the existence of periodic solutions to the nonhomogeneous linear
case of Equation (5.1). In some of their results, the linear part was assumed to generate a
compact C0 -semigroup. The authors obtained several criteria on the existence of periodic
solutions. The followed method is based on the perturbation theory of semi-Fredholm
operators and a …xed point theorem for a¢ne maps which has been obtained by Chow and
Hale [33].
In Chapters 2 and 3, we have considered Equation (5.1) with A being a not necessarily
densely de…ned operator satisfying the Hille-Yosida condition. Precisely, since D(A) is
not densely de…ned, we have addressed the problem of existence, uniqueness, regularity,
existence of a global attractor and local stability... by means of the integrated semigroups
theory.
It is our aim in this chapter to investigate the existence of periodic solutions for Equation
(5.1) with A satis…es the Hille-Yosida condition without being densely de…ned. Mainly,
we recuperate the announced results in [58]. Then, using a key property of the solution
operator, we prove the existence of a periodic solution under slightly di¤erent conditions.
Finally, in the case where B is a (uniform) fading memory space, we exhibit a Massera type
criterion for the nonhomogeneous linear case. Our main results in this chapter would be
considered as some extensions of [33], [58] and [114].
Note that, in the previous chapters, we have been considering a variation of constants
formula in the sense of integrated semigroups’ theory. Here, we use the formula given by
Thieme [117] to see that the two formulas are equivalent.
In Section 5.2, we establish some results on existence of periodic solutions in the non-
linear case. However, in Section 5.3, we deal with the nonhomogeneous linear case , we
prove the Massera criterion : the existence of a bounded solution implies the existence of a
periodic solution. The remaining section is devoted to an application of our abstract results
on a partial functional di¤erential equation with in…nite delay.
86
5.2 Existence of periodic solutions in nonlinear case
Suppose that (H2:1) is veri…ed and that (B; k:kB ) is a seminormed abstract linear space of
functions mapping (¡1; 0] into E; which satis…es the fundamental axioms (A); (A1) and
(B) introduced in Chapter 1 and used in the other previous chapters.
De…ne the part A0 of A in D(A) by
8 n o
< D(A ) = x 2 D(A) : Ax 2 D(A) ;
0
: A0 x = Ax for x 2 D(A0 )
It is known from [117] and Chapters 2 and 3 that for Á 2 B such that Á(0) 2 D(A), if
the associated integral solution x(:; Á) exists, then it is given by the following formula :
8 Z t
< T (t)Á(0)+ lim T0 (t ¡ s)B¸ F (s; xs (:; Á))ds; for t ¸ 0;
0
x(t; Á) = ¸!+1 0
:
Á(t); for t 2 (¡1; 0] ;
where B¸ = ¸R (¸; A) :
Set n o
X := Á 2 B : Á(0) 2 D(A) ;
and
Proposition 5.2.2 Assume that (H2:1) holds. If Á 2 E such that xp (:; Á) = Á then x(:; Á)
is a p-periodic solution of Equation (5.1).
Proof. Let x(:) := x(:; Á) and y(:) := x(: + p): Then for t ¸ 0
t+p
Z
y(t) = T0 (t + p)Á(0)+ lim T0 (t + p ¡ s)B¸ F (s; xs )ds;
¸!+1
0
Zp
y(t) = T0 (t)T0 (p)Á(0)+ lim T0 (t + p ¡ s)B¸ F (s; xs )ds
¸!+1
0
87
t+p
Z
+ lim T0 (t + p ¡ s)B¸ F (s; xs )ds
¸!+1
p
2 3
Zp
= T0 (t) 4T0 (p)Á(0)+ lim T0 (p ¡ s)B¸ F (s; xs )ds5
¸!+1
0
Zt
+ lim T0 (t ¡ s)B¸ F (s + p; xs+p )ds
¸!+1
0
Zt
= T0 (t)x(p)+ lim T0 (t ¡ s)B¸ F (s; ys )ds
¸!+1
0
Zt
= T0 (t)y(0)+ lim T0 (t ¡ s)B¸ F (s; ys )ds:
¸!+1
0
Since y0 = xp = Á; by the uniqueness property of x(:; Á) we have y(t) = x(t) for all t 2 IR+ :
Which proves our claim.
De…ne the Poincaré map Pp : E ! X by Pp (Á) := xp (:; Á), where x(:; Á) is the integral
solution of Equation (5.1). By Proposition 5.2.2, it su¢ces to prove that Pp has a …xed
point in order to deduce existence of a p-periodic solution to Equation (5.1). To this end,
we study the continuity of Pp .
We make the following hypothesis.
(H5:1) The semigroup (T0 (t))t¸0 is compact on D(A); this means, T0 (t) is compact in D(A)
whenever t > 0:
Proposition 5.2.3 Assume that (H2:1) holds. Then each of the following conditions im-
plies the continuity of Pp :
(i) (H2:1) and (H2:4) are satis…ed.
(ii) (H2:1); (H2:3) and (H5:1) are satis…ed and for each à 2 E, there exists rà > 0 such
that the set fxs (:; Á) : s 2 [0; p]; Á 2 B(Ã; rà )g is bounded, where
B(Ã; rà ) := fÁ 2 E : kÁ ¡ ÃkB · rà g.
