Track synthesis, materials and catalysis as it is published.
Search papers, labs, and topics across Lattice.
This work develops DFT-trained machine-learning force-fields (ML-FF) to overcome the computational limitations of ab initio molecular dynamics for simulating the martensitic transformation in Ni$_2$MnGa. The ML-FF accurately reproduces key physical phenomena, including the soft TA$_2$ phonon and nanotwinned martensite formation. Classical molecular dynamics simulations using these ML-FF reveal that adaptive nanotwinning significantly facilitates the austenite-martensite transformation dynamics, with a negative twin boundary energy supporting ordered modulated martensites at finite temperatures.
Nanotwinned martensites in Ni$_2$MnGa are not just a structural preference but a dynamic facilitator of the austenite-martensite phase transition.
The martensitic transformations in stoichiometric Ni$_2$MnGa and related Heusler alloys are governed by a delicate interplay between electronic structure, magnetism, and lattice dynamics. While density functional theory (DFT) has successfully mapped the energy landscape of modulated martensites, it remains computationally prohibitive for capturing finite-temperature dynamics across relevant length and time scales. Here, we bridge this gap by constructing DFT-trained machine-learning force-fields (ML-FF) interpolated from ab initio molecular dynamics. We demonstrate that ML-FF trained on DFT data can reproduce key features of Ni$_2$MnGa, including the soft TA$_2$ phonon and a preference for nanotwinned martensites. Using classical molecular dynamics simulations, we explore the impact of adaptive nanotwinning on the austenite-martensite transition, which facilitates the dynamics of both, forward and backward transformation. We demonstrate, that the negative twin boundary energy along the entire transformations path predicted by DFT for $T=0$, supports the formation of ordered modulated martensites at finite temperatures. Our results pave the way for large-scale simulations of magnetic shape memory alloys with near-DFT accuracy.