Item

Abstract

In the present work, we have evaluated the performance of different embedded atom method (EAM) and second-nearest neighbour modified embedded atom method (2NN-MEAM) potentials based on their predictive capabilities for modelling fracture in single-and bicrystalline tungsten. As part of the study, a new 2NN-MEAM was fitted with emphasis on reproducing surface, unstable stacking fault and twinning energies as derived from density functional theory (DFT) modelling. The investigation showed a systematic underestimation of surface energies by most EAM potentials, and a significant variation in unstable stacking and twinning fault energies. Moreover, the EAM potentials in general lack the ability to reproduce the DFT traction-separation (TS) curves. The shorter interaction length and higher peak stress of the EAM TS curves compared to the 2NN-MEAM and DFT TS curves result in one order of magnitude higher lattice trapping than for cracks studied with 2NN-MEAM. These differences in lattice trapping can lead to significant qualitative differences in the fracture behaviour. Overall, the new 2NN-MEAM potential best reproduced fracture-relevant material properties and its results were consistent with fracture experiments. Finally, the results of fracture simulations were compared with analytical predictions based on Griffith and Rice theories, for which emerging discrepancies were discussed.