Structured invariant spaces of vector valued rational functions, Hermitian matrices, and a generalization of the Iohvidov laws (Q919073)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 4158862
Language Label Description Also known as
default for all languages
No label defined
    English
    Structured invariant spaces of vector valued rational functions, Hermitian matrices, and a generalization of the Iohvidov laws
    scientific article; zbMATH DE number 4158862

      Statements

      Structured invariant spaces of vector valued rational functions, Hermitian matrices, and a generalization of the Iohvidov laws (English)
      0 references
      0 references
      0 references
      1990
      0 references
      The theory of finite dimensional reproducing kernel spaces with an indefinite inner product is applied to study the relations between inertia of principal submatrices of Hermitian solutions to certain matrix equations. Let be given matrix equations \[ (1)\quad P-(Z_ w)^*PZ_{w^*}=V^*JV \] \[ (2)\quad PZ_{w^*}- (Z_{w^*})^*P=(2\pi i)^{-1}V^*JV \] where \(Z_ w\) is the \(n\times n\) upper triangular Jordan block with the eigenvalue \(w\in {\mathbb{C}}\), J is a given \(2\times 2\) matrix which is indefinite Hermitian and unitary, and the \(2\times n\) matrix V is given as well. Hermitian Toeplitz and Hankel matrices are, in particular, solutions of equations of these types. One of the main results (in a simplified form): Let \(P=P^*\) be a singular solution to (1), where \(| w| \neq 1\) and the first column in V is nonzero. Let \(P_{[j]}\) be the upper left \(j\times j\) block of P, and let r be the maximal index j such that \(P_{[j]}\) is invertible \((r=0\) if there is no such j). Denote by \(k\times k\) the size of the largest all zero block in the upper left hand corner of P [the Schur complement of \(P_{[r]}]\) if \(r=0\) \([r>0]\). Then \[ \mu_{\pm}(P_{[r+j]})=\mu_{\pm}(P_{[r]}),\quad i=1,...,k \] \[ \mu_{\pm}(P_{[r+j]})=\mu_{\pm}(P_{[r]})+j-k,\quad j=k+1,...,n-r. \] Here by \(\mu_+(A)\) (resp. \(\mu_ -(A))\) we denote the number of positive (resp. negative) eigenvalues of a Hermitian matrix A, counted with multiplicities. Analogous results are proven for \(| w| =1\) and for Hermitian solutions P of equation (2). The proofs are based on the identification of P as the Gram matrix of an appropriately chosen basis in some finite dimensional vector space of rational column functions. This vector space is invariant under generalized shifts, carries an indefinite (generally degenerate) inner product, and satisfies a suitable identity. There are stated (with proofs appearing elsewhere) some results in the same spirit concerning the relations between ranks of principal submatrices of solutions P to equations \[ P-(Z_{w^*})^*PZ_{w^*}=U^*JV \] \[ PZ_{w^*}- (Z_{w^*})^*P=(2\pi i)^{-1}U^*JV, \] where \(Z_ w\) and J are as above.
      0 references
      0 references
      Iohvidov laws
      0 references
      finite dimensional reproducing kernel spaces
      0 references
      indefinite inner product
      0 references
      inertia
      0 references
      matrix equations
      0 references
      Hermitian Toeplitz and Hankel matrices
      0 references
      Schur complement
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references
      0 references