Uniformly bounded limit of fractional homomorphisms (Q4674753)

Theorem A given below, due to \textit{J.\ Kisyński} [``The Widder spaces, representations of the convolution algebra \(L({\mathbb{R}}^+)\), and one parameter semigroups of operators'' (Preprint No.\ 558, IMPAN) (1998; http://basilo.kaist.ac.kr/preprint/IMPAN/kisynski.pdf)], has a wide range of applications. Theorem A. Let \(A\) be a Banach algebra, let \(\omega\geq 0\), and let \((r_\lambda)_{\lambda\in(\omega,+\infty)}\) be a pseudo-resolvent in \(A\) and let \(M= \sup\{(\lambda-\omega)^n\| r^n_\lambda\|:\;n\in \mathbb N,\;\lambda\in (\omega, + \infty)\}.\) Then the following conditions are equivalent: (i) \(M< +\infty\). (ii) There exists a continuous homomorphism \(T :L^{1}_{\omega}(R^+)\rightarrow A\) such that \(t(\varepsilon_\lambda) =r_\lambda\) for each \(\lambda\in(\omega,+\infty)\). Further, if a continuous homomorphism \(T : L^1_{\omega}(R^+) \rightarrow A\) satisfying \(t (\varepsilon_\lambda)= r_\lambda\) for each \(\lambda\in (\omega, + \infty)\) exists, then it is unique and \(\| T\| = M\). Alternative proofs of this theorem were given by many authors, notably \textit{A.\ Bobrowski} [Stud.\ Math.\ 124, 281--290 (1997; Zbl 0876.44001)] using the Yosida approximation, and \textit{W.~Chojnacki} [Proc.\ Am.\ Math.\ Soc.\ 126, No. 2, 491--497 (1998; Zbl 0893.46044); Ann.\ Sc.\ Norm.\ Super.\ Pisa, Cl.\ Sci.\ (4) 28, 287--322 (1999; Zbl 0970.46032)]. In the present paper, the authors present a new proof of this theorem, based on a different technique involving \(\alpha\)-times integrated semigroups for any \(\alpha > 0\) and the limit as \(\alpha\) goes to zero. The \(\alpha\)-times integrated semigroups for \(\alpha= 1\) were used earlier to establish the above theorem with \(\omega= 0\) by \textit{J.~Kisyński} [op.\ cit., Theorem 4.2]. Its improvement for the case \(\alpha>0\) is presented by the author by proving the following Theorem B, which forms the main result of this paper. Theorem B. Let \(A\) be a Banach algebra and \(\omega\geq 0\). (i) If there exists a continuous homomorphism \(T : L^1_\omega(\mathbb{R}^+)\rightarrow A\), then there exists a family of continuous homomorphisms \(T_\alpha : AC^{(\alpha)}_\omega(\mathbb{R}^+) \rightarrow A\) for each \(\alpha > 0\) so that \(T_\alpha (\varepsilon_{-\lambda})= T(\varepsilon_{-\lambda})\) for each \(\lambda \in (\omega, + \infty)\) and \(\| T_\alpha\| \leq\| T\|\) for any \(\alpha > 0\). (ii) Conversely, if for each \(\alpha >0\) there exists a continuous homomorphism \(T_\alpha : AC^{(\alpha)}_{\omega} (\mathbb{R}^+) \rightarrow A\) such that \(T_\alpha(e_{-\lambda})\) does not depend on \(\alpha\) for each \(\lambda\in (\omega, + \infty)\) and \(\limsup_{\alpha \rightarrow 0+}\| T_\alpha\|< +\infty\), then there exists a unique continuous homomorphism \(T_\alpha : L^1_{\omega}(\mathbb{R}^+) \rightarrow A\) such that \(T_\alpha(\varepsilon_{-\lambda}) = T(\varepsilon_{-\lambda})\) for each \(\lambda\in (\omega, +\infty)\) and \(\| T\|\leq \limsup_{\alpha\rightarrow 0+}\| T_\alpha\|\).