88
© ª
compact in C([0; p] ; E): Using Axiom (A ¡ (ii)0 ); we can see that Ãk (0) : k 2 IN is rela-
° ° © ª
tively compact in E: Suppose that °Ãk ¡ à ° · rà . It is clear that T0 (t)à k (0) : k 2 IN
B
is relatively compact in E. Let 0 < t · p and 0 < " < t: Then
Z t Z t¡"
k
lim T0 (t ¡ s)B¸ F (s; xs )ds = T0 (") lim T0 (t ¡ " ¡ s)B¸ F (s; xks )ds
¸!+1 0 ¸!+1 0
Z t
+ lim T0 (t ¡ s)B¸ F (s; xks )ds:
¸!+1 t¡"
Since T0 (") is compact, there exists a compact set W" such that
½ µ Z t¡" ¶ ¾
k
T0 (") lim T0 (t ¡ " ¡ s)B¸ F (s; xs )ds : k 2 IN µ W" :
¸!+1 0
The semigroup (T0 (t))t¸0 is compact for t > 0, it follows from [97] that
89
¯ ¯
Which implies that lim ¯xk (t) ¡ xk (t0 )¯ = 0; uniformly in k 2 IN: Using a similar argument
t!t0
t>t0
¯ ¯
for 0 · t < t0 · p, we prove that lim ¯xk (t) ¡ xk (t0 )¯ = 0; uniformly in k 2 IN: It
t!t0
t<t0
¡ ¢
follows that xk (:) k2IN is equicontinuous. By Arzéla-Ascoli’s theorem, we deduce that
© k ª
x (:) : k 2 IN is relatively compact in C([0; p] ; E). Therefore, if ´n is a subsequence of
Ãk ; there exists a subsequence x(:; ´ nk ) of x(:; ´ n ) which converges to a certain function
u(:) 2 C([0; p] ; E) with u(0) = Ã(0): Set u(µ) = Ã(µ) for µ < 0, using Axiom (A ¡ iii), we
infer that for all s 2 [0; p] ; xs (:; ´nk ) ! us as k ! 1: Since fxs (:; ´ nk ) : 0 · s · p; k 2 INg
is a bounded set of B and F takes bounded sets into bounded sets, then we have
Z t
u(t) = T0 (t)Ã(0)+ lim T0 (t ¡ s)B¸ F (s; us (:; Ã))ds; t ¸ 0:
¸!+1 0
Theorem 5.2.4 Assume that (H2:1); (H2:3); (H5:1) and (H5:2) hold. If
In order to prove this theorem, we recall the following …xed point theorem.
Theorem 5.2.5 (Sadovskii’s theorem [110]) Let Z be a Banach space, D0 a convex closed
and bounded set of Z and P : D0 ! D0 an ®-condensing map, that is P is continuous and
for every bounded set D µ D0 with ®(D) > 0; where ®(:) is the Kuratowski’s measure of
noncompactness,
®(P (D)) < ®(D):
Proof of Theorem 5.2.4. It is clear that the condition (ii) in Proposition 5.2.3 is
veri…ed for E = D. Then the map Pp (Á) = xp (:; Á) is continuous from D to D. Moreover,
90
b!D
there exists an induced application P^p : D b that satis…es the condition P^p (^ [
') = P p (')
for all ' b
^ 2 D and every ' 2 ' ^
^ : We shall prove that Pp is condensing. Since for all C µ D; b
there exists V µ D such that C = Vb and ® (C) = ® (V) ; it su¢ces to estimate ®(P^p (G))
b only
for each subset G µ D with ® (G) > 0 . If we denote G [¾; p] ; 0 · ¾ · p; the set de…ned by
© ª
G [¾; p] := x(:; ')=[¾;p] : ' 2 G ;
then, using (H5:2) and Axiom (A ¡ ii); we can see that the conditions of Theorem 2.1 in
[112] are veri…ed. Hence
Let d > 0 such that ®(G(0)) < d: By de…nition of ®(:); G(0) has a …nite cover (Ci )1·i·n µ E;
such that diam(Ci ) < d: Henceforth, for a subset C of E; we denote by C ¤ the set
© ª
C ¤ = T0 (:)u=[0;¾] : u 2 C ;
91
and
n
G(0)¤ µ [ Ci¤ :
i=1
From this together with (5.7) and the choice d := H®(G) + "; " > 0; we infer that
Since n o
G [0; ¾] µ G(0)¤ + w(:; ')j[0;¾] : ' 2 G ;
Z t
where w(t; ') = lim T0 (t¡s)B¸ F (s; xs (:; '))ds; for t 2 [0; ¾] : Using similar arguments
¸!+1 0 n o
as in the proof of Proposition 5.2.3, we can see that w(:; ')j[0;¾] : ' 2 G is relatively
compact in C([0; ¾] ; E). In fact, let 0 < t · ¾ and 0 < " < t: Then
Z t
lim T0 (t ¡ s)B¸ F (s; xs (:; '))ds
¸!+1 0
Z t¡"
= T0 (") lim T0 (t ¡ " ¡ s)B¸ F (s; xs (:; '))ds
¸!+1 0
Z t
+ lim T0 (t ¡ s)B¸ F (s; xs (:; '))ds:
¸!+1 t¡"
Since T0 (") is compact, there exists a compact set W" such that
½ µ Z t¡" ¶ ¾
T0 (") lim T0 (t ¡ " ¡ s)B¸ F (s; xs (:; '))ds : ' 2 G µ W" :
¸!+1 0
92
There exists a positive constant b such that
¯ ¯
¯ Zt ¯
¯ ¯
¯ lim ¯
¯¸!+1 T0 (t ¡ s)B¸ F (s; xs (:; '))ds¯ · b(t ¡ t0 ); uniformly in ' 2 G:
¯ ¯
t0
½ Z t0 ¾
Moreover, W0 = lim T0 (t0 ¡ s)B¸ F (s; xs (:; '))ds : ' 2 G is relatively compact
¸!+1 0
and it’s well known that
Which implies that lim jw(t; ') ¡ w(t0 ; ')j = 0; uniformly in ' 2 G: Using a similar
t!t0
t>t0
argument for 0 · t < t0 · ¾, we prove that lim jw(t; ') ¡ w(t0 ; ')j = 0; uniformly in
t!t0
t<t0
n o
' 2 G: It follows that w(:; ')j[0;¾] : ' 2 G is equicontinuous. By Arzéla-Ascoli’s theorem,
n o
we deduce that w(:; ')j[0;¾] : ' 2 G is relatively compact in C([0; ¾] ; E). We complete
the proof of (5.6) by the properties of ®(:).
