More on Electrostatic Models for Zeros of Orthagonal Polynomials (Q4484804)

Recall that a matrix \(c=(c_{ij})_{i,j=1}^n\) is called a (Toeplitz-plus-Hankel-) \(r\)-Bézoutian if the \((n+2)\times(n+2)\) matrix \((c_{i-1,j}-c_{i,j-1}-c_{i-2,j- 1}+c_{i-1,j-2})_{i,j=1}^{n+2}\) (where \(c_{ij}:=0\) if \(i\) or \(j\) is \(<0\) or \(>n\)) has rank \(\geq r\). The inverse of a nonsingular Toeplitz-plus-Hankel matrix is a 4-Bézoutian, and all nonsingular 4-Bézoutians arise in this way. The authors exhibit representations of \(r\)-Bézoutians involving only discrete Hartley transforms (DHTs) and diagonal matrices. Using these representations, it follows that the multiplication of a real \(r\)-Bézoutian of size \(n\times n\) by a column vector can be accomplished with the help of \(r+2\), if \(r\) is even, or \(r+3\), if \(r\) is odd, DHTs, plus \(2r\) DHTs for preprocessing. In~particular, multiplication of a column vector by an inverse of a real \(n\times n\) Toeplitz-plus-Hankel matrix requires only 6 Hartley transforms plus \(O(n)\) operations. This significantly improves previously known complexity estimates, also due to the present authors [Linear Algebra Appl. 284, No.~1-3, 157-175, (1998; Zbl 0938.65073)].