Spatial decomposition of functionally commutative matrices (Q920182)

An \(n\times n\) matrix-valued function F defined on an interval I is said to be proper on I if F can be represented in the form f(t) for some fixed matrix A; here f(t,) is a scalar function for each t with f(t,A) defined by the usual functional calculus for \(n\times n\) matrices. The matrix valued function F is said to be semi-proper if its values commute, i.e. \(F(s)F(t)=F(t)F(s)\) for all s and t in I. An earlier result of \textit{J. F. P. Martin} [SIAM J. Appl. Math. 15, 1171- 1183 (1967; Zbl 0155.356)] characterizes bounded, piecewise continuous semiproper matrix functions F as those of the form \(F(t)=\sum^{M}_{k=1}\alpha_ k(t)A_ k\) where \(\{A_ k:\) \(k=1,...,M\}\) is a set of mutually commutative constant matrices and \(\{\alpha_ k(t)\}\) is a set of bounded piecewise continuous scalar functions. This result is important theoretically but has limited practical use. The present paper obtains a structural characterization of proper functions, obtains a characterization of semiproper matrix functions in terms of proper ones, presents a systematic procedure for decomposing a semiproper matrix function into a finite sum of mutually commutative proper ones. A brief survey of some applications of these results in control engineering, linear systems theory and the theory of linear differential equations is also included.