Behavior decompositions and two-sided diophantine equations (Q5947648)

The authors present the possibility of expressing a behavior \({\mathcal B}\) as the sum of two sub-behaviors \({\mathcal B}_1\) and \({\mathcal B}_2\), endowed with a finite-dimensional and hence autonomous intersection \({\mathcal B}_1\cap{\mathcal B}_2\), where \({\mathcal B}_1\) is a priori chosen. For the description of \({\mathcal B}_1\), a full row rank polynomial matrix \(B\) is chosen. If \(\ker R\) is any kernel representation of \({\mathcal B}\), and \({\mathcal B}_1=\ker B\) is included in \({\mathcal B}\), then \(R=AB\) for some polynomial matrix \(A\). If we set \({\mathcal B}_2= \ker H\), then \({\mathcal B}_1\cap{\mathcal B}_2=\ker\left [\begin{smallmatrix} B\\ H\end{smallmatrix} \right]\). If \(\left[ \begin{smallmatrix} B\\ H\end{smallmatrix} \right]\) is of full column rank (which is the condition for \({\mathcal B}_1\cap{\mathcal B}_2\) being autonomous) and \(\Delta(B,H)\) denotes one of its greatest right divisors, \({\mathcal B}_1\cap {\mathcal B}_2\) can also be described as \(\ker\Delta (B,H)\). Next, \({\mathcal V}(\Delta(B,H))\) denotes the variety of (the determinant of) a greatest right divisor \(\Delta(B,H)\) of \(\left[\begin{smallmatrix} B \\ H\end{smallmatrix}\right]\), as \(H\) varies over the set of matrices satisfying \({\mathcal B}_1+ \ker H=\ker(AB)\) with \({\mathcal B}_1\cap\ker H\) autonomous, and \[ {\mathcal D}^*: =\bigl\{\delta \in\mathbb{R} [\xi]: \exists\;X,Y\text{ polynomial matrices s.t. }XA+BY =\delta I_r\bigr\}. \] A first result states that having given \({\mathcal B}\in {\mathcal L}^q\) and \({\mathcal B}_1:=\ker B\), if there exists \(B_2= \ker H\in {\mathcal L}^2\) s.t. \({\mathcal B}\) can be expressed as \({\mathcal B}={\mathcal B}_1+ {\mathcal B}_2\), with \({\mathcal B}_1\cap {\mathcal B}_2\) an autonomous behavior, then \(\delta :=\det (\Delta(B,H))\) belongs to \({\mathcal D}^*\). Conversely, if \(\delta \in{\mathcal D}^*\), and hence there exist polynomial matrices \(X\) and \(Y\), of suitable dimensions, that solve the two-sided diophantine equation \(XA+BY =\delta I\), then there exists \({\mathcal B}_2= \ker H\in {\mathcal L}^2\) s.t. \({\mathcal B}= {\mathcal B}_1+ {\mathcal B}_2\), with \({\mathcal B}_1\cap {\mathcal B}_2\) autonomous, and \({\mathcal V}(\Delta (B,H))\subseteq V(\delta)\). Moreover, it is proved that a decomposition \({\mathcal B}={\mathcal B}_1 +{\mathcal B}_2\), with autonomous and hence finite-dimensional intersection, is a ``minimal intersection'' decomposition, in terms of the distinct zeros of \(\delta^*\), a generator of \({\mathcal D}^*\), where the ``minimality of the intersection'' means the minimality of \({\mathcal V}(\det \Delta (B,H))\). Another result, in the case \(1=\delta (=\delta^*)\), states a necessary and sufficient condition for the existence of a direct sum decomposition; namely, having given \({\mathcal B}\in {\mathcal L}^2\) and \({\mathcal B}_1: =\ker B\), with \(B\) an \(r\times 2\) full row rank matrix, a sub-behavior of \({\mathcal B}\), then the following facts are equivalent: (i) \({\mathcal B}={\mathcal B}_1\oplus {\mathcal B}_2\), for some \({\mathcal B}_2\in {\mathcal L}^2\), (ii) for every polynomial matrix \(A\) s.t. \({\mathcal B}=\ker(AB)\), the pair \((A,B)\) is internally skew-prime . Moreover, if the above condition holds, \({\mathcal B}_2\) is autonomous (controllable) if and only if \(\ker A\) is. In particular, if \({\mathcal B}\) is a behavior in \({\mathcal L}^2\) and \({\mathcal B}_c\) is its controllable part, then \({\mathcal B}={\mathcal B}_c\oplus {\mathcal B}_a\), for some autonomous behavior \({\mathcal B}_a\in {\mathcal L}^2\).