Core-EP decomposition of finite potent endomorphisms and applications (Q6876150)

This paper generalizes the core-EP decomposition, core-EP inverse, and core-EP order from complex matrices to finite potent endomorphisms on arbitrary Hilbert spaces and vector spaces with inner products. The work extends finite-dimensional matrix theory to infinite-dimensional settings while addressing complications arising from the fact that finite potent endomorphisms do not form an ideal.\N\NIn 2010, \textit{O. M. Baksalary} and \textit{G. Trenkler} [Linear Multilinear Algebra 58, No. 5--6, 681--697 (2010; Zbl 1202.15009)] introduced the core inverse for matrices with \(i(A) \le 1\) as the unique matrix \(X\) satisfying: \N\[\NA \cdot X = P_{\mathcal{R}(A)} \quad \text{and} \quad \mathcal{R}(X) \subseteq \mathcal{R}(A), \N\] \Nwhere \(P_{\mathcal{R}(A)}\) is the orthogonal projection onto the range space of \(A\).\N\NFor matrices of arbitrary index, the core-EP inverse is the unique matrix \(X\) such that: \N\[\NX \cdot A \cdot X = X \quad \text{and} \quad \mathcal{R}(X) = \mathcal{R}((X)^*) = \mathcal{R}(A^m) \N\]\Nfor some \(m \ge i(A)\).\N\NFor an arbitrary \(n \times n\) matrix \(A\) with complex entries, let \(\operatorname{rk}(A)\) denote the rank. The index \(i(A) \ge 0\) is the smallest integer such that \(\operatorname{rk}(A^{i(A)}) = \operatorname{rk}(A^{i(A)+1})\). \textit{H. Wang} in Theorem 2.1 of [Linear Algebra Appl. 508, 289--300 (2016; Zbl 1346.15003)] gives the following core-EP decomposition:\N\NAny \(A \in \operatorname{Mat}_{n \times n}(\mathbb{C})\) can be uniquely written as \(A = A_1 + A_2\) where:\N\begin{itemize}\N\item \(A_1 \in \operatorname{Mat}_{n \times n}(\mathbb{C})\) with \(i(A_1) \le 1\);\N\item \(A_2\) is nilpotent;\N\item \(A_1^* \cdot A_2 = 0 = A_2 \cdot A_1\).\N\end{itemize}\NHere \(A_1^*\) denotes the conjugate transpose of \(A_1\).\N\NLet \(V\) be an infinite-dimensional vector space over an arbitrary ground field \(k\). An endomorphism \(\phi \in \operatorname{End}_k(V)\) is finite potent if there exists \(n \in \mathbb{N}\) such that \(\dim_k \operatorname{Im}(\phi^n) < \infty\).\N\NA finite potent endomorphism \(\phi\) admits a \(\phi\)-invariant decomposition:\N\[\NV = U_\phi \oplus W_\phi\N\]\Nsuch that:\N\begin{itemize}\N\item \(\phi|_{U_\phi}\) is nilpotent;\N\item \(W_\phi\) is finite-dimensional;\N\item \(\phi|_{W_\phi}: W_\phi \xrightarrow{\sim} W_\phi\) is an isomorphism.\N\end{itemize}\NThe nilpotency index of \(\phi|_{U_\phi}\) is called the index of \(\phi\).\N\NIn Section~3, where bounded finite potent operators are studied in connection with generalized Drazin invertibility, it is shown that every bounded finite potent endomorphism on a Hilbert space is generalized Drazin invertible (Koliha--Drazin invertible). Consequently, the adjoint of a bounded finite potent endomorphism is also generalized Drazin invertible.\N\NIn Section~4, where the core-EP decomposition for bounded finite potent operators is developed, a bounded finite potent endomorphism \(\phi\) on a Hilbert space is considered. The key observation is that \(W_\phi\) is finite-dimensional and hence closed, so that it admits an orthogonal complement,\N\[\NV = W_\phi \oplus [W_\phi]^\perp.\N\]\NThis induces the orthogonal projection \(P_{W_\phi}\), which is fundamental to both the core-EP decomposition and the construction of the core-EP inverse. An example shows that the core-EP decomposition does not coincide with the core-nilpotent decomposition. A characterization of the cases in which these two decompositions differ is provided in terms of the AST-decomposition.\N\NIn Section~5, where the core-EP generalized inverse is analyzed, finite potent endomorphisms on arbitrary vector spaces endowed with an inner product over fields of arbitrary characteristic are investigated.\N\NIn Section~6, where the core-EP order is examined, the analysis is carried out within the framework of bounded finite potent operators.