Phase discovery with active learning: Application to structural phase transitions in equiatomic NiTi
Jonathan Vandermause, Anders Johansson, Yucong Miao, Joost J. Vlassak, Boris Kozinsky
TL;DR
The paper addresses displacive phase transitions in equiatomic NiTi by developing four on-the-fly trained, machine-learned force fields based on $\text{LDA}$, $\text{PBE}$, $\text{PBEsol}$, and $\text{SCAN}$ DFT functionals. An automated active-learning protocol drives MD simulations and updates the models with DFT data whenever local-energy uncertainty crosses a threshold, providing ab initio-accurate sampling at a fraction of the cost. Across functionals, SCAN uniquely captures the reversible $B2$ to $B19'$ martensitic transition, while $\text{LDA}$, $\text{PBE}$, and $\text{PBEsol}$ predict a distinct low-volume phase $M2$ upon cooling; large-scale MD confirms these pathways and reveals a pressure-stabilized $M2$ region, suggesting a phase diagram where $M2$ is distinct from $B19'$. Overall, the work demonstrates a scalable framework to study complex displacive transformations with high fidelity, and highlights meaningful functional-dependence in NiTi that can only be unveiled through large-scale, data-driven simulations.
Abstract
Nickel titanium (NiTi) is a protypical shape-memory alloy used in a range of biomedical and engineering devices, but direct molecular dynamics simulations of the martensitic B19' -> B2 phase transition driving its shape-memory behavior are rare and have relied on classical force fields with limited accuracy. Here, we train four machine-learned force fields for equiatomic NiTi based on the LDA, PBE, PBEsol, and SCAN DFT functionals. The models are trained on the fly during NPT molecular dynamics, with DFT calculations and model updates performed automatically whenever the uncertainty of a local energy prediction exceeds a chosen threshold. The models achieve accuracies of 1-2 meV/atom during training and are shown to closely track DFT predictions of B2 and B19' elastic constants and phonon frequencies. Surprisingly, in large-scale molecular dynamics simulations, only the SCAN model predicts a reversible B19' -> B2 phase transition, with the LDA, PBE, and PBEsol models predicting a reversible transition to a previously uncharacterized low-volume phase, which we hypothesize to be a new stable high-pressure phase. We examine the structure of the new phase and estimate its stability on the temperature-pressure phase diagram. This work establishes an automated active learning protocol for studying displacive transformations, reveals important differences between DFT functionals that can only be detected in large-scale simulations, provides an accurate force field for NiTi, and identifies a new phase.
