Publications of the Research Institute for Mathematical Sciences · 1975 · 10 citations · 8 references
Natural SciencesMathematicsDifferential EquationsDynamical SystemsRepresentation TheoryOperator TheoryRegular SolutionsCauchy ProblemOperatorsRegularity TheoryResolvent OperatorsResolvent Theory
CAUCHY PROBLEM 53 it is densely defined closed and the resolvent set is nonvoid. We shall denote by G(k, T) (or Gw(/c)) the set of all regular operators satisfying (I) and (II) (or (I J and (II J). It is proved in [10] that, A e G(/c, T) (or A e Gw(/c)), the exponential formula (E) u(t'9x)=\\ml-—A x H-»OO n / converges for x e D(A) and t e [0, T) (or t 0), and for x E D(A), u(t x) of (E) gives the unique solution to (ACT; A, T, x) (or (ACPm A, x)). Therefore, we have: Let A be a regular operator. Then (ACP; A,T,D(A})(or(ACPco',A,D(A J)) is well posed and only AeG(k,T)(or AeGm(k)). The relationship between classes G(/c, T) and Gw(/c), and the class of infinitesimal generators of distribution semigroups is studied in [10] and [11]: (\) A is the infinitesimal generator of a regular distribution semigroup (R. D. S. G.) and only for every Te(0, oo) there is a fc(T)eZ such that A e G(/c(T), T). (2) A is the infinitesimal generator of an exponential distribution semigroup and only A e Gw(/c) for some co e R and k e Z. Our second purpose is to reform if part of (1). Theorem 1.3. // AeG(kJT) for some /ceZ + and Te(0, oo], then A is the infinitesimal generator of an R. D. S. G. Detailed explanations concerning R. D. S. G. will be seen in [3], [4], [8], [10], [11], [16], and [17]. This paper consists of five sections. Section 2 contains fundamental lemmas which are based on the closed graph theorem. These lemmas are used throughout this paper. In Section 3, we prove Theorem 1.2. The proof is easy because of (1.1). We prove (Iro) and then we show (•oo the relation &(A; A)x = e~^u(t x)dt. (IIW) follows from this relation. Jo However, in the proof of Main Theorem, this relation may not be employed. Because it is not assumed that ACP is well posed in the sense of semigroup (see [13; p. 396] and [7; Chap. 1]) or ACP is Hadamard correct (see [9] and Theorem 1.2 of this paper), and so the growth order of the solution u(t) of ACP is not necessarily estimated 54 NOBUHIRO SANEKATA as (1.1). Also, ACP considered in Main Theorem does not necessarily become well posed in the sense of semigroup (see [12] and [14]). Therefore, we must employ another approach. In Section 4, we show that the resolvent set of the regular operator A contains a logarithmic region ACP is well posed. An estimate for the resolvent of A is given under the same assumptions. See Proposition 4.1. As a consequence of Proposition 4.1 and [8], we obtain Theorem 1.3. Proposition 4.1 is also used in the proof of Main Theorem. Finally, in Section 5, we complete the proof of Main Theorem. In the preparation of this paper, the author received many valuable suggestions and constant encouragement from Professor H. Sunouchi and Professor I. Miyadera. My thanks also go to Professor K. Kojima and Professor S. Oharu. Without their advice and interests in this work this paper could not be accomplished. 2. Preliminaries In this paper, R denotes the real line, and C denotes the complex plane. Re A and ImA denote the real part and the imaginary part of AeC respectively. N = {1, 2,...