Synergistic coupling in ab initio-machine learning simulations of dislocations

Review this paper

Submitted by jameskermode

Feb. 11, 2022, 11:25 a.m.

Synergistic coupling in ab initio-machine learning simulations of dislocations

Petr Grigorev, Alexandra M. Goryaeva, Mihai-Cosmin Marinica, James R. Kermode, Thomas D. Swinburnea

Brief Description
Ab initio simulations of dislocations are essential to build quantitative models of material strength, but the required system sizes are often at or beyond the limit of existing methods. Many important structures are thus missing in the training or validation of interatomic potentials, whilst studies of dislocation-defect interactions must mitigate the effect of strong periodic image interactions along the line direction. We show how these restrictions can be lifted through the use of linear machine learning potentials in hybrid simulations, where only a subset of atoms are governed by ab initio forces. The linear form is exploited in a constrained retraining procedure, qualitatively expanding the range of training structures for learning and giving precise matching of dislocation core structures, such that lines can cross the quantum/classical boundary.

We apply our method to fully three dimensional studies of impurity segregation to edge and screw dislocations in tungsten. Our retrained potentials give systematically improved accuracy to QM/ML reference data and the three dimensional geometry allows for long-range relaxations that qualitatively change impurity-induced core reconstructions compared to simulations using short periodic supercells. More generally, the ability to treat arbitrary sub-regions of large scale simulations with ab initio accuracy opens a vast range of previously inaccessible extended defects to quantitative investigation.

The code for refitting
potentials is written in Python and is provided in the dataset
accompanying the paper. The updated potentials can then be used with
Why should we reproduce your paper?
Systematically improvable machine learning potentials could have a significant impact on the range of properties that can be modelled, but the toolchain associated with using them presents a barrier to entry for new users. Attempting to reproduce some of our results will help us improve the accessibility of the approach.
What should reviewers focus on?
We suggest using the initial linear machine learning (LML) potential and QM datasets provided to reproduced the updated potentials and then to use them with LAMMPS to compute the 'Original LML' (blue plus signs) and 'Segregation retrained LML total energy' (orange crosses) vacancy-dislocation interaction energy results shown in Figure 2 in the paper. The QM/ML configurations are available from This includes retraining on dislocations as well as on the QM/ML vacancy segregation. There are also the files to generate constraints. Basically one needs to try to run the notebook. We would not suggest trying to reproduce the QM datasets themselves or the QM/ML results due to time constraints and code availability restrictions (we used VASP, which has a restricted usage license).


Associated event