Determinantal representations of invariant hyperbolic plane curves (Q1668990): Difference between revisions

A homogeneous real polynomial \(f\) in three indeterminates \(t\), \(x\), \(y\) is called \textit{hyperbolic} with respect to a point \(e \in \mathbb{P}^2(\mathbb{R})\) if every line through \(e\) intersects the real curve \(\mathcal V_{\mathbb R}(f)\) in \(\deg f\) points. It is well-known that there exists a definite matrix \(M = tM_0 + xM_1 + yM_2\) such that \(f = \det M\). Conversely, given a \(A \in {\mathbb C}^{n \times n}\) one can associate the polynomial \(f_A := \det(tI_n + x\Re(A) + y\Im(A))\) to \(A\). Here \(I_n\) is the identity matrix of dimension \(n \times n\) and \(\Re(A)\) and \(\Im(A)\) denote is real and imaginary part, respectively. If \(A\) is a real cyclic weighted shift matrix, then \(f_A\) is hyperbolic with respect to \((1, 0, 0)\) and it is invariant under the action of the dihedral group \(D_n\). Answering a question by \textit{M. T. Chien} and \textit{H. Nakazato} [Appl. Math. Comput. 258, 172--181 (2015; Zbl 1338.15048)], the authors prove converse statements: Any plane curve that is hyperbolic with respect to \((1,0,0)\) and invariant under the action of the cyclic group of order \(n\) admits a determinantal representation \(f_A\) with a cyclic weighted shift matrix \(A\). If the curve is even invariant under the action of the dihedral group, a real matrix with this property exists. Their proof yields a computer algebra construction for \(A\). Due to the necessity of computing the intersection of \(f\) with \(\partial f/\partial t\) it is unfortunately not feasible for large~\(n\). Above main result is related to the \(k\)-higher rank numerical range of a matrix \(A \in \mathbb{C}^{n \times n}\), defined as \[ \mathcal W_k(A) := \Bigl\{ \sum_{j=1}^k x_j^\star A x_j \Bigr\} \] where \(\{x_1,x_2,\dots,x_k\}\) ranges over all orthonormal sets in \(\mathbb{C}^n\). In particular, if the classical numerical range \(\mathcal W_1(A)\) of a matrix \(A \in \mathbb C^{n \times n}\) exhibits an \(n\)-fold rotational symmetry, it is also the numerical range \(\mathcal W_1(B)\) of some cyclic weighted shift matrix \(B \in \mathbb C^{n \times n}\).