On doubly structured matrices and pencils that arise in linear response theory. (Q1428608): Difference between revisions

In linear response theory, consider the problem of finding the eigenvalues of the pencil \((\lambda E_0 - A_0)(x) = 0,\) \(x \in \mathbb{C}^n,\) where: \[ E_0 = \left[ \begin{matrix} C & Z\\ -Z & -C \end{matrix} \right] ,\quad A_0 = \left[ \begin{matrix} A & B\\ B & A \end{matrix} \right], \] \(A,B,C,Z \in \mathbb{C}^{n \times n},\) \(A=A^*,\) \(B = A^*,\) \(C = A^*,\) \(Z = - A^*.\) If \(H \in \mathbb{C}^{n \times n}\) is a nonsingular and Hermitian or skew-Hermitian matrix , then a matrix \(A\in\mathbb{C}^{n \times n}\) is called \(H\)-self-adjoint (respectively \(H\)-skew-adjoint, \(H\)-unitary) if \(A^*H = HA\) (respectively \(A^*H =-HA\), \(A^*HA = H\)). With: \[ \Sigma=\left[ \begin{matrix} I_n & O\\ O & -I_n \end{matrix} \right],\quad \Gamma = \left[ \begin{matrix} O & I_n\\ I_n & 0 \end{matrix} \right],\quad J = \left[ \begin{matrix} O & I_n\\ -I_n & O \end{matrix} \right], \] \[ \mathbb{G}\mathbb{P}_{2n} = \{U \in \mathbb{C}^{2n \times 2n}; J^{-1}U^*J = \Gamma^{-1}U^*\Gamma, det(U) \neq 0\}; \] \(A \in \mathbb{C}^{n \times n}\) is called Hamiltonian, skew-Hamiltonian, and symplectic if \(A\) is \(J\)-skew-adjoint, \(J\)-self-adjoint, \(J\)-unitary, respectively. A pencil \(\lambda M - N \in \mathbb{C}^{2n \times 2n}\) is called \(\Gamma\)-self-adjoint, if \(M\) and \(N\) are \(\Gamma\)-self-adjoint; \(\lambda M - N\) is called skew-Hamiltonian/Hamiltonian, if \(M\) is skew-Hamiltonian and \(N\) is Hamiltonian. Among other results, the authors prove: Let \({\mathcal B}=\lambda {\mathcal E}-{\mathcal A} \in \mathbb{C}^{2n \times 2n}\) be an \(\infty\)-regular, \(\Gamma\)-self-adjoint, and skew-Hamiltonian/Hamiltonian pencil. 1) If there exists a matrix \(P \in \mathbb{G}\mathbb{P}_{2n}: J^{-1}P^*J {\mathcal B}P\) is in anti-triangular form, then there exists an unitary matrix \(Q \in \mathbb{G}\mathbb{P}_{2n}:J^{-1}Q^*J {\mathcal B}Q\) is in anti-triangular form. 2) If \(\lambda_0\) is a non-real and not purely imaginary eigenvalue having algebraic multiplicity \(r\), then there exists an unitary matrix \[ {\mathcal P} = diag(P_1,P_2) \in \mathbb{G}\mathbb{P}_{2n}: J^{-1}{\mathcal P}^*J {\mathcal B}{\mathcal P} = \lambda \left[ \begin{matrix} {\hat E} & O\\ O & {\hat E}^* \end{matrix} \right] -\left[ \begin{matrix} O & {\hat G}\\ {\hat H} & O \end{matrix} \right]; \] where the matrices \({\hat E}\), \({\hat G}\), \({\hat H}\) have the form: \[ X = \left[ \begin{matrix} O & O & X_{13}\\ O & X_{22} & X_{23}\\ X_{31} & X_{32} & X_{33} \end{matrix} \right] ; \] with \(X_{13} , X_{31} \in \mathbb{C}^{r \times r}\) lower anti-triangular. It is interesting that the existence of transformation associated with anti-triangular form implies the existence of associated unitary transformation, and numerically stable algorithms .