Eunice Y. S. Chan,Robert M. Corless,Laureano Gonzalez-Vega,J. Rafael Sendra,Juana Sendra
Abstract:We show how to construct linearizations of matrix polynomials $z\mathbf{a}(z)\mathbf{d}_0 + \mathbf{c}_0$, $\mathbf{a}(z)\mathbf{b}(z)$, $\mathbf{a}(z) + \mathbf{b}(z)$ (when $\mathrm{deg}\left(\mathbf{b}(z)\right) < \mathrm{deg}\left(\mathbf{a}(z)\right)$), and $z\mathbf{a}(z)\mathbf{d}_0\mathbf{b}(z) + \mathbf{c_0}$ from linearizations of the component parts, $\mathbf{a}(z)$ and $\mathbf{b}(z)$. This allows the extension to matrix polynomials of a new companion matrix construction.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the linearization problem of matrix polynomials, especially how to construct linearizations of matrix polynomials in specific forms. Specifically, the authors show how to construct linearizations of more complex matrix polynomials from the basic parts of linearizations \(a(z)\) and \(b(z)\), such as:
1. \(e_1(z)=zd_0a(z)+c_0\)
2. \(e_2(z)=za(z)d_0 + c_0\)
3. \(a(z)b(z)\)
4. \(a(z)+b(z)\) (when \(\deg(b(z))<\deg(a(z))\))
Through these constructions, the higher - order matrix polynomial eigenvalue problems can be transformed into "simple" linear matrix eigenvalue problems, so that existing efficient numerical algorithms can be used for solution. For example, for the generalized eigenvalue problem, that is, to find the complex number \(z\) that makes the matrix polynomial \(P(z)\) singular (i.e., solve \(\det P(z) = 0\)), if \(P(z)\) is a first - order matrix polynomial, then the already very mature and reliable generalized eigenvalue solution methods (such as the QZ iteration method) can be directly used.
The main contribution of the paper lies in providing a new method for constructing companion matrices and extending it to the case of matrix polynomials, thus providing an effective tool for dealing with higher - order matrix polynomial eigenvalue problems. This is very important in many applications because many practical problems require the calculation or approximate solution of polynomial eigenvalues.
### Formula display
To ensure the correctness and readability of the formulas, the following are some key formulas involved in the paper:
- **Define the spectrum of a matrix polynomial**:
\[
\Lambda(a(z)):=\{z\mid\det(a(z)) = 0\}
\]
- **The linearization matrices in Theorem 1**:
\[
E_1=\begin{bmatrix}0&c_0&X_A\\-Y_A&A\end{bmatrix},\quad D_{E_1}=\begin{bmatrix}d_0\\D_A\end{bmatrix}
\]
\[
E_2=\begin{bmatrix}A&Y_A&c_0\\-X_A&0\end{bmatrix},\quad D_{E_2}=\begin{bmatrix}D_A\\d_0\end{bmatrix}
\]
- **The linearization matrices in Theorem 2**:
\[
F_1=\begin{bmatrix}A&0\\Y_BX_A&B\end{bmatrix},\quad D_{F_1}=\begin{bmatrix}D_A&0\\0&D_B\end{bmatrix}
\]
\[
F_2=\begin{bmatrix}B&Y_BX_A\\0&A\end{bmatrix},\quad D_{F_2}=\begin{bmatrix}D_B&0\\0&D_A\end{bmatrix}
\]
These formulas show how to construct linearizations of more complex matrix polynomials by combining the linearizations of basic parts.