Thus the inequalities (5.4), (5.5) and (5.6) prove that for all 0 < ¾ · p;
(Ppn ')n is a Cauchy sequence of B. Which implies that there exists à 2 D such that
Ppk ' ! Ã; as k ! +1: Consequently Pp à = Ã: This completes the proof by use of
Proposition 5.2.2.
Remark 5.2.1 Assumption (H5:2) seems to be strong, but in nonhomogeneous linear case,
we will prove that the existence of a bounded solution implies that (H5:2) is true.
93
where x(:; Á) is the integral solution of Equation (5.1).
The following result on the solution operator U(t); t ¸ 0; is a key property to the
existence of periodic solutions.
Proposition 5.2.6 Assume that (H2:1) ; (H2:4) and (H5:1) hold. Then U(t) is decom-
posed as follows
U(t) = U1 (t) + U2 (t); t ¸ 0;
In order to prove that for all µ 2 (¡1; 0] ; f(U2 (t)Ã n )(µ) : n ¸ 0g is relatively compact in
E. Let 0 < t + µ and choose " such that " < t + µ: Then
Z t+µ
(U2 (t)Ã n )(µ) = lim T0 (t + µ ¡ s)B¸ F (s; xs (:; Ã n ))ds
¸!+1 0
Z t+µ¡"
= T0 (") lim T0 (t + µ ¡ " ¡ s)B¸ F (s; xs (:; Ã n ))ds
¸!+1 0
Z t+µ
+ lim T0 (t + µ ¡ s)B¸ F (s; xs (:; Ã n ))ds:
¸!+1 t+µ¡"
94
Which gives that f(U2 (t)Ãn )(µ) : n ¸ 0g is totally bounded and therefore relatively compact:
To establish the second assertion, it is su¢cient to show that (U2 (t)Ã n )n¸0 is equicontinuous
in (¡1; 0] : Let µ 0 2 (¡1; 0] : For µ 2 (¡1; 0] enough close to µ 0 such that µ0 < µ; we see
that
8
>
< (U2 (t + µ)Ã n ) (0) ¡ (U2 (t + µ0 )Ãn ) (0) if µ 0 > ¡t
>
(U2 (t)Ã n )(µ) ¡ (U2 (t)Ã n )(µ 0 ) = (U2 (t + µ)Ã n ) (0) if µ0 = ¡t
>
>
: 0 if µ < ¡t:
0
95
we get kÁn ¡ ÁkB ! 0 as n ! 1 because of kÁn ¡ ÁkB · K(t) sup jÁn (µ) ¡ Á (µ)j by
¡t·µ·0
Axiom (A ¡ iii): In de…nitive, we have proved that the image of any bounded sequence
contains a converging subsequence in B with respect to the seminorm.
We need to suppose that
(H5:3) There exists a nonempty closed bounded and convex subset D in X such that
Pp (D) µ D:
Theorem 5.2.7 Assume that (H2:1); (H2:4) ; (H5:1) and (H5:3) hold. If there exists a
positive constant ´ such that
kU1 (t)k · e¡´t ; t ¸ 0: (5.9)
Proof. As in the proof of Theorem 5.2.4, we use Sadovskii’s …xed point theorem. Since
U2 (p) is compact, for each subset G µ D with ® (G) > 0; one has
cp (G))
®(P b = ®(U(p)
b G) b = ®(U
c1 (p)G)
b
b
· e¡´p ®(G)
b
< ®(G):
cp is ®-condensing on D,
From what we deduce that P b and Equation (5.1) has a p-periodic
solution.
In relation with the condition (5.9), we recall the following result.
Lemma 5.2.8 ( Chapters 1 and 3) The space C° ; ° > 0 de…ned in the previous chapters
with the norm kÁk° :=sup e°µ jÁ(µ)j ; Á 2 C° , satis…es Axioms (A), (A1); and (B). More-
µ·0
over, if kT0 (t)k · e¡±t ; t ¸ 0; where ± is a positive constant. Then Condition (5.9) is
true.
In the remaining of this section, we give some su¢cient conditions ensuring assumption
(H5:2) and, in particular, (H5:3). Consider the phase space Cr £ L1 (g); r > 0; de…ned in
Chapter 1, with the seminorm
Z
¡r
96
From Chapter 1, if we suppose that g satis…es :
(i) g is integrable on (¡d; ¡r) for any d ¸ r; and
(ii) there exists a locally bounded function G : (¡1; 0] ! [0; +1) such that
Proposition 5.2.9 Assuming that F : [0; +1) £ (Cr £ L1 (g)) ! E is continuous and
locally Lipschitzian with respect to the second variable and there exist N1 ; N2 > 0
If there exists !0 > 0 and k0 > 0 such that the following conditions hold :
(i) kT (t)k · exp(¡!0 t); t ¸ 0;
Z 0t¡r
(ii) g(s ¡ t) exp(! 0 (t ¡ r ¡ s))ds · k0 for t ¸ r;
0
(iii) N1 exp(! 0 r)(1 + k0 ) < ! 0 :
Then there exists R > 0 and ¾ > 0 such that P¾ (B(0; R)) µ B(0; R) and
the set fxt (:; Á) : Á 2 B(0; R); 0 · t · ¾g is bounded in Cr £ L1 (g):
The proof of this proposition is omitted here, it can be done similarly as in [58].
where L is a continuous function form IR+ £ B into E, linear with the second argument
and p-periodic in t; f is a continuous p-periodic function.
