Every atom in a magnet is like a tiny compass needle. These needles can point in different directions, and their collective dance determines the material's properties. But calculating this dance directly, atom by atom, is a task too heavy even for supercomputers.
The new model, named STEP, takes on the role of a conductor. It doesn't try to recalculate every move but learns to grasp the rules by which magnetic needles interact with the crystal scaffold (its lattice). STEP treats magnetic moments as continuous quantities and builds a special mathematical description for them. It is independent of the rotation of the whole system, just like a good compass always points north, no matter how you turn it.
STEP doesn't just repeat what it has memorized. It captures subtle magnetic interactions that previously eluded calculations. For instance, for the compound Fe₂Mo₃O₈, the model for the first time quantitatively described how magnetic waves mix with lattice vibrations to form hybrid particles — magnon polarons. Simulations of heating also allowed precise calculation of the Curie temperature (the point where a magnet loses its properties) for chromium and iron — the numbers matched experiments.
This approach saves computational resources and opens the door to designing new magnetic materials without endless trial and error. STEP is not just a fast calculator but a model with physical intuition that understands how atomic arrows obey each other and the surrounding crystal.
The Curie temperature for iron is 1043 Kelvin (about 770 °C). Above this point, iron ceases to be magnetic, though it remains solid.