Improved atomistic Monte Carlo models based on ab-initio-trained neural networks Application to FeCu and FeCr alloys
Abstract
We significantly improve the physical models underlying atomistic Monte Carlo (MC) simulations, throughthe use of ab initio fitted high-dimensional neural network potentials (NNPs). In this way, we can incorporateenergetics derived from density functional theory (DFT) in MC, and avoid using empirical potentials that arevery challenging to design for complex alloys. We take significant steps forward from a recent work whereartificial neural networks (ANNs), exclusively trained on DFT vacancy migration energies, were used to performkinetic MC simulations of Cu precipitation in Fe. Here, a more extensive transfer of knowledge from DFTto our cohesive model is achieved via the fitting of NNPs, aimed at accurately mimicking the most importantaspects of the ab initio predictions. Rigid-lattice potentials are designed to monitor the evolution during thesimulation of the system energy, thus taking care of the thermodynamic aspects of the model. In addition,other ANNs are designed to evaluate the activation energies associated with the MC events (migration towardsfirst-nearest-neighbor positions of single point defects), thereby providing an accurate kinetic modeling. Becauseour methodology inherently requires the calculation of a substantial amount of reference data, we design as welllattice-free potentials, aimed at replacing the very costly DFT method with an approximate, yet accurate andconsiderably more computationally efficient, potential. The binary FeCu and FeCr alloys are taken as sampleapplications considering the extensive literature covering these systems.