97
For ' 2 B, t ¸ 0 and µ · 0; we de…ne
(
'(0) if t + µ ¸ 0
[W (t)'] (µ) =
'(t + µ) if t + µ < 0:
Let C00 be the set of continuous functions ' : (¡1; 0] ! E with compact support and
recallÃthat a sequence!of functions ('n )n2IN : (¡1; 0] ! E; is uniformly bounded if
Theorem 5.3.1 Assume that (H2:1); (H5:1); (H5:4); (C) and (D2) are satis…ed. If
Proof. Set D := co fynp : n 2 INg where co denotes the closure of the convex hull: Then
D is a nonempty closed convex subset of X . D is bounded and xp (:; Ã) 2 D for all à 2 D.
In fact, let à 2 D; then there exists a sequence (à k )k ½ fynp : n 2 INg such that
nk nk ° °
° °
à k = § ®ki ynk p ; ®ki ¸ 0; § ®ki = 1 and °Ã ¡ Ãk ° ! 0 as k ! 1.
i=1 i i=1 B
98
Therefore, we get
° ° ° °
° ° ° °
kÃkB · °Ãk ° + °Ãk ¡ Ã °
B
nk ° ° °B °
k° ° ° k °
· § ®i °ynk p ° + °Ã ¡ ð ;
i=1 i B B
and (5.12) yields kÃkB < N1 . On the other hand, one has
nk
xp (:; Ãk ) = § ®ki xp (:; ynki p )
i=1
nk
= § ®ki ynki p+p ;
i=1
° °
then xp (:; Ãk ) 2 D: Since °xp (:; Ã) ¡ xp (:; Ãk )°B ! 0 as k ! +1; and D is closed we
deduce that xp (:; Ã) 2 D for all à 2 D: Which implies that assumption (H5:3) is true and
by Theorem 5.2.5, we conclude that Equation (5.11) has a p-periodic solution.
Corollary 5.3.2 Assume that B = C° ; ° > 0. Under the conditions (H2:1); (H5:1) and
(H5:4); Equation (5.11) has a mp-periodic solution, for a certain m 2 IN ¤ .
Proof. It was noted in Chapter 1 that if ° > 0; the phase space C° satis…es (C) and (D2):
Moreover one has K(¾) = 1 and M(¾) = e¡°¾ : Hence
½ ¾
inf M(p ¡ ¾)[HK(¾) sup kT0 (t)k + M(¾)]
0<¾<p 0·t·¾
½ ¾
= inf H sup kT0 (t)k e¡°(p¡¾) + e¡°p :
0<¾<p 0·t·¾
From what we can see that for a certain m 2 IN ¤ , (5.13) in Theorem 5.3.1 is satis…ed with
mp instead of p. Consequently, all conditions of Theorem 5.3.1 are veri…ed and Equation
(5.11) has a mp-periodic solution.
Next, we suppose that B is a normed uniform fading memory space, that means, B is
normed and satis…es the extra axioms (C) and
(D3) kW0 (t)k ! 0 as t ! +1:
Theorem 5.3.3 Suppose that B is a normed uniform fading memory space. Under condi-
tions (H2:1); (H5:1) and (H5:4); Equation (5.11) has a p-periodic solution.
99
Lemma 5.3.4 Let conditions (H2:1) and (H5:1) hold. Then, for any " > 0; there exists
a constant C" > 0 such that ®(U1 (t)) · C" M(t ¡ ") for t > ":
Proof. Let - be a bounded set in X such that ® (-) = ¯: For ± > 0; there exists a …nite
cover (-i )1·i·n of - such that diam(-i ) < ¯ + ±: Using Axiom (A ¡ iii), we have for t > "
and '; Á 2 -;
kU1 (t)' ¡ U1 (t)ÁkB · K(t ¡ ") sup jT0 (s)' (0) ¡ T0 (s)Á (0)j (5.14)
"·s·t
+M(t ¡ ") kU1 (")' ¡ U1 (")ÁkB :
kU1 (")' ¡ U1 (")ÁkB · K(") sup jT0 (s)' (0) ¡ T0 (s)Á (0)j + M(") k' ¡ ÁkB ;
0·s·"
kU1 (t)' ¡ U1 (t)ÁkB · K(t ¡ ") sup jT0 (s)' (0) ¡ T0 (s)Á (0)j + M(t ¡ ")C" k' ¡ ÁkB :
"·s·t
(5.16)
- is bounded in B it follows by Axiom (A¡ (ii)0 ) that fà (0) : à 2 -g is bounded in E. Using
the compactness of the semigroup (T0 (t))t¸0 , we can prove that fT0 (:)Ã (0) : Ã 2 -g is rela-
tively compact in C (["; t] ; E) ; then there exists a …nite cover (¡i )1·i·m of fT0 (:)Ã (0) : Ã 2 -g
in C (["; t] ; E) such that diam(¡i ) < ±: Let -i;j = fà 2 -i : T0 (:)à (0) 2 ¡jg : Then - ½
S
i;j -i;j : For '; Á 2 -i;j by (5.16) we have
Theorem 5.3.5 [33] Let Z be a Banach space and P : Z ! Z a linear a¢ne map, that is
P x = Lx + y where L is a bounded linear map and y 2 Z is …xed. If Im(I ¡ L) is closed
and there exists an x0 2 Z such that (P n x0 )n¸0 is bounded, then P has at least one …xed
point.
