8.7. Miscellaneous Comments
507
theorem and the Lax equivalence theorem) have been extended to the
general equation (8.4.1) (see the author [1982: 2]).
Hereditary Differential Equations.
The systems described shortly
after Theorem 8.3.1 do not in general satisfy Huygens' principle; the state of
the system at a single time s < t does not suffice to determine the state at
time t (compare with Section 2.5(b)). That this type of situation is common
in physical phenomena was well known more than a century ago, although
its mathematical modeling had to wait for some time. Hereditary equations
were introduced (and named) by Picard [1907: 1]. Their importance in
physical, and especially biological phenomena was recognized by Volterra
[1909: 11, [1928: 1], [1931: 11, who undertook a systematic study of what
would be called now ordinary integro-differential equations. We quote from
[1931: 1, p. 142]: "L'etat d'un systeme biologique a un moment donne
semble donc bien devoir dependre des rencontres ayant eu lieu pendant une
periode plus on moins longue precedant ce moment; et dans les chapitres
qu'on vient de developper, on a, somme toute, neglige la duree de cette
periode. Il convient de tenir compte maintenant de l'influence du passe."
"On rencontre, en physique,
dans
l'etude
de
1'elasticite,
du
magnetisme, de l'electricite, bien des phenomenes analogues de retard,
trainage ou histeresis. On peut dire que dans le monde inorganique it existe
aussi une memoire du passe, comme la memoire du fil de torsion dont la
deformation actuelle depend des etats anterieurs."
A few lines below, Volterra remarks: "Mais lorsqu'en physique
1'heredite entre en jeu, les equations differentielles et aux derivees partielles
ne peuvent pas suffire; sinon les donnees initiales determineraient 1'avenir.
Pour faire jouer un role a la suite continue des etats anterieurs (infinite de
parametres ayant la puissance du continu) it a fallu recourir a des equations
integrales et integro-differentielles oiz figurent des integrales sous lesquelles
entrent les parametres caracteristiques du systeme fonctions du temps
pendant une periode anterieure a l'instant considers; on a meme introduit
des types plus generaux d'equations aux derivees fonctionnelles."
After these investigations, interest in hereditary differential equations
faded for almost two decades, except for work such as that of Minorski
[ 1942: 11, where delay differential equations x'(t) = f (t, x (t), x (t - h)) are
examined. Beginning in the late forties, an intense revival occurred, stimu-
lated in part by control theory; the objects of study were functional
differential equations, in general of the form
x'(t) = f(t,x1),
(8.7.4)
where for each t, f is a vector-valued functional defined in a suitable
function space of vector-valued "histories" xt = (x(s), - oc < s s t). As-
suming the histories are continuous functions, the functional is continuous,
linear and time independent, and the system has finite memory, one obtains