In this project machine learning (ML) interatomic potentials are trained and used in molecular dynamics (MD) simulations to predict the physical properties of total energy, mean squared displacement (MSD) and specific heat capacity for systems of bulk Aluminium and Silicon. The interatomic potentials investigated are potentials trained using the ML models kernel ridge regression (KRR) and moment tensor potentials (MTPs). The simulations using these ML potentials are then compared with results obtained from ab-initio simulations using the gold standard method of density functional theory (DFT), as implemented in the Vienna ab-intio simulation package (VASP). The results show that the MTP simulations reach comparable accuracy compared to the DFT simulations for the properties total energy and MSD for Aluminium, with errors in the orders of magnitudes of meV and 10-5 Å2. Specific heat capacity is not reasonably replicated for Aluminium. The MTP simulations do not reasonably replicate the studied properties for the system of Silicon. The KRR models are implemented in the most direct way, and do not yield reasonably low errors even when trained on all available 10000 time steps of DFT training data. On the other hand, the MTPs require only to be trained on approximately 100 time steps to replicate the physical properties of Aluminium with accuracy comparable to DFT. After being trained on 100 time steps, the trained MTPs achieve mean absolute errors in the orders of magnitudes for the energy per atom and force magnitude predictions of 10-3 and 10-1 respectively for Aluminium, and 10-3 and 10-2 respectively for Silicon. At the same time, the MTP simulations require less core hours to simulate the same amount of time steps as the DFT simulations. In conclusion, MTPs could very likely play a role in accelerating both materials simulations themselves and subsequently the emergence of the data-driven materials design and informatics paradigm.

Une approche donnant de meilleurs résultats pour les potentiels interatomiques en apprentissage-machine à la volée est proposée en comparant trois approches pour la recherche de processus activés par la technique d'activation-relaxation. Tout d'abord, nous discutons de l'intérêt et des enjeux de l'utilisation des potentiels en apprentissage-machine et justifions l'utilisation de l'apprentissage à la volée pour la recherche de processus activés. Cela nous mène à présenter la forme générale des potentiels en apprentissage-machine, quelques modèles via leurs descripteurs de configuration atomique, paramètres et hyperparamètres ainsi que la méthode de l'apprentissage à la volée. Ensuite, nous présentons les méthodes d'exploration utilisées et les détails d'intégration du potentiel à la volée. Enfin, nous menons une étude comparative des trois approches pour un système de Si et de SiGe avec diffusion de lacune. La méthodologie proposée de potentiel de haute précision permet d'étendre la gamme de problèmes pouvant être étudiés par la technique d'activation-relaxation. / An approach giving better results for on-the-fly machine learning interatomic potential proposed by comparing three approaches for exploration of activated processes by the activationrelaxation technique. We first discuss the interest and challenges of on-the-fly machine learning potential and justify the use of on-the-fly learning for the search for activated processes. This leads us to present the general form of machine learning potentials and some models via their atomic configuration descriptors, parameters and hyperparameters as well as the on-the-fly learning method. Then, the exploration methods used are defined and the details of the integration of the potential are presented. Finally, a study is conducted comparing the three approaches for a Si and SiGe system with vacancy diffusion. The proposed methodology of high-precision potential allows to extend the range of possible problems to be studied by the activation-relaxation technique.