Constructing matrices with prescribed off-diagonal submatrix and invariant polynomials (Q886155)

The authors analyze the following problem which is called off-diagonal submatrix problem. Given: (a) an arbitrary field \(F\), (b) \(s\) monic nonconstant polynomials \(f_1,f_2 \dots, f_s \in F[x]\) with \(f_s | \cdots | f_1\), (c) two nonnegative integers \(p,q\) such that \(p+q=\sum_{i=1}^s \text{deg} f_i\), (d) a matrix \(B \in F^{q \times p}\). They are interested in the explicit construction of a matrix in \(F^{(p+q) \times (p+q)}\), \[ A=\left( \begin{matrix} A_{11,p} & A_{12,p} \\ A_{21,p} & A_{22,p} \end{matrix} \right), \] with invariant polynomials \(f_s | \cdots | f_1\) and such that \(A_{21,p}=B\) The authors introduce the notion of \(F\)-admissible set, that permits to state the off-diagonal submatrix problem in a compact way. An \(F\)-admissible set for which the off-diagonal submatrix problem has a solution is called \(F\)-solvable, and any solution is called a solution matrix. The main idea of the work is the following: to split an \(F\)-solvable set into smaller \(F\)-solvable sets such that a solution matrix can be found for each one and all these matrices can be used to construct a solution matrix for the original \(F\)-solvable set. This idea allows the authors to provide a diagram with an algorithm that constructs the solution, when it exists, of the off-diagonal submatrix problem for fields of characteristic different from two. They also solve several subcases of the off-diagonal submatrix problem.