The micro-iterative method divides the whole system in two parts, a core zone where an accurate second order search (a Rational function optimization method in our case) of the transition state structure is done, and an environment that is kept minimized with a cheap first order method (an L-BFGS method in this case). Our results show that the core has to include at least all the atoms that participate significantly in the reaction coordinate of the corresponding reaction step. Otherwise, the right transition state structures are not reached. Indeed, this threshold size of the core depends a lot on each particular reaction step. Beyond this minimum core size, there is an interval of medium core sizes that minimizes the total CPU time. This arises from a compromise among a lower core/environment coupling, the efficiency of the Rational function optimization method in handling a progressively increasing number of core atoms, and the monotone augment of the CPU time required to calculate the initial Hessian matrix, as the core size grows. As a consequence, the use of a core as great as possible is not advised. In other words, above that threshold size of the core, which depends of the relevant motions taking place during the corresponding chemical step, it is not true that the larger the core size the more efficient the TS search.
A considerable amount of CPU time can be saved if only one SCF cycle is performed to evaluate the potential energy during the environment minimization. This option is clearly faster than a complete QM/MM energy calculation and leads to the right transition state structures. Conversely, the use of the ESP charges to simplify the calculation of the interaction energy between the QM and the MM regions seems to be less efficient.
Finally, we have to remark that the location of the transition state structures in enzyme catalysis is needed to define reliable reaction pathways along which a set of generalized free energies can be calculated. In this sense, an extreme accuracy in the geometrical or energetic parameters of the transition state structures is not required. We just need that the right pathway can be built up starting from the transition state structure, so avoiding the use of reaction paths that run on wrong potential energy valleys. However, when a QM(ab initio)/MM scheme is used, a configurational sampling becomes prohibitive. Therefore, an accurate location of stationary points will be the only strategy to provide meaningful results.