An unconditionally convergent method for computing zeros of splines and polynomials

Abstract

We present a simple and efficient method for computing zeros of spline functions. The method exploits the close relationship between a spline and its control polygon and is based on repeated knot insertion. Like Newton’s method it is quadratically convergent, but the new method overcomes the principal problem with Newton’s method in that it always converges and no starting value needs to be supplied by the user.