On the Smith Normal Form of Structured Polynomial Matrices

From MaRDI portal





Let \(D\) be an \(m\times n\) matrix with entries in the polynomial ring \(K[s]\), where \(K\) is a field and \(s\) an indeterminate over \(K\). The author is interest in the efficient computation of the Smith normal form of \(D\) when \(D\) has structure of a specified kind. Let \(F\) be a fixed subfield of \(K\). It is required tht \(D\) have the form \(Q+T\), where every determinantal minor (and so, in particular, every entry) of \(Q\) is a monomial \(\lambda s^ k\) \((\lambda\in F)\) and where the nonzero coefficients of the elements of \(T\) are algebraically independent over \(F\). For example, when \(K\) and \(F\) are the real and rational fields respectively, \(D\) could be \[ \begin{pmatrix} 1 & 2s\\ 3s^ 2 & 0\end{pmatrix}+\begin{pmatrix} p_ 1s & p_ 2+p_ 3s^ 3\\ 0 & p_ 4\end{pmatrix}, \] where \(p_ 1,\dots,p_ 4\) are real numbers algebraically independent over \(F\). Roughly speaking, the motive for taking \(D\) in the above form is to take account of ``accurate values (such as theoretical constants) and ``inaccurate values (such as measurements of physical quantities) that arise in physical and engineering problems. The accurate values are represented by elements of \(F\) and the inaccurate ones by elements of \(K\) algebraically independent over \(F\). The practical problem is to calculate the determinantal divisors \(d_ i(s)\) \((1\leq i\leq r)\), where \(r\) is the rank of \(D\). It turns out that, for \(i<r\), \(d_ i(s)\) is simply a power of \(s\). The situation for \(i=r\) is more complicated. Let \(\Delta_ m=-\hbox{diag}(t_ 1,\dots,t_ m)\), where the \(t_ i\) and the nonzero coefficients of the elements of \(T\) are together algebraically independent over \(F\), and form the \((2m)\times(m+n)\)-matrix \(\tilde D=\begin{pmatrix} I_ m & Q\\ \Delta_ m &T\end{pmatrix}\). By certain admissible transformations one converts \(\tilde D\) into a block triangular matrix called its combinatorial canonical form. The polynomial \(d_ r(s)\) can then be read off from the diagonal blocks. The paper concludes with a description of efficient matroid-theoretic algorithms for calculating the \(d_ i(s)\) in practice.











This page was built for publication: On the Smith Normal Form of Structured Polynomial Matrices

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3980617)