Key points are not available for this paper at this time.
Liquid metal embrittlement (LME) is a bottleneck for applications such as welding of galvanised steels in the automotive industry. The mechanistic origin of this phenomenon is not fully understood. In this work, we address some of the key LME hypotheses by developing and thoroughly validating a machine learning interatomic potential for the Fe-Zn binary system. The interatomic potential is suited for reproducing with good accuracy relevant benchmark properties, dislocation- and crack-related characteristics, as well as intermetallic phases of the Fe-Zn binary system. Based on the validated potential, we conduct a series of numerical fracture toughness tests (K-tests) on (110)1¯10 single crystals and Σ3, as well as Σ5 grain boundaries in Fe with varying Zn solute concentrations. The K-test results demonstrate the general applicability of the potential for studying complex fracture phenomena at the atomic scale. Furthermore, the conducted simulations suggest that for the prediction of LME, detailed analyses, including realistic Zn distributions, possible formation of intermetallic phases at the crack tip, and combination with microscale modelling, are required. We therefore discuss how machine learning interatomic potentials, like the one presented here, can be used in follow-up fracture studies to aid a more profound understanding of fracture and embrittlement phenomena in the Fe-Zn system.
Brunner et al. (Mon,) studied this question.