Three panels on a dark ground: an optical micrograph of the device with the FGT/O-FGT and CrPS4 regions arrowed across eight gold contacts; a greyscale cross-sectional electron micrograph showing three stacked bands labelled thick O-FGT, thin O-FGT and FGT with a 5 nm scale bar; and a schematic of exchange bias against temperature divided into three shaded regions marking the blocking temperatures of CrPS4, FeO and Fe3O4.

Research · spintronics

Surface oxidation as the dominant source of exchange bias

The rust layer everyone treats as damage turns out to run the magnetic coupling, and can reverse its direction.

The short version

Fe3GeTe2, one of the magnetic crystals in this stack, grows a thin oxide layer within half an hour of air exposure, and that layer is usually treated as damage. At these thicknesses an unintended interface layer can be a large part of the device, so what it does to the magnetic coupling between the two crystals had to be measured.

Three panels. Left: an optical micrograph of the device, a red and green flake stack sitting across eight yellow gold contacts on a dark blue substrate, with arrows labelling the FGT/O-FGT and CrPS4 regions. Centre: a greyscale cross-sectional electron micrograph with a 5 nm scale bar, showing three stacked bands labelled thick O-FGT, thin O-FGT and FGT. Right: a schematic plot of exchange bias against temperature divided into three shaded regions, with the blocking temperatures of CrPS4, FeO and Fe3O4 marked in sequence. Each magnetic phase in the stack stops pinning at its own blocking temperature, which is why the bias is non-monotonic in temperature. Adapted from ACS Nano 18, 8383–8391 (2024), © American Chemical Society.

How

  • Let the Fe3GeTe2 (FGT) oxidise on purpose and measured exchange bias across the FGT/CrPS4 stack against a pristine, hBN-capped control device.
  • Resolved the native oxide by cross-sectional electron microscopy with elemental spectroscopy (STEM-EELS) and X-ray photoelectron spectroscopy (XPS): two magnetically ordered sublayers, which explain the bias minima near 20 K and 70 K.
  • Set the sign with a ±1 T preset at fixed temperature, with no field cooling.
  • Stated the limits: the layer-by-layer assignment rests on a three-region spin model rather than depth-resolved magnetometry, and the effect is cryogenic.

Shared first authorship with Aravind Puthirath Balan. Published in ACS Nano 18, 8383–8391 (2024), 10.1021/acsnano.3c13034. The analysis code is public on GitHub as EB-in-FGT-OFGT-CrPS4-vdW-heterostructure.