PHYSICAL REVIEW MATERIALS 5, 063607 (2021) Angular-dependent interatomic potential for large-scale atomistic simulation of iron: Development and comprehensive comparison with existing interatomic models Sergei Starikov , 1, * Daria Smirnova, 1, 2 Tapaswani Pradhan , 1 Yury Lysogorskiy, 1 Harry Chapman, 3 Matous Mrovec , 1 and Ralf Drautz 1 1 The Interdisciplinary Centre for Advanced Materials Simulation (ICAMS), Ruhr-Universität Bochum, 44801 Bochum, Germany 2 Joint Institute for High Temperatures of RAS, 125412 Moscow, Russia 3 Department of Materials, University of Oxford, Oxford OX1 3PH, United Kingdom (Received 26 March 2021; revised 29 May 2021; accepted 15 June 2021; published 30 June 2021) The development of classical interatomic potential for iron is a quite demanding task with a long history background. A new interatomic potential for simulation of iron was created with a focus on description of crystal defects properties. In contrast with previous studies, here the potential development was based on force-matching method that requires only ab initio data as reference values. To verify our model, we studied various features of body-centered-cubic iron including the properties of point defects (vacancy and self-interstitial atom), the Peierls energy barrier for dislocations (screw and mix types), and the formation energies of planar defects (surfaces, grain boundaries, and stacking fault). The verification also implies thorough comparison of a potential with 11 other interatomic potentials reported in literature. This potential correctly reproduces the largest number of iron characteristics which ensures its advantage and wider applicability range compared to the other considered classical potentials. Here application of the model is illustrated by estimation of self-diffusion coefficients and the calculation of fcc lattice properties at high temperature. DOI: 10.1103/PhysRevMaterials.5.063607 I. INTRODUCTION An accurate prediction of thermodynamic and mechanical properties of iron is of vital importance for materials engineer- ing as Fe-based alloys are widely used as structural materials. Both thermodynamic and mechanical properties are to a great extent governed by crystal defects (from point defects to dislocations and planar defects). The scale of these objects and understanding of their fundamental properties, such as the formation, migration, and interaction energies, demands employment of atomistic modeling techniques such as molec- ular dynamics (MD). However, the description of interatomic forces in Fe is highly challenging and there exists a broad range of models with different levels of sophistication [1–5]. The diversity of the observed structural and magnetic phases of Fe stems from the complexity of atoms interac- tions, namely, the mixed metallic and covalent bonding due to overlapping d orbitals and the itinerant magnetism [6]. A subtle mixture of these two phenomena gives rise to a complex dependence of the potential energy and local magnetic order on the atomic environment. Even a precise description of elastic moduli is a quite challenging task demanding special approaches in first-principle calculations [3,4]. The peculiar character of Fe bonding is also reflected in the properties of its crystal defects. For instance, the most stable configuration of self-interstitial atom (SIA) in the body-centered-cubic (bcc) iron (α-Fe) is the 〈110〉 dumbbell (SIA-D110) [5,7,8]. This contrasts to other nonmagnetic bcc * sergei.starikov@icams.rub.de metals, where SIA stabilizes in the 〈111〉 dumbbell (SIA- D111) orientation. Moreover, SIAs in iron tend to cluster together and form C15 phase inclusions, which is again atypical for other bcc metals [9–11]. Another outstanding property of iron is a strong dependence of vacancy forma- tion and migration energies on temperature, which leads to a marked non-Arrhenius behavior of the self-diffusion co- efficient [12–14]. A proper modeling of this phenomenon requires to take into account both the vibrational and magnetic degrees of freedom, for instance, by coupling molecular and spin dynamics (SD) [15–17]. It is important to note that such coupled simulations can be successful only when the atom- istic and spin models describe accurately the atomic and spin interactions, respectively. There exist a number of classical interatomic potentials for atomistic simulation of iron [1,18–29]. Mostly, they have been developed by fitting the potential functions to reproduce fun- damental properties of Fe bulk phases (e.g., lattice parameters, elastic constants, or thermal expansion) and, in some cases, also properties of simple defects (e.g., formation energies of point defects, surfaces, or grain boundaries). However, these simplified models, which usually do not contain any explicit treatment of magnetism, suffer from a limited transferability and exhibit various deficiencies. Some of the well-known problems are an underestimation and incorrect shape of the Peierls energy barrier for the 1 2 〈111〉 screw dislocation, under- estimated surface energies, or spurious phase transformations [7,30]. Some of the deficiencies disappear when more sophis- ticated models are used. For instance, an improvement in dislocation properties was achieved by the magnetic 2475-9953/2021/5(6)/063607(23) 063607-1 ©2021 American Physical Society