Interatomic potentials for irradiation-induced defects in iron

Abstract

Empirical potentials using embedded atom method are developed for Fe, mainly to study irradiation-induced defects such as self-interstitial atom clusters or dislocation loops. The potentials are fitted using experimental values of solid-state properties, ab initio formation energies of basic point defects and ab initio forces acting on the atoms in the liquid or random state configurations. Various bulk and defect properties are compared to validate the transferability of the new potential. In this paper, we also investigate the energetic landscape of C15 self-interstitial atom clusters. In order to simplify and to facilitate the construction of lowest energy configurations in the complex energy landscape of C15 clusters, we test and propose three selection rules.

Efficient and transferable machine learning potentials for the simulation of crystal defects in bcc Fe and W

“…Moreover, for larger clusters, GAP18 [93] predicts less stable C15 clusters. These results are not consistent with the DFT calculations [72,83]. The EAM AM04 potential poorly reproduces the energy landscape of C15, while MA07, is much closer to the DFT results [72,112,118].…”

“…These results are not consistent with the DFT calculations [72,83]. The EAM AM04 potential poorly reproduces the energy landscape of C15, while MA07, is much closer to the DFT results [72,112,118]. In contrast to the 2D loops, the tested GAP18 potential [93] exhibits a limited transferability for the C15 clusters and provides results close to those from AM04.…”

“…As in the case of vacancies, the characteristics of SIA clusters in Fe and W differ considerably from each other. For single SIAs, resistivity recovery experiments [69,107], DFT calculations [66,71,[108][109][110], EAM [72,77,81,[110][111][112][113], ML-GAP [93,94] and the present ML potentials consistently indicate that the I 1 dumbbells with 110 orientation in Fe and 111 orientation in W are the most stable. In W, the energy landscape of SIAs predicted by DFT calculations or accurate physicsinformed models, such as the discrete-continuum model (DC) [71], suggests that variants of 1 2 111 dislocation loops [67,114] are the lowest energy configurations at all sizes.…”

“…These clusters are known to be immobile [82]. Moreover, I Gao 2 represents an elementary building brick for complex 3D SIA clusters with C15 structure [72,83]. Until now, there are no empirical potentials that are able to predict the Gao configuration as SIAs with the lowest energy.…”

“…In this work we aim to design ML force fields that are suitable for modeling radiation-induced defects in Fe and W and allow performing large-scale calculations. The energy landscape of defects in bcc Fe and W is extremely complex [65][66][67][68][69][70][71][72][73] and its accurate description at the atomic scale requires using appropriate force-field models that provide a correct description of atomic systems beyond the equilibrium conditions. The new potentials should have higher accuracy than traditional potentials for the essential properties of defects and, in addition to that, correctly reproduce the peculiar behavior of some small defects known from ab initio calculations, such as negative binding energy of the di-vacancies in W [74][75][76][77][78][79][80], the distinct energy landscape of C15 interstitial clusters in Fe [68,72,[81][82][83], dislocation core structures as well as the shape and the magnitude of the Peierls barrier [84][85][86][87].…”