} and Z = Nu{0}. Let / be an interval and let Y be a Banach space. We denote by C(I: Y) the set of all 7-valued j-times continuously differentiate functions defined on /. C(/; 7) denotes the set of all continuous functions. If there is no ambiguity, we write C(Y). Let A be a closed linear operator in a Banach space X. p(A) and R(k A) denote the resolvent set and the resolvent of A respectively. D(A) is often considered as a Banach space j with respect to the norm \\x\\j= ]£ M'x||; A°x = x. This Banach space will be denoted by [DG4-0]. Let Te(0, oo] and coeR. We shall consider C([0, T); [D(A)J) as a Frechet space Z endowed with the system of seminorms { sup ||u(0lli; ?e(0, T)}. Also, we denote by O^f^T Z0, the Banach space consists of all elements weC([0, oo); [D(A)1) such that supe~\\u(i)\\1 0 whenever it is assumed that ACPW is well posed. This remark is used in the proof of Lemma 2.1. Lemma 2.1. Let A be a closed linear operator. If (ACP; T9 D(AJ) or (ACPm D(A J) is well posed, then for every jeN and xeD(AJ), we have U(f)xeD(A^ U(-)xe&(X) and Proof. We shall prove the case j = l. The general case is proved by induction. For every xeD(A), we have (d/dt)U(t)Ax = AU(t)Ax. Integrating this equation on [0, f], and using the closedness of A, we have (2.1) V(t)Ax = Ax+ AU(s)Axds = Az(t), Jo rt where z(t) = x+ U(s)AxdsED(A). It is easy to see that (dldi)z(i) = Jo U(t)Ax = Az(t) and z(0) = x. Therefore, z(t) is the solution to ACP. In the case of ACP^, we have to show that the solution z(t) satisfies (1.1). But, since co>0 and \\U(s)Ax\\^Me° for some M>0, we have = 0(e> as t > o o . Thus z(t) is the solution to ACPm. Therefore, in either case, we have z(i)=U(t)x, by the uniqueness of the solution, and by (2.1), we have U(t)Ax = AU(t)x. Lemma 2.2. Let A be a closed linear operator with (a) // (ACP r, D(A)) is well posed, then for every ie(0, T), there is an M t>0 such that 56 NOBUHIRO SANEKATA (2.2) for every Je[0, T] and xeD(A). (b) // (ACP^ D(A)) is well posed, then there is an M>0 such that (2.3) || 17(0*11 M*'«||x||k for every t^Q and xeD(A). Proof. We prove (a). It is easy to see that the linear operator x^U(-)x9 defined on [D(A J] with range Z, is a closed operator. Therefore, x*-+U(-)x is continuous by the closed graph theorem. Let A.0Ep(A) and xeD(A ), and put y = R(A0i A)x. Then, by Lemma 2.1, we have 11*7(0*11 = \\VoA)U(t)y ^ const. \\U(t)y\\i, where constant is independent of t and x. Since £/(•) is continuous, for every te(0, T), we have for *e[0, t]. This shows (2.2). We can prove (b) in a same way as (a), we replace Z with Zw. Lemma 2.3. Let A be a regular operator. Then D(A) is dense in [D(^4)] for every m>n, m, neN. Lemma 2.3 is proved in [10; Lemma 2.7]. Lemmas 2.2 and 2.3 show that, A is a regular operator, we can extend {17(0} uniquely to a family of operators defined on D(A) with range X satisfying (2.2) (or (2.3)) for all xeD(A) and *e[0, T] (or * ^0); this extended family will be denoted by the same symbol {U(i)}, also it will be called the solution operators for ACP (or ACP^). Moreover, we see that (7(0* is strongly continuous in t e [0, T) (or t^ 0) for every xeD(A). ABSTRACT CAUCHY PROBLEM 57CAUCHY PROBLEM 57 3. Proof of Theorem 1.2 Throughout this section, let A be a regular operator, and let coeR and keZ. Suppose that (ACP^ A, D(A)) is well posed, and let {17(0} be the solution operators for ACP^. We define for Re A > CD and xeD(A). Lemma 3.1. K0(A) has a unique extension to a bounded linear operator •; £/ns extended operator will be denoted by Proof. Let A0ep(^) and xeD(A and put K* (f ) = (A0 yl) w C/(OJR(A0 ; A) x for 0 m fc . Then we have
8
Functional Analysis and Semi-Groups
Richard G. Cooke · Nature · 1949 · 1.9K citations
On the abstract Cauchy problem
Richard Beals · Journal of Functional Analysis · 1972 · 47 citations