On the pole assignability property over commutative rings (Q1107583): Difference between revisions

If \(F\in R^{n\times n}\) and \(G\in R^{n\times m}\) are matrices over a commutative ring R, the system (F,G) is said to be reachable if the columns of the matrix \([G,FG,...,F^{n-1}G]\) span R n. The system (F,G) is pole-assignable if for each \(r_ 1,...,r_ n\in R\), there exists \(K\in R^{m\times n}\) such that the matrix \(F+GK\) has \((X-r_ 1)...(X- r_ n)\) as its characteristic polynomial, and the ring R is said to have the PA-property if every reachable system over R is pole-assignable. The authors call a matrix \(G\in R^{n\times m}\) good if there exists an \(F\in R^{n\times n}\) such that the system (F,G) is reachable, and they say that the ring R has the GCU-property if every good matrix has a unimodular vector in its image. The authors prove that if R has the GCU- property, then R has the PA-property. They note that the converse is not true in general, but prove the converse does hold provided that all rank one projective modules over R are free. It is also shown that if R has the PA-property and G is a good matrix over R, then there exists a matrix V over R such that the entries in GV generate the unit ideal in R while all \(2\times 2\) minors from GV are zero. The authors conjecture that the converse of this result is also true, that is, if the above property holds for every good matrix over R, then R has the PA-property.