General Learning of the Electric Response of Inorganic Materials

We present MACE-Field, a field-aware $O(3)$-equivariant interatomic potential that provides a compact, derivative-consistent route to dielectric properties (such as polarisation $\mathbf P$, Born effective charges $Z^*$ and polarisability $\boldsymbolα$) and finite-field simulations across chemistry for inorganic solids. A uniform electric field is injected within each message-passing layer via a Clebsch-Gordan tensor-product which couples the field to latent spherical-tensor features, and perturbs them via an equivariant residual mixing. This plug-in design preserves the standard MACE readout and can inherit existing MACE foundation weights, turning pretrained models into field-aware ones with minimal change. To demonstrate, we train: (i) a cross-chemistry ferroelectric polarisation model (2.5k nonpolar$\!\to$polar polarisation branches), (ii) a cross-chemistry BECs/polarisability model ($\sim$6k Materials Project dielectrics spanning 81 elements), and (iii-iv) single-material molecular dynamics on BaTiO$_3$ and $α$-SiO$_2$. The models recover polarisation branches and spontaneous polarisation, predict $Z^*$ and $\boldsymbolα$ (hence $\varepsilon_\infty$) across diverse chemistries, and reproduce BaTiO$_3$ hysteresis and the IR/Raman and dielectric spectra of $α$-quartz, benchmarking comparatively with Allegro-pol.