My research is centered on the development and application of large-scale atomistic simulation methodologies. These computational approaches are essential for advancing the understanding of complex atomic-scale phenomena occurring at material interfaces, with a particular focus on processes relevant to nano-tribology.
In the initial phase of my work, I investigated and adopted Green's function techniques to accurately reproduce the elastic response of solids subjected to vibrational excitations within interfacial regions. This methodology enables a substantial reduction in the number of atoms required to model the bulk regions of interfaces, thus optimizing the use of computational resources.
More recently, with the advent of machine learning techniques within the domain of atomistic simulations, my research has shifted towards the deployment of machine learning force fields (ML-FFs). These models employ advanced machine learning strategies to parameterize the intricate interatomic interactions. In particular, deep neural network architectures have demonstrated remarkable success: once trained on high-fidelity ab initio datasets (approximate solutions to the many-body electronic Schrödinger equation), these models are capable of reproducing atomic interactions with orders-of-magnitude improvements in efficiency—up to 10,000 times faster—compared to conventional methods.
My current research activities focus on the design and implementation of automated pipelines and workflows for the generation of ab initio data and the subsequent training of ML-FF models. These models are ultimately employed in large-scale atomistic simulations aimed at exploring the fundamental mechanisms governing interfacial phenomena.
