RFO method avoids the Hessian matrix inversion required by a quasi-Newton-Raphson method. It also introduces in an automatic way a shift that preserves the current behavior of the optimization process. This algorithm has been successfully tested for a variety of chemical and biochemical systems from small to medium size. The good behavior of the algorithm presented here has encouraged us to extend it to locate minima and transition states on QM/MM potential-energy surfaces corresponding to real reactive biochemical systems including thousands of atoms.
In the next sections we will modify this algorithm in order to take into account the special problems derived from the large size of those systems, but still handling the information contained in a full Hessian matrix. In section 3.2 the method is implemented in the micro-iterative scheme. In section 3.4 the problem of the storage and diagonalization of big matrices is studied.