scientific article; zbMATH DE number 4040663

From MaRDI portal
Publication:3779310





[For the entire collection see Zbl 0585.00006.] This is a survey of a body of work by T. Iwamiya, S. Oharu, and T. Takahashi which is yet to appear. Of concern is the Cauchy problem \[ (*)\quad du(t)/dt\in A(t)u(t)\quad (s<t<T),\quad u(s)=x, \] where \(\{\) A(t)\(\}\) is a family of (possible multivalued) operators on a Banach space X and \(0\leq s\leq t<T<\infty\). If (*) is well-posed for initial data x in D(s) where \(D\subset [0,T)\times X\) and D(ds)\(\in \{x\in X:\) (s,x)\(\in D\}\), then \(U(t,s)x=u(t)\) satisfies \[ U(t,s): D(s)\to D(t),\quad U(t,s)\quad U(s,r)=U(t,r),\quad U(t,t)=I, \] and U(\(\cdot,s)x\in C([s,T,X))\) for each \(x\in D(s)\). Assume these conditions; \(U=\{U(t,s):\) \(0\leq s\leq t\leq T\}\) is called an evolution operator constrained in D. An evolution operator constrained in D is said to be of class \({\mathcal E}(D,\omega,f)\) if \[ \| U(t+s,s)x- U(t+r,r)y\| \leq e^{\omega t}\| x-y\| +\int^{t}_{0}e^{\omega (t-s)}\quad f(\xi +s,\quad \xi +r)d\xi \] holds for \(0\leq r\leq s<T\), \(0\leq t<T-s\), \(x\in D(s)\), \(y\in D(r)\), and some \(\omega\in {\mathbb{R}}\) and a suitable real function f. Such evolution operators automatically satisfy various continuity conditions. Using a method of discretization in time inspired by the work of \textit{M. G. Crandall} and \textit{L. C. Evans} [Isr. J. Math. 21, 261-278 (1975; Zbl 0351.34037)], the authors obtain sufficient conditions for U to be of class \({\mathcal E}(D,\omega,f)\). This gives rise to well-posedness for (*), so that mild solutions exist and are unique. Moreover the authors relate their results to the usual work on iterated resolvents (so that U(t,s)x can be obtained as a limit of \[ \prod^{k_ n}_{j=1}[I-(t^ n_ j- t^ n_{j-1})A(t^ n_ n)]^{-1}x_ n). \] A third approach to (*) is an extension of work of \textit{M. Pierre} [Isr. J. Math. 23, 189-199 (1976; Zbl 0343.34050)] on envelopes of quasicontractive semigroups. This is technical stuff but very interesting nevertheless. The domain D(A(t)) of A(t) is allowed to vary substantially with t.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3779310)