Extensions of Perron-Frobenius theory (Q395590): Difference between revisions

In this paper, the author is interested in two results of the theory of Perron-Frobenius. The first one says that if \(A>0\) is an irreducible matrix, then its spectral radius \(r(A) >0\) is a simple root of its characteristic polynomial and \(\ker (r(A) -A) =\mathrm{span}(x_{0})\) for some strictly positive vector \(x_{0}\). The second one says that if \(A\) and \(B\) are two matrices satisfying \(0\leq B\leq A\) and the spectral radius of \(A\) coincides with that of \(B\), i.e., \( r(A)=r(B)\) with \(A\) being irreducible, then \(A=B\). The author gives different references of some authors who have generalized these results of Perron-Frobenius to positive operators on Banach lattices of infinite dimension, in particular, \textit{V. Caselles} [Isr. J. Math. 58, 144--160 (1987; Zbl 0643.47002), Theorem 4.3], \textit{E. A. Alekhno} [in: Proceedings of the conference ``Positivity IV -- Theory and Applications'', Dresden, Germany, July 25--29, 2005. Dresden, TU Dresden, Institut für Analysis. 5--14 (2006; Zbl 1112.47305), Theorem 5; Positivity 16, No. 1, 143--176 (2012; Zbl 1259.46013), Theorem 4.11 (e)] and \textit{J. Bernik} et al. [J. Lond. Math. Soc., II. Ser. 86, No. 1, 214--234 (2012; Zbl 1259.47050), Theorem 2.8 and Corollary 2.12]. The author begins by generalizing some results on positive eigenvectors of irreducible operators and their adjoints. After that, he applies these results to get some extensions of the second result of Perron-Frobenius to positive operators on arbitrary Banach lattices, i.e., he gives sufficient conditions under which, if \(S\) and \(T\) are operators on Banach lattices such that \(0\leq S\leq T\) and \( r(T) =r(S)\), then \(S=T\). As an important consequence, he obtains a result (Corollary 3.5) showing that Theorem 2.8 and Corollary 2.12 of Bernik et al. [loc.\,cit.], which were proved for positive operators on \(L^{p}(\mu)\), for \(1<p<\infty\) and \(\mu\) being \(\sigma\)-finite, remains valid for general Banach lattices. Also, he shows (Theorem 3.9) that one can weaken the hypothesis of Theorem 2.8 and Corollary 2.12 of Bernik et al. [loc.\,cit.] by replacing the condition ``\(T\) or \(S\) is power compact'' by a weaker condition, namely, ``the spectral radius \(r(T)\) is a pole of the resolvent \(R(\cdot,T) \)'', which is just the condition of Theorem 4.3 of Caselles [loc.\,cit.]. Therefore, Theorem 4.3 of Caselles [loc.\,cit.], Theorem 5 of Alekhno [Zbl 1112.47305] and Theorem 4.11 (e) of Alekhno [Zbl 1259.46013] become special cases of Theorem 3.9 of the author.