niversitÀ di adova eminario atematico...
TRANSCRIPT
![Page 1: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/1.jpg)
RENDICONTIdel
SEMINARIO MATEMATICOdella
UNIVERSITÀ DI PADOVA
G. DA PRATO
M. IANNELLILinear abstract integro-differential equationsof hyperbolic type in Hilbert spacesRendiconti del Seminario Matematico della Università di Padova,tome 62 (1980), p. 191-206<http://www.numdam.org/item?id=RSMUP_1980__62__191_0>
© Rendiconti del Seminario Matematico della Università di Padova, 1980, tousdroits réservés.
L’accès aux archives de la revue « Rendiconti del Seminario Matematicodella Università di Padova » (http://rendiconti.math.unipd.it/) implique l’accordavec les conditions générales d’utilisation (http://www.numdam.org/conditions).Toute utilisation commerciale ou impression systématique est constitutived’une infraction pénale. Toute copie ou impression de ce fichier doit conte-nir la présente mention de copyright.
Article numérisé dans le cadre du programmeNumérisation de documents anciens mathématiques
http://www.numdam.org/
![Page 2: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/2.jpg)
Linear Abstract Integro-Differential Equationsof Hyperbolic Type in Hilbert Spaces.
G. DA PRATO - M. IANNELLI (*)
Introduction.
This paper is concerned with the study of the problem:
where for t:>O, A(t) is the infinitesimal generator of a strongly con-tinuous semi-group in a Hilbert space H and X is of the form
where B(~) is self-adjoint and semi-bounded; K is then a vectorialgeneralization of a completely positive kernel.
We study this problem with the same methods of sum of linearoperators as in [8]. We are able, under suitable hypotheses, y to showexistence and uniqueness of a continuous strong solution u for every
and x E g; moreover u is a classical solution if
T; K), where K is a Hilbert space densely and continuouslyembedded in H and x E K.
(*) Indirizzo degli AA.: Dipartimento di Matematica, Universita di Trento -Povo.
Lavoro svolto nell’ambito del G.N.A.F.A.
![Page 3: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/3.jpg)
192
Similar problems have been studied by several authors by differentmethods, in the autonomous case. We remark that we do not assumethat the domains of A(t) and B(~) are constant.
For the proofs we need an « energy equality » (see formula (21))that we think will be useful to study asymptotic properties of u.
1. Notations.
We note by H a real Hilbert space (1) (inner product ( , ), norm(.) and by L2(o, T; H) the Hilbert space of the measurables map-pings ~c : [0, T] - H such that JU 12 is integrable in [0, T], endowedwith the inner product:
We put:
It is well known that every u E T] ; H) can be identified witha continuous function (2); in the following we always make such anidentification.
We note also by 0([0, T]; H) (resp. 01([0, T]; H)) the set of map-pings continuous (resp. continuously derivable); it is
~’] ~ H) c C( Co, z’] ~ .g) .Finally we put:
We write, for brevity, L2(g), 7 Wo(.t1), C(H), 01(H). We studythe problem:
(1) We suppose H real for simplicity.(2) See for exemple [3].
![Page 4: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/4.jpg)
193
We assume:
B is and 3WB E R such that B - WB 0, 7
~ c : [0, l ] -~ measurable and bounded such that
We write (P) in the following form:
x} is given in and yo is defined by:
We consider also the approximating problem:
where Bn = n2R(n, B) - n. It is well known ([8]) that (Pn) has a
unique strong solution un E C(.8’) ; moreover if f E and x e j9~it is:
because 7~ E 01([0, T] ~ .
(3) If L: is a linear operator we note by (resp. the resolvent set (resp. the spectrum) of L and by L) the resolventoperator of L.
(4) It is known that DA is dense in H.(5) This hypothesis implies that the operator Lu = k * u is completely
positive in L2(0, T; H); for the existence of a solution of (P) it will be suf-ficient to assume L positive.
(s ) D~ and D, are endowed with the graph norm.
![Page 5: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/5.jpg)
194
2. A priori bound.
PROPOSITION 1. Assume (H) and let Un be the solution of (Pn), thenit is :
PROOF. Choose f E E D,4; put exp [- t~~ ~ ~cn = it is
u~ = vn~ -E- multiply by then it is :
integrating from 0 to t we get:
![Page 6: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/6.jpg)
195
and the conclusion follows from the Gronwall lemma for f E W((H);in the general case we use the density of w’1(g) in L2(H).
COROLLARY 2. Under the hypotheses of the Proposition 1, if u is
a classical solution of (P) (7) it is :
where K is a suitable constant.
PROOF. It is :
using (3) we obtain
and the conclusion follows by dominated convergence.
3. Strong solutions.
PROPOSITION 3. Assume (H) and suppose DA n DB dense in H;then yo is pre-closed.