100
Proof of Theorem 5.3.3. Let x(:; '; f ) be the solution of Equation (5.11). Then
x(:; '; f) = x(:; '; 0) + x(:; 0; f ); where x(:; '; 0) and x(:; 0; f) are, respectively, the solutions
of Equation (5.11) with f = 0 and ' = 0: It follows that the Poincaré map Pp is given by
Pp ' = L' + Ã; with L' = xp (:; '; 0) and à = xp (:; 0; f ). According to Proposition 5.2.6, L
is decomposed as follows L = U1 (p) + U2 (p); where (U1 (t))t¸0 is the solution semigroup of
Equation (5.8) and U2 (p) is compact: By Proposition 1.1.2, K(:) and M(:) can be chosen
such that K(:) is bounded and M(t) ! 0 as t ! +1: Let " > 0; and t0 > 0 such that
C" M(t0 ¡ ") < 1: Then, by Lemma 5.3.4, we have
Consequently we obtain that Im(I ¡ L) is closed: Let y(:; Á; f ) be the bounded solution.
¡ ¢
Then Ppn Á n¸0 = (ynp (:; Á; f))n¸0 is bounded. Consequently, by Theorem 5.3.5, we deduce
that the map Pp has a …xed point in B, which gives a p-periodic solution to Equation (5.11).
5.4 Application
In order to apply the abstract result of the previous section, we consider the following
partial functional di¤erential equation with in…nite delay
8 Z 0
>
> @ @2
>
> w(t; ») = w(t; ») + b(t) G(µ)w(t + µ; »)dµ + g(t; »);
>
> @t @» 2
¡1
<
t ¸ 0; 0 · » · ¼; (5.17)
>
>
>
> w(t; 0) = w(t; ¼) = 0; t ¸ 0;
>
>
: w(µ; ») = w (µ; »); ¡ 1 < µ · 0; 0 · » · ¼;
0
101
where b : IR+ ! IR is a continuous p-periodic function, g : IR+ £ [0; ¼] ! IR is a continuous
function, p-periodic with respect to time, G is a positive function on (¡1; 0] and w0 :
(¡1; 0] £ [0; ¼] ! IR is a continuous function.
Equation (5.17) can be written as the abstract equation (5.11). Indeed, we choose E =
C ([0; ¼] ; IR) endowed with the uniform norm topology and we consider the operator A :
D(A) ½ E ! E de…ned by
( © ª
D(A) = y 2 C 2 ([0; ¼] : IR) : y(0) = y(¼) = 0 ;
Ay = y 00 :
It is known from [66] that the space C° ; ° > 0; which is introduced in Lemma 5.2.8, is a
normed uniform fading memory space.
We suppose that :
(i) G(:)e¡°: is integrable on (¡1; 0], Ã !
(ii) w0 2 C((¡1; 0] £ [0; ¼] : IR), with lim e°µ sup jw0 (µ; »)j exists, and w0 (0; 0) =
µ!¡1 0·»·¼
w0 (0; ¼) = 0:
Let
8 Z 0
>
< L(t; Á)(») = b(t) G(µ)Á(µ)(»)dµ; t ¸ 0; » 2 [0; ¼] ; Á 2 C° ;
¡1
>
: f(t)(») = g(t; »); t ¸ 0; » 2 [0; ¼] :
Then L is continuous from IR+ £ C° into E, linear with respect to the second argument
and p-periodic in t, f : IR+ ! E is continuous p-periodic. If we put
(
x(t)(») = w(t; »); t ¸ 0; » 2 [0; ¼] ;
:
'(µ)(») = w0 (µ; »); µ · 0; » 2 [0; ¼] :
102
We can see that assumption (ii) implies that ' 2 C° and '(0) 2 D(A):
Let A0 be the part of the operator A in D(A) given by
( © ª
D(A0 ) = y 2 C 2 ([0; ¼] : IR) : y(0) = y 00 (0) = y(¼) = y 00 (¼) = 0 ;
A0 y = y 00 :
A0 generates a compact C0 -semigroup (T0 (t))t¸0 in D(A): Moreover kT0 (t)k · e¡t ; t ¸ 0.
Assume further that Z 0
(iii) there exists d 2 ]0; 1[ such that 0 < l G(µ)dµ < 1 ¡ d;
¡1
where l = sup jb(t)j :
0·t·p
m
Let ½ = 1 + ; where m = sup jf(t)j : Then we have
d 0·t·p
Lemma 5.4.1 Under the above assumptions, let ' 2 C° such that j'(µ)j < ½ for all
µ 2 (¡1; 0] : Then the integral solution of Equation (5.18) is bounded by ½ on IR+ :
Proof. Let x (:; ') be the integral solution of Equation (5.18). Then jx (t; ')j < ½ for t ¸ 0.
We proceed by contradiction. Suppose that there exists t0 > 0 such that jx (t0 ; ')j > ½.
Let
t1 = inf ft > 0 : jx (t; ')j > ½g :
By continuity, it follows that
jx (t1 ; ')j = ½; (5.19)
and there exists ± > 0 such that jx (t; ')j > ½ for t 2 (t0 ; t0 + ±) : Since
Zt
x (t; ') = T0 (t)'(0)+ lim T0 (t ¡ s)B¸ (L(s; xs (:; ')) + f(s)) ds; for t ¸ 0:
¸!+1
0
We have
Zt1 Zt1
¡t1 ¡(t1 ¡s)
jx (t1 ; ')j · e ½+ e jL(s; xs (:; '))j ds + e¡(t1 ¡s) mds;
0 0
103
This gives by assumption (iii)
¡ ¢
jx (t1 ; ')j · e¡t1 ½ + (½ ¡ ½d + m) 1 ¡ e¡t1 :
Therefore
¡ ¢
jx (t1 ; ')j · ½ ¡ d 1 ¡ e¡t1
< ½:
Which contradicts (5.19) by de…nition of t1 . This completes the proof of the lemma.
