Three algorithms for optimization of minimum energy conical intersections (MECI) are implemented inside an ONIOM(QM:MM) scheme combined with microiterations. The algorithms follow the composed gradient (CG), composed gradient-composed steps (CG-CS), and double Newton-Raphson-composed step (DNR-CS) schemes developed previously for purely QM optimizations. The CASSCF and UFF methods are employed for the QM and MM calculations, respectively. Conical intersections are essential to describe excited state processes in chemistry, including biological systems or functional molecules, and our approach is suitable for large molecules or systems where the excitation is well localized on a fragment that can be treated at the CASSCF level. The algorithms are tested on a set of 14 large hydrocarbons composed of a medium-sized chromophore (fulvene, benzene, butadiene, and hexatriene) derivatized with alkyl substituents. Thanks to the microiteration technique, the number of steps required to optimize the MECI of the large molecules is similar to the one needed to optimize the unsubstituted chromophores at the QM level. The three tested algorithms have a similar performance, although the CG-CS implementation is the most efficient one on average. The implementation can be straightforwardly applied to ONIOM(QM:QM) schemes, and its potential is further demonstrated locating the MECI of diphenyl dibenzofulvene (DPDBF) in its crystal, which is relevant for the aggregation induced emission (AIE) of this molecule. A cluster of 12 molecules (528 atoms) is relaxed during the MECI optimization, with one molecule treated at the QM level. Our results confirm the mechanistic picture that AIE in DPDBF is due to the packing of the molecules in the crystal. Even when the molecules surrounding the excited molecule are allowed to relax, the rotation of the bulky substituents is hindered, and the conical intersection responsible for radiationless decay in solution is not accessible energetically.
