OpenFerro: a unified framework for mesoscale ferroic dynamics and enhanced sampling

Overview & Synopsis

Modeling ferroic materials requires connecting microscopic fluctuations to collective dynamics over mesoscales. We introduce OpenFerro, a unified lattice-field dynamics framework that combines flexible Hamiltonians, automatic differentiation, parallel execution, and enhanced sampling. The same framework can simulate ferroelectric domain motion, magnetic disordering, and coupled spin-lattice thermodynamics. On four graphics processors, a bulk BaTiO$_3$ benchmark with long-range dipolar interaction reaches $840^3$ unit cells, representing 3 billion atoms through the retained lattice variables. We simulate polarization textures in a PbTiO$_3$/SrTiO$_3$ superlattice spanning $0.8~μ\mathrm{m}$ laterally and a BaTiO$_3$ nanoparticle 200~nm in diameter, retaining unit-cell resolution. We also demonstrate enhanced-sampling simulations that reconstruct the polarization free-energy surface of bulk BaTiO$_3$ without a prescribed polynomial form. With moderate computational resources, OpenFerro achieves atomistic-resolution simulations at mesoscopic scales traditionally studied with continuum models.