Geometry of matrix polynomial spaces (Q2189395)

Let \(P(\lambda)\) be an \(m \times n\) matrix polynomial defined by \[ P(\lambda) = \lambda^d A_d + \dots + \lambda A_1 +A_0 \] where \(A_i \in {\mathbb C} ^{m\times n}\) for \(i = 0, \dots, d\), and \(A_d \neq 0\). Let \(E(\lambda)\) be an \(m\times n\) matrix polynomial with \(\deg P(\lambda) \ge \deg E(\lambda)\). A matrix polynomial \(\widetilde {P(\lambda)} := P(\lambda) +E(\lambda)\) is called a perturbation of \(P(\lambda)\). The first companion form of \(P(\lambda)\) can be expressed as the matrix pencil \[ \mathscr C^{1}_{P(\lambda)} = \lambda \begin{bmatrix} A_d & \\ & I_n&\\ & & \ddots \\ &&&I_n \end{bmatrix} + \begin{bmatrix} A_{d-1} & A_{d-2}& \dots & A_0\\ -I_n & 0& \dots & 0\\ & \ddots& \ddots&\vdots \\ 0 &&-I_n&0 \end{bmatrix}, \] which is also one of the best known Fiedler linearizations. Matrix polynomials play an important role in some applications, resulting in rapid developments of corresponding theories and computational techniques. The canonical structure information (elementary divisors and minimal indices of the matrix polynomial) is the actual object of interest, which is usually computed via linearizations, typically Fiedler linearizations. However, the canonical structure information is sensitive to perturbations in the coefficient matrices of the polynomial. In this paper, the authors are interested in small perturbations, i.e., when \(\| \widetilde {P(\lambda)} - P(\lambda) \| \) is small compared to \(\| P(\lambda)\|\). They study how small perturbations may change the canonical structure information by constructing the stratification graphs of the orbits and bundles of matrix polynomial and their Fiedler linearization. They construct the stratification graphs for the first companion Fiedler linearization. By representing the canonical structure information as integer partitions, the authors express the cover relations between two orbits by using minimal coin moves and combinatorial rules on these integer partitions and then they provided the rules to obtain neighbouring pencils for any Fiedler linearizations. Their main theorem (Theorem 9) generalizes the result from [\textit{S. Johansson} et al., Linear Algebra Appl. 439, No. 4, 1062--1090 (2013; Zbl 1281.15019)] by relating block-structure-preserving perturbations and full perturbations of matrix pencils for any Fiedler linearization. Moreover, they show that the stratification graphs do not depend on the choice of Fielder linearization, which means that all the spaces of the matrix polynomial Fiedler linearizations have the same geometry.