AtlAS: Accessing longer time and length scales in atomistic simulations

High performance computing (HPC) centers are moving towards the exascale transition. In recent years a strong effort has been devoted to enable atomistic simulations to efficiently scale on these novel hybrid architectures. Despite significant advances, accurate modeling of complex reactive phenomena in materials such as oxidation, degradation, and interface growth, remains an open challenge both in terms of obtaining a physico-chemical understanding of these phenomena, as well as in the field of scientific software development. A particular issue in this respect is how to address rare events, i.e., events that are associated with energy barriers much larger than kT. A method that is particularly suited to model such events is known as kinetic Monte Carlo (kMC). In this method, the time evolution of a system is governed by the probability of the transition between the various states the system can occupy. In order to obtain this probability it relies on transition state theory, which states that the system occupies minima on the potential energy surface (PES), whereas the rate of transitions between minima is governed by the height of the energy barrier, characterized by a saddle point on the PES, connecting the two minima. Therefore, in order to perform a successful kMC simulation one needs to know all the critical points on the PES of the system under study. However, the description of critical points on the PES depends on the employed atomic interaction model, therefore addressing complex phenomena requires a kMC, interaction model and an algorithm that explores the PES. In other words, the software performing the kMC simulation requires interoperability and the possibility to be easily extended. One of the main issues, in terms of software development, is that the vast majority of software capable of performing kMC simulations has mainly been developed with a particular system in mind, often imposing the so-called lattice assumption. The main objective of the proposed research project is therefore to develop and establish a flexible computational framework that allows the simulation of complex reactive phenomena based on kMC and apply it on a prototypical and scientifically interesting case, the formation of a thin oxide film on the surfaces of aluminum. Specifically, we will incorporate and extend the recently developed implementation of the Activation Relaxation Technique (ART), that is capable of identifying the critical points on the PES by utilizing the well-established Quantum ESPRESSO (QE) and LAMMPS software packages. Additionally, we will include a recently developed shape matching algorithm, named Iterative Rotations and Assignments (IRA). By implementing these tools our aim is to provide a framework capable of performing an on-the-fly kMC simulation, based both on empirical and first-principle interaction potentials. The framework itself will be designed in a modular way, as to enable seamless incorporation of alternative methods and algorithms as well as be capable of utilizing the full parallel capabilities of modern HPC centers. Finally, an additional goal is to make the data generated by this framework adhere closely to the FAIR principles, i.e., to be Findable, Accessible, Interoperable, and Reusable.