Neural network molecular dynamics of ferroelectric domain boundary

Oxide perovskites like PbTiO₃ (PTO) are ferroelectric materials which exhibit spontaneous
polarization at ambient conditions. Formation of nanometer-size domains with opposite
polarizations separated by domain walls (DWs) essentially control the design and function of
ferroelectric devices. However, the underlying physics is not well understood and ab-initio
molecular dynamics (MD) simulation is computationally too costly to study DW structures.
Neural network MD (NNAIMD) is an emerging approach to study large-scale atomistic systems
with quantum mechanics accuracy. In this study, I will discuss the development of NNAIMD force
field to study PTO crystal, along with simulation results on complex DW dynamics using it.