On the classification and quantification of crystal defects after energetic bombardment by machine learned molecular dynamics simulations

Abstract

The analysis of the damage on plasma facing materials (PFM), due to its direct interaction with the plasma environment, is needed to build the next generation of nuclear machines, where tungsten has been proposed as a candidate. In this work, we perform molecular dynamics (MD) simulations using a machine learned inter-atomic potential, based on the Gaussian Approximation Potential framework, to model better neutron bombardment mechanisms in pristine W lattices. The MD potential is trained to reproduce realistic short-range dynamics, the liquid phase, and the material recrystallization, which are important for collision cascades. The formation of point defects is quantified and classified by a descriptor vector (DV) based method, which is independent of the sample temperature and its constituents, requiring only modest computational resources. The locations of vacancies are calculated by the k-d-tree algorithm. The analysis of the damage in the W samples is compared to results obtained by EAM Finnis-Sinclair and Tersoff-ZBL potentials, at a sample temperature of 300 K and a primary knock-on atom (PKA) energy range of 0.5-10 keV, where a good agreement with the reported number of Frenkel pair is observed. Our results provide information about the advantages and limits of the machine learned MD simulations with respect to the standard ones. The formation of dumbbell and crowdion defects as a function of PKA is discussed.

Machine learning driven simulated deposition of carbon films: From low-density to diamondlike amorphous carbon

“…Together with the previously evidenced high quality of the deposition simulations, i.e., the good agreement with experimental observables observed in initial work [24], this suggests that the carbon GAP is indeed able to capture the deposition process correctly. In this context, we mention the recently demonstrated usefulness of GAP simulations for radiation damage in elemental tungsten and silicon, where the impact of (very) highly energetic ions must be correctly described as well [41][42][43].…”

Computational study of crystal defect formation in Mo by a machine learning molecular dynamics potential

“…Point defects can be identified by comparing their DVs to those of the defect-free sample thermalized to 300 K. In order to show this new feature, we perform MD simulations at PKA energies of 10 keV and In figure 2(a), we present results for the quantification of point defects formed as a function of the simulation time at 1 (empty symbols) and 10 (solid symbols) keV of PKA with the GAP and EAM potentials. FaVAD is applied to identify and quantify the point defects with a distance difference threshold of d M = 0.6 for all cases, for which value is observed that Mo atoms have the highest probability to be considered as actual defects [11]. Although the profiles presented by the MD simulations at 10 keV are similar, the maximum number of displaced atoms is located at 0.8 and 1.1 ps for the GAP and EAM respectively.…”

“…In figure 3(b) we show results for the average number of dumbbells and crowdions as a function of the PKA energy. Here, the number of crowdions are quantified by identifying four Mo atoms sharing three lattice positions, while dumbbells are detected when two Mo atoms share one lattice position by FaVAD [11], with a d M (T ) > 0.2. We notice that the formation of crowdions is more favorable for collision cascades simulated by the GAP potential, while dumbbells are formed for all the PKA energy values with the EAM potential.…”

“…Transition metals and alloys are traditionally modeled by the embedded atom method (EAM) potentials in MD simulations [5,[8][9][10], reproducing many material properties in good agreement with those measured experimentally. However, EAM and other traditional potentials are limited to fixed functional forms [1,8] and can wrongly model some point defects that are energetically unstable, or lack physical meaning in material damaging processes [11]. For this reason, interatomic potentials developed by using machine learning (ML) methods are now increasingly used to perform MD simulations with an accuracy close to density functional theory (DFT) [12][13][14].…”

“…The goal of the present work is to numerically model the damage in crystalline Mo samples due to irradiation in a fusion reactor by utilizing the recently developed ML interatomic potential by Byggmästar et al [15]. Hence, we perform MD simulations to emulate neutron bombardment at intermediate primary knock-on atom (PKA) energies, providing an understanding about the modeling of the re-crystallization process after the collision cascade, which has been an issue for numerical simulations based on fixed functional forms [11,12].…”