Finally, all conditions of Theorem 5.3.3 are satis…ed, and consequently, Equation (5.18)
has a p-periodic solution.
104
CONCLUSION ET PERSPECTIVES
Dans ce travail, nous avons étudié des aspects quantitatif et qualitatif pour une classe
d’équations di¤érentielles à retard in…ni et de type neutre dans un espace de Banach. Nous
nous sommes intéressés dans un premier lieu aux problèmes d’existence, d’unicité et de
régularité des solutions. Puis, nous avons étudié des problèmes de stabilté et d’attractivité.
En…n, nous avous traité l’existence de solutions périodiques.
An terme de ce travail, nous avons pensé à un certain nombre de problèmes ouverts.
Dans la suite nous donnons quelques questions que nous projetons étudier dans le futur.
Dans le chapitre 3, nous avons étudié l’existence de solutions intégrales dans le cas où F
est continue et localement Lipschitzienne par rapport à la deuxième variable. Nous avons
l’intention de traiter l’existence sous les conditions de Carathéodory: F est mesurable par
rapport à la première variable et continue par rapport à la deuxième variable.
Nous avons aussi formulé un critère de stabilité dans le cas où la part de A dans D(A)
engendre un C0 semi-groupe compact (T0 (t))t¸0 . Nous projetons généraliser le résultat
obtenu dans le cas où (T0 (t))t¸0 n’est pas compact mais uniformément continu, à savoir
lim kT0 (t) ¡ T0 (t0 )kL(D(A)) = 0; pour tout t0 > 0:
t!t0
Nous avons aussi donné des conditions su¢santes pour l’existence d’un attracteur global
pour une classe d’équations di¤érentielles à retard in…ni dissipatives. Nous comptons ex-
ploiter cette propriété pour étudier l’existence et la stabilté des solutions périodiques.
Nous avons étudié, dans le chapitre 4, l’existence de solutions intégrales et strictes pour
une classe d’équations di¤érentielles de type neutre à retard in…ni. Dans le cas autonome,
nous avons démontré que la solution dé…nit un semi-groupe (U(t))t¸0 sur l’espace de phase
B. Auquel cas, nous avons formulé un critère de stabilité. Par la suite, la question que nous
nous posons est la suivante: peut on décomposer le semi groupe (U(t))t¸0 sous la forme
suivante : U(t) = V (t) + W (t); t ¸ 0 où (V (t))t¸0 est un C0 semi-groupe exponentiellement
stable et W (t) est un opérateur compact pour t > 0? Une réponse positive à cette question
nous permettra d’examiner la stabilité du semi groupe solution dans le cas linéaire par
une équation caractéristique. Dans le cas non linéaire, nous étudierons l’existence d’un
attracteur global. Un autre volet auquel nous nous intéressons, est celui de l’existence de
solutions bornés et périodiques.
Dans le chapitre 5, nous avons étudié l’existence de solutions périodiques dans le cas où
F est périodique en t. Nous nous sommes intéressés par l’existence de solutions presque-
périodiques dans le cas où F est presque-périodique en t uniformément par rapport à la
deuxième variable. Nous avons l’intention de traiter une situation intermédiaire, à savoir, le
105
cas où F est quasi-périodique, c’est à dire, F = F1 + F2 , avec F1 et F2 sont respectivement,
p1 -périodique et p2 -périodique en t et satisfaisant la condition de noncommensurabilité,
p1
c’est à dire 2
= IQ:
p2
106
Bibliography
[1] M. Adimy, Abstract semilinear functional di¤erential equations with non-dense do-
main, Publications internes de l’Université de Pau et des Pays de L’Adour, URA
1204 Pau 95/18, (1995).
[2] M. Adimy, Integrated semigroups and delay di¤erential equations, J. Math. Anal.
Appl., Vol. 177, N. 1, 125-134, (1993).
[3] M. Adimy, Semi-groupes intégrés et équations aux dérivées partielles non locales en
temps, Thèse d’Habilitation, Université de Pau et des Pays de L’Adour, (1999).
[4] M. Adimy, H. Bouzahir and K. Ezzinbi, Existence for a class of partial functional
di¤erential equations with in…nite delay, Nonlinear Analysis, Theory, Methods and
Applications, to appear, (2001).
[5] M. Adimy, H. Bouzahir and K. Ezzinbi, Local existence and stability for a class
of partial functional di¤erential equations with in…nite delay, Nonlinear Analysis,
Theory, Methods and Applications, to appear, (2001).
[6] M. Adimy, H. Bouzahir and K. Ezzinbi, Existence and stability for some neutral
functional di¤erential equations with in…nite delay, (in progress).
[7] M. Adimy and K. Ezzinbi, Local existence and linearized stability for partial func-
tional di¤erential equations, Dynamic Systems and Applications, Vol. 7, N. 3, 389-
404, (1998).
107
[10] M. Adimy and K. Ezzinbi, A class of linear partial neutral functional di¤erential
equations with nondense domain, J. Di¤erential Equations, Vol. 147, N. 2, 285-332,
(1998).
[11] M. Adimy and K. Ezzinbi, Strict solutions of nonlinear hyperbolic neutral di¤er-
ential equations, Applied Mathematics Letters., Vol. 12, N. 1, 107-112, (1999).
[12] M. Adimy and K. Ezzinbi, Existence and linearized stability for partial neutral
functional di¤erential equations with nondense domains, Di¤erential Equations and
Dynamical Systems, Vol. 7, 371-417, (1999).
[17] E. Ait Dads and K. Ezzinbi, Pseudo almost periodic solutions for some delay
di¤erential equations, J. Math. Anal. Appl., Vol. 201, N. 3, 840-850, (1996).
[18] W. Arendt, Resolvent positive operators and integrated semigroups, Proc. London
Math. Soc. (3), Vol. 54, N. 2, 321-349, (1987).
[19] W. Arendt, Vector valued Laplace transforms and Cauchy problems, Israelien J.
Math., Vol. 59, N. 3, 327-352, (1987).
108
[21] O. Arino and My. L. Hbid, Existence of periodic solutions for a delay di¤eren-
tial equation via the Poincaré procedure, J. Di¤erential Equations and Dynamical
Systems, Vol. 4, N. 2, 125-148, (1996).
[23] O. Arino and E. Sanchez, Linear theory of abstract functional di¤erential equa-
tions of retarded type, J. Math. Anal. Appl., Vol. 191, N. 3, 547-571, (1995).
[24] O. Arino and E. Sanchez, A variation of constants formula for an abstract func-
tional di¤erential equations of retarded type, Di¤erential and Integral Equations, Vol.
6, N. 9, 1305-1320, (1996).
[25] D. Bainov and P. Simeonov, Integral Inequalities and Applications, Kluwer Aca-
demic Publishers, (1992).
[27] L. L. Bonilla and A. Linan, Relaxation oscillations, pulses, and travelling waves
in the di¤usive Volterra delay-di¤erential equations, SIAM J. Appl. Math., Vol. 44,
N. 2, 369-391, (1984).
[28] H. Bouzahir and K. Ezzinbi, Global attractor for a class of partial functional
di¤erential equations with in…nite delay, in “Topics in Functional Di¤erential and
Di¤erence Equations”, eds. P. Freitas and T. Faria, Fields Inst. Commun., Vol. 29,
Amer. Math. Soc., Providence, RI, in press (2001).
[31] T. A. Burton, Stability and Periodic Solutions of Ordinary and Functional Di¤er-
ential Equations, Academic Press, Orlando, (1985).
109
[32] S. Busenberg and B. Wu, Convergence theorems for integrated semigroups, J.
Di¤erential and Integral Equations, Vol. 5, N. 3, 509-520, (1992).
[35] B. C. Coleman and V. J. Mizel, Norms and semigroups in the theory of fading
memory, Arch. Rational Mech. Anal., Vol. 23, 87-123, (1966).
[36] B. C. Coleman and V. J. Mizel, On the general theory of fading memory, Arch.
Rational Mech. Anal., Vol. 29, 18-31, (1968).
[39] G. Da Prato and A. Lunardi, Solvability on the real line of a class of linear
Volterra integro-di¤erential equations of parabolic type, Annali Mat. Pura Appl. (4),
Vol. 150, 67-117, (1988).
[43] K. Engel and R. Nagel, One Parameter Semigroups for Linear Evolution Equa-
tions, Graduate Texts in Math., Vol. 194, Springer Verlag, (2000).
110
[44] K. Ezzinbi, Contribution à l’étude des équations di¤érentielles en dimension …ne et
in…nie : existence et aspects qualitatifs; application à des problèmes de dynamique de
populations, Thèse d’état, Faculté des sciences Semlalia, Université Cadi Ayyad de
Marrakech, N. d’ordre 167, (1997).
[45] K. Ezzinbi and H. Tamou, Abstract functional di¤erential equations, Dynam. Dis-
crete Continuous and Impulsive Systems, to appear, (2001).
[49] J. Haddock and J. Terjeki, On the location of positive limit sets for autonomous
functional di¤erential equations with in…nite delay, J. Di¤erential Equations, Vol. 86,
N. 1, 1-32, (1990).
[52] J. K. Hale, Dynamical systems and stability, J. Math. Anal. Appl., Vol. 26, 35-59,
(1969).
[53] J. K. Hale, Partial neutral functional di¤erential equations, Rev. Roumaine Math.
Pure Appl., Vol. 39, N. 4, 339-344, (1994).
111
[54] J. K. Hale, Coupled oscillators on a circle, Dynamical phase transitions (São Paulo,
1994), Resenhas, Vol. 1, N. 4, 441-457, (1994).
[55] J. K. Hale, and J. Kato, Phase space for retarded equations with in…nite delay,
Funkcial. Ekvac., Vol. 21, N. 1, 11-41, (1978).
[57] My. L. Hbid, Contribution à l’étude des perturbations des équations di¤érentielles
à retard, Thèse d’état, Faculté des sciences, Université Mohamed V, Rabat, (1993).
[61] E. Hernandez and H. R. Henriquez, Existence results for partial neutral func-
tional di¤erential equations with unbounded delay, J. Math. Anal. Appl., Vol. 221,
N. 2, 452-475, (1998).
[63] E. Hernandez, Regular solutions for partial neutral functional di¤erential equations
with unbounded delay, (preprint).
112
[66] Y. Hino, S. Murakami and T. Naito, Functional Di¤erential Equations with
In…nite Delay, Lecture Notes in Mathematics, Springer-Verlag, Berlin, Vol. 1473,
(1991).
[70] S. Kato, Almost periodic solutions of functional di¤erential equations with in…nite
delays in a Banach space, Hokkaido Math. J., Vol. 23, N. 3, 465-474, (1994).
[71] H. Kellermann and M. Hieber, Integrated semigroups, J. Fun. Anal., Vol. 15,
N. 1, 160-180, (1989).
[72] K. Kunisch and W. Schappacher, Variation of constants formula for partial dif-
ferential equations with delay, Nonlinear Analysis, Theory, Methods and Applications,
Vol. 5, N. 2, 123-142, (1981).
[75] J. Liang and T. Xiao, Functional di¤erential equations with in…nite delay in Ba-
nach spaces, Int. J. Math. Sci., Vol. 14, N. 3, 497-508, (1991).
[76] X. Lin, J. W.-H. So and J. Wu, Centre manifolds for partial functional di¤erential
equations with delays, Proceedings of the Royal Society of Edinburgh, Sect. A 122, N.
