On Matrices with Elliptical Numerical Ranges (Q4818360)

For an \(n\)-by-\(n\) complex matrix \(A\), its numerical range \(W(A)\) is by definition the subset \(\{x^* Ax: x\in \mathbb{C}^n,\,x^* x=1\}\) of the plane. It has long been known that, for \(n= 2\), \(W(A)\) is a (closed) elliptic disc. In this paper, it is first shown that if \(A\) is unitarily equivalent to a matrix of the form \[ \left[\begin{matrix} a_1 I_{n_1} & X\\ Y^* & a_2 I_{n_2}\end{matrix}\right] \] with \(XY^*\) and \(Y^*X\) both normal, then \(W(A)\) is the convex hull of at most \(\min\{n_1, n_2\}\) ellipses. If, moreover, \(Y= bX\) for some scalar \(b\), then \(W(A)\) is an elliptic disc. This latter assertion generalizes the corresponding result for quadratic matrices \(A\) (\(A\) satisfying \(A^2+ c_1A+ c_2I_n= 0\) for some scalars \(c_1\) and \(c_2\)) obtained before by \textit{S.-H. Tso} and this reviewer [Rocky Mt. J. Math. 29, No. 3, 1139--1152 (1999; Zbl 0957.47005)]. The assertion is then applied to the class of tridiagonal matrices of the form \([a_{ij}]^n_{i,j=1}\), where \(a_{11}= a_{33}=\cdots\), \(a_{22}= a_{44}=\cdots\), \(a_{i+1,i}= b\overline a_{i,i+1}\) or \(a_{i,i+1}= b\overline a_{i+1,i}\) for some fixed scalar \(b\), and \(a_{ij}= 0\) for \(| i-j|> 1\), to infer that \(W(A)\) is also always an elliptic disc.