Quadratically normal matrices of type 1 and unitary similarities between them (Q1048577)

Specht's Theorem [cf. \textit{W. Specht}, Jahresber. Dtsch. Math.-Ver. 50, Abt. 1, 19--23 (1940; Zbl 0023.00203)] gives a criterion for Unitary Similarity between two complex matrices \(A\) and \(B\) which is not practical due to verification of an infinite set of conditions. Later on, \textit{C. Pearcy} [Pac. J. Math. 12, 1405--1416 (1962; Zbl 0144.37801)] modified this theorem to use a finite number of calculations. Still, huge computational work is required to apply the Pearcy method. This work can be reduced if \(A\) and \(B\) considered to be normal matrices. The author of the paper under review uses this idea and gives an effective method to determine whether \(A\) and \(B\) are unitarily similar for special matrix classes. Let \(M_{n}\) denote the set of all complex matrices of order \(n\). A matrix \(A\in M_{n}\) is a polynomially normal matrix of degree \(d\) if there exists a monic polynomial \(p(z)\) of degree \(d\) such that \(p(A)\) is a normal matrix, while \(q(A)\) is not normal if the degree of \(q(z)\) is less than \(d\). \(p(z)\) is called the normal polynomial of this matrix. When \(d=2\), the matrix is called quadratically normal matrix. A quadratically normal matrix \(A\) is called of type \(k\) if the rank of the normal matrix \(p(A)\) is equal to \(k\). The author of the paper under review formulates and proves the following theorem: Theorem. Let \(A\) and \(B\in M_{n}\) be quadratically normal matrices of type 1, then these matrices are unitarily similar if and only if \(A\) and \(B\) have the same eigenvalues and the same singular values.