Towards 3D hydro-chemical models of AGB outflows with MACE

Modelling the chemistry of cool stellar environments in 3D hydrodynamical simulations is essential for capturing the complex interplay between chemistry and dynamics, especially in AGB outflows. However, directly solving detailed chemical networks in 3D simulations is computationally infeasible. To overcome this, we aim to couple MACE (Machine-learning Approach to Chemistry Emulation) with the smoothed particle hydrodynamics (SPH) code Phantom. We detail the architecture of MACE, optimised to enable fast and accurate chemical evolution and achieving an average 300× speed-up over classical methods. We show early results of post-processing simulations with MACE, and the prospects of a full coupling with 3D hydrodynamics solvers.