An eigenvector-following technique [24] is used to locate minima and transition states on the potential energy surface (PES). All stationary points are characterised by normal mode analysis using analytic derivatives [9]. Local minima have only real normal mode frequencies, and we follow Murrell and Laidler [25] in defining a true transition state as a stationary point with precisely one imaginary normal mode frequency. Details of the precise algorithms employed in these geometry optimisations and reaction pathway calculations have been given elsewhere [9, 26] and numerous previous applications have been described [9, 10, 11, 26]. Here we simply note that eigenvector-following provides a means to systematically raise the energy for one degree of freedom, while simultaneously minimising in all the conjugate directions. We have implemented eigenvector-following in ORIENT 3.2 [27] so that we can suspend both Monte Carlo and Molecular Dynamics runs at pre-determined intervals, implement a minimisation or transition state search from the instantaneous geometry, and then resume the simulations from where they left off.