PROOF. Let c DYo such that :
![Page 7: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/7.jpg)
196
we have to show that f = 0, x = 0. From (5) it follows
therefore is a Cauchy sequence in O(H) and furthermore Ui -+ 0in C(H); then .r = == 0. We, now, go to show that f = 0.
i - oo
Remark that, since A - WA, I (B - CùB) k * are positive operators inL2(H) it is:
Choose g in n which is dense in L2(H), from(7) it follows
Then f or i - o0
and
Finally for
which implies f = 0.In the following we denote by y the closure of yo. We call u E L2(H)
a strong solution of (P) if it is :
![Page 8: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/8.jpg)
197
i.e. if there exists E r1 r1 L2(DB) such that
It is not easy in general to characterize the domain D(y) of y; how-ever we can show that D(y) c C(H) and that if u E D(y), y - u = ff, t x}then n ( o ) = x.
The following proposition is straightforward:
PROPOSITION 4. Under the hypotheses of the Proposition 3 if u E D(y~.and = (f, x~ then (5) and (6) hold.
Moreover y is one-to-one and has a closed range; consequently (P)has, at most, one strong solution.
PROPOSITION 5. Assume that the hypotheses of Proposition 3 arefulfilled. Let u E D(y), = (f, xl; then u E O(H) and it is:
PROOF. Let c D(yo) such that
from (6) it follows
Thus is a Cauchy sequence in C(H) which implies vi -’7- U
in C(H). Finally put then z is a strong solution of theproblem:
![Page 9: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/9.jpg)
198
from (5) it follows
and in C(H).
4. Existence.
If Z : is a linear mapping and g a sub-space of H wedenote by LK the following mapping in K:
It is easy to see that if A E e(L) r1 e(LK) then it is R(A, L)(K) c Kand RK(Â, L) = R(A, LK) (8).
THEOREM 6. Assume that the hypothesis (H) holds and that there,exists a Hilbert space K (inner products ((,)), norm II densely em-bedded in H such that:
a) K n DA is dense in H (9),
b) such that + cxJ[ ((A. y, 2
c) BK is self -adjoint and such that
Then b’ f E L2(.H), b’x E H, the problem (P) has a unique strong solu-tion u such that u E C(H), u(O) = x.
lVloreover ’BIf E Wi(K) and x E K r1 DA the solution u belongs to
i.e. it is a classical solution.
(8) L) is the restriction of L) to K.(9) K4DB means that .K is continuously and densely embedded in DB,
![Page 10: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/10.jpg)
199
PROOF. By virtue of the closed graph theorem it is B E H)put
It is clear that DAn DB is dense in 1~, therefore yo is pre-closed(Proposition 3). Finally, due to the Corollary 4, to show existenceit is sufficient to prove that y has a range dense in L2 (H) O H.
Take and let un be the classical solution in H
of the problem ( lo ) :
from Proposition 1 there exists N > 0 such that
it follows
due to (18) ~N’ > 0 such that
It is
then if T E L2(DB) it is
By virtue of (19) is bounded in L2(H), it follows
because L2(DB) is dense in L2(g).
(lo) See inclusion (2).(11) - means weak convergence.
![Page 11: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/11.jpg)
200
Consequently the range of y, is weakly dense in y is onto.
We prove now the regularity result. Recall that Un - u in
(Proposition 5); moreover due to (19) 3 a sub-sequence suchthat is weakly Cauchy; consequently u E DB and by virtueof (18) u E LOO(K).
Consider now the problem:
and the approximating one
It is in = ~n and vn 2013~ (Proposition 5); it follows v = u’ and
because
Finally it is easy to see that u E DA and Au = u’ - f - E
E L2(H). ~
5. Generalizations.
We generalize now the problem (P). We consider two familiesA = > of linear operators in H. We put:
and the study the problem:
![Page 12: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/12.jpg)
201
We write ( P ‘ ) in the following form:
where yo is defined by:
We consider also the approximating problem:
If u E multiplying (P’ ) for u(t) and putting v~ = exp [- t~] ~ uwe get the energy equality:
We prove the theorem:
THEOREM 7. Assume that:
![Page 13: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/13.jpg)
202
Then E L2(H), x E H, 3a unique strong solution C(H). More-
over if U E _L2(.g), x E K then it is u E W1(.g) f1 We can write the problem (P[) in the following f orm :
where Z(t,8) are the evolutions operators attached to the family A.If we can solve (22) by the contraction principle (1~) ~therefore + f E L2(K) which implies Un E n ([8]).Using the energy equality in the I~ space we get:
Via e L2(K), x c- > 0; Proceeding as in previous sections we are able to show:
a) yo is preclosed and its closure y’ is one-to-one and with aclosed range.
b ) 3N’ > 0 such that
c)Itis
It follows that y’ is onto and 3 a unique strong solution u of (P’ ) forevery f c-L2 (H); moreover u in L2(H).
Concerning regularity we remark that if x E K, f E L2(g), then byvirtue of (24) Bu belongs to L2(H) and recalling (23) u E L°°(.K),consequently Au E L2(H) and
REMARK 8. For sake of simplicity we have assumed K c D(A(t)) r)r1 D(B(~)) instead of K c D(B($)) as in previous section; actuallysimilar results can be proved with this latter assumption.
