Ab initio molecular dynamics (AIMD) methods consider the nuclear motions under the potential generated by electronic wavefunctions which are determined from ab initio quantum mechanical calculations on-the-fly. AIMD methods allow researchers to investigate chemical processes without prior knowledge or assumptions about the shape of the potential energy surface (PES). In this thesis, we applied AIMD methods to study silicon nanocrystals with dangling bond defects (DB-SiNCs). DB defects on SiNCs have been known as nonradiative (NR) decay centers. However, the atomistic mechanism for the decay process is unclear. Previously, researchers considered a pyramidalization mode surrounding the DB site involved in the process. Based on our AIMD calculations on the first excited state and the static analysis of the PESs of SiNC systems, we discovered that asymmetrical Si-Si bond stretching modes surrounding DB sites are important, in addition to pyramidalization. Most importantly, we found a low-lying defect-induced conical intersection (DICI) in the neutral DB system. The minimum energy conical intersection (MECI) is estimated to be 1.74 eV above the ground state minimum energy geometry by application of multi-state complete active space second-order perturbation theory (MS-CASPT2) to a small cluster model system. In addition, the roles of charged DBs on NR decay process are investigated. We found DICIs for both positively and negatively charged DB systems. The MECI energies are 2.10 eV and 2.65 eV respectively. The rationalization of the existence of conical intersections and detailed dynamics after excitation of these systems are discussed in the thesis. Additionally, to study the possible defect-defect interactions during the NR recombination process, we considered slab models with two DB defects at short (4 0303A) and long (100303 A) separations. According to our simulations, the NR recombination process is localized on a single DB site, regardless the defect-defect distances. However, energy transfer between defect sites with short separations is possible.For the defective SiNC systems, we demonstrated the power of the AIMD method to investigate the dynamics after excitations. However, the applications of AIMD to high-lying states are much more challenging, due to the dense manifold of states that cause immense computational effort. In the thesis, we developed several methods toward the application to such systems. First, we developed a time-dependent configuration interaction (TD-CI) method that can simulate the electron dynamics under a strong field efficiently. The method is based on the direct scheme to form the vector, , which can be accelerated by a graphical processing unit. A TD-CI calculation with 853776 determinants requires only 20.1 hours to propagate to 100 fs with 1 attosecond (10-18 second) time steps. On the other hand, when the field is strong enough, the electrons can be driven to the boundary of the basis set, which would cause unphysical effects such as reflection. To account for this, we developed an analytical expression for a molecule-centered complex absorbing potential which can be evaluated efficiently to remove the unwanted effects. Finally, for the nuclear dynamics, we developed an Ehrenfest dynamics method based on the TD-CI wavefunction. In this approach, the nuclear motions are propagated under the averaged potential generated by TD-CI wave function, thus the approach is promising for application to systems with dense manifolds of states.
Read