3-4, 237-254, (1992).
113
[77] X. Liu and D. Y. Xu, Uniform asymptotic stability of abstract di¤erential equations,
J. Math. Anal. Appl., Vol. 216, N. 2, 626-643, (1997).
[80] M. Memory, Stable and unstable manifolds for partial functional di¤erential equa-
tions, Nonlinear Analysis, Theory, Methods and Applications, Vol. 16, N. 2, 131-142,
(1991).
[82] J. Milota, Stability and saddle point property for a linear autonomous functional
parabolic equations, Comment. Math Univ. Carolia., Vol. 27, N. 1, 87-101, (1986).
[83] J. Milota, Stability in models with long memories, in Biomathematics and Related
Computational Problems, ed. L. M. Ricciardi, Kluwer Academic Publishers, 523-527,
(1988).
[84] S. Murakami, Stability for functional di¤erential equations with in…nite delay in
Banach space, (unpublished).
[85] S. Murakami, Stable equilibrium point of some di¤usive functional di¤erential equa-
tions, Nonlinear Analysis, Theory, Methods and Applications, Vol. 25, N. 9-10, 1037-
1043, (1995).
[86] S. Murakami and T. Naito, Some properties of phase spaces for functional di¤er-
ential equations with in…nite delay, in Di¤erential Equations, Lecture Notes in Pure
and Applied Mathematics, Vol. 118, Marcel Dekker, Inc., 507-514, (1989).
114
[88] S. Murakami and T. Yoshizawa, Asymptotic behavior in a system of Volterra
integrodi¤erential equations with di¤usion, Dynam. Systems Appl., Vol. 3, N. 2,
175-188, (1994).
[92] T. Naito, J. S. Shin and S. Murakami, The generator of the solution semigroup
for the general linear functional-di¤erential equation, Bull. Univ. Electro-Comm., Vol.
11, N. 1, 29–38, (1998).
[94] R. D. Nussbaum, The radius of essential spectrum, Duke Math. J., Vol. 37, 473-478,
(1970).
[95] M. E. Parrott, Positivity and a principle of linearized stability for delay di¤erential
equations, Di¤erential Integral Equations, Vol. 2, N. 2, 170-182, (1989).
[99] H. Petzeltova, Solution semigroup and invariant manifolds for functional di¤eren-
tial equations with in…nite delay, Mathematica Bohemica, Vol. 118, N. 2, 175-193,
(1993).
115
[100] H. Petzeltova, The Hopf bifurcation theorem for parabolic equations with in…nite
delay, Mathematica Bohemica, Vol. 116, N. 2, 181-190, (1993).
[101] H. Petzeltova and J. Milota, Resolvent operator for abstract functional di¤er-
ential equations with in…nite delay, Numer. Funct. Anal and Optimiz., Vol. 9, N. 7-8,
779-807, (1987).
[103] S. G. Ruan and J. Wu, Reaction-di¤usion equations with in…nite delay, Canad.
Appl. Math. Quart. Vol. 2, N. 4, 485-550, (1994).
[110] B. N. Sadovskii, On a …xed point principle, Funct. Anal. Appl., Vol. 1, N. 2, 74-76,
(1967).
116
[111] K. Schumacher, Existence and continuous dependence for di¤erential equations
with unbounded delay, Arch. Rational Mech. Anal., Vol. 64, N. 4, 315-335, (1978).
[113] J. S. Shin, Comparison theorems and uniqueness of mild solutions to semilinear func-
tional di¤erential equations in Banach spaces, Nonlinear Analysis, Theory, Methods
and Applications, Vol. 23, N. 7, 825-847, (1994).
[114] J. S. Shin and T. Naito, Semi-Fredholm operators and periodic solutions for linear
functional di¤erential equations, J. Di¤erential Equations, Vol. 153, N. 2, 407-441,
(1999).
[116] H. Thieme, Integrated semigroups and integrated solutions to abstract Cauchy prob-
lems, J. Math. Anal. Appl., Vol. 152, N. 2, 416-447, (1990).
[118] C. C. Travis and G. F. Webb, Existence and stability for partial functional dif-
ferential equations, Trans. Amer. Math. Soc., Vol. 200, 395-418, (1974).
[121] G. F. Webb, Asymptotic stability for abstract nonlinear functional di¤erential equa-
tions, Proc. Amer. Math. Soc., Vol. 54, 225-230, (1976).
117
[123] G. F. Webb, Regularity of solutions to an abstract inhomogeneous linear di¤erential
equation, Proc. Amer. Math. Soc., Vol. 62, N. 2, 271-277, (1977).
[124] J. Wu, Theory and Applications of Partial Functional Di¤erential Equations, Applied
Mathematical Sciences, Vol. 119, Springer-Verlag, New York, (1996).
[126] J. Wu and H. Xia, Rotating waves in neutral partial functional di¤erential equa-
tions, J. Dynamics and Di¤erential Equations, Vol. 11, N. 2, 209–238, (1999).
“Ce à quoi l’un s’était failli, l’autre y est arrivé et ce qui était inconnu à un siècle, le
siècle suivant l’a éclairci, et les sciences et les arts ne se jettent pas en moule mais se forment
et …gurent peu à peu, en les maniant et polissant à plusieurs fois [:::] : ce que ma force ne
peut découvrir, je ne laisse pas de le sonder et essayer: et en retastant et pétrissant cette
nouvelle matière, la remuant et l’echaufant, j’ouvre à celui qui me suit quelque facilité et la
lui rends plus souple et plus maniable. Autant en fera le second au tiers qui est cause que
la di¢culté ne me doit pas désepérer, ni aussi peu mon impuissance...”
118