Ab-initio and large-scale simulations of the migration of impurities and defects in solids under the influence of femtosecond laser excitations

This thesis addresses a relevant problem in materials science: elucidating the atomic mechanisms of impurity diffusion in silicon under the influence of femtosecond laser pulses. As sulfur (S) doping is crucial for advanced optoelectronic applications, a precise understanding of its spatial distribution is essential, yet hindered by the complexity of its migration under laser-induced non-equilibrium conditions. To address this problem, our work developed a multiscale simulation approach, starting with ab initio molecular dynamics simulations using the Code for Highly- excIted Valence Electron Systems (CHIVES) which models the behaviour of matter under intense electronic excitations. These simulations generated accurate data on silicon-silicon and silicon-sulfur interactions as well as the dynamic and structural response of silicon under various electronic excitations. These data then served as the basis for training and validating our Deep Potential Molecular Dynamics (DeepMD) Machine Learning Interatomic Potential (MLIP), whose accuracy and transferability were confirmed by close agreement with benchmark Mean Square Displacements (MSDs) and Velocity Autocorrelation Functions (VACFs). Using the validated MLIP, we conducted large-scale molecular dynamics simulations with LAMMPS (Large-scale Atomic Molecular Massively Parallel Simulator) on 10648-atom supercells, exploring electronic temperatures up to 18946 K. Theses temperatures represents the energy femtosecond laser pulses deposit in the electronic subsystem. Our results reveal the atomic behaviour: while at 300 K the system maintains a stable crystalline state, in the high electronic excitation regime (9473 K, 12631 K, 18946 K), we observed an ultrafast and massive diffusion of silicon and sulfur atoms. MSDs grew dramatically compared to thermal regimes, reaching thousands of Å2 over 12 ps, a signature of extreme mobility. Structural analyses using Radial Distribution Functions (RDFs) and Centrosymmetry Parameter (CSP) confirmed a near-complete collapse of the crystalline order, creating an environment conducive to this accelerated diffusion, while the VACF corroborated this dynamics by damping almost instantaneously. Quantification of these phenomena by calculating the diffusion coefficient revealed an exponential increase with effective electronic temperature, demonstrating the existence of laser-induced non-thermal diffusion mechanisms that allow atomic mobility inaccessible by conventional heating. This work thus makes a contribution to the physics of laser induced doped solids, validates the MLIP potential for non-equilibrium materials simulations, and opens concrete perspectives for precision doping and optimized manufacturing of black silicon in advanced optoelectronics.