(12) In fact Z is strongly measurable and bounded in .K [8].
![Page 14: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/14.jpg)
203
6. Resolvent family.
Assume that the hypotheses either (H’ ) or (H) and (HE) hold.Consider the problem:
where K = Bk if (H) and (HE) hold.Due to Theorem 7 (PS ) has a unique strong solution. Moreover
if (H’) (resp. (H) and (HE)) hold and x c K (resp. then u
is classical. Put:
it is G(t, s) E C(H). The mapping:
is called the evolution operator for the problem (P~) and the family~G(t, the resolvent family.
Consider also the approximating problem:
and put Un(t) = Gn(t, s) x.
PROPOSITION 9. For every x E H it is :
Therefore (7( -, -) .r E H).
PROOF. It is sufhcient to prove (27) for (resp. DA r1 ~)~Put v = ~ - un, then it is:
![Page 15: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/15.jpg)
204
from which
and the thesis follows from dominated convergence. #
PROPOSITION 10. Assume that the hypotheses (H’) (resp. (H) and(H~)) hold. If strong solution of (P’)(resp. (P)) it is:
PROOF. Go to the limit in the equality:
REMARK 11. By similar arguments we can study the problem:
where
and p : [0, T]x[0,1] - R+ is continuous.The energy equality is in this case:
![Page 16: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/16.jpg)
205
where
is the solution of the problem
EXAMPLE 12. Let H = L2(R), 99 E Co (R2 , R) and
Then (P’ ) is equivalent to
If U, E L2(R) and f E .L2 ( [o, T] X R) then (35) has a unique strong solu-tion T]; .L2(ll~)) ; if moreover T] X R) and Uo E W2(R)then u is classical and it is
REFERENCES
[1] V. BARBU, Nonlinear Volterra equations in a Hilbert space, SIAM J.
Math. Anal., 6 (1975), pp. 728-741.[2] V. BARBU, Nonlinear Volterra integro-differential equations in Hilbert
space, Sem. Mat. Bari, 143 (1976).
![Page 17: NIVERSITÀ DI ADOVA EMINARIO ATEMATICO ENDICONTIarchive.numdam.org/article/RSMUP_1980__62__191_0.pdfDipartimento di Matematica, Universita di Trento - Povo. Lavoro svolto nell ambito](https://reader033.vdocuments.us/reader033/viewer/2022041416/5e1bf8a06105b25d7b19e554/html5/thumbnails/17.jpg)
206
[3] H. BRÉZIS, Opérateurs maximaux monotones et sémi-groupes de con-
tractions dans les espaces de Hilbert, Math. Studies, 5, North Holland(1973).
[4] M. G. CRANDALL - S. O. LONDEN - J. A. NOHEL, An abstract nonlinearVolterra integro-differential equation, J. Math. Anal. Appl., 64 (1978),pp. 701-735.
[5] C. M. DAFERMOS, An abstract Volterra equation with applications to linearviscoelasticity, J. Diff. Eq., 7 (1970), pp. 554-569.
[6] C. M. DAFERMOS - J. A. NOHEL, Energy methods for nonlinearhyperbolicintegro-differential equations, Comm. PDE, 4 (1979), pp. 219-278.
[7] G. DA PRATO - P. GRISVARD, Sommes d’opérateurs linéaires et équationsdifferentielles opérationnelles, J. Math. Pures et Appl., 54 (1975), pp. 305-387.
[8] G. DA PRATO - M. IANNELLI, On a method to study abstract evolution equa-tions in the hyperbolic case, Comm. PDE, 1 (1976), pp. 585-608.
[9] G. DA PRATO - M. IANNELLI, Linear integro-differential equations in Banachspaces, Rend. Sem. Mat. Padova, 62 (1980), pp. 207-219.
[10] S. O. LONDEN, On an integral equation in a Hilbert space, SIAM J. Math.Anal., 8 (1977), pp. 950-970.
[11] S. O. LONDEN, On an integro-differential Volterra equation with a maximalmonotone mapping, J. Diff. Eq., 27 (1978), pp. 405-420.
[12] R. C. MAC CAMY, Approximations for a class of functional differentialequations, SIAM J. Appl. Math., 23 (1972), pp. 70-83.
[13] R. C. MAC CAMY, Stability theorems for a class of functional differentialequations, SIAM J. Appl. Math., 30 (1976), pp. 557-576.
[14] R. C. MAC CAMY - J. S. W. WONG, Stability theorems for some functiona.equations, Trans. AMS, 164 (1972), pp. 1-37.
[15] R. K. MILLER - R. L. WHEELER, Asymptotic behavior for a linear Volterraintegral equation in Hilbert space, J. Diff. Eq., 23 (1977), pp. 270-284.
Manoscritto pervenuto in redazione il 5 giugno 1979.