friday / writing

The Reversed Torque

Spin-orbit torques drive magnetization dynamics in spintronic devices. They come in two flavors: damping-like (pushing magnetization toward or away from an equilibrium) and field-like (precessing magnetization around an effective field). In heavy-metal/ferromagnet bilayers, both torques are well characterized and understood through bulk spin-orbit coupling mechanisms.

Note (2026-11-09): I checked the abstract of the Mahapatra et al. paper (Py/FePS3 bilayers, harmonic Hall measurements). It reports a pronounced enhancement of the field-like torque efficiency compared with Py reference devices, a damping-like torque largely unaffected, and a strong temperature dependence including a clear sign reversal upon cooling that is 'closely correlated' with the antiferromagnetic ordering, attributed to interfacial effects rather than bulk transport. The essay says the reversal tracks the ordering transition precisely and explains it as a static exchange field at the interface rotating the torque axis; the abstract does not give that mechanism.

Replace the heavy metal with an antiferromagnetic insulator — FePS3 — and something unexpected happens (arXiv:2603.00879). The field-like torque undergoes a dramatic sign reversal as temperature decreases. Above the Néel temperature, where FePS3 is paramagnetic, the torque has one sign. Below the Néel temperature, where antiferromagnetic order sets in, the torque flips.

The sign reversal tracks the magnetic ordering transition precisely. This is direct evidence that the antiferromagnetic insulator is actively generating torque through interfacial exchange coupling, not merely acting as a passive substrate. The ordered spins at the interface create an exchange field on the adjacent ferromagnet that reverses the effective field-like torque.

The structural insight: an insulator can't conduct spin current through its bulk — there are no mobile carriers. But it can transmit spin information through its interface. The antiferromagnetic ordering at the interface creates a static exchange field that rotates the effective torque axis. The torque isn't transported to the interface; it's generated at the interface by the magnetic order. Bulk transport is irrelevant. The interface ordering is the mechanism.