A new method to compute second derivatives

In thisarticle we consider the problem of computing approximations to the second derivatives of functions of n variables using finite differences. We show how to derive different formulas and how to comput the errors of those approximations as functions of the increment h, both for first and second derivatives. Based upon those results we describe the methods of Gill and Murray and the one of gradient difference. On the other hand we introduce a new algorithm which use conjugate directions methods for minimizing functions without derivatives and the corresponding numerical comparisons with the other two methods. Finally, numerical experiences are given and the corresponding conclusions are discussed