Fig. 3: Qualitative differences between configurations in the training sets (TSs) of the here-developed MLIPs suitable only for the atomic scale ML-MD tensile tests (MLIP-[1]) or for both atomic and nanoscale ML-MD tensile tests (MLIP-[4]). | npj Computational Materials

Fig. 3: Qualitative differences between configurations in the training sets (TSs) of the here-developed MLIPs suitable only for the atomic scale ML-MD tensile tests (MLIP-[1]) or for both atomic and nanoscale ML-MD tensile tests (MLIP-[4]).

From: Machine-learning potentials for nanoscale simulations of tensile deformation and fracture in ceramics

Fig. 3

a Radial distribution function (RDF, with 5.5 Å cutoff) for B–B, Ti–B, and Ti–Ti bonds (integrated over all configurations). b Stress components (in-plane and in the loaded direction) vs. total energy of all configurations in the training set. c Snapshots of representative structures from the two training sets